The Syntax of Surprise: Expletive Negation and the Left Periphery [1 ed.] 1527540928, 9781527540927

Citation preview

The Syntax of Surprise

The Syntax of Surprise: Expletive Negation and the Left Periphery By

Matteo Greco

The Syntax of Surprise: Expletive Negation and the Left Periphery By Matteo Greco This book first published 2020 Cambridge Scholars Publishing Lady Stephenson Library, Newcastle upon Tyne, NE6 2PA, UK British Library Cataloguing in Publication Data A catalogue record for this book is available from the British Library Copyright © 2020 by Matteo Greco All rights for this book reserved. No part of this book may be reproduced, stored in a retrieval system, or transmitted, in any form or by any means, electronic, mechanical, photocopying, recording or otherwise, without the prior permission of the copyright owner. ISBN (10): 1-5275-4092-8 ISBN (13): 978-1-5275-4092-7

CONTENTS

Acknowledgments .................................................................................... vii Introduction ................................................................................................ 1 Chapter 1 .................................................................................................. 12 Expletive negation: the case study of Surprise Negation Sentences (Snegs) in Italian 1.1 Standard Negation: the syntactic structure 1.1.1 On the negation head status 1.1.2 On the negation maximal projection status 1.2 Italian negation structures 1.2.1 On the syntactic nature of the negative morpheme “non” (not) 1.3 Expletive Negation (EN): a general discussion 1.3.1 Some previous analyses of EN 1.4 Focusing on the Italian EN clauses 1.5 Italian EN is not a uniform phenomenon 1.6 An unnoticed case of Italian EN: Surprise Negation Sentences Chapter 2 .................................................................................................. 66 Snegs from a descriptive point of view 2.1 Snegs is an instance of the strong EN 2.2 Some other features of Snegs 2.3 Diagnostics for Snegs 2.4 Snegs and the puzzling incompatibilities Chapter 3 .................................................................................................. 85 What Snegs are not: a justification of a dedicated structure 3.1 Snegs are not Negative Rhetorical Questions 3.2 Snegs are not Expletive Negative Exclamatives 3.3. Snegs are not Bi-clausal Structures

Contents

vi

Chapter 4 .................................................................................................. 99 A preliminary step: a view of the Left Periphery 4.1 The syntactic status of the left periphery-CP field 4.1.1 The Complementizer phrase and the Cartographic Approach 4.1.2 The Complementizer phrase and The Minimalist Program 4.2 An overview of Topicalized structures 4.3 An overview of Focalized structures 4.4 The left periphery in other XPs Chapter 5 ................................................................................................ 134 Snegs and the Left Periphery: toward a syntactic hypothesis 5.1 A crucial observation 5.2 CP and neg features 5.3 A syntactic proposal for Snegs 5.4 Advantages and consequences of the proposed analysis 5.5 A comparative support in favor of the special status of expletive negation Chapter 6 ................................................................................................ 179 An exploratory psycholinguistic approach to Snegs 6.1 Introduction 6.2 The present study 6.3 Method 6.4 Results 6.5 General Discussion Concluding Remarks .............................................................................. 202 References .............................................................................................. 212

ACKNOWLEDGMENTS

I’m very grateful to my Ph.D. supervisor, Prof. Andrea Moro, who showed me the amazing world of Linguistics for the first time. He taught me the fundamental spirit of this discipline and shared with me his curiosity and passion. Special thanks also to Prof. Valentina Bambini, who has patiently illustrated experimental aspects of linguistic studies to me, and to Prof. Cristiano Chesi and to Dr. Paolo Canal for their precious support in understanding the computational and psycholinguistic aspects of the present research. I would also like to thank all the members of the Center for Neurocognition, Epistemology and theoretical Syntax (NEtS) of the University School for Advanced Studies of Pavia, who provided me with the necessary scientific and cultural environment in which my ideas could grow. Many other people have provided their scientific and technical contributions to the realization of this book. I’m very grateful to Prof. Raffaella Zanuttini and Prof. Robert Frank, who hosted me as Visiting Assistant in Research in their department at Yale University. They have shared with me their deepest linguistics knowledge as well as their everyday lives. I would also like to thank Prof. Denis Delfitto for his patient attention to my studies, and for his advice. Many thanks also to Prof. Alessandra Tomaselli and Prof. Luigi Rizzi for the precious discussions they had with me. A special thanks to Prof. Cecilia Poletto for the support she always gave me. Last, but not least, I thank Cindy Farlino, Marta Frapoli, Roberto Lobrace and Laura Schofield for their great help. Of course, I’m fully responsible for the content of this book and any mistakes are only mine. The ideas expressed here only partially overlap with those developed in my doctoral thesis; substantial changes differ this work from the previous one, and new parts have been added. Finally, I want to thank my friends and my family for their un-silent and endless support both for this book and, especially, for my life. My last thought–and needless to say also my first–is for Martina, Tommaso, Chiara and Lucia: you all prevented me from believing that the last years of my life

viii

Acknowledgments

were meant solely for working on this book and other scientific issues. You remind me that I’m destined for happiness in both this life and in the next (you also prevent me from sleeping during the night, but this “fa parte del gioco” [is part of the game]).

INTRODUCTION

Negation is a linguistic universal (Horn 1989; Bernini and Ramat 1996; Horn and Kato 2000; Speranza and Horn 2012). From a semantic point of view, negation is a one-place operator reversing the truth-value conditions of the sentence in which it occurs. Consider, for example, the following sentences: (1) a. Luca is smart b. Luca is not smart The sentence in (1b) is true if and only if the sentence in (1a) is false, and vice versa. However, even if the logical meaning of negation is clear in this simple pair, its contribution to the natural languages is not always so clear and transparent (cfr. Taglicht 1984). According, for example, to Jespersen (1924) and Horn (2010), an individual does not often interpret linguistic negation as a pure logic operator. To illustrate this, consider the sentences in (2): (2) a. Rimarrò alla festa finché arriva Gianni stay.1stsing.fut to the party as long as arrives Gianni ‘I will stay at the party until Gianni arrives’ b. Rimarrò alla festa finché non arriva Gianni stay.1stsing.fut to the party as long as neg arrives Gianni ‘I will stay at the party until Gianni arrives’ The sentences in (2a-b) minimally differ in the occurrence of the negative marker non in the temporal clause. However, their meaning does not change (as the English translation shows): the arrival of Gianni is the event that marks the last moment in which the speaker stays at the party. Whereas the form in (2a) is unproblematic because its affirmative polarity matches the truth-value conditions of the proposition, the one in (2b) is not. In fact, the occurrence of negation should change its meaning (like in 1b) but that is not the case: the temporal clause is still affirmative regardless of the occurrence of the negative marker non. This particular use of negation is generally called expletive or pleonastic negation (cfr. Jespersen 1917), since the negative operator does not affect the propositional meaning of the sentence.

2

Introduction

The fact that an apparently easy true-operator shows an opposite value to its nature is undoubtedly puzzling. The main aim of this work is to shed light on expletive negation (EN) (or, at least, on a piece of it). In the past fifty years, there has been a great effort in this field to identify the pragmatic, semantic and syntactic conditions governing the use of expletive negation (cfr. Horn 2010 and the references therein). In this book I want to focus on an unnoticed case (to the best of my knowledge) in Italian, which shows some unique features possibly assisting us in understanding this phenomenon. Consider, for instance, the following sentence: (3) E non mi è scesa dal treno Maria?! and neg Cl.to me is got off-the train Maria ‘Maria got off the train!’ The meaning of the sentence (3) is affirmative regardless of the occurrence of the negative element non and, therefore, it has to be considered an expletive negation clause. Pragmatically, (3) conveys a strong sense of surprise since the speaker is struck by an unexpected fact and s/he wants to communicate it (in our case, the fact that Maria got off the train). Starting from this particular pragmatic requirement, I will call such structures Surprise Negation Sentences (Snegs). Before discussing the other features of Snegs, I want to introduce a general fact affecting the Italian negation. Italian displays the same negative marker non both for standard and for expletive negation as the following sentences show: (4) a. Non è scesa dal treno Maria neg is got off-the train Maria ‘Maria did not get off the train’ b. Non è scesa dal treno Maria?! neg is got off-the train Maria ‘Maria got off the train!’

(Standard Negation) (Sneg)

In (4a), non realizes a case of standard negation (SN); in (4b), it realizes a case of expletive negation. Crucially, Snegs display several peculiar features forcing the expletive interpretation of the negative marker:

The Syntax of Surprise: Expletive Negation and the Left Periphery

3

i. Snegs show a marked intonation blending acoustic features pertaining to both questions and exclamatives (hence the ?! combined diacritic). Such an intonation specifically selects the Sneg reading, ruling out the standard one1: (5) a. Non è scesa dal treno Maria?! neg is got off-the train Maria .‘ #Maria did not get off the train’ .‘Maria got off the train!’

(#SN) (Sneg)

This particular intonation is strongly and uniquely associated with Snegs; other kinds of market prosodies such as, for example, the simple interrogative (6a) and exclamative (6b), fail to select Snegs.: (6) a. Non è scesa dal treno Maria? neg is got off-the train Maria .‘Did not Maria get off the train?’ .‘#Maria got off the train!’ b. Non è scesa dal treno Maria! neg is got off-the train Maria .‘Maria did not get off the train!’ .‘#Maria got off the train!’

(SN) (#Sneg) (SN) (#Sneg)

ii. Another element specifically associated with Snegs is the Ethical Dative (ED). ED is a non-argumental clitic picking out a person, either the speaker or the hearer, who emotionally participates in the event expressed by the sentence. Italian displays two forms of ED: mi (to me) and ti (to you). Consider the following sentences: (7) a. Laura ha vinto il primo premio alla lotteria Laura has won the first prize to the lottery ‘Laura won the first prize in the lottery’ b. Laura mi/ti ha vinto il primo premio alla lotteria Laura Cl.to me/you has won the first prize to the lottery ‘Laura won the first prize in the lottery’ The sentences in (7a-b) only differ for the occurrence of the ED mi and ti; nevertheless, their propositional meaning is the same. However, the 1 I will indicate the unavailability of one of the two readings of negation–i.e. expletive or standard–by means of the special diacritic #

4

Introduction

sentence in (7b) displays a different pragmatic nuance: the speaker, in case of mi, or the hearer, in case of ti, is particularly involved in the described fact. Crucially, something unexpected happens when the ED co-occurs with the negative marker: only the expletive reading is available (8a), whereas the standard one is ruled out (8b): (8) a. Non mi/ti è scesa dal treno Maria?! neg Cl.to me/you is got off-the train Maria ‘Maria got off the train!’ b. *Non mi/ti è scesa dal treno Maria neg Cl.to me/you is got off-the train Maria

(Sneg) (*SN)

This particular relation with the ED represents one of the most characteristic properties of Snegs. As seen so far, at least three elements define Snegs: the unexpectedness of the fact described by them, the particular intonation, and the co-occurrence with the ED. I will show that these features also distinguish Snegs from other types of expletive negation structures, such as negative exclamatives and negative rhetorical questions, making Snegs a unique instance of EN. I want now to investigate three unrelated phenomena in which the distribution of Snegs appears puzzling. I will discuss them in a rather informal way, leaving a more detailed analysis to the next chapters2. I. It is well known that Romance languages syntactically mark discourse-related phenomena, such as topicalization and focalization, by displaying a different word order than the standard. More specifically, languages such as Italian dislocate focalized and topicalized phrases in a fronted position, anticipating other elements. Consider, for example, the case of the direct object. The standard word order of an Italian assertive sentence is subject-verb-direct object- like in (9a); when the direct object is focalized or topicalized, it occurs in the fronted position, giving the scrambled order direct object- subject-verb (9bc).

2

This is also the reason why I will not indicate bibliographic voices.

The Syntax of Surprise: Expletive Negation and the Left Periphery

5

(9) a. Gianni ha dato il libro a Luca (Assertive clause) Gianni has given the book to Luca ‘Gianni gave the book to Luca’ b. LO ZAINO Gianni ha dato t a Luca (non il libro)3 (Foc.) the backpack Gianni has given to Luca (not the book) ‘THE BACKPACK Gianni gave to Luca (not the book)’ c. Il libro Gianni lo ha dato t a Luca (Topic.) the book Gianni Cl.it has given to Luca ‘The book, Gianni gave it to Luca’ To take into account the fronted position of the focalized and topicalized elements in (9b-c), it has been proposed that the syntactic structure of a sentence displays some dedicated functional heads in the Complementizer Phrases (CP): the Foc° and the Top°4. Elements like LO ZAINO/Il libro are generated in the thematic position, as any other direct object; they then move to those heads in order to take on a specific interpretative value appearing in the fronted position. Coming to Snegs, they show a puzzling asymmetry: they only allow fronted topicalized phrases by rejecting the focalized ones. Such an asymmetry appears evident when comparing Snegs with SN sentences: both allow a fronted topicalized phrase (10), but only the latter accepts the focalized one (11). (10) a. Il libro Gianni non lo ha dato a Luca (SN) the book Gianni neg Cl.it has given to Luca ‘The book, Gianni did not give to Luca’ b. E il libro Gianni non me lo ha dato a Luca?! (Sneg) and the book Gianni neg ED Cl.it has given to Luca ‘The book, Gianni gave to Luca!’ Context (i): ‘Maria did not get off the train’ (11) a. GIANNI non è sceso dal treno (non Maria)?! Gianni not is got off-the train (not Maria) ‘Gianni did not get off the train! (not Maria)’

3

(SN)

I use the uppercase to highlight the focalized interpretation of a phrase and the notation ti to indicate the position in which the fronted phrases LO ZAINO/il libro have been generated before being moved to the CP domain. 4 A more detailed discussion on the dislocation of elements in fronted positions is not crucial at the moment and I postpone it to the body of thebook.

6

Introduction

Context (ii): ‘Maria got off the train’ (11)b. *E GIANNI non mi è sceso dal treno (non Maria)?! (*Sneg) and Gianni EN ED is got off-the train (not Maria) The reason why Snegs do not allow fronted focalized elements does not have an immediate and intuitive explanation. Since the structure of CP plays a crucial role in the distribution of discourse-related phenomena, its relation with Snegs is the probable key to understanding the asymmetry in (10-11). II. Another particular fact concerning Snegs is their relation to the polarity-sensitive elements usually associated with standard negation. As is well known, languages display some grammatical elements licensed only by negation: if negation is absent, their occurrence becomes ungrammatical. Consider, for example, the distribution of the Italian adverb affatto (at all) in negative and affirmative sentences: (12) a. Non è affatto scesa dal treno Maria neg is at all got off-the train Maria ‘Maria did not get off the train at all’ b. *È affatto scesa dal treno Maria is at all got off-the train Maria

(Neg.) (*Aff.)

The sentence in (12a) hosts the adverb affatto because of the negative marker non; in (12b), the negative marker is absent and the sentence becomes ungrammatical. Like the adverb affatto, several other elements are uniquely licensed by negation; consider, among others, the conjunction neanche (not also) in (13) and the negative pronouns nessuno (nobody) in (14)5: (13) a. Non è scesa dal treno Maria e neanche Gianni neg is got off-the train Maria and neither Gianni ‘Maria did not get off the train and Gianni either’ b. *È scesa dal treno Maria e neanche Gianni is got off-the train Maria and neither Gianni

5

(Neg.) (*Aff.)

The pronoun nessuno can also appear in affirmative sentences when it realizes a subject or when it occurs both in questions and in protasis of conditionals. I specifically chose an example in which the clause is assertive and the pronoun realizes the direct object of the main verb; in such a contest nessuno requires the cooccurrence with the negative marker non to be correctly licensed.

The Syntax of Surprise: Expletive Negation and the Left Periphery

(14) a. Non è sceso dal treno nessuno neg is got off-the train nobody ‘Nobody got off the train’ b.*È sceso dal treno nessuno is got off-the train nobody

7

(Neg.) (*Aff.)

To explain the pattern in (12-13-14), it has been proposed that the negative polarity-sensitive elements have to be in the scope of the negative marker non, syntactically represented as the head of NegP dominating TP. When it is absent, those polarity-sensitive elements cannot be licensed. However, just the occurrence of negation is not sufficient to legitimate them. Consider the following cases: (15) a. *Non è bello che sia affatto scesa dal treno Maria neg is nice that be.3rdsing.subj at all got off-the train Maria b. *Non è bello che sia scesa dal treno Maria neg is nice that be.3rdsing.subj got off-the train Maria e neanche Gianni and neither Gianni c. *Non è bello che sia sceso dal treno nessuno neg is nice that be.3rdsing.subj got off-the train nobody The sentences in (15) display the negative marker non; even so, they are ungrammatical. Since the negative marker in those sentences occurs in the root clause and the polarity-sensitive elements in the embedded one, it has been proposed that the scope-relation is interrupted by the edge of the clause. Graphically, this fact can be visualized by using the following representations: (16) a. [CP1 [NegP non [TP …. affatto/nessuno/neanche… ]]] b. *[CP1 [NegP non … [CP2 [TP …. affatto/nessuno/neanche… ]]]] The negative marker in (16a) takes TP as its argument and it scopes over it, licensing the polarity-sensitive elements as in (12a-13a-14a); in (16b), the negative marker cannot scope over the embedded TP because the CP2 represents a barrier (or a phase putting in minimalist words) as in (15a-b-c). When these polarity-sensitive elements occur in Snegs, they unexpectedly still raise ungrammaticality, thus differing Snegs from standard negative sentences:

8

Introduction

(17) a. Non è affatto scesa dal treno Maria (SN) neg is at all got off-the train Maria ‘Maria did not get off the train at all’ b.*E non mi è affatto scesa dal treno Maria?! (*Sneg) and EN ED.to me is at all got off-the train Maria (18) a. Non è scesa dal treno Maria e neanche Gianni (SN) neg is got off-the train Maria and neither Gianni ‘Maria did not get off the train and Gianni either’ b.*E non mi è scesa dal treno Maria e neanche and EN ED.to me is got off-the train Maria and neither Gianni?! (*Sneg) Gianni (19) a. Non è sceso dal treno nessuno neg is got off-the train nobody ‘Nobody got off the train’ b.*E non mi è sceso dal treno nessuno?! and neg ED.to me is got off -the train nobody

(SN) (*Sneg)

Apparently, there is no reason explaining this pattern, since the negative marker and the polarity-sensitive elements belong to the same clause, both in the standard negation sentences (17a-18a-19a) and in Snegs (17a-18b19b). Neither can this effect only be related to the expletive nature of the negative marker, since some of the polarity-sensitive elements are perfectly fine when they occur in other EN structures. Consider, for example, the cases of the negative pronoun nobody in until-clauses: (20) Rimarrò alla festa finché non arriva nessuno stay.1stsing.fut to-the party as long as neg arrives nobody ‘I will stay at the party until somebody arrives’ The sentence in (20) is perfectly grammatical even though negation is expletive. Again, the reason why Snegs depart from this pattern does not find an immediate explanation. Certainly, it cannot only be a semantic issue, since both standard and expletive negations can legitimate such elements. III. Finally, there is another aspect making Snegs a peculiar case of EN: they do not have any affirmative correspondent form. More specifically, many instances of EN clauses display a correspondent affirmative form–as I discussed in (2) (repeated here as 21)–and the

The Syntax of Surprise: Expletive Negation and the Left Periphery

9

occurrence of the negative marker is optional from the propositional point of view. (21) a. Rimarrò alla festa finché arriva Gianni stay.1stsing.fut to-the party as long as arrives Gianni ‘I will stay at the party until Gianni arrives’ b. Rimarrò alla festa finché non arriva Gianni stay.1stsing.fut to the party as long as neg arrives Gianni ‘I will stay at the party until Gianni arrives’ Crucially, Snegs depart from this pattern because they are strictly related to the negative marker non. In fact, there is not an affirmative correspondent form for Snegs displaying all the pragmatic, syntactic and prosodic features usually associated with them: (22) a. E non mi è scesa dal treno Maria?! and neg Cl.to me is got off-the train Maria ‘Maria got off the train!’ b. *E mi è scesa dal treno Maria?! and Cl.to me is got off-the train Maria Snegs, therefore, constitute a paradox: the negative marker non does not inverse the polarity of the sentence but, at the same time, its absence leads to the exclusion of the Sneg reading. In summarizing, so far, I showed three unrelated facts that do not allow an immediate and intuitive explanation: the asymmetry in licensing topicalized and focalized elements (section I); the ungrammaticality of polarity-sensitive elements (section II); the obligatory nature of the negative marker non (section III). The aim of this book is to give an analysis of Snegs from which all those properties, and many others, are derived in a unitary way. In order to do so6, I will propose to change a basic assumption on the syntactic status of the negative marker non. I will propose that the head non, generally assumed to be merged in TP area, can also be externally merged in the CP-domain. Such a syntactic configuration does not allow negation interacting with the predicative core of the sentence, recreating a similar situation to (16b) in which the scope of negation is blocked by a barrier (or 6

The following discussion can only be brief and incomplete because of its introductory nature. See Chapter 5 for a more detailed analysis.

10

Introduction

a phase). I will show that this specific change in syntactic theory will correctly predict both the obligatory and expletive nature of negation and its impossible co-occurrence with polarity-sensitive elements in Snegs. As a consequence of this syntactic hypothesis, I will also argue that Snegs display the whole TP in the focalized position, not allowing other focalized elements. From this fact follows the asymmetry between the topicalized and focalized phrases (as seen above). Even though my proposal pertains to the syntax domain, it can also be used to shed light on the on-line processing of Snegs. In fact, I will suggest that the introduction of negation in the CP domain is the cause of its affirmative polarity. Such a syntactic representation raises a clear-cut prediction on the elaboration of Snegs: they reproduce the processing of affirmative clauses rather than that of negative clauses regardless of the occurrence of the negative marker non. To investigate this idea, I built a psycho-linguistic experiment using the eye-tracker methodology in a visual-world paradigm. The selection of the eye-tracker is related to its very high temporal accuracy, which is essential in the study of negative sentences. In fact, it is traditionally assumed that negative sentences require two mental representations, temporally subsequent, to be elaborated. Consider, for example, the negative sentence “The window is not open”. According to this, a subject first would elaborate the mental representation of the window open (contrafactual states of affairs) and then, in a successive moment, the mental representation of the closed window (the negative meaning of the sentence). These two steps are not required for affirmative sentences that need just one mental representation equivalent to the meaning of the sentence. My prediction was that, like affirmative clauses, Snegs require only one mental representation. I consequently examined Snegs in comparison to affirmative and negative sentences. Thirty-four subjects participated in the experiment, during which they were seated in a quiet room in front of a computer screen. A narrative voice told them a short real-life story (e.g., Laura invited some friends to her home. When they arrived, she showed her domestic animal to them) while 4 pictures were projected on the screen: two related to the story they were listening to (e.g., a dog and a snake), and two which were not (e.g., a backpack and an air-conditioner). One of the two related pictures represented an object highly expected in the story (e.g., the dog) because of its prototypic value compared to the described situation, and one highly unexpected (e.g. the snake), because of its non-prototypic nature. Each story ended with a target sentence, which might be affirmative (e.g., La ragazza

The Syntax of Surprise: Expletive Negation and the Left Periphery

11

ha mostrato un serpent, translated as ‘the girl showed a snake’), negative (e.g., La ragazza non ha mostrato un serpente, translated as ‘The girl did not show a snake’) or Sneg (e.g., La ragazza non ha mostrato un serpente?! translated as ‘The girl showed a snake!’). Throughout the whole experiment I recorded the participants’ eye movements. Just a small road map. I will start this book with a general discussion on the syntax of standard and expletive negation (Ch. 1). I will then introduce the case of Surprise Negation Sentences from a descriptive point of view (Ch. 2). Since Snegs share some semantic, syntactic and prosodic features with other kinds of clauses, such as exclamatives and negative rhetorical questions, I will discuss their possible common membership (Ch. 3). This hypothesis will be rejected because Snegs display certain different and innovative elements that exclamatives and negative rhetorical questions do not. Pursuing the reasonable alternative that Snegs form a distinct class of EN clause, I will discuss the structure of the CP, from both cartographic and minimalist points of view, starting from their asymmetrical interaction with focalized and topicalized phrases (Ch. 4). After these preliminary sections, I will introduce the syntactic analysis of Snegs (Ch. 5) discussing the consequences and the advantages of such a proposal. I will also highlight that Italian displays other structures that share some elements with Snegs (like the focalization of TP) and, moreover, I will discuss some comparative observations further confirming my analysis. I will dedicate the last chapter (Ch. 6) to the experimental aspects of this book, showing that the syntactic configuration proposed for Snegs is also compatible with psycho-linguistic and cognitive data obtained by an eye-movements experiment.

CHAPTER 1 EXPLETIVE NEGATION: THE CASE STUDY OF SURPRISE NEGATION SENTENCES (SNEGS) IN ITALIAN

In this chapter I am going to discuss the distinction between standard and expletive negation (EN). Consider, for example, the following sentence: (1) Che cosa non ha fatto Gianni (!) what not has done Gianni a. ‘What has not done Gianni’ b. ‘What has done Gianni!’ As the English translations show, the sentence in (1) can be interpreted as a case of standard negation (1a), when it is uttered with an assertive prosody, or as a case of expletive negation (1b), when it is uttered with an exclamative prosody. The tension between these two readings is not only pragmatic, but involves the syntactic structure in which they occur. In fact, the constant feature among several different strategies to realize a negative sentence7 is an extra negative morpheme (absent in the correspondent affirmative form) displaying its own syntactic physiognomy. I will focus on the implications of such a fact. More specifically, I will summarize the rich debate around the syntax of negative elements in order to obtain the essential knowledge to analyze Snegs. Since Italian negative sentences are the core subject of this study, I will consider them more carefully in this chapter.

7

I choose to neglect the constituent negation, because, as Klima (1964) points out, it follows a different computational procedure than the standard one. See De Clercq (2013) for a unifying approach to standard negation and the constituent one.

Expletive negation

13

1.1 Standard Negation: the syntactic structure Languages display different strategies to express standard negation. This range of variation is limited and has been used to investigate the structure of the Universal Grammar underlying linguistic negation (cfr. Zanuttini 2001). According to Payne (1985), there are 4 strategies8: a) A negative marker carrying auxiliary features (like person, number, aspect and tense) is added to a non-finite participial form. This strategy emerges in some Siberian dialects like Evenki in (2): (2) Bi ˜-˜-w dukuwun-ma duku-ra I neg-PAST-1SG letter-OBJ write-PART ‘I didn’t write a letter’

(Zanuttini 2001: 513)

b) A negative marker selecting a sentence as its argument. This kind of negative element resembles the main verbs inasmuch as it displays morphosyntactic features. For example, in Tongan, a Polynesian language, the negative marker ‘ikai selects a CP like in (3): (3) Na’e ‘ikai [CP ke ‘alu ‘a Siale] ASP neg [ASP go ABS Charlie] ‘Charlie didn’t go’

(Zanuttini 2001: 513)

c) A negative marker added to a main verb as a prefix, infix or suffix. For example, Turkish negative marker –me (4) is adjoined to the verb similarly to other tense and number affixes; (4) gör-ül-mc-yecek-ler see-PASS-NEG-FUT-3PL ‘They will not be seen'

(Bernini and Ramat 1996:9)

d) A negative marker displaying a particle form, which can be sensitive to verbal modality (Hungarian ne/nem) or not (Russian ne), preceding or following the main verb. This strategy is the most widespread in the human languages. Negation can be realized as a single element, like the English not (5), or as two elements in a combination of a negative particle and an adverb,

8

Cfr. Kahrel and Van Den Berg (1994), Bernini and Ramat (1996) and Miestamo (2007) for a typological overview.

14

Chapter 1

like the French ne...pas (6). In the latter case, the second element has a reinforcing function (see Jespersen 1917 and Dahl 1979). (5) John does not like pizza

(English)

(6) Je ne parle pas chinois

(French)

The shared characteristic of the strategies in (a-d) is the addition of an extra negative particle reversing the polarity of the sentence. This is the reason why it is crucial to investigate its syntactic nature, because it affects the processing of the sentence in which it occurs. To pursue this goal, I will restrict my interest on the last of these typologies, since Italian falls under that category. The first question arising on the syntactic nature of negation is whether negative markers have a proper syntactic characterization, or whether they are instances of some general categories, such as adverbs or polarity elements (see Zanuttini 1991,1996 and Zeijlstra 2004 for a detailed discussion). With regard to this question, Kitagawa (1986) firstly hypostasizes the existence of a negative head for Japanese negation. On the same line, Pollock (1989) suggested that the English negative marker n’t and the French ne are expressions of a specific functional projection NegP (Kayne 1989b extends this idea for all Romance languages). The introduction of a specific phrase for negative markers has followed the splitInfl hypothesis proposed by Pollock (1989), Belletti (1990) and, for independent reason, by Moro (1988). Since this idea is central to the discussion on linguistic negation, I need to recall it briefly. Pollock’s work starts from the well-known observation that English finite clauses display two different syntactic positions for auxiliary verbs and main verbs respectively. This phenomenon becomes evident if we put them in relation to the English negative marker not and the VP-adverb often: the auxiliary verbs can precede them (7) whereas main ones cannot (8). (7) a. John does not play football / *John not does play football b. John is often late (8) a. *John plays not football / John does not play football b. *John eats often pizza / John often eats pizza

Expletive negation

15

To explain such asymmetry, it has been proposed that the auxiliary verb moves to the head of IP while the main verb realizes the head of VP; since IP is higher than VP in the structure, auxiliary verbs precede the main ones9, as well as negation and adverbs. Pollock proposes the same idea for French non-finite verb clauses. He notes that French non-finite auxiliary verbs could either stay in situ or move to I, whereas lexical ones can only stay inside VP. For example, the non-finite auxiliary verb être can follow the negative adverb pas (9a) or can precede it (9b); the main verb sembler can only follow negation (10a) and cannot precede it (10b)10. (9) a. Ne pas être heureux est une condition pour écrire des romans Ne pas to be happy is a prerequisite to-write of the novel ‘Not to be happy is a prerequisite to write novels’ b. N’être pas heureux est une condition pour écrire des romans (10) a. Ne pas sembler heureux est une condition pour écrire Ne pas to seem happy is a prerequisite to-write des romans of the novel ‘Not to seem happy is a prerequisite to write novels’ b. *Ne sembler pas heureux est une condition pour écrire des romans Surprisingly, if we add the French form for “often”, souvent, the main verb can occur before it, adding a new puzzling piece to the discussion: (11) Paraitre souvent triste pendant son voyage de noce, c’est rare to-look often sad during one’s honeymoon that is rare ‘To often look sad during one’s honeymoon is rare’ In order to explain the asymmetry of main verb behavior with negation and with adverbs, Pollock suggests the presence of different functional projections between IP and VP. He splits the traditional Inflectional Phrase into a complex system composed by three different syntactic projections:

9

A possible alternative is the movement of the adverb. Edmonds (1978) excludes this hypothesis arguing that adverbs usually do not move (except when they raises in CP-field for interpretative reasons; see Chapter 4). 10 All the sentences are taken from Zanuttini 2001: 515-517.

16

Chapter 1

tense (TP), negation (NegP) and agreement (AgrP). According to Pollock, TP dominates NegP and AgrP11: (12) [TP [NegP [AgrP [VP ]]]] To explain the French pattern, he proposes that both the negative elements ne and pas are generated in NegP: pas is [Spec, NegP] and ne is the Neg0. He argues that the auxiliary verbs move to the head of TP12, as in English, whereas the main verb can only move to the head of AgrP, therefore following the negative marker and preceding the adverb13: (13) [TP nek-Aux [NegP pas [Neg0 tk [AgrP verbi [VP souvent [VP ti]]]]] The configuration in (12) takes into account the linear order of finite clauses in English as well as non-finite clauses in French. English auxiliary verbs move to the head of TP and, therefore, precede both negation and the adverb; instead, the main verb is at the head of VP and, consequently, follows them: (14) [TP Aux [NegP n’t [AgrP [VP often [VP Verb ]]]]] Pollock (1989) introduces valid arguments in favor of the hypothesis that negation has a specific syntactic projection in the split-IP domain. This view has become the standard theory in generative grammar (among others, see Chomsky 1989, 1991; Zanuttini 1991, 1997, 2001; Haegeman 1995; Potsdam 1997). Starting from Pollock’s works, many scientists propose different hypotheses in order to comprehend the complex distribution of negative elements. For example, Poletto (2008) expands the edge of NegP from an unitary projection to a complex set of sub-projections (this operation is like the one seen above for IP). She wants to compress the variation of the negative markers in Romance languages in a unique syntactic projection. Each subprojection carries a peculiar semantic feature selecting a coherent negative 11

Belletti (1990) proposes a contrary order of the functional projections, i.e. [AgrP [NegP [TP [VP ]]]]. Ouhalla (1991) tries to take into account both the Pollock and Belletti’s hypotheses arguing for a parametric choice. 12 The head ne cliticizes onto the auxiliary verb and, therefore, it moves with it. 13 Both after and souvent are VP-adverbs and they are interpreted as adjoined to VP (see Zanuttini 2001).

Expletive negation

17

element. Such a feature is related to the etymological origin of the negative element. The final result is a NegP-field with five different positions: [NegP [focus/operator [ScalarP [MinQ [QP [Existential IP]] A contrary view is found in Laka (1990) 14 . Laka (1990) proposes that negation is not an independent syntactic category but rather a realization of a specific value of an abstract projection, the ™ Phrase, which can host different operators. Compare, for example, some English sentences displaying negative and emphatic structures: (15) a. Mary didn’t leave b. Mary did leave15 c. *Mary did not leave

(negative clause) (emphatic clause)

English displays do-support both in negative (15a) and affirmative (15b) cases. In order to take into account (15b), Chomsky (1957) proposed that an affirmative head projects an affirmative phrase requiring the do-support element. Unfortunately, this does not predict the impossible co-occurrence of two structures requiring the do-support element (like in 15c). In order to take into account their complementary distribution, Laka (1990) suggests that the ™ Phrase can instantiate, alternatively, a negative or an emphatic affirmative feature. Since there is only one head, only one element can be hosted. This is the reason why the negative clause cannot co-occur in emphatic structures. From this point of view, negation is no more a specific projection, but just a value that an operator can assume. However, Laka’s hypothesis shares with Pollock the idea that negation is generated in a projection sensitive to the sentential polarity. It can be unique, as in NegP, or shared, as in ™ Phrase. I leave aside the discussion on which of these two alternatives is more correct because it is beyond the aim of this work. The Snegs analysis is coherent with both these two hypotheses. For convenience, I follow Pollock’s theory since it is the most widespread. Let us assume that a negative marker is generated in NegP; we can ask where such a projection is merged in the syntactic structure. More specifically, NegP can cross-linguistically occupy the same position, or it can vary depending on the language. As we saw before, Pollock (1989)

14

Frascarelli (2000) discusses a similar proposal for Italian negative marker non; see §1.4 for a detailed analysis. 15 I use the underlining to indicate the emphatic nature of the element.

18

Chapter 1

suggests that NegP is located in an IP-split system, specifically between TP and AgrP (TP…NegP…AgrP…VP). Laka (1990) shows that negation in Basque is generated above TP and, more generally, she proposes that the ™ Phrase occurs in a position between CP and TP, selecting TP as its argument16. Again, Zanuttini (1997, 2001) emphasizes the different natures of the negative markers,17 identifying different cross-linguistic positions for NegP. For example, she argues that even if the French negative marker ne and the Italian non occur in a pre-verbal position, they show a syntactic different behavior due to the different locations occupied by NegP in those languages. According to Zanuttini (2001), French ne displays a pronominal clitic nature and it precedes both the finite (16a) and non-finite verbs (16b) as other pronominal clitics. The Italian non, instead, behaves as a pronominal clitic only in finite verb clauses (17a); in non-finite verb clauses non precedes the verb whereas the pronominal clitic follows it (17b). (16) a. Jean ne les mange pas (Zanuttuni 2001:519) Jean neg Cl.them eats neg ‘Jean doesn’t eat them’ b. Jean voudrait ne pas les manger Jean would-want neg neg them to.eat ‘Jean would want not to-eat them’ (17) a. Gianni non le mangia Gianni not Cl.them eats ‘Jean doesn’t eat them’ b. Gianni preferisce non mangiar-le Gianni prefers not to-eat-Cl.them ‘Gianni prefers not to eat them’ In order to explain (16)–(17), Zanuttini (1997) proposes that Italian NegP is higher than French NegP and, therefore, the two negative phrases behave differently 18 . More specifically, Italian non is the head of NegP located above TP (NegP…TP…VP) and French ne is a head of a lower NegP (TP…NegP…VP) which raises in the structure adjoining to a functional 16

She proposes that only the pre-verbal negative markers are generated in ™ Phrase while post-verbal adverbs like pas occur in an independent phrase lower in the structure. From this point of view, negation should occupy two different positions depending on its pre-verbal or post-verbal nature. 17 See next section for a more detailed discussion. 18 For a similar hypothesis see Haegeman (1995).

Expletive negation

19

projection (often Vo) in virtue of its clitic nature. Such structural difference is the reason why these negatives markers exhibit different syntactic behavior. Following this kind of reasoning, Zanuttini (1997) discusses four different positions for NegP in Romance languages19: a) A negative marker in a pre-verbal position that can deny a sentence by itself. We can find such a negation in Italian non, Spanish no, Portuguese nao, Catalan no, etc. (18) a. Gianni non ha telefonato a sua madre b. Juan no ha llamado a su madre c. El Joan no a trucat a sa mare d. Joao nao ligou para sua mae 'John hasn't called his mother'

(Zanuttini 1997:3)

b–c) A negative marker in a post-verbal position which precedes the past participle. This case is split in two sub-cases in relation to their linear order with some “low adverbs” (see Cinque 1995, 199920): i) a negative marker that follows adverbs such as already and no more. This negative marker occurs in some northern Italian and southeastern French languages as Piedmontese nen, Occitane pas, etc. (19) a. A 1'avia (*nen) gia nen vulu 'ntlura s.cl. s.cl'had neg already neg wanted then 'Already at that time he had not wanted to.'

(Zanuttini 1997:70)

ii) A negative marker that precedes adverbs such as already and no more. Piedmontese displays this kind of negation as well: (20) A 1'avia pa gia (*pa) vulu 'ntlura s.cl s.cl'had neg already neg wanted then 'Already at that time he had not wanted to.'

19

(Zanuttini 1997:70)

It has been attempted to reconstruct the original Indo-European form of standard negation. According to Fortson (2004), the original form was nê (or mê) and it usually occurs in a pre-verbal position. 20 Cinque proposes that the structural order of the adverbs between IP and VP is cross-linguistically fixed. The sequence is: already > no more > always > completely > all > well. See Chapter 4.

20

Chapter 1

d) A negative marker in a post verbal position which follows the past particle. This is the lowest position of a negative marker. Consider the Milanese no, a northern Italian dialect: (21) a. El l’ha scrivuu no s.cl s.Cl’has written neg ‘He hasn’t written’ b. *El l’ha no scrivuu s.cl s.Cl’has neg written

(Zanuttini 1997: 88)

In summary, the four typologies of negative markers represent four different NegPs located in as many places in the sentential structure. Each language using a negative marker to deny a sentence adopts, at least, one of these strategies. According to Zanuttini, I progressively number the four NegPs starting from the highest, NegP1, to the lowest, NegP4: (22) (Zanuttini 1997: 101)

I follow Zanuttini (1997) in considering the four different locations that a NegP can display in a specific language21. To summarize, we saw that a negative marker cannot be assimilated to other syntactic categories, such as adverbs, since it exhibits unique and specific behavior (Pollock 1989). From this fact, it has been proposed that there is a 21 As we saw before, Poletto (2008) proposes the existence of a “bigger” projection that absorbs every kind of negation. See also Moscati (2010), and the references therein, for an overview on the cross-linguistic realization of the NegP positions.

Expletive negation

21

functional negative phrase, NegP, located in the IP-split field. Finally, according to Zanuttini (2001), we have individuated four possible positions for NegP depending on its location in the linear word order22. Focusing on NegP, we can now ask if a negative marker occupies a head position or a more specific one. The alternation between these two positions generates some interesting facts.

1.1.1 On the negation head status Treating a negative marker as the realization of a specific projection is not sufficiently accurate to catch the heterogeneity of negative phenomena. This is the reason why, in the last thirty years, a lot of research has been carried out to define correctly a X-bar schema for NegP (above all, see Zeijlstra 2004 and Moscati 201023). I will start with some proposals on the head status of the negative markers. According to Kayne (1989b), the French ne is the head of NegP (as I already discussed). As a proof, he discusses some cases in which negation blocks the head-to-head movement. Consider, for example, the pronominal clitic climbing phenomenon. According to him, clitics are heads that can raise from the subordinate clause to the main clause, as in (23a). However, when the negative marker ne occurs between the clitic and its trace, as in (23b), the structure is ungrammatical: (23) a. Jean lai fait manger ti par/à Paul (Zanuttini 2001:524) John it makes to-eat by/to Paul ‘John makes Paul eat it’ b. *Jean l’ai fait ne pas manger ti à l’enfant John it has made neg neg to-eat to the-child ‘John made the child not eat it’24 22 I do not refer to those works treating negation as a criterial operation, leaving this task to Haegeman and Zanuttini (1991) and to Haegeman (1995). 23 He suggests that the distribution of negation is parametric (Moscati 2010:8): A) Negation is signalled by an overt adverbial in the specifier or a bound morpheme in the head of NegP. B) Both the specifier and the head positions of NegP may be phonologically realized. I do not take a position on this proposal because it is beyond the aim of this book. However, in the next two sections I will present some elements in favour of it. 24 As we saw, the adverb pas is usually considered as the specifier of NegP headed by ne [NegP pas [ne]]. See Kayne (1989b) and the discussion below.

22

Chapter 1

This blocking effect rises from a violation of the head-movement constraint proposed by Travis (1984)25 (or, as is considered today, from a relativized minimality violation26). More specifically, according to Kayne (1989b), in (24b) clitic climbing is impossible because ne blocks the antecedent governing relation between the clitic and its trace27. Since clitics are heads, Kayne proposes that the French ne is a head too. An additional test for the head status of negative markers is in Zanuttini (1997). In Padoan, a northern Italian dialect, interrogative structures show that V0 moves to C0 in order to satisfy the polar criterion. Since Padoan usually displays pronominal clitics in the subject position, both in affirmative (24a) and in negative clauses (24b), in interrogatives verbs compulsorily precede pronominal clitics–at least in affirmative form (24c)– yielding ungrammaticality if such a movement does not take place (24d). (24) a. El vien” Cl.he comes ‘He’s coming’ b. No el vien neg Cl.he comes ‘He isn’t coming’ c. Vien-lo? comes-Cl.he ‘Is he coming?’ d. *El vien? Cl.he comes

(Zanuttini 1997:40)

25 The head movement constraint states that a head Xo can move into a head Y0 only if Y0 properly governs X0. 26 See Rizzi (2001). 27 In RM’s view, negation blocks clitic climbing as the result of an intervention effect. Since clitic movement forms a head-chain crossing over negation, the resulting configuration [Clitic…Neg…Trace] is a violation of the RM assumptions. In fact, the minimal configuration required by the probe-trace chain in RM is given in (i)

(i) Y is in a Minimal Configuration with X if there is no Z such that: i. Z is of the same structural type as X, and ii. Z intervenes between X and Y. (Rizzi 2001: 225) In my example, the negative marker is Z and, therefore, Y and Z cannot stay in the required minimal configuration. See Rizzi (1990, 2001) for a detailed analysis. An example of RM’s application and negation is in Ouhalla (1990).

Expletive negation

23

Crucially, the subject-clitic inversion is ruled out in negative interrogatives and verbs cannot pass over the clitic: (25) a. *No vien-lo? neg comes-Cl.he b. No el vien? neg Cl.he comes ‘Isn’t he coming?’ Zanuttini (1997) explains the pattern in (25) by proposing that negation is the head that moves to C°; such a movement is driven by the criterial requirement of the polar question. Since C° is already occupied by no, verbs cannot move there in (25a). Such a conclusion perfectly matches with Kayne’s (1989b) analysis. Finally, Merchant (2001) shows that if a negative marker can appear in a question like “Why not?”, it displays a maximal projection (XP) status; on the contrary, if a negative marker cannot appear in such a form, it displays a head (X°) status. It follows from the fact that why is a phrasal adverb (an XP) and, therefore, only XPs can be adjoined to it. When the negative marker is a head, languages usually display other negative elements to build the question “Why + XP-negation”28. (26) a. *Giati dhen? vs. Giati oxi? (Modern Greek, Merchant 2001) why not why no b. *Sud-esh?” vs. Sud gä’ä? (Bezhta) why not why no c. Warum nicht? vs. *Warum nein? (German) why not? *Why no In (26a-b), the head status of the negative markers dhen and esh appears in their impossibility to occur in why-not questions. Consequently, modern Greek and Bezhta resort to an alternative form, oxi and gä’ä, to build the alternative why-no form. On the contrary, giving the phrasal nature of the German nicht, it can occur in the “Why not?” form. Zeijlstra (2004, 2006) definitely shows that Romance negative markers which are not adverbs are heads regardless of their morphological status 28 Some languages exhibit hybrid behavior or exceptions; see Merchant (2001) for a detailed analysis.

24

Chapter 1

(clitics, morphemes headed a NegP, morphemes adjoined to V0, etc.). All of these elements carry an uninterpretable negative feature [UNeg] that projects a functional projection, NegP. He proposes that a negative operator Op¬ that carries a [INeg] is merged in NegP in order to eliminate such a [UNeg] feature under Spec-head Agreement. This mechanism works for both preverbal negative markers, which are directly merged in Neg° (such as the modern Greek dhen), and negative markers that are found in some lower positions (such as the French ne, which is adjoined to VP). To summarize, we reviewed at least 3 tests to identify the head status of a negative marker: i) it blocks the clitic climbing phenomenon (as in French); ii) it blocks verb movement (as in Padoan); it cannot occur in “Why not” questions and, more in general, be an adverb. Now we can see what happens when a negative marker is a maximal projection.

1.1.2 On the negation maximal projection status According to Zanuttini (2001), a simple way to identify the maximal projection status of a negative marker is by exclusion: if a negative marker does not behave as a head (it does not interfere with a head movement, it can occur in “Why not” form, etc.), then it displays maximal projection status. For example, since two elements belonging to the same syntactic category usually interact with each other, we expect that a negative marker that is a maximal projection will only interact with some maximal projection movements. Zeijlstra (2004) highlights that in French, only ne blocks clitic climbing, thus displaying a head status; on the contrary, pas does not interact with the clitic and, therefore, must be considered a maximal projection29. (27) Jean ne l1’a pas fait manger t1 à Paul (Zeijlstra 2004: 153) Jean neg Cl.it .has neg made eat to Paul ‘Jean hasn’t made Paul eat it’ A specular observation is found in Zanuttini (2001); she discusses some interesting effects in V2-languages in which verbs can raise in the structure 29

Rizzi (1990) gives an ulterior proof of the maximal projection status of pas, showing that it interacts with the extraction of an adjunct (which is a maximal projection). See Rowlett (1998) for an opposite view.

Expletive negation

25

to fill the C0. In Swedish, the negative marker inte, which usually precedes a verb in linear word order (28a), does not block the raising of a verb that can pass over it (28b). This lacking of interference is taken as a proof of the different syntactic status between verbs and negation: the former is a head; the latter is a maximal projection30. (28) a. … om Jan inte köpte boken that John neg bought books ‘… if John didn’t buy books’ b. Jan köptei inte ti boken John bought neg libri ‘John didn’t buy books’

(Zanuttini 2001: 527)

As far as Merchant’s test is concerned, we saw above that negative markers building “Why not?” questions are considered maximal projections. This is what happens, for example, with the English not, the German nicht and the Icelanding ekki: (29) Warum nicht? ‘Hverfor ekki?’ Why not?

vs. vs. vs.

*Warum nein? *Hvarfor nej? *Why no?

(German) (Icelanding)

In summary, we can synthetically point out two tests in order to distinguish the syntactic nature of a negative marker by gathering together the observations of the last two paragraphs: (30) a. If it interacts with the head-movement phenomena, then it is a head (otherwise it is a maximal projection) b. If it occurs in a “Why not?” frame, then it is a maximal projection (otherwise it is a head) I will adopt these two tests to investigate the nature of the Italian marker non, since it is the characterizing element of Snegs.

30

Zeijlstra (2004) also shows that the Swedish negative marker can be topicalized moving in [Spec, CP]. Since such a position can only be filled by a XP, this proves that Swedish negation is a maximal projection.

26

Chapter 1

1.2 Italian negation structures Standard Italian displays, at least, two ways to deny a sentence: i) with the negative morpheme non (not) and ii) with indefinite quantifiers (like nwords and pre-verbal negative polarity items). The negative marker non is always in a preverbal position and it denies the sentence by itself31: (31) Luca non ha mangiato la pizza Luca neg has eaten the pizza ‘Luca has not eaten pizza’ As we saw above, non precedes all the clitics and, according to Zanuttini’s classification, it belongs to the NegP1 class. Coherently, non always precedes both the auxiliary and main verbs: (32) Luca non ha *(non) mangiato *(non) la pizza Luca neg has neg eaten neg the pizza ‘Luca has not eaten the pizza’ According to Payne (1985), I consider non as an instance of standard (or, alternatively, propositional) negation, since it occurs in a declarative main clause and it does not involve quantification. Italian also shows alternative ways to deny a sentence: it displays n-words, such as nessuno (nobody) and niente (nothing), and NPIs, such as mica32, in pre-verbal positions: (33) a. Nessuno è venuto nobody is come ‘Nobody has come’ b. Niente ha mangiato Luca Nothing has eaten Luca ‘Luca has not eaten anything’ 31

For an historical perspective on the evolution of non, see Hansen & Visconti (2012). 32 According to Cinque (1976), mica is a presuppositional negative marker. For example, in (ii) mica denies a presupposition like “You think I eat the cake”: (ii) Mica l’ho mangiata io la torta Mica Cl.has eaten I the cake ‘I did not eat the cake (as you think)’

Expletive negation

27

c. Mica l’ho mangiato io NPI Cl.it have.1stsing eaten I ‘I have not eaten it (as you think)’ Negative constituents can negate a sentence by themselves only if they command the verb, otherwise they are ungrammatical, and require the negative marker non to be rescued: (34) a. *E’ venuto nessuno is come nobody a’. Non è venuto nessuno neg is come nobody ‘Nobody has come’ b. *Ha mangiato niente has eaten nothing b’. Non ha mangiato niente neg has eaten nothing ‘S/he has eaten nothing’ c. *L’ho mangiato mica io Cl.it have.1stsing eaten mica I c’. Non l’ho mangiato mica io Neg Cl.it have.1stsing eaten mica I ‘I have not eaten it (as you think)’ In a’, b’ and c’ the co-occurrence of two negative elements, the negative marker and the n-word / NPI, acts as a unique instance of negation (as the English translation shows). This phenomenon is traditionally called “negative concord” (cfr., among others, Zanuttini 1991; Haegeman 1995; Haegeman and Zanuttini 1991, 1996; Giannakidou 1998, 2000; Zeijlstra 2004; Ovalle and Guerzoni 2004; Chierchia 2013 and the references therein) and it contrasts with “double negation”, in which each negative element realizes a singular instance of negation (for example, in English “I did not know nothing about Maria” means “I know something about Maria” because two negations give an affirmation). According to Giannakidou (1997, 2000), there are, at least, two types of negative concord languages: the strict type, in which the n-words in preverbal and post-verbal positions always require a negative marker (ex. Czech in 35), and the non-strict type, in which the n-words in a post-verbal position obligatorily require the negative marker, and the n-words in a preverbal position must occur alone. Italian belongs to the latter type, as (34) and (36) show.

28

Chapter 1

(35) a. Nikdo *(ne)volá n-body neg-calls ‘Nobody is calling’ b. *(Ne)volá nikdo neg-calls n-body ‘Nobody is calling’

(Zeijlstra 2004: 64)

(36) *[Nessuno] [non] mangia nobody neg eats A further proof that Italian belongs to the non-strict type is that n-words can legitimize other n-words (whereas it is forbidden in the strict negative concord languages, which used the negative marker to build a multiple nword sentence): (37) Nessuno ha regalato niente a nessuno nobody has given nothing to nobody ‘Nobody has given anything to anyone’ According to van Der Wouden and Zwarts (1993), since the “negative feature is ‘spread’ or distributed over any number of indefinite expressions within its scope”, Italian is considered a non-strict negative concord language of the Spread-type33. Essentially, Italian is a non-strict negative concord language of the spreadtype, and it displays the standard negation marker, non, that is generated in NegP1. Leaving aside the discussion on the n-words34, I will focus on the syntactic structure of negative markers, since it is the element occurring in Snegs.

1.2.1 On the syntactic nature of the negative morpheme “non” (not) In §1.1 we saw that a negative element can be either a head or a maximal projection, and we also individuated various tests to discriminate between 33

They use this classification to distinguish languages like Italian, in which a second negative element optionally appears in a sentence, from languages like French, in which “a distinguished negative element shows up in all sentences that contain a negative expression” (Van Der Wouden and Zwarts 1993: 202). They label the latter type Negative Doubling. 34 See Giannakidou (1997, 2000) and Zeijstra (2004) for a detailed discussion.

Expletive negation

29

those two options. I am now going to employ those tests on the Italian non in order to identify its syntactic nature. As is well known, Italian displays a rich system of clitics, and allows their long climbing from subordinate to matrix clauses. According to Rizzi (1982) and Kayne (1989b), clitic climbing is allowed in affirmative clauses (38a-a’), but is forbidden in negative ones (38b-b’), following the French pattern seen in §1.1.1: (38) a. Gianni vuole veder-li Gianni wants see-Cl.them ‘Gianni wants to see them’ a’. Gianni lii vuole vedere ti Gianni Cl.them wants see ‘Gianni wants to see them’ b. Gianni vuole non veder-li Gianni wants neg see-Cl.them ‘Gianni does not want to see them’ b’. *Gianni lii vuole non vedere ti Gianni Cl.them wants Neg see

(Kayne 1989b: 243)

Crucially, non does not interact with the raising of a WH-element from VP to CP: (39) a. [Quali libri]i vuole vedere ti Luca? which books wants see Luca ‘Which books does Luca want to see?’ b. [Quali libri]i non vuole vedere ti Luca? which books neg wants see Luca ‘Which books doesn’t Luca want to see?’ According to Kayne (1989b) and Zanuttini (1996, 2001), a simple way to take into account (38–39) is to assume that the Italian non is the head of NegP. It therefore blocks long clitic climbing because it is a headmovement, but it allows the raising of a WH-element because it is a maximal projection movement. A further clue of the head status of the Italian non comes from Merchant’s test: (40) *Perchè non? vs. Perchè no? why not why no?

30

Chapter 1

Non cannot appear in the “Why not?” question as result of its head status. Coherently, Italian displays a negative element no, fitting the requirement to be a maximal projection. However, other Italian data seem to contradict the head status of non, suggesting a maximal projection configuration. According to Rizzi (1981, 1982), non shows a pronominal behavior following verbs in C0 in Aux-toComp structures: (41) Non essendo arrivato Gianni, siamo venuti senza di lui neg being come Gianni are.10plu come without of him ‘Not being Gianni come, we come without him’ Cinque (1999) proposes that negation carries along verbs because it is leftadjoin to it35. Moreover, he suggests that non is generated in [Spec, NegP] instead of Neg0 given that non does not cliticize onto verbs leaving the linear order Neg-Verb (41). He uses a similar proposal to explain a particular scope effect between negation and some adverbs36: (42) a. Gianni fortuatamente non è riuscito a corrompervi Gianni luckily neg is managed to corrupt-CL.you ‘Luckily, Gianni hasn’t managed to corrupt you’ (Adv > Neg) ‘*Non-luckily Gianni has managed to corrupt you’ (*Neg > Adv) b. Gianni non è fortunatamente riuscito a corrompervi Gianni neg is luckily managed to corrupt-CL.you ‘Luckily, Gianni hasn’t managed to corrupt you’ (Adv > Neg) ‘*Non-luckily Gianni has managed to corrupt you’ (*Neg > Adv) In (42a) and (42b) the different linear order of the adverb fortunatamente and the negative marker non does not impact the meaning of the sentence: the adverb always scopes over negation, regardless of linear position. While (42a) is not problematic because the adverb precedes negation, (42b) needs some clarifications. Cinque proposes that in such a case the “base” position of NegP is to the right of the adverb; it then follows the raising of the verb (left-adjoined to it) in a position immediately dominating NegP, i.e. AdvP: (43)

35

… [AdvP[AdvP[Adv non Vi –Adv] [NegP [ [non] ] [NegP ti [XP … ti]]]]

For a detailed discussion on the clitics as left-adjoin elements, see Kayne (1994) and Cinque (1996). 36 All data are taken from Cinque (1999: 123).

Expletive negation

31

The final linear word order after the movement displays negation before the adverb, but, as Cinque states, its scope domain is given by the structural position of the elements (before the Spell-out), in which the adverb ccommands negation. A further clue of the maximal projection nature of non is provided by Rizzi (1990). He shows that the negative marker non blocks the antecedentgovernment chain between the adverb perchè (why) and its trace: (44) Mi chiedo perchè hai detto che Luca è arrivato Cl.to me wonder.1SG why have.2nd.sing. said that Luca is arrived a. ‘I wonder why you have said that Luca has arrived’ b. ‘I wonder whyi you have said [ti that Luca has arrived]’ (45) Mi chiedo perchè non hai detto che Luca è Cl.to me wonder.1SG why neg have.2ndsing said that Luca is arrivato arrived a. ‘I wonder why you do not say that Luca has arrived’ b. ‘#I wonder why you do not say [ti that Luca has arrived]’ In (44a) perchè can either refer to the matrix verb “say” or to the embedded “arrive” (even if the former reading is more natural). Structurally, it is explained by assuming that perchè can be generated either in the CP of the matrix clause37, as in (44a), or in the CP of the subordinate one, as in (44b). The occurrence of non in (45) rules out the interpretation in which perchè refers to the embedded verb (45b), leaving as the only option the reference to the matrix verb (45a). Rizzi interprets this fact as a case of relativized minimality. Given that perchè is a maximal projection (see. Rizzi 2001), it seems that non is a maximal projection too, because it blocks the government chain between perchè and its trace in the subordinate CP. However, there is, at least, a good reason to reject this hypothesis. As Moro (2011) points out, the same intervening effect occurs with negare (to deny): (46) Mi chiedo perché neghi che Luca è arrivato Cl.to me wonder.1SG why deny.2nd.sing that Luca is arrived ‘I wonder why you deny that Luca have arrived’ ‘# I wonder whyi you deny [ti that Luca have arrived]’

37

See Chapter 5.

32

Chapter 1

The reading in which perchè refers to ‘arrive’ is ruled out by the occurrence of deny. Since the verb to deny is not a maximal projection, we should look for another element with the correct syntactic category in order to explain the blocking effect in (46). Coherently, this element cannot be the negative marker non because the minimality effect persists regardless of its occurrence. According to Rizzi (1991), Haegeman and Zanuttini (1991), Haegeman (1995) and Moro (2001), I assume that a negative null operator in [Spec, NegP] causes the relativized minimality violation of the XP-type for both the non and the verb negare38. From this point of view, the Italian NegP shows the following structure: (47) [NegP Neg Op [Neg non [+neg]]] A negative feature is licensed in Neg0 and a negative operator is found in the specifier position. The distribution of these elements is regulated by the NEG-criterion (see Haegeman 1995 and references therein). This proposal takes into account the mixed-behavior of the Italian negation: non is a X0 and, therefore, it interacts with head-chains and cannot appear in “Why not” questions. At the same time, the negative operator constitutes a XP blocking the maximal projection chains and, as Cinque points out, carrying along verbs both in Aux-To-Comp structures and in the adverbial ones. Reasonably, we can reject Cinque’s hypothesis on the maximal projection status of non, and can continue to treat non as a head (with a negative operator in its specifier). An aspect related to the syntactic characterization of non is its spatial position. As we saw in §1.1, Zanuttini (1997) proposes that Italian NegP is projected above TP (as NegP1 usually is). We can further investigate this point because it is controversial. Belletti (1990) proposes that the order of the Italian functional projection is AgrP-NegP-TP-VP. According to the Mirror Principle (Baker 1985), she assumes that the morphologic make up of an inflected verb is given by its syntactic derivation. Given that Italian inflected verbs are made up of a root + tense and agreement clitic morphemes, she proposes that it results from the head-to-head verb movement from V0 to T0 and Agr0:

38

As Moro (2011: 396) suggests, the negative operator can be “independently realized as non ‘not’ or inherently licensed by a verb such as negare ‘deny’ ”.

Expletive negation

33

(48) Loro non [Agr mangi-avai-no[T ti [V ti ] they neg eat-Cl.T-Cl.Agr ‘They did not eat’ Since the negative sentence precedes the inflected verb, we are tempted to assume that negation selects AgrP as its argument, giving the order NegPAgrP-TP-VP. However, Belletti rejects this configuration and proposes that NegP is lower than AgrP: leaving the order AgrP-NegP-TP-VP 39 . She follows Moritz (1989) in proposing that Agr0 requires two elements: a verb with the tense inflection and the negative marker. A Verb+T moves to Agr0 (it skips the Neg0) given the expected order verb-T-Agr; for an independent mechanism, the negative marker moves from Neg0 to Agr0 giving the order Neg-verb-T-Agr40 (from this point of view, negation is assumed to be clitic). According to this proposal, (48) displays the structure (49)41: (49)

[AgrP Loro [Agr nonk [Agr mangi-avai-no [NegP [Neg tk [TP [T ti [VP ti ]]]]]]]]

Zanuttini (1996, 1997) reinforces the hypothesis that Italian non selects TP, arguing that i) negation must be adjacent to verbs and nothing, except elements already adjunct to verbs such as pronominal clitics, can intervene between them (50); ii) the structure lacking TP, such as true imperatives42 and past participle clauses, cannot be denied by non (51)43: (50) a. Maria non ha telefonato Maria neg has phoned ‘Maria did not phone’ 39

The same sequence is proposed in Chomsky (1993). Such a suggestion can take into account the Italian order position of the subject, which always precedes negation (Sub–neg–verb). 40 See Belletti (1990) for all the theoretical reasons for this derivation. 41 Belletti also proposes that Italian negative adverbs, such as più (anymore) and mai (ever), are located in [Spec, NegP] and they do not move in AgrP0 as the negative head does. According to her, a sentence such as (v) shows the structure (v): (v) Loro non mangiavano più they neg ate anymore ‘They did not eat anymore’ (vi) [AgrP Loro [Agr nonk [Agr mangi-avai-no [NegP più [Neg tk [TP [T ti [VP ti ]]]]]]]] 42 Zanuttini (1997) rethinks her proposal arguing that in imperatives clauses NegP selects MoodP instead of TP. She follows Kayne (1992). I leave the discussion on these works to Riviero (1994), Riviero and Terzi (1995) and Zeijlstra (2004). 43 See Zanuttini (1996) for detailed analysis of true imperative and past participle as structures without TP.

34

Chapter 1

b. Maria non gli ha telefonato Maria neg Cl.to him has phoned ‘Maria did not phone to him’ c. *Maria non ieri ha telefonato Maria neg yesterday has phoned (51) a. Telefona! (Imperative) call.Imp ‘Call!’ b. *Non telefona! neg call c. Raggiunta Maria, siamo andati al bar (Past Participle) reached Maria are.1st.plu went to the bar ‘Having reached Maria, we have gone to the bar’ d. *Non raggiunta Maria, siamo andati al bar neg reached Maria are.1st.plu went to the bar A contrary proposal is found (again) in Cinque (1999). He starts from the interaction between negation and the adverbs, suggesting that NegP can be base generated on the top of a set of high adverbs and not in a specific and rigid position. For example, when negation co-occurs with the adverb deliberatamente (deliberately), two readings are available: one in which the adverb scopes over negation and one in which negation scopes over the adverb: (52) Luca non ha deliberatamente lasciato cadere la sua candidature Luca neg has deliberately let drop the his candidature ‘Deliberately Luca has not dropped his candidature’ (Adv > Neg) ‘Luca has dropped his candidature not deliberately’ (Neg > Adv) Cinque explains this ambiguity by saying that NegP can be originated either in a position lower than the adverb deliberatamente (the adverb scopes over non) or in a position higher than the adverb deliberatamente (non scopes over the adverb)44. This double option is available only for adverbs included in a contiguous portion of the adverbial universal hierarchy between the

44

The fixed liner order Neg>Adv is given by the movement of the Neg+verb in a position dominating the adverb (as we saw before).

Expletive negation

35

Moodspeech act and the Modepistemlc (the highest part of the hierarchy) 45 . Interestingly, even Cinque (1999) highlights the close relation between NegP and TP; proposing, as we saw before, that non is left adjoined to verbs. Even though I have discarded his analysis for various syntactic reasons, I accept this suggestion. Frascarelli (2000) deepens the clitic nature of non. He proposes that non is a phonetic clitic, not a syntactic one. He starts from the observation that Italian negation differs from clitics insofar as it always precedes verbs both in finite and non-finite clauses (as we saw above). Non displays a defective status because it is a monosyllable; therefore, it requires a host word (a verb) with which it forms a Phonological Phrase46. Interestingly, Frascarelli preserves the syntactic order of the functional projections seen in Belletti and Zanuttini: AgrP dominates NegP, which dominates TP. Starting from different observations and coming to different conclusions, all the analyses sketched above share the idea that Italian non must be adjacent to verbs. It can be because non selects TP (as in Zanuttini’s hypothesis), or because it is adjunct to the Agr0, landing site of verbs as well (as in Belletti’s hypothesis), or because non is left adjoined to verbs (as Cinque proposes) or, finally, because it forms a phonological phrase with verbs (as in Frascarelli). 45 He proposes that the sequence of adverbial projections is cross-linguistically fixed according to their universal hierarchy, i.e. [frankly Moodspeech act [fortunately Moodevaluative [ allegedly Moodevidential [ probably Modepistemlc [ once T(Past) [ then T(Future) [ perhaps Moodirrealis [ necessarily Modnecessity [ possibly Modpossibility [ usually Asphabitual [ again Asprepetitive(I) [ often Aspfrequentative(I) [ intentionally Modvolitional [ quickly Aspcelerative(1) [ already T(Anterior) [ no longer Aspterminative [ still Aspcontinuative [ always AsPperfect [just Aspretrospective [ soon Aspproximative [ briefly Aspdurative [ characteristically(?) Aspgeneric/progressive [ almost Aspprospective [ completely AspSgCompletive(I) [ tutto AspPlCompletlve [ well Voice [fast/early Aspcelerative(II) [ again Asprepetitive(II) [ often Aspfrequentative(II) [ completely AspSgCompletive(II) ] (Cinque 1999: 106). 46 He also proposes that the verb and the negative head non are syntactically independent. Both raise in the structure, but for different reasons: the verb raises in Agr0 in order to achieve agreement with the subject; the negative head non raises in the Foc0 in order to match its negative feature. In fact, he proposes that negation requires a features-checking mechanism generating a [+neg] feature in Foc0. Both a Spec-head configuration and a Spec-Argument (with negation in situ) can check and delete the feature. For detail, see Frascarelli (2000). Choosing between a criterial system or a syntactic configuration as analysis for the linguistic negation is beyond the aim of this book and, therefore, I leave aside this problem.

36

Chapter 1

Since choosing between these systems is beyond the aim of this book, I will not discuss this problem any further; I will focus instead on their common aspects. According to Belletti, Zanuttini and Frascarelli, I assume by now that Italian Neg0 non selects a TP as its argument, realizing one of the two parametrical values proposed by Ouhalla (1990)47. In summary, in this paragraph we saw that the Italian negative marker non is the head of NegP: it blocks head-movements (such as clitic climbing) and it does not build the “Why not?” question. However, I have shown that non displays some syntactic behaviors typically associated with the maximal projection status (it interacts with scope phenomena and with adverbs like perchè). Since these facts not only affect the negative marker non but also inherently negative verbs (such as “to deny”), I had to assume that a null negative operator fills the [Spec, NegP] giving the configuration [NegP Neg Op [Neg non [+neg]]]. Furthermore, we focused on the spatial characterization of the Italian NegP, and saw several proposals. Among others, I chose the standard hypothesis that Neg0 selects TP; from this follows the obligatory adjacency between them (this is also the reason why non cannot occur in clauses TP-lacking like negative imperatives and past participle clauses). We now have all the tools necessary to investigate some aspects of the expletive use of negation.

47

According to Ouhalla (1990), standard negation cross-linguistically selects VP or TP. This choice is parametrical (see also Puskás 2000 for a detailed discussion). Zeijlstra (2004) criticizes this proposal. According to Nilsen (2003), he assumes that if the relation between the negative elements and other elements in the sentence, such as the adverbs, can be explained by a semantic analysis, then there is no reason to propose a syntactic selection. Therefore, he rejects Belletti and Zanuttini’s analysis of NegP dominating TP and he proposes that “the locus of negation in the clause follows from semantic properties of negation” (p. 179). From this point of view, negation simply introduces an existential quantifier in the structure binding the free variables: (vi) [Op ¬ ‫׌‬e,x [...(e)...(x)...]]

Since the v domain is the locus in which a free event variable is introduced, he proposes that negation dominates vP. Even though this proposal is persuasive, I opt for the syntactic explanation. The syntactic view can easily explain the phenomena related to negation (the intervening effect with head-movements, the “why not?” restrictions, etc.) and I am interested in it.

Expletive negation

37

1.3 Expletive Negation (EN): a general discussion According to Jespersen (1917, 1924) a negative marker can either deny a sentence (standard negation) or not because it has “no longer its full negative force” (Jespersen 1917:75). He first analyzed some complex sentences built by a matrix verb of denying or doubting, and negative subordinate clauses: (53) I drede not pat ne pe curs of God [...] wolde brynge me into a I doubt not that neg the curse of God would bring me into a ful yitel eende if I contynuedepus very evil end if I continued thus ů ‘I do not doubt that God's curse would bring me to a very evil end if I continued like this’ (Old English, Testimony of William Thorpe 482 in Wim van der Wurff (1999) In (53), the negative marker ne in the subordinate clause does not reverse the polarity of the sentence, as the (modern) English translation shows. Jespersen called this paratactic negation because negation in an embedded clause is, in a way, an independent clause expressing the positive content of the matrix verb (what is doubted or denied). Some languages display this kind of negation that does not deny a proposition in matrix clauses too: (54) How often have I not watched him

(Jespersen 1924: 323)

Generally, linguists refer to this negation as pleonastic or expletive48, given its apparent vacuous semantics. The distribution of EN shows some cross-linguistic similarities (Van Der Wouden 1994a): it is never obligatory; it often selects subjective and conjunctive mood; it can occur after verbs such as “to fear”, “to prohibit” and “to prevent”; it can occur in comparative constructions and in subordinate clauses introduced by conjunctions such as before, without and unless49. However, not all languages display two different morphemes, as the Old English in (53), for expletive and standard negation. More often, the same negative marker performs both expletive and standard negation. Italian is an example of this: 48

See Yoon (2011) and Makri (2013) for an overview on expletive negation. I do not discuss examples here because we will see them in the paragraph dedicated to the Italian EN.

49

38

Chapter 1

(55) Che cosa non ha fatto Gianni!50 what neg has done Gianni ‘What has Gianni done!’ / ‘What has not Gianni done!’ Fortunately, researchers individuated various tests to distinguish between the two readings. Let us briefly consider at least four of these tests: a. EN does not reverse the polarity of the sentence in which it occurs (Horn 1989, Horn and Kato 2000), whereas the SN does: (56) Gianni è più alto di quanto non lo sia Lorenzo Gianni is more tall of much neg Cl to be.3rdsing.Subj Lorenzo ‘Gianni is taller than Lorenzo’ b. EN does not license strong negative polarity items (NPIs) but it legitimates positive polarity items (PPIs); SN does the contrary51. Before going any further, we need to offer a few words on this point, because the definition of a strong/weak NPI will be essential for the discussion. According to Giannakidou and Yoon (2010) and to Giannakidou (2011), a NPI can be strong or weak depending on whether it is only licensed by a 50

Negative exclamatives are a case of the Italian EN. As Delfitto and Fiorin (2014) discuss, this kind of sentence has an expletive reading when it is uttered with a strong exclamative prosody, hence the diacritic !. Such a marked prosody is often indicated as an element discriminating between SN and EN (see Yoon 2011). This is true in many cases, but Italian shows some examples of EN without marked prosody (above all, see the comparative structure in 56). Therefore, I will not use the prosody as a test for the EN structures even if it is often associated with it. 51 Makri (2013) shows that the non-negative-concord languages, like Latin, can display EN and the SN at the same time, without giving a double negation reading (viib), as usually they do (viia): (vii) a. non nulla in Hirtio. [Double Negation] NEG nothing in Hirtius . . . ‘Some few points in Hirtius’ [Cic. Ad Brut. 1.10 in Makri 2013:25] b. Nullum-ne interea nactu ’s, qui possettibi remissum quem dixti imperare ‘Meanwhile, have you found no one to command for you the army that you mentioned as disbanded?’ [Plaut. Capt. 154-155] I leave aside this possibility because Italian is a negative concord language and, therefore, it usually does not show the double reading phenomenon.

Expletive negation

39

negative marker in the scope of an anti-veridical operator or not. The veridicality of an operator is definite in the following way (cfr. Giannakidou 2006): (57) i. A propositional operator F is veridical if Fp entails or presupposes that p is true in some individual’s epistemic model ME(x); otherwise F is nonveridical; ii. A nonveridical operator F is antiveridical if Fp entails that not p in some individual’s epistemic model: Fpĺ¬p in some ME(x). Put in different words, a veridical/non-veridical/anti-veridical operator measures the speaker’s epistemic attitude toward the truth of an expression: respectively, s/he can be sure of the true of it (ex. factive structure), uncertain (ex. questions or conditionals) or sure of the falsity of it (as with negation). From this point of view, strong NPIs can only occur in negative sentences, because they require an anti-veridical context. Coherently, they cannot occur in questions or in a protasis of a conditional clause because they are non-veridical operators. We can see that the Italian minimizer un tubo (a tube) and the adverb affatto (at all) fall in the category of the strong NPI; the idiomatic expression alzare un dito (to lift a finger) and aver la più pallida idea (to have the faintest idea) do not and, in fact, they can occur in questions and in the protasis of the conditional: (58) a. Luca non capisce affatto / un tubo Luca neg understands at all a tube ‘Luca understands nothing’ b. *Luca capisce affatto / un tubo? Luca understands at all a tube c. *Se Luca capisse affatto / un tubo… if Luca understand.3rdsing.Subj at all a tube (59) a. Luca non ha la più pallida idea di come fare Luca neg has the most faintest idea of how to.do ‘Luca does not have the faintest idea of how he could do it’ b. Luca ha la più pallida idea di come fare? ‘Does Luca have the faintest idea of how he could do it?’ c. Se Luca avesse la più pallida idea di come fare… ‘if Luca had the faintest idea of how he could do it…’ (60) a. Luca non ha alzato un dito per aiutare Maria Luca neg has lift a finger to help Maria ‘Luca has not lifted a finger to help Maria’

40

Chapter 1

b. Luca ha alzato un dito per aiutare Maria? ‘Has Luca lifted a finger to help Maria?’ c. Se Luca avesse alzato un dito per aiutare Maria… ‘If Luca had lifted a finger to help Maria…’ If we consider other elements that are sensible to the polarity of a sentence, we can find the opposite pattern than in NPIs: some elements, such as the Italian già (already), can only occur in affirmative contexts yielding ungrammaticality when they undergo the scope of negation. We can consider such elements as positive polarity items (cfr. Espinal 1997; Israel 1997; Giannakidou 2011): (61) a. Luca ha già finito i compiti (Aff. Context) Luca has already finished the homework ‘Luca has already finished the homework’ b.*Luca non ha già finito i compiti (Neg. Context) *Luca neg has already finished the homework We have now some examples of strong NPI and PPI; the crucial point in this paragraph is that EN clauses license the latter but only the former: (62) a. *Che cosa non ha affatto visto Gianni! what neg has at all seen Gianni b. Che cosa non ha già visto Gianni what neg has already seen Gianni! ‘What has Gianni already seen!’

(Strong NPI) (PPI)

c. EN sentences do not license a conjunction e (and) + the negative word neanche (not-also), but only a conjunction e (and) + the affirmative word anche (also)52–as affirmative sentences usually do– whilst SN shows the contrary behavior: (63) a. Che cosa non ha visto Gianni e anche Luca! what neg has seen Gianni and also Luca What has Gianni seen and also Luca! a’. *Che cosa non ha visto Gianni e neanche Luca! what neg has seen Gianni and not-also Luca b. Gianni ha visto un film horror e anche Luca. Gianni has seen a movie horror and also Luca ‘Gianni has seen a horror movie and Luca too’ 52

(EN)

(Aff. clause)

A similar observation is found in DelFitto and Fiorin (2014) and Chierchia (2013).

Expletive negation

41

b’. *Gianni ha visto un film horror e neanche Luca Gianni has seen a movie horror and not-also Luca c. *Gianni non ha visto un film horror e anche Luca (Neg. clause) Gianni neg has seen a movie horror and also Luca c’. Gianni non ha visto un film horror e neanche Luca Gianni neg has seen a movie horror and not-also Luca ‘Gianni has not seen a horror movie and Luca either’ d. ENS do not license n-words (Laka 1990) 53 following the behaviour of affirmative sentences; again, negative sentences show the contrary pattern: (64) a. *Che cosa non ha dato a nessuno Gianni! what neg has given to nobody Gianni b. *Luca ha dato il libro a nessuno Luca has given the book to nobody c. Luca non ha dato il libro a nessuno Luca neg has given the book to nobody ‘Luca has not given the book to anybody

(*EN) (*Aff. clause) (Neg. clause)

According to (a-d) we have four different tests to distinguish between EN and SN: EN does not change the sentence polarity and it does not license strong NPIs, not-also conjunctions and n-words. I can now investigate how EN has been analyzed in literature.

1.3.1 Some previous analyses of EN Since EN displays several specific properties (as we saw above), many proposals have been advanced in order to catch the peculiarity of EN structures, distinguishing them from SN ones. One of the most popular lines of research is to consider EN both as a member of the negative concord phenomenon and also as subjected to the same licensing conditions as NPIs and n-words. EN would share with them the same “vacuous” semantics or, alternatively, the permeability to the semantic value of the operator selecting it. For example, Fischer (1992) starts his work from the well known fact 53

For a detailed discussion on the semantic and syntactic nature of n-words, see Rizzi (1982); Zeijlstra (2004, 2006) and Chierchia (2013).

42

Chapter 1

that some languages display a specific morpheme for EN structures and another for standard ones54 (for example, in 53, repeated here as 65, the Old English not realizes the standard reading and ne realizes an expletive one; the same is true for the Latin non vs. ne/quin55 and for the Catalan no vs. n’: (65) a. I drede not pat ne pe curs of God [...] wolde brynge me into a I doubt not that neg the curse of God would bring me into a ful yitel eende if I contynuedepus very evil end if I continued thus b. Abans que passi res, jo me n’aniria (Espinal 2000: 54) before that happens anything I meCL EN.go.Cond.1sg ‘Before anything happens, I would leave’ c. Non timeo ne mutent te, neg fear.1stsing.Pres EN change.3rdplu.Subj cl.you, timeo ne impediant fear.1stsing.Pres EN impede.3rdplu.Subj ‘I’m not afraid they may change you, I’m afraid they may impede you’ (Epistulae morales ad Lucilium, Seneca) Fischer (1992) proposes that EN ne in (65a) obtains its vacuous semantics by being a member of a negative concord construction with not. From this point of view, EN looks like a n-word inasmuch as it requires an adequate negative context where another negative element selects it. Along the same line, van Der Wouden (1994a, 1994b, 1997) proposes that negation in paratactic constructions (see above Jespersen 1917, 1924) is licensed by a negative operator realized by the verb of the matrix clause, as in the case of a NPI56. From this point of view, a sentence such as (66) displays a vacuous negation (no) because it is licensed by a negative operator realized by denied (to deny):

54

In order to adapt this hypothesis to languages displaying a unique morpheme for both standard and expletive negation (such as Modern English or Italian), it has been proposed that the same negative marker realizes two different negations, Neg1 and Neg2; standard and expletive negation respectively (viii). Non lo farà a meno che Luca non lo voglia (Italian) neg1 cl.it make.3rdsing.fut unless that Luca neg2 Cl.it wants.3rdsing.Subj ‘S/he will not make it unless Luca wants it’ 55 56

See Ernout and Thomas (1953). See Van der Wurff (1999) for a critique on this proposal.

Expletive negation

43

(66) First he denied [you had in him no right] (Shakespear quoted in van Der Wouden 1994b: 30) Such a mechanism has been extended to all those elements introducing a downward monotonicity context (see Ladusaw1979 and Fauconnier 1975) like unless, if-clauses, etc. Unfortunately, this approach does not explain what happens when EN is realized in a matrix clause where, usually, other negative elements do not occur and, therefore, EN cannot represent the second element of a negative concord phenomenon. Espinal (1992, 2000) seems to face the problem of EN in matrix contexts. She still proposes that EN works as a negative concord phenomenon (NC). According to her, both EN and NC display a cover feature-checking movement in PF57. The difference is that in EN structures a non-veridical feature (FNeg) moves from Neg0 to a higher non-veridical X0, located above CP, where it checks off its feature: (67)

(Espinal 2000: 55 ) The NC phenomenon is different inasmuch as it displays a negative X0 giving a negative meaning to the sentence. Espinal suggests that the XP in EN structures is a DegreeP (DegP) and, since it selects CP, EN can occur in structures that already have CP occupied by a WH-element, as in an exclamative: (68) ¡[CP A cuántas personas] no habrá matado este dictador! to how many people not have+FUT.3sg killed this dictator ‘So many people must have been killed by this dictator!’

57

According to Zanuttini (1991), Haegeman-Zanuttini (1991) and Haegeman (1995), the negative concord phenomenon is a consequence of a [Spec, Head] agreement between a negative operator and a quantifier. This phenomenon can happen covertly or overtly depending on the language. See these sources for a detailed analysis.

44

Chapter 1

Since these kinds of exclamatives cannot be embedded, she assumes that the DegP is the cause: a matrix verb can select, at most, a CP; but there is no space for something bigger. Even though this proposal offers some good predictions, as Makri (2013) points out, it is unclear why negation, and not another element, should be the target of a non-veridical operator instead of a negative one. A similar line is taken by van der Wurff (1999). He starts from Fisher and Espinal’s idea and proposes that EN is syntactically realized as a NegP, in which the Specifier position is filled by a true-value operator, and the head by a negative element: (69)

As per Haegeman and Zanuttini (1991), van der Wurff assumes that both the specifier and the head of NegP must be filled by negative elements in order to give a negative meaning. Since in EN only the head displays a negative element and the specifier is a true-value operator, NegP is no longer negative. He also suggests that the head of such a NegP can be nonovert, as the Specifier, and this is the reason why the presence of the negative marker in EN is optional. He insists that EN only involve the negative marker with a head status because the true-value operator always fills the specifier position. As a proof, French EN is realized only by ne, which, as we saw, represents the Neg0, and not by pas, which is generated in a Specifier position. This also predicts that Modern English does not involve EN: not is a negative adverb generated in [Spec, NegP] and it can only be in complementary distribution with the true-value operator, which is generated in the same position. Even in this analysis the reason why a NegP is specifically involved remains opaque, since negation is completely emptied of its negative meaning.

Expletive negation

45

Yoon (2011) notes that some languages, for example Japanese, display EN in sentences with verbs such as hope, ask, demand, etc. Such verbs are neither negative nor non-veridical, and, therefore, the previous analyses do not catch this particular distribution of EN structures: (70) John-wa Mary-ga ko-nai-ka(-to) kitaisi-te iru (Joon 2011: 66) John-Top Mary-Nom come-Neg-NFcomp hope-Asp ‘John hopes that Mary might come (although it is unlikely to happen).’ She proposes that EN is an evaluative mood marker licensed by a nonveridical operator (exactly like the NPI seen in §1.3). A speaker uses EN because s/he is not sure about the truth of the proposition. In this respect, EN looks like the subjunctive modality, in which a speaker expresses an evaluative uncertainty. EN, which she calls evaluative negation, marked the subjunctive use for politeness, emphatic/rhetorical effects, dubitatives, etc. She proposes that EN is not an imperfection of language, but it is another function of negation working on the conventional implicatures (see Potts 2005). Makri (2013) critics Yoon’s hypothesis that EN is a subjunctive mood marker, showing that it can also occur in indicative, optative and infinitival clauses: (71) Anisiho min tu simvi tipota (Greek, Makri 2013:29) worry EN him.CL happen anything ‘I am worried that something might happen to him’ She suggests that EN is limited to semantic contexts denoting a set of possible propositional alternatives (see below Delfitto and Fiorin 2014) exactly as epistemic modal verbs do. From this point of view, EN would be an epistemic modal marker with a specific feature: it introduces a set of equally probable doxastic alternatives. Interestingly, she uses some Greek, Latin and Hebrew examples in which EN is introduced by a negative complementizer (as min in the previous example); this fact leads her to propose that syntactically, EN clearly lays in the CP domain (moved there, from a lower position, or directly merged there). Unfortunately, both Yoon and Makri’s works do not consider the cases in which EN is in matrix clauses. An alternative line of research is to analyze EN as a scope-marker of an operator. Donati (2000) notes that Italian and French comparative clauses display a combination of EN and focalized elements; in fact, a comparative

46

Chapter 1

is well formed only when two elements are contrasted by the focus mechanism: (72) a. Maria mangia più biscotti in soggiorno di quanti non Maria eats more cookies in living-room of how-many EN ne mangi in cucina (Donati 2000: 63) CL.of-them eats in kitchen ‘Maria eats more cookies in the living room than she does in the kitchen’ b. *Maria mangia più biscotti di quanti non ne mangi Maria eats more cookies of how-many EN Cl.of-then eats in cucina in kitchen In (a), the focalized pair is formed by in soggiorno (in the living room) and in cucina (in the kitchen); in (b), the lack of the first element of the pair makes the sentences ungrammatical. In order to take into account this fact, she proposes an extreme idea: negative morphemes (f.e. the Italian non or the French ne) are not negative at all, but they are a scope-marker of an intentional operator. The operator can be negative, and, consequently, the morpheme obtains a negative meaning, or it can be focal, and, consequently, the morpheme obtains a contrastive semantics. The comparatives show the latter type of negation: the negative morpheme is in the scope of a focal operator and, therefore, negation is considered vacuous. As a proof, she discusses the island blocking effect caused by standard negation (see Ross 196758) but not by EN: (73) a. *Marie mange plus biscuits que Paul n’en mangie pas *Marie mange plus biscuits que [ Opi … Neg0 … [e]i] b. Marie mange plus biscuits que Paul n’en mangie Marie mange plus biscuits que [ Opi …… [e]i] (Donati 2000:66) The blocking effect is excluded with EN because there is no negative feature intervening in the chain Op-trace. As in the previous case, it remains unclear why specifically negation is the scope marker of a focalized structure. Certainly, the syntactic relation between EN and focus is strong, and has to be taken in consideration.

58 Ross (1980) suggests both that comparatives display a WH-movement and that negation provides an island configuration blocking movements.

Expletive negation

47

Delfitto (2013) pursues the same idea as Donati (2000), that EN is a scope marker; however, he proposes that it only marks the scope of a negative operator. He specifically studies the Germanic bevor-clauses (before-clause), in which two negations co-occur: one realizes standard negation in the matrix clause, the other realizes EN in subordinate bevorclauses: (74) a. Mozart vollendete sein Requiem nicht bevor er nicht nach Wien umzog ‘Mozart did not finish his Requiem before he did not move to Vienna’ b. Mozart vollendete sein Requiem nicht bevor er nach Wien umzog ‘Mozart did not finish his Requiem before he moved to Vienna’ The expletive nature of the second nicht in (74a) is evident because it can be omitted (74b) without changing the meaning of the sentence. He proposes that the lower negation in (74a) marks the lower scope of the matrix standard negation during its semantic derivation59. According to Delfitto, the role of EN in temporal clauses is conceptually the same as the that of a negative quantifier (for example, nobody) in a subordinate clause, which is triggered by negation in the matrix clause: (75) Non credo che la polizia abbia arrestato nessuno (Delfitto 2013: 79) For no x, I believe that the police has arrested x Adopting the analysis proposed in Krifka (2010), Delfitto assumes that EN in bevor-clauses is structurally higher than the usual standard negation and, therefore, they do not behave alike. Unfortunately, it remains unclear which structural position it fills. Abels (2002, 2005) tries to find a syntactic characterization for EN in Russian. He starts from the different behaviors of Russian standard and expletive negation: both license the Genitive of Negation (76 a-a’), a structural case depending on the presence of the negative marker ne, but only the SN is able to license ni-phrases (76 b-b’), a particular class of negative concord items such as niþego (nothing): (76) a. Ivan ne þitaet zurnal / zurnala Ivan neg reads journalACC journalDAT ‘Ivan does not read the journal/a journal

59

(Abels 2002:5)

I refer to Delfitto (2013) and Krifka (2010) for a detailed semantic analysis.

48

Chapter 1

a’. Ja bojus' kak by Petr ne narusil èksperimenta I fear how Mod Petr neg ruined experimentGEN I fear Peter might ruin the experiment b. Ivan niþego ne znaet. Ivan NI-what neg knows ‘Ivan does not know anything’ b’. *Ja bojus' kak by þtoby nikto ne opozdal I fear how Mod that Ni-who neg was-late He shows that ni-words are triggered by ne in a minimal syntactic context within TP; on the contrary, the Genitive of Negation does not need such a local constraint. From this fact, he proposes that EN is only a descriptive term and that Russian displays a single negative marker ne differently interpreted depending on its syntactic configuration and derivation. According to Brown and Franks (1995), ne is the head of NegP and it is generated between TP and vP ([CP[TP[NegP[vP[VP]]]]]). Abels proposes that EN ne moves from NegP to CP in LF and, since it does not reconstruct by assumption, it cannot legitimize the ni-words, which require local government. On the contrary, the Genitive on Negation does not require such a constraint and, therefore, it can occur in EN clauses. Moreover, in order to explain the relation with the subjective in subordinate clause, he suggests that EN occupies a high functional projection of modality and, consequently, it negates the positive evaluation of a proposition rather than the proposition in itself. This syntactic explanation is very appealing because it considers expletive and standard negation as two aspects of a unique negation60. A different line of thought is found in Benincà (1996), Portner and Zanuttini (2000) and Zanuttini and Portner (2003). They study Padoan exclamatives displaying EN. These exclamatives show, at least, two unexpected facts: i) the subject clitic inversion in (77) (whereas it is ruled out in standard negative sentences as we saw in §1.1.1 61 ); ii) the cooccurrence of EN and the presuppositional negative element miga in (78) (the Padoan variant of mica): (77) a. *Cossa no ga-la magnà! what neg has-s.Cl eaten 60

(SN)

See Makri (2013) for a criticism of this proposal. Remember the contrast “*No vien-lo?” vs. “No el vien?” in which el is the subject clitic that in affirmative correspondent forms occurs after the verb (Vien-lo? vs. *El vien?) but in negative ones remains before.

61

Expletive negation

b. Cossa no ghe dise-lo! what EN him say-s.Cl. ‘What things he’s telling him!’

49

(EN)

(78) No vien-lo miga? neg comes-s.cl neg ‘He’s not coming?’ Benincà (1996) proposes that EN is presuppositional, and, therefore, it loses its ability to deny a sentence, but it still interacts with the other presuppositional elements, like miga. Portner and Zanuttini (2000) and Zanuttini and Portner (2003) state that the Padoan no is the morphological realization of two different elements: the first is the standard negative marker; the other is a scalar implicature element. The different syntactic nature of those two elements, one the head of NegP and the other an adjunct to verbs, takes into account the distribution of the subject clitic inversion in negative sentences and in expletive ones: standard negation no raises in CP in order to satisfy some criterial requirements leaving intact the order [subject clitic–verb]; the expletive no cliticizes onto the verb and moves with it in CP showing the subject clitic inversion. They call the Padoan EN clitic-no because they refuse the idea that EN is not a negative element at all; they link its meaning to the implicature mechanism. More specifically, the meaning of a negative exclamative with EN is associated with a scalar implicature composed by a set P of true negative propositions 62 . The alternative propositions are displaced on a scale of expectedness/likelihood from the less expected to the most expected. The role of EN is to affirm that only the leftmost proposition, the most expected, is true. Consider, for example, (77b). The set P could be composed in the following way: (79) P = (‘he didn’t tell him he committed a murder’ < ‘he didn’t tell him he is having trouble in his marriage’ < ‘he didn’t tell him he dislikes his neighbor’ < ‘he didn’t tell him it is a nice day outside') In this case, the clitic-type no imposes on the structure that only the leftmost proposition is true, i.e. ‘he didn’t tell him he committed a murder’. All the other alternatives are false and, for a double negation mechanism, the exclamative means: ‘He tell him he is having trouble in his marriage, he dislikes his neighbor and it is a nice day outside'. According to the authors, this is exactly the meaning of (77b) for which the unexpectedness of the 62

For a similar idea see Villalba (2004).

50

Chapter 1

sentence is the result of a “likelihood” mechanism. Since all the presuppositions in P are negative, the uttered sentence shows a negative marker, even though the meaning is expletive. An undesirable consequence of this proposal is that the EN structures should entail both n-words and NPIs because they are realized by negative propositions. However, this is not the case (see Delfitto and Fiorin 2014). Working on the clitic-type negation proposed by Portner and Zanuttini (2000, 2003), Delfitto and Fiorin (2014) proposes another sematic account for EN. They start pointing out that negative exclamatives and negative rhetorical questions express a sort of “universal quantification”. In fact, both (80a) and (80b) state that Gianni has done everything he could do: (80) a. Che cosa non ha fatto Gianni! (Exclamative) what neg has done Gianni ‘What has Gianni done!’ (Gianni has done everything he could do) b. Che cosa non ha fatto Gianni? (Negative Rhetorical Question) what neg has done Gianni ‘What hasn’t Gianni done?’ (G. has done everything he could do) In order to give a common account for both exclamative and rhetorical clauses, they propose that negation reverses the order of informativity usually denoted by exclamatives and questions. More specifically, according to Hamblin (1973), they assume that a question denotes a set of propositions correspondent to every possible answer. Following Han (2002) they also assume that this set of alternative answers is structured within a Boolean’s algebra framework, ordered by an entailment mechanism: the topmost proposition is the most informative (and therefore, the least likely to be true) because it entails the truth of each inferior proposition; the lowest proposition is the less informative (and therefore, the most likely to be true) because it does not have any inferior proposition. For example, suppose that (80a-b) denotes a set of three objects {a,b,c} corresponding to the things that Gianni could do. Boolean’s algebra ordered by an entailment mechanism could be represented as (81):

Expletive negation

51

(81)

(Delfitto and Fiorin 2014: 286) The proposition “Gianni did a, b, and c” is the less likely to be true because it entails the truth of all the other propositions; the contrary goes for propositions such as “Gianni did a”, “Gianni did b”, or ‘Gianni did c”. They propose that the role of EN is to reverse this order of informativity and, therefore, of likelihood. With EN, the topmost proposition “Gianni did a, b, and c” becomes the less informative and, therefore, the most likely to be true; on the contrary, the lowest proposition “Gianni did a” is the most informative and, therefore, the least likely to be true. Put in other words, EN selects the topmost proposition as the meaning of the rhetorical question (or the exclamative). From this special semantics it also follows that such a question/exclamative conveys the “universal quantification” that Gianni has done everything he could do. In summary, we saw several approaches to EN: according to Fischer (1992), Espinal (1992, 2000) and van der Wurff (1999), EN looks like a negative concord phenomenon in which the negativity of the element in the matrix clause licenses the vacuous negation in the subordinate one; similar to the NC approach, van Der Wouden (1994a-1994b-1997) treats EN as a NPI; it has also been proposed that EN is a scope marker of an operator that can be either negative (Delfitto 2013) or not (Donati 2000); alternatively, Yoon (2011) proposes that EN realizes the subjunctive modality and Makri (2013) hypothesizes that it is a epistemic operator; furthermore, EN can be considered a case of presuppositional negation, as mica, with some particular syntactic and pragmatic features (Benincà 1996, Portner and Zanuttini 2000, Zanuttini and Portner 2003, Delfitto and Fiorin 2014). Finally, Abeil (2002, 2005) suggests that EN is a syntactic phenomenon, and negation in those cases occupies a higher structural location than standard one. Even though the goal of this book is not to propose a unified approach to all the EN instances, I am going to use the elements seen above in order to study the structure of Snegs. I will start by discussing EN structures in Italian.

52

Chapter 1

1.4 Focusing on the Italian EN clauses Like other Romance languages, Italian displays a rich system of EN structures. They can be root or subordinate clauses. Let us consider some examples of the former. We can find four different typologies of root clauses instantiating cases of EN: as the second element of the negative concord phenomenon; in expletive 63 negative exclamatives (Delfitto and Fiorin2014); in negative rhetorical questions (see Han 2002) and in not-that clauses (unnoticed case)64. In all these structures, negation is vacuous and it does not reverse the polarity of the proposition (as the English translation shows) in which it occurs: (82) a. Maria non ha visto niente (NC) Maria EN has seen nothing ‘Maria did not see anything’ / ‘#Maria did not see nothing’ b. Che cosa non ha fatto Gianni!” (Expletive Negative Exclamative) what EN has done Gianni ‘What has Gianni done!’ / ‘#What has not Gianni done’ c. Che cosa non ha fatto Gianni per Maria? (Rhetorical questions) what EN has done Gianni for Maria ‘What has Gianni done for Maria!’ / ‘#What has not Gianni done for Maria’ d. Maria non ha pianto che all’inizio (Not-that clauses) Maria EN has cried that at the beginning ‘Maria has only cried at the beginning’ / ‘#Maria has not only cried at the beginning’ EN can also appear in subordinate clauses given the paratactic effect seen in Jespersen (1917, 1924). It can be introduced by a verb (as in interrogative; see Donati 2000), by a conjunction (like rather than (Espinal 2000), until (Renzi and Salvi 1991), unless (Makri 2013) and before (Donati 2000)), by

63

I introduce the adjective expletive in order to mark a specific instance of negative exclamatives. See the Ch. 3 for more details. 64 For sake of clarity, I repeat here the example of each case even though I have already mentioned them above.

Expletive negation

53

the emphatic element “chissà” (‘who knows’; see Renzi and Salvi 1991), and by a comparative (Napoli and Nestor 1976)65: (83) a. Paolo si chiede se Maria non abbia Paolo Cl.to himself wonders if Maria EN have.3rdsing.Subj mangiato troppo (Interrogative clauses) eaten too much ‘Paolo wonders whether Maria has eaten too much’ / ‘#Paolo wonders whether Maria has not eaten too much’ b. Preferisco uscire con te piuttosto che non lavorare prefer.1stsing to go out with you rather than EN to work tutta la sera (Rather than-clause) all the night ‘I prefer going out with you rather than working all night long’ / #I prefer going out with you rather than not working all night long’ c. Rimarrò qui finchè Gianni non arriva (Until-clause)66 st stay.1 sing.fut here until Gianni EN comes ‘I will stay here until Gianni comes’ / ‘#I will stay here until Gianni does not come’ d. Me ne andrò a meno che tu non Cl Cl go away.1stsing.fut unless you EN mangi (Unless-clause) eat.2ndsing.subj ‘I will go away unless you eat’ / ‘#I will go away unless you do not eat’ 65

I leave aside the case of the metalinguistic negation (see Horn 1989, 2010, Ovalle and Guerzoni 2002 and Martins 2014) because negation in such a structure is standard. See, for example, the following speech: (ix) A: Ieri sera ho visto Luca al cinema con una donna ‘Yesterday night I saw Luca with a woman in a movie theatre’ B: No, non era con una donna. Era con sua moglie! ‘No, he was not with a woman. He was with his wife’

According to Horn (1989), the metalinguistic negation is “a device for objecting to a previous utterance”. In such a case negation maintains its “denying” power and the pragmatic re-arrangement of the speech makes it look vacuous. 66 Interestingly, EN is obligatory in the correspondent cleft-structure: (x) E’ finchè (*non) arriverà Gianni che rimarrò qui EN come.3rdsing.fut Gianni that stay.1stSing.Fut here be.3rdsing until ‘It is until Gianni comes that I will stay here’

54

Chapter 1

e. Avvertila prima che non le advise.2ndsing.Imp-CL.her before EN Cl.to her succeda qualcosa di brutto (Before-clauses) happen.3rdsing.Subj something of bad ‘Advise her before some troubles happen to her’ / ‘#Advise her before some troubles do not happen to her’ f. Chissà che non piova! (Chissà-clause) who knows that EN rain.3rdsing.Subj ‘Who knows whether it will rain!’ / ‘#Who knows whether it will not rain’ g. Maria è più intelligente di quanto non lo sia Maria is more smart than EN Cl. be.3rdsing.Subj Carlo (Comparative-clause) Carlo ‘Maria is smarter than Carlo’ / ‘#Maria is not smarter than Carlo’ The contribution of the negative marker non is vacuous even in these cases and, from a purely semantic point of view, we can judge EN in matrix and subordinate clauses as a unique phenomenon. However, if we look closer, we can distinguish two different classes within the set of Italian ENs, showing different syntactic and semantic behaviours.

1.5 Italian EN is not a uniform phenomenon In §1.4, I selected several tests to distinguish EN from SN: EN does not reverse sentence polarity and it does not license strong NPIs, not-also conjunctions and n-words. However, not all Italian EN clauses pass those tests. For example, concerning n-words, the until-clauses license them (84a) whereas expletive negative exclamatives do not (see 63 repeated here as 84b) even though they both are instances of EN: (84) a. Rimarrò qui finché non arriva nessuno stay.1stsing.fut here until EN comes nobody ‘I will stay here until someone comes’ b. *Che cosa non ha fatto a nessuno! what neg has done to nobody Such a fact requires a more detailed analysis. We will see that Italian ENstructures always do not legitimize strong-NPIs and not-also conjunctions, but, concerning weak NPIs and n-words, they show two patterns: those that allow them and those that do not allow them. We can specifically consider

Expletive negation

55

this case by case, starting with SN clauses in order to have a comparison with EN ones: . Standard negation structures: as we saw, they select weak NPIs; strong NPIs, not-also conjunctions and n-words. (85) a. Luca non ha alzato un dito per aiutare Maria (weak NPI) Luca neg has lifted a finger to help Maria ‘Luca did not lift a finger to help Maria’ b. Luca non ha capito un tubo della lezione (strong NPI) Luca neg has understood a tube of-the class ‘Luca did not understand anything about the class’ c. Luca non ha visto Gianni e neanche/*anche Lorenzo (not-also) Luca neg has seen Gianni and neither Lorenzo ‘Luca did not see Gianni and neither he saw Lorenzo’ d. Luca non ha letto niente a nessuno (n-words) Luca neg has read nothing to nobody ‘Luca did not read anything to anybody’ . Expletive Negative Exclamatives: such a structure is not available for weak NPIs; strong NPIs, not-also conjunctions and n-words67. (86) a. Chi non ha alzato un dito per aiutare Maria! who neg has lifted a finger to help Mary . #‘Who has lifted a finger to help Mary!’ (#EN) . ‘Who has not lifted a finger to help Mary!’ (SN) b. *Che cosa non ha affatto fatto Gianni per aiutare Maria!68 what neg has at all done John to help Mary c. Che cosa non ha fatto Gianni e neanche Luca! what neg has done John and not-also Luke . #‘What has John done and Luke either!’ (#EN) . ‘What hasn’t John done and Luke either!’ (SN) d. *Che cosa non ha fatto nessuno! what neg has done nobody . #‘What has someone done!’ (#EN) . ‘What has nobody done!’ (SN) 67 I will indicate the unavailability of one of the two readings of negation-i.e. expletive or standard-by means of the special diacritic #: 68 There is not an immediate explanation for the fact that negative exclamatives do not allow strong-NPIs in a standard negation reading. I will leave this point open for future research.

56

Chapter 1

. Rhetorical Questions: as in Expletive Negative Exclamative, they do not allow NPIs; strong NPIs, not-also conjunctions and n-words. (87) a. Dopo tutto, chi non ha alzato un dito per Maria? after all who neg has lifted a finger for Mary . #‘After all, who has lifted a finger to help Mary?’ (#EN) . ‘After all, who has not lifted a finger to help Mary?’ (SN) b. Dopo tutto, chi non ha affatto aiutato Maria? after all who neg has at all helped Mary . #‘After all, who has helped Maria?’ (#EN) . ‘After all, who has not helped Mary at all?’ (SN) c. Dopo tutto, chi non ha aiutato Maria e neanche Luca? after all who neg has helped Mary and not-also Luke . #‘After all, who has helped Mary and Luke either?’ (#EN) . ‘After all, who has not helped Mary and Luke either?’(SN) d. Dopo tutto, chi non ha aiutato nessuno? after all who neg has helped nobody . #‘After all, who has helped someone?’ (EN) . ‘After all, who has not helped anyone?’ (SN) . “Not…That”- clauses: This kind of clause does not have the correspondent SN form, and the negative marker non can only realize EN; as with the previous two, it does not allow NPIs; strong NPIs, not-also conjunctions and n-words. (88) a. *Luca non ha alzato un dito che all’inizio Luke EN has lifted a finger that at-the beginning b. *Luca non ha affatto pianto che all’inizio Luke EN has at all cried that at-the beginning c. *Luca non ha pianto che all’inizio e neanche Gianni Luke EN has cried that at-the beginning and not-also Gianni d. *Luca non ha fatto niente che all’inizio Luke EN has done nothing that at-the beginning . Rather than-clauses 69 : again, we can see that such a clause does not legitimize NPIs; strong NPIs, not-also conjunctions and n-words.

69

Negation in this clause can be standard or expletive. Interestingly, when it is standard it legitimizes strong NPI but not not-also conjunctions (as the SN usually does). I have no explanation for such a pattern.

Expletive negation

57

(89) a. Preferisco studiare tutta notte piuttosto che non avere la prefer.1stSG.PRS to.study all night rather than neg to.have the più pallida idea di come risolvere il problema most faintest idea of how to.solve the problem . #‘I prefer studying all night long rather than having the faintest idea of how to solve the problem’ (#EN) . ‘I prefer studying all the night rather than not having the faintest idea of how to solve the problem’ (SN) b. Preferisco uscire con te piuttosto che non fare un tubo prefer.1stSG.PRS to.go out with you rather than neg to.do a tube . #‘I prefer going out with you rather than doing something’ (#EN) . ‘I prefer going out with you rather than not doing anything’ (SN) c. *Preferisco uscire con te piuttosto che non lavorare prefer. 1stSG.PRS to go out with you rather than neg to.work tutta la sera con Gianni e neanche con Luca all the night with John and not-also with Luke d. Preferisco uscire con te piuttosto che non lavorare prefer. 1stSG.PRS to.go out with you rather than neg to.work tutta la sera con nessuno all the night with nobody . #’I prefer going out with you rather than working all night with someone’ (#EN) . ‘I prefer going out with you rather than working all night long with nobody’ (SN) . Before-clauses: as the latter cases, they do not allow NPIs; strong NPIs, not-also conjunctions and n-words. (90) a. Spiega-le prima che non abbia la explain.3rdSG.IMP-CL.to her before that neg has.3rdSG.SBJV the più pallida idea di come fare most faintest idea of how to.do . #‘Explain her before she has the faintest idea of how to do it’ (#EN) . ‘Explain her before she has not the faintest idea of how to do it’ (SN) b. Spiega-le prima che non capisca explain.3rdSG.IMP-CL.to her before that neg understood. 3rdSG.SBJV un tubo di come fare a tube of how to.do . #‘Explain her before she understands at all of how do it’ (#EN)

58

Chapter 1

. ‘Explain her before she did not understand at all of how to do it’ (SN) c. Avverti-la prima che non aiuti Gianni advise.3rdSG.IMP-CL.to her before that neg helped. 3rdSG.SBJV John e neanche Luca and not-also Luke . #‘Let her know before she helped John and Luke too’ (#EN) . ‘Let her know before she did not help John and Luke either’(SN) d. Avverti-la prima che non trovi nessuno advise.3rdSG.IMP-CL.to her before that neg find.3rdSG.SBJV nobody . #‘ Let her know before she finds someone’ (#EN) . ‘Let her know before she did not find nobody’ (SN) . Until-clauses: such a structure only refuses the strong NPIs and not-also conjunctions; the weak NPIs and n-words are allowed. (91) a. Rimarrai qui finché non avrai alzato un dito per stay.2rdSG.FUT here until neg have.2rdSG.FUT lifted a finger for aiutar-mi help-CL.me . ‘You will stay here until you will have lifted a finger to help me’ (EN) . #‘You will stay here until you will have not lifted a finger to help me’ (#SN) b. *Rimarrai qui finché non avrai fatto un tubo di stay.2rdSG.FUT here until neg have.2rdSG.FUT done a tube of quello che ti ho detto what that Cl.to you have said c. *Rimarrai qui finché non avrai finito gli esercizi stay.2rdSG.FUT here until neg have.2rdSG.FUT ended the exercises e neanche la traduzione del libro and not-also the translation of-the book d. Rimarrai qui finché non arriverà nessuno stay.2rdSG.FUT here until neg come.3rdSG.FUT nobody . ‘You will stay here until someone will have come’ (EN) . #‘You will stay here until nobody will have come’ (#SN) . Who knows-clauses: as with the until-cluases, they allow the weak NPIs and n-words, but refuse strong NPIs and not-also conjunctions.

Expletive negation

59

(92) a. Chissà che Luca non alzi un dito per aiutar-mi who knows that Luke neg lift.3rdSG.SBJV a finger to help-CL.me .‘Who knows whether Luke lifted a finger to help me’ (EN) .#‘Who knows whether Luke did not lift a finger to help me’(#SN) b. *Chissà che Luca non abbia capito un tubo who knows that Luke neg have.3rdSG.SBJV understood a tube della storia of-the tale c. *Chissà che Luca non abbia fatto i compiti who knows that Luke neg have.3rdSG.SBJV done the homework di matematica e neanche di inglese of math and not-also of English d. Chissà che Luca non aiuti nessuno all’esame who knows that Luke neg help.3rdSG.SBJV nobody at-the exam .‘Who knows whether Luke may help someone for the exam’(EN) .#‘Who knows whether Luke may help nobody for the exam’(#SN) . Unless-clauses: they allow weak NPIs and n-words, but not strong NPIs or not-also conjunctions. (93) a. Il vaso si romperà completamente a meno che Luca non the vase CL.it break.3rdSG.FUT completely unless Luke neg alzi un dito per aggiustarlo lift.3rdSG.SBJV a finger to fix-CL.it . ‘The vase will break completely unless Luke lifted a finger to fix it’ (EN) . #‘The vase will break unless Luke did not lift a finger to fix it’ (#SN) b. *Il vaso si romperà completamente a meno che Luca the vase CL.it break.3rdSG.FUT completely unless Luke non faccia affatto qualcosa per aggiustarlo neg do.3rdSG.SBJV at all someting to fix-CL.it c. *Il vaso si romperà completamente a meno che Luca the vase CL.it break.3rdSG.FUT completely unless Luke non usi un cacciavite e neanche la colla per aggiustarlo neg use a screwdriver and not-also the glue to fix-CL.it d. Il vaso si romperà a meno che non lo aggiusti the vase CL.it break.3rdSG.SBJV unless neg CL.it fix.3rdSG.SBJV nessuno prima nobody first

60

Chapter 1

. ‘The vase will break unless someone fixed it first’70 . #‘The vase will break unless nobody fixed it first’

(EN) (#SN)

. Indirect Interrogative-clauses: they allow weak NPIs and n-words but not strong NPIs or not-also conjunctions. (94) a. Paolo si chiede se Maria non abbia Paul CL.to himself wonders whether Mary neg have.3rdSG.SBJV alzato un dito per aiutare Gianni lifted a finger to help John . ‘Paul wonders whether Mary lifted a finger to help John’ (EN) . #‘ Paul wonders whether Mary did not lift a finger to help John’ (#SN) b. Paolo si chiede se Maria non capisca Paul CL.to himself wonders whether Mary neg understood.3rdSBJ un tubo di come fare a tube of how to.do . #‘Paul wonders whether Mary understood how to do it’ (#EN) . ‘Paul wonders whether Mary did not understand at all how to do it’ (SN) c. Paolo si chiede se Maria non abbia Paul CL.to himself wonders whether Mary neg have.3rdSG.SBJV aiutato Gianni e neanche Luca helped John and not-also Luke . #‘Paul wonders whether Mary helped John and Luke either (#EN) . ‘Paul wonders whether Mary did not help John and Luke neither’ (SN) d. Paolo si chiede se Maria non abbia Paul CL.to himself wonders whether Mary neg have.3rdSG.SBJV aiutato nessuno durante l’esame helped nobody during the exam .‘Paul wonders whether Mary helped someone during the exam’(EN) .#‘Paul wonders whether Mary did not help nobody during the exam’ (#SN)

70

See Moscati (2010) note 17 for the grammaticality judgments on the expletive interpretation of the n-word nessuno in unless-clauses.

Expletive negation

61

. Comparative-clause: they allow weak NPIs71 and n-words but not strong NPIs or not-also conjunctions. (95) a. Maria ha mangiato più mele di quante non ne Mary has eaten more apples of than neg CL.them abbia mai mangiate Gianni be.3rdSG.SBJV never eaten John . ‘Mary has eaten more apples than the ones that John has ever eaten’ (EN) .#‘Mary has eaten more apples than the ones that John has never eaten’ (#SN) b. *Maria ha mangiato più mele di quante non ne Mary has eaten more apples of than neg CL.them abbia affatto mangiate Gianni have.3rdSG.SBJV at all eaten John c. *Maria è più alta di quanto non sia Gianni e Mary is more tall of than neg be.3rdSG.SBJV John and notneanche Luca also Luke d. Maria ha mangiato più mele di quante non ne Mary has eaten more apples of than neg CL.them abbia mangiate nessun altro have.3rdSG.SBJV eaten nobody also . ‘Mary has eaten more apples than the ones that everybody also has eaten’ (EN) . #‘Mary has eaten more apples than the ones that nobody also has eaten’ (#SN) In order to have a shorter overview of different patterns in (85–95), for each Italian EN clause I will summarize the relation with the four elements seen 71 I use a different weak-NPI, mai (never). As other weak-NPIs, mai needs either a negative context or a downward entailing one; on the contrary, it refuses an affirmative context: (xi) a. *Luca ha mai visto Roma (Affirmative clause) Luca has never seen Roma b. Luca non ha mai visto Roma (Negative clause) Luca neg has never seen Roma ‘Luca has never seen Roma’ c. Luca ha mai visto Roma? (Interrogative clause) Luca has never seen Roma ‘Has Luca ever seen Roma?’

Chapter 1

62

above (weak-NPIs, strong-NPIs, not-also conjunctions, and N-words) in the table A: the diacritic + indicates that the syntactic element at the top of the column is legitimized in a specific type of clause; the diacritic–indicates that it is not.

Standard Negation Until-clauses Who knowsclauses Unless-clauses IndirectInterrogatives Comparativeclauses Expl. Negative Exclamatives Rhetorical Questions Not…thatclauses Rather thanclauses Before-clauses

Weak-NPIs

Strong-NPIs

Not-also

N-words

+

+

+

+

+

-

-

+

+

-

-

+

+

-

-

+

+

-

-

+

+

-

-

+

-

-

-

-

-

-

-

-

-

-

-

-

-

-

-

-

-

-

-

-

TABLE 1-1: THIS MATRIX COMBINES 4 SYNTACTIC CONSTRUCTIONS WITH A STANDARD NEGATIVE SENTENCE AND 10 TYPES OF EN CLAUSES

As the table A shows, all the EN clauses reject strong-NPIs and notalso-conjunctions, crucially diverging from the standard negation pattern (first line). I will thus consider these two elements as distinctive in the distribution of EN clauses (beside the fact that all EN clauses share an affirmative polarity regardless of the occurrence of negation). However, with regards to other features (weak-NPIs and N-words), it is possible to discriminate two groups within EN-clauses: one in which weak-NPIs and N-words are legitimate (Until-clauses, Who-knows–clauses, Unless–clauses, Indirect Interrogative-clauses, Comparative-clauses) and one in which they are not (Negative Exclamatives, Rhetorical Questions, Not…that-clauses, Rather than-clauses, Before-clauses). I will label the first group weak EN (light grey in the table) and the second one strong EN (dark

Expletive negation

63

grey in the table), since the former maintains some features of standard negation (i.e. allowing weak-NPIs and N-words) whereas the latter does not. Crucially, the strong EN group realizes a natural class within the EN one, in which each member is defined by the two binary features [-weak NPIs] and [-Nwords]. Any EN clause with both the features [-weak NPIs] and [-Nwords] belongs to this natural class, realizing a proper subgroup in which, clearly, each member conserves specific properties as well. The detection of a natural class within the EN one is as original as puzzling for a correct definition of the EN phenomenon. Of course, some questions may arise: (i) are the distributional differences consequences of different syntactic and semantic configurations? (ii) Why do strong-NPIs and not-also conjunctions always go together like weak-NPIs and the nwords? Interestingly, the complexity of the EN structures is even higher than the distinction in the two categories weak and strong. For instance, an unnoticed case of Italian EN witnesses puzzling and unique properties, involving some peculiar pragmatic and prosodic aspects as well as semanticsyntactic ones.

1.6 An unnoticed case of an Italian EN: Surprise Negation Sentences As we saw above, Italian displays several EN structures both in matrix and in subordinate clauses. They show different semantic, syntactic and pragmatic features even though they share the vacuousness of negation and the incompatibility with strong NPIs and not-also conjunctions. Another difference is that some EN clauses, such as comparatives and untilsentences, are grammaticalized and occur in both written and spoken language; some others only occur in the spoken one (ex. emphatic negative exclamative). In the latter class we find a previously unnoticed case that I will label Surprise Negation (Sneg) sentences: (96) E non mi è scesa dal treno Maria?! and EN CL.to me is got off-the train Mary ‘Mary got off the train!’ As the English translation shows, the sentential meaning is affirmative regardless of the occurrence of the negative marker non. Pragmatically, Snegs are limited to a restricted context in which a speaker is struck by an surprise fact (hence, the label “Surprise”) and s/he wants to communicate it (hence, the assertive force). Snegs display a marked intonation blending

64

Chapter 1

acoustic features pertaining to both questions and exclamatives (hence the ?! combined diacritic). We will see a deepened description of Snegs in the next chapter; for this moment, I just wish to show some examples of Snegs in different semantic and syntactic environments. Firstly, Snegs are not affected by verbal valency and, in fact, verbal valency does not limit their distribution: we find Snegs in zero-argumental, monoargumental (unergative and unaccusative forms as proposed by Burzio 1981, 1986), bi-argumental (transitive) and tri-argumental (ditransitive) verbs72: (97) a. E non ti viene a piovere?! (Zero-argument) and EN ED.to you comes to rain ‘It rained!’ b. E non ti è scesa dal treno Maria?! (Unaccusative) and EN ED.to you is got off-the train Maria ‘Maria got off the train!’ c. E Luca non mi ha telefonato ieri?! (Unergative) and Luca EN Cl.to me has called yesterday ‘Yesterday Luca called me!’ d. E Luca non mi ha mangiato tutta la verdura?! (Transitive) and Luca EN ED has eaten all the vegetables ‘Luca ate all the vegetables!’ e. E Luca non mi ha regalato un libro a Maria?! (Ditransitive) and Luca EN ED has given a book to Maria ‘Luca gave a book to Maria!’ Secondly, Snegs display sentential subjects both in the preverbal and postverbal positions (even though there is a different pragmatic interpretation; see Chapter 2) as Italian usually allows by being a pro-drop language: (98) a. E Maria non mi è scesa dal treno?! (Pre-verbal subject) ‘Maria got Maria got off the train!’ b. E non mi è scesa dal treno Maria?! (Post-verbal subject) ‘Maria got off the train!’

72 All the following sentences start with an expletive E and display the Ethical Dative (ED) mi (to me) or ti (to you). Such facts are very common in Snegs. See the next chapter for a detailed analysis.

Expletive negation

65

Moreover, Snegs are not sensitive to the alternation of the case assignment to the subject as well. More specifically, they can host subjects in both the nominative and dative cases73: (99) a. E io non ho pianto di fronte a tutti?! (Nom.) and I.Nom EN have.1stsing cried in front to everyone ‘I cried in front of everyone!’ b. E a me non sono piaciute le ostriche?! (Dat.) and to me.Dat EN are.3rdplu liked the oysters ‘I liked the oysters!’ Finally, Snegs are not sensitive to verbal diathesis and they can occur in active, as the previous sentences show, and passive structures: (100) E la mela non mi è stata rubata da Luca?! and the apple EN ED.to me is been stolen by Luca ‘The apple has been stolen by Luca!’ Sentences (97-100) show that Snegs are widespread even though they pertain to spoken language. Snegs also display many other properties that make them a very peculiar case of EN. I will dedicate the next chapter to discussing them.

73

See Belletti and Rizzi (1988) for a detailed analysis on Italian psych-verbs clauses.

CHAPTER 2 SNEGS FROM A DESCRIPTIVE POINT OF VIEW

In the previous chapter I added an unnoticed case of EN to the Italian rich system. I called it Surprise Negation and I sketched its distribution in different syntactic contexts. I want now to focus on Snegs in order to understand their semantic, syntactic and pragmatic nature. I will start by showing the vacuousness of the negative marker non in Snegs by means of the tests seen in §1.4 I will also suggest that Snegs belong to the strong-EN class because they do not allow either weak-NPIs or nwords. I will further investigate the semantic and pragmatic properties of Snegs and I will wonder whether such features are enough to define the complexity of the phenomenon. I will finish by showing some puzzling elements related to the Snegs distribution.

2.1 Snegs is an instance of the strong expletive negation As with other EN structures, Snegs do not reverse sentential polarity (1) and do not trigger strong NPIs (2) or not-also conjunctions (3): (1) E non mi è scesa dal treno Maria?! and EN ED.to me is got off-the train Maria ‘*Maria did not get off the train!’ / ‘Maria got off the train!’ (2) *E non mi è affatto scesa dal treno Maria?! and EN ED.to me is at all got off-the train Maria (3) *E non mi è scesa dal treno Maria e neanche Gianni?! and EN ED.to me is got off-the train Maria and neither Gianni and neither have.3rdplu slept Starting from these facts, Snegs have to be considered an instance of EN. Coherently, Snegs legitimize PPIs (4) and not-also conjunctions (5):

Snegs from a descriptive point of view

67

(4) E non mi è già scesa dal treno Maria?! and EN ED.to me is already got off-the train Maria ‘Maria already got off the train!’ (5) E non mi è scesa dal treno Maria e anche Gianni?! and EN ED.to me is got off-the train Maria and either Gianni ‘Maria got off the train and Gianni too!’ We can now wonder whether Snegs belong to the strong or the weak EN classes. As we saw in the previous chapter, only the weak forms of EN allows weak NPIs and n-words; the strong ones do not: (6) a. *E Maria non mi ha alzato un dito per and Maria neg ED.to me has lifted a finger to aiutare Luca?! (weak-NPI) help Luca b. *E Luca non ha la più pallida idea di quello che ha fatto?! and Luca neg has the most faintest idea of what that has done (7) a. *E non mi è sceso dal treno nessuno?! and neg ED.to me is got off-the train nobody b. *E Luca non mi ha mangiato niente ?! and Luca neg ED.to me has eaten nothing

(n-words)

Since Snegs do not allow either weak NPIs (6) or n-words (7), they must be classified as a strong EN; schematically, we should add a new line to the table 1:

Snegs

Weak-NPIs

Strong-NPIs

Not-also

N-words

-

-

-

-

Interestingly, Snegs do not also allow al-words, which, according to Moscati 2010, are polarity items licensed by negative and downward environments. An example of Italian al-words is “alcunché” that we can roughly translate as “anything”: (8) *E Luca non ti ha trovato alcunchè di importante nel cassetto?! and Luca neg ED has found anything of important in-the drawer

68

Chapter 2

Such a fact suggests that Snegs are not downward entailing operators at all. As a proof, in (9b) the inferential relation, usually allowed in negative clauses (9a), cannot occur in Snegs: (9) a. Luca non ha mangiato una pizza ĺ Luca non ha mangiato una pizza ai peperoni ‘Luca has not eaten a pizza’ ĺ Luca has not eaten a pizza with peppers b. E Luca non ha mangiato una pizza?! -/ĺ E Luca non ha mangiato una pizza ai peperoni?! ‘Luca has eaten a pizza!’ -/ĺ ‘Luca has eaten a pizza with peppers’ Moreover, Snegs are perfectly fine when they occur with p-words, such as qualcuno (‘someone’) and qualcosa (‘something’), confirming, again, their affirmative nature: (10) a. E non mi è sceso dal treno qualcuno?! and EN ED.to me is got off-the train someone ‘Someone got off the train!’ b. E Luca non mi ha mangiato qualcosa?! and Luca EN ED.to me has eaten something ‘Luca ate something!’ The vacuous value of negation, the classification into the EN strong class and the non-downward entailing operator nature are only some of the sematic-syntactic features of Snegs. Let us investigate some other properties.

2.2 Some other features of Snegs In the first chapter we saw that Italian displays, at least two tools to deny a sentence: the negative marker non and preverbal n-words. Snegs, like some other EN structures, only allow the former and completely reject the latter: (11) a. *E nessuno mi è sceso dal treno?! and nobody ED.to me is got off to the train b. *E nessuno mi ha mangiato niente?! and nobody ED.to me has eaten nothing

Snegs from a descriptive point of view

69

This basic observation suggests that Snegs are strictly related to the negative marker non, but not to linguistic negation in itself. Since negation is vacuous, we can wonder what its semantic value is; let us investigate the pragmatic properties of Snegs in order to answer to this question. From a pragmatic point of view, Snegs convey completely new information. The whole sentence asserts something completely new for the hearer. This appears evident when Snegs are used as answers to questions: (12) A: Che cosa è successo? ‘What’s happened?’ B: Non mi è scesa dal treno Maria?! EN ED.to me is got off-the train Maria ‘Maria got off the train!’ The sentence in (12B) carries the information that Maria got off the train and nothing of the sentential meaning is presupposed in the speech context. In fact, when some information is already present in the speech context, the same Sneg sentence is not felicitously uttered: (13) A: Chi è sceso dal treno? ‘Who got off the train?’ B: #Non mi è scesa dal treno Maria?! EN ED.to me is got off-the train Maria To understand this pattern better, let me introduce the difference between what I call “Propositional Questions” and “Entity Questions”. With the first label I indicate a question in which the content of the variable is an event and not just an entity (grammaticalized as a PP, DP, etc.); this is the reason why the answer to such a question can only be a whole sentence and not a bare DP/PP. Consider, for example, the following dialogue74: (14) A: What’s happened? B: Maria è appena scesa dal treno ‘Maria got off the train right now’ B’: *Maria Since this kind of question selects propositions as answers, I label it propositional.

74

Italian allow bare PPs and DPs to be answers to questions.

70

Chapter 2

On the contrary, I introduce the label “Entity Questions” to indicate the question in which the content of the variable is a single entity (grammaticalized as a DP, PP, etc.). For example, the congruent answer for the question (15A) is realized as a DP (15B): (15) A: Who got off the train? B: Maria. B’: Maria è scesa dal treno ‘Maria got off the train’ Entity Questions legitimize a whole sentence as well (15B’). However, the new information remains related to the entity (in the example, Maria) and the rest of the sentence can be deleted (15B). This fact also establishes a syntactic difference between the two typologies of question: Entity Questions accept the deletion of some parts of TP (with the exception of the element encoding the value of the variable introduced by the question); Propositional Questions cannot, because the whole sentence is new information. As proof of such a difference, we can consider other grammatical phenomena, such as the fact that only Entity questions legitimize Clitic right dislocation (CLRD), Propositional questions do not. In fact, according to Cecchetto (1999), CLRDs are only allowed when an element is already present in the context of speech (like topics): since the answer to a Propositional question carries new information, we expect the CLRD to be excluded; on the contrary, the CLRD should be possible in Entity questions because they carry some information already presented in the question, as well as new information. Compare the following dialogues: (16) A: A chi ha rubato un orologio Laura? (Entity Questions) to who has stolen a clock Laura ‘Whom did Laura steal a clock?’ B: Laura lo ha rubato a Marco un orologio (CLRD) Laura Cl.it has stolen to Marco a clock ‘Laura stole a clock from Marco’ (17) A: Sembri sconvolto. Cos’è successo? (Prop. Questions) ‘You seem shocked, what’s happened?’ B: #Laura lo ha rubato a Marco un orologio (#CLRD) Laura Cl.it has stolen to Marco a clock

Snegs from a descriptive point of view

71

In (16B), the clitic object lo refers to the direct object un orologio. Since the object is already present in the context of speech, the dislocation is allowed. On the contrary, it is ruled out in (17B) since the whole sentence is new information and the CLRD cannot refer to any element. I will therefore use the CLRD as the diagnostic for entity questions. Crucially, Snegs can only be coherent answers to the Propositional questions, not to Entity ones: (18) A: Sembri sconvolto, cos’è successo? (Propositional Questions) ‘You seem shocked, what’s happened?’ B: Non mi è scesa dal treno Maria?! EN ED.to me is got off-the train Maria ‘Maria got off the train!’ (19) A: A chi ha rubato un orologio Laura? (Entity Questions) ‘Whom did Laura steal a clock?’ B: *E Laura non me lo ha rubato a Marco un orologio?! and Laura EN ED Cl.it has stolen to Marco a clock The occurrence of the CLRD in Entity Questions indicates that a part of the information is not new, and Snegs are ruled out in these cases. On the other hand, the fact that Snegs are congruent answers to questions selecting whole CPs proves that they only carry new information. This is not the only interpretative property of Snegs: they also carry a strong effect of unexpectedness and surprise. From this point of view, Snegs look like the mirativity phenomenon. According to DeLancey (1997, 2001) and Aikhenvald (2005), some languages display a specific morpheme to mark the unexpected nature of the information, something shocking or surprising the speaker. For example, Turkish shows the suffix –mis as a mark of unexpectedness: (20) Ecevit istifa et-miú (DeLancey 1997: 37) Ecevit resignation make-PAST.MIR “ ‘Ecevit has resigned!’ Even though Snegs cannot be considered a mirativity phenomenon because they do not involve a specific morpheme, the pragmatic effect is very alike: the hearer of the utterance is struck by the unexpectedness of the

72

Chapter 2

information 75 . Coherently, Snegs cannot occur with some expressions usually associated with the repetition of an action or with some habits, such as come al solito (as usual): (21) E come al solito, non mi è scesa dal treno Maria(?!) and as to the usual neg ED.to me is got off-the train Maria ‘#As the usual, Maria got off the train’ / ‘As the usual, Maria did not get off the train’ (#Sneg/SN) The context required by Snegs further confirms the unexpected nature of the information. Snegs are successfully uttered when the described fact is out of the blue or it contrasts with some previous expectations. For example, comparing the contexts A and B, Snegs are allowed only in the latter case because the fact that the speaker met Maria on the train appears unexpected: (22) Context A: Every morning Maria gets the same train that I get. As any other days, this morning I saw her. A: Ciao, come va? ‘Hi, how are you?’ B: *Non ti ho incontrato Maria sul treno?! EN ED.to you have.1stsing met Maria on the train Context B: I had not seen Maria for 10 years. This morning I saw her getting the same train that I got. A: Ciao, come va? ‘Hi, how are you?’ B: Non ti ho incontrato Maria sul treno?! EN ED.to you have.1stsing met Maria on the train ‘I met Maria on the train!’ The same Sneg sentence is fine with the context B, but it is not with the context A. The difference between A and B is the likelihood that the event denoted by Sneg happens: in the former is high, in the latter is low. Snegs are, therefore, associated with a surprise effect because it denotes unlikely events. In summary, we have seen that the whole sentence in Snegs carries new information and that such information is unexpected and surprising. 75

See Rett and Murray (2013) for a semantic analysis of mirativity.

Snegs from a descriptive point of view

73

Finally, Snegs show a peculiar intonational pattern. According to Cresti and Firenzuoli (1999), an utterance realizes the fundamental frequency of sounds, F0, to which the speakers are sensitive. Such a fundamental frequency describes the illocutionary intentions of the utterance (Cresti 2000) and is set up by changes of tone heights (pitch accents and edge tones) (Hart et al. 1990, Beckman et al. 2005). Roughly speaking76, pitch accents describe two fundamental tonal layers: high tonal (H*) and low tonal (L*) 77 . They are associated with tonal syllables: H* represents high tone and L* represents the lowest one 78 . According to Avesani (1990), Italian displays 4 types of pitch accent: H*, L*, [L+H*] and [H+L*]. The former two are mono-tonal pitches; the latter two are bi-tonal formed by two F0 patterns: [L+H*] describes a sequence of a pretonic syllable with a minimum value of F0 followed by a stressed syllable with a peak of F0 (raising pattern); [H+L*] describes a sequence of a syllable with peak of F0 followed by a stressed syllable with a minimum value of F0 (falling pattern). Moving on the supra-syllabic segmentation we find the phrase accents [L-/H-] and the boundaries tones [L%/H%]. The former describe intermediate intonational phrases set up by syllabic and phrasal unites; the latter describe the final boundary tones. All these elements (pitch accents, phrase accents and boundaries tones) establish the tonal events forming the basic F0 for every utterance. Let me clarify with some examples. According to Chapallaz (1979), Avesani (1990) and Sorianello (2011), Italian declaratives (figure 2-1) are usually described as a sequence of “H* (or H*+L) L- L%” (falling pattern); yes-no questions (figure 2-2) as a sequence of “H* H- H%” (raising pattern) and exclamatives (figure 2-3) as a sequence of “%H H H* L%” (fallen pattern with an initial high start)79: 76

See Savino et al. (2003) for a deeper description. Cfr. Pierrehumbert (1980). The diacritic * indicates the syllable to which the tonal event is associated. 78 Put in other words, H* is the maximum of F and L* is the minimum. 0 79 The Italian sentence in picture A is: I Rossi partiranno domani probabilmete the Rossi leave.3rdplu.Fut tomorrow probably ‘The Rossi’s family will probably leave tomorrow’ The Italian sentence in picture B is: Fate un errore? do.2ndplu.Pres an error ‘Do you do an error?’ 77

74 Figure 2-1

Chapter 2 (Avesani 1999:6)

Figure 2-2:

Figure 2-3:

(Sorianello 2011: 320)

Picture 2-1 displays two intermediate phrases separated by a short break: ‘The Rossi family will leave tomorrow’ and ‘Probably’. The F0 in the first phrase starts with a pitch accent (above the DP subject); it then decreases until the adverb domani; the second phrase shows a pitch accent with a minimum value of the Fo at the beginning and, then, it ends with a low boundary tone. The F0 pattern is clearly fallen. On the contrary, picture 2-2 shows a clear rising pattern in which the direct object errore displays a high pitch accent followed by a high post-tonic syllable and a high final boundary tone80. The exclamative intonation in picture 2-3 looks like the declarative The Italian sentence in Picture C is: Giuseppe è tornato! Giuseppe is come back ‘Giuseppe has come back!’ 80 See Rossano (2010) and Canepari (2012) for a detailed discussion of the intonational pattern in polar questions.

Snegs from a descriptive point of view

75

one (falling pattern) with the major difference that F0 starts with a higher value at the beginning of the utterance (indicated by %H)81. Now we have all the elements to investigate Snegs’ intonational pattern. I have already said that Snegs displays a marked intonation but I have not sharply defined it. Let me start with the visual representation of their F082: Figure 2-4:

F0 OF THE SENTENCE “IL RAGAZZO NON HA VINTO UN MAIALE?!” (‘THE GUY HAS WON A PIG’).

The F0 in Snegs starts with a high value– as in exclamatives–and ends with a rising pattern–as in questions. This fact is graphically more evident if we compare Snegs with affirmative and negative declaratives83: Figure 2-5:

F0 IN SNEG, NEGATIVE AND AFFIRMATIVE SENTENCES 81

See Sorianello (2011) and the references therein. I collected and analyzed the following data that realize an original part of this book. 83 I collected this data from the experiment described in Ch. 6. The following feature shows the average of F0 in 60 Snegs, 60 affirmative declaratives and 60 negative declaratives. 82

76

Chapter 2

After a high start (%H), the F0 falls and then rises on the verb (H- H*); in the last part it falls again (L*) and it ends with a final rise (H%) which is not present in declaratives. The final part of Snegs looks like the intonational pattern seen in picture 2.2 for interrogatives. Since the Sneg F0 puts together some features of exclamative and interrogatives, at least at the beginning and at the ending of the utterance, we are legitimized in using the double diacritic ?! to indicate their peculiar prosody. To sum up, we saw that Snegs have specific semantic, pragmatic and prosodic features. They carry new information associated with the effect of surprise and they blend some acoustic features of interrogatives and exclamatives. Crucially, Snegs also show some syntactic features that need to be deeply investigated as well.

2.3 Diagnostics for Snegs To investigate the syntactic nature of Snegs we will focus on their interaction with some independent syntactic phenomena and structures. We can see, at first, the relation with the Ethical Dative (ED)84. ED is a non-argumental clitic picking out a person, either the speaker or the hearer, who emotionally participates in the event expressed by the sentence. For this reason, it takes the dative case in languages with a rich case system (like Latin): (23) Quid mihi Celsus agit?’ how me.Dat Celsus.Nom act.3SG ‘How does Celsus do?’

(Roberge and Troberg 2009)

Its distribution in Italian is restricted to the 1st (mi) and 2nd (ti) singular person indicating respectively the emotional relation with the speaker and the hearer. Since Italian displays a rich clitic system (Russi 2008), very often it is difficult to determine if the clitic is Ethical or if it is something else (ex. benefactives). Fortunately, ED can be uniquely associated with some grammatical phenomena representing a sort of diagnostics; for example, ED displays a clitic nature and it cannot be realized by means of a correspondent noun, pronoun or prepositional phrase differing itself from other kinds of clitics, such as benefactive85: 84

For a detailed analysis of ED see Cuervo (2003); Boneh and Nash (2012), Roberge and Troberg (2009) and Franco and Huidobro (2008). 85 I anticipate this point here because I will use it for the Snegs analysis.

Snegs from a descriptive point of view

77

(24) a. Maria mi è scesa dal treno (Ethical Dative) Maria ED.to me is got off-the train ‘Maria got off the train’ a’. *Maria è scesa dal treno a me Maria is got off-the train to me b. Maria mi ha stirato le camice (Benefactive) Maria Benefactive.To me has iron the shirts ‘Maria ironed the shirts for me’ b’. Maria ha stirato le camice a me Maria has iron the shirts to me ‘Maria ironed the shirts for me’ Crucially, ED shows a particular relation with Snegs. As we saw in Chapter 1, in Italian the same negative marker non can express both standard and surprise negation. However, ED can only be associated with the Sneg reading, ruling out a standard one86: (25) Non mi/ti è scesa dal treno Maria [*a me/te] (?!) neg ED.to me/you is got off-the train Maria [to me/you] ‘Maria got off the train!’/‘#Maria did not get off the train’(Sneg/#SN) Since the co-occurrence of ED and the negative marker non always selects the Snegs reading, we can consider ED as a diagnostics for Sneg. Consider now another syntactic property of Snegs: their interaction with some discourse-orientated elements moved in the left periphery of the sentence87. Since Rizzi (1997), it has been proposed that the complementizer phrase is set up by an array of functional projections assembling a layer (CPlayer). CP is the landing site for discourse-orientated elements, such as topicalized or contrastive focalized phrases, and some operator-like elements such as relative or interrogative pronouns88. Italian displays a rich system of such movements and it can realize more than one at the same time:

86

To mark the Ethical Dative nature of the clitic, I use the ungrammaticality of the correspondent PP as in (24). 87 See Chapter 4 for a detailed analysis of the left periphery. As for the ED, for the time being I only give a very brief explanation. 88 The reasons why a phrase should move from its base position to the CP-layer are still debatable; see Chapter 4 for a synthesis of the most prominent positions.

78

Chapter 2

(26) [CP [TopPIl libro]i [FocPA LUCA]k Mario dovrebbe darloi tk] (non a G.) the book to Luca Mario should to give.Cl.it (not to G.) ‘Mario should give the book to Luca (not to Gianni)’ In (26) a topicalized element co-occurs with a focalized one. Both are in a fronting position as the result of a movement or, according to the minimalist approach, of an internal merge: the topic is signaled by the co-reference with the resuptive clitic –lo in a left dislocated structure (CLLD)89; contrastive focus is signaled by the uppercase. Usually, such discourse-oriented elements occur in both affirmative and negative sentences: (27) a. GIANNI è sceso dal treno (non Maria) (Focus in affirmative) Gianni is got off-the train (not Maria) ‘Gianni got off the train, not Maria’ b. GIANNI non è sceso dal treno (non Maria) (Focus in negative) Gianni not is got off-the train (not Maria) ‘Gianni did not get off the train but Maria did’ c. Il libro Gianni lo ha dato a Luca (Topic in affirmative) the book Gianni Cl.it has given to Luca ‘The book, Gianni gave it to Luca’ d. Il libro Gianni non lo ha dato a Luca (Topic in negative) the book Gianni neg Cl.it has given to Luca ‘The book, Gianni did not give it to Luca’ According to Belletti (2001, 2004), Italian also displays a CP-like periphery inside IP immediately above VP. Therefore, it is possible to find topicalized and focalized elements even in a non-fronting position: (28) a. E’ sceso dal treno GIANNI (non Maria) (Foc. in affirmative) is got off-the train GIANNI (not Maria) ‘Gianni got off the train, not Maria’ b. Non è sceso dal treno GIANNI (non Maria) (Foc. in negative) neg is got off-the train GIANNI (not Maria) ‘Gianni did not get off the train, but Maria did’ c. Gianni lo ha dato a Luca il libro. (Topic in affirmative) Gianni Cl.it has given to Luca the book ‘Gianni gave the book to Luca’

89 See Cecchetto (1999) Bianchi and Frascarelli (2010) for a detailed analysis of different Italian topics.

Snegs from a descriptive point of view

79

d. Gianni non lo ha dato a Luca il libro (Topic in affirmative) Gianni neg Cl.it has given to Luca the book ‘Gianni did not give the book to Luca’ Like the fronting case, the topicalized element is signaled by the coreference with the resuptive clitic –lo forming a right dislocated structure; the focused one is again in uppercase90. Crucially, Snegs only allow topicalization, but not focalization, in both fronting (29) and IP cases (30): (29) a. *E GIANNI non mi è sceso dal treno (non Maria)?! (Foc.) and Gianni EN ED is got off-the train (not Maria) b. E il libro Gianni non me lo ha dato a Luca?! (Topic)91 and the book Gianni neg ED Cl.it has given to Luca ‘The book, Gianni gave it to Luca!’ (30) a. *E non mi è sceso dal treno GIANNI (non Maria)?! (Foc.) and EN ED is got off-the train Gianni (not Maria) b. E Gianni non me lo ha dato a Luca il libro?! (Topic) and Gianni EN ED.to me Cl.it has given to Luca the book ‘The book, Gianni gave it to Luca!’ The reason why focalization is ruled out will have to find an explanation in each coherent analysis for Snegs. Finally, there is another element characterizing the syntax of Snegs: they are naturally introduced by an expletive “E” (and) (Poletto 2005). The morpheme E is the logic operator corresponding to and, and it usually coordinates two phrases (see Johannessen 1998). However, in some cases, it does not connect anything, like in the Old Italian sentence in (31):

90

See Samek-Lodovici (2006) for the different sentential positions that a contrastive focalized element can occupy. 91 I leave the subject Gianni in a pre-verbal position even though it sounds unusual for an Italian speaker. I will show that the pre-verbal position in Snegs is a topic position and, to have a more natural effect, the subject should stay in the post-verbal position. See Chapter 5.

80

Chapter 2

(31) e quando avea forbiti i piedi ed elli tornava fuori e and when had cleaned the feet and he came back outside and rinfangavalisi vie piu` e tornava a ricalpitareů il letto. got mudded more and more and came back to step on the bed E partisi e disse a Platone and went away and said to Plato … (Poletto 2005: 227) ‘and when he had cleaned his feet went back outside, put mud on them, came back inside and went up onto the bed. He left and told Plato’ As Poletto argues, when more than two XPs are coordinated, only the last member of such a chain is introduced by e; in (31) all sentences are preceded by e (regardless of their matrix or subordinate nature) and she proposes that it is a case of the expletive E (EE)92. Modern Italian displays EE as well: (32) E che cosa dovrei fare? and what should to do ‘What should I do?’ Interestingly, the occurrence of EE with Snegs is very natural and spontaneous for an Italian speaker (I repeat here the sentence in 1 as 32): (33) a. E non mi è scesa dal treno Maria?! and EN ED.to me is got off-the train Maria ‘Maria got off the train!’ b. E Luca non mi ha mangiato la verdura?! and Luca EN ED.to me has eaten the vegetable ‘Luca ate vegetables!’ c. E non mi hanno cenato a casa di Luca i ragazzi?! and EN ED.to me have.3rdplu dined at home of Luca the guys ‘The guys had a dinner at Luca’s home!’ Since the naturalness of the sentence is higher when it is introduced by EE, I will use it from now on to mark Snegs. As for the left periphery and for the ED, the syntactic codification of EE (see Poletto 2005) further confirms the syntactic peculiarity of Snegs. In summary, the relation between ED, left periphery and EE suggests that Snegs are syntactically encoded, as well as semantically and pragmatically. 92

See §5.5 for a detailed analysis.

Snegs from a descriptive point of view

81

A coherent analysis for Snegs must take into account all those elements. We can start wondering whether Snegs are instances of some other grammatical structures, such as exclamatives and interrogatives, or whether they constitute their own class of Italian EN clauses. In fact, Snegs share with exclamatives and interrogatives some of their features and, moreover, they share a similar intonation. In the next chapter I will investigate this point but, before that, we still have to consider some final points about Snegs.

2.4 Snegs and a puzzling incompatibility As we saw above, Snegs can occur in several types of clauses. Interestingly, there are some cases in which Snegs are not allowed without clear reason; such as, for example, in some copular sentences. According to Moro (1988, 1997, 2000), copular sentences are small clauses (SC) in which one of the two DP raises to the higher subject position in the matrix clause. Let me take, for example, a SC set up by two DPs: the first is the subject of the SC (Gianni) and the second one is its predicate (the alleged culprit). If the subject of the small clause moves, then the sentence is labeled canonical, giving the order Subject–copula–predicate (34a); if the predicate of the small clause moves, then the sentence is labeled inverse, given the order predicate–copula- subject (34b)93: (34) a. Giannii è [SC ti il presunto colpevole] Gianni is the alleged culprit ‘Gianni is the alleged culprit’ b. [Il presunto colpevole]i è [SC Gianni ti] the alleged culprit is Gianni ‘The alleged culprit is Gianni’

(Canonical form) (Inverse form)

Italian copular sentences also display a form in which (apparently) both elements forming the SC remain in situ (Moro 2009): (35) a. pro E’ [SC [DP Gianni] [DP il presunto colpevole]] is Gianni the alleged culprit ‘It is Gianni the alleged culprit’ b. pro E’ [SC [DP il presunto complevole] [DP Gianni]] is the alleged culprit Gianni ‘It is Gianni the alleged culprit’ 93

See Moro for a detailed analysis.

82

Chapter 2

As is known, Italian is a pro-drop language. The preverbal subject can be phonologically omitted and, in order to satisfy the EPP, a null pronoun pro is inserted94. The order of constituents in (35) is apparently free, even though their informational conditions are different (Belletti 2001): in (36a) the subject Gianni is interpreted as the new information (the focus) and the predicate the alleged culprit is the old one (the topic); in (35b) the subject Gianni is interpreted as the old information (the topic) and the predicate the alleged culprit is the new one (the focus). As a proof, compare the following speeches: (36) A: Chi è il presunto compevole? ‘Who is the alleged culprit?’ B: E’ Gianni il presunto colpevole ‘Gianni is the alleged culprit’ B’: *E’ il presunto colpevole Gianni is the alleged culprit Gianni (37) A: Chi è Gianni? ‘Who is Gianni?’ B: E’ il presunto colpevole Gianni ‘Gianni is the alleged culprit’ B’: *E’ Gianni il presunto colpevole is Gianni the alleged culprit’ In (36B), the subject Gianni is the new information and is coherently in the focus position; in (36B’), the focus position is occupied by the predicate, which carries the old information giving a crush into the interpretation. The pattern in (37) is the contrary. Each DP position is associated with a specific discourse-oriented feature: roughly speaking, the closest to the verb is the focus; the most distant is the topic95. Coming back to Snegs, we can see that they allow both canonical and inverse copular sentences: (38) a. E Giannii non è [SC ti il presunto colpevole]?! (Canonical Form) and Gianni EN is the alleged culprit ‘Gianni is the alleged culprit!’

94 See Moro (1997) for an alternative proposal on the pro in copular sentences. He suggests that pro is a null predicate. 95 I do not deepen this issue here because it is beyond the aim of this book. See Belletti (2001) and the references cited therein.

Snegs from a descriptive point of view

83

b. E il presunto colpevole non è [SC Gianni ti]?! (Inverse form) and the alleged culprit EN is Gianni ‘The alleged culprit is Gianni!’ However, when we come to copular sentences with DPs in situ, only the form To be–subject–predicate is allowed; the one To be–predicate–subject is ruled out: (39) a. E non è Gianni il presunto colpevole?! and EN is Gianni the alleged culprit ‘Gianni is the alleged culprit!’ b. *E non è il presunto colpevole Gianni?! and EN is the alleged culprit Gianni Interestingly, such a restriction is valid only for the to be verb, which, according to Moro (1997, 2017), is not predicative96. In fact, both forms become available, replacing to be with a predicative and raising verb such as to become: (40) a. E non è diventato Gianni il presunto colpevole?! and EN is became Gianni the alleged culprit ‘Gianni became the alleged culprit!’ b. E non è diventato il presunto colpevole Gianni?! and EN is became the alleged culprit Gianni ‘Gianni became the alleged culprit!’ There are not apparent reasons why Snegs cannot occur in sentences such as To be–predicate–subject, with the exception of the non-predicative nature of the copular verb. Surprisingly, ED appears to be sensitive to this feature too: it occurs in predicative structures but not in copular sentences (regardless of the typology): (41) a. Giannii mi è diventato [SC ti il presunto colpevole] Gianni ED.to me is became the alleged culprit ‘Gianni has become the alleged culprit’ b. [Il presunto colpevole]i mi è diventato [SC Gianni ti] the alleged culprit ED.to me is became Gianni “The alleged culprit has become Gianni’

96

In his analysis the verb to be is considered a tense realization.

84

Chapter 2

c. Mi è diventato [SC Gianni il presunto colpevole] ED.to me is became Gianni the alleged culprit ‘It has happened that Gianni has become the alleged culprit’ d. Mi è diventato [SC il presunto colpevole Gianni] ED.to me is became the alleged culprit Gianni ‘It has happened that the alleged culprit has become Gianni’ (42) a. *Giannii mi è [SC ti il presunto colpevole] Gianni ED.to me is the alleged culprit b. *[Il presunto colpevole]i mi è [SC Gianni ti] the alleged culprit ED.to me is Gianni c. *Mi è [SC Gianni il presunto colpevole] ED.to me is Gianni the alleged. culprit d. *Mi è [SC il presunto colpevole Gianni] ED.to me is the alleged culprit Gianni The pattern in (41) shows that ED can appear in any combination with the to become verb. On the contrary, ED cannot occur in copular clauses (42). Both these patterns lead us to think that the non-predicative nature of to be is the cause of the ungrammaticality in (39b). However, it still remains puzzling why the non-predicative nature of a verb interacts with ED and Snegs. I will leave this point for future research because it is beyond the aim of this book, and I will go further with my analysis of Snegs, wondering whether they are an instance of another EN class (such as exclamatives and interrogatives) or whether they instantiate their own class.

CHAPTER 3 WHAT SNEGS ARE NOT: A JUSTIFICATION OF A DEDICATED STRUCTURE

In Chapter 2, I sketched some semantic, prosodic and syntactic properties of Snegs. Among other things, we saw that Snegs’ intonational pattern displays both the raising of the tonic function at the end of the utterance– following the interrogative pattern–and a high tonic function at the beginning of the utterance–following the exclamatives pattern. Such similarity leads us to wonder whether Snegs instantiate either a case of interrogatives or a case of exclamatives. For each of these structures I will indicate the elements that are shared with Snegs, and we will see that Snegs do not completely overlap with any of them. For this reason I will pursue the possibility that Snegs form their own class of EN structure, calling for its own syntactic analysis. I will also reject the idea that Snegs are bi-clausal structures (as has been proposed for some expletive negation structures).

3.1 Snegs are not Negative Rhetorical Questions The class of interrogatives includes different types of sentences. As is well known, interrogatives can be polar questions or wh-questions; they can display the interrogative particles in situ or in a dislocated position, etc. Among all the different types of interrogatives, the closest types to Snegs are the Negative Rhetorical Questions (NRQs). In fact, both NRQs and Snegs display the negative marker non even though their meaning is affirmative (expletive negation) and both realize the illocutionary force of an affirmative assertion 97 (Schmidt-Radefeldt 1977; Reese 2006; Wang 2014):

97

I repeat Sneg sentences in the following comparisons even though I have already presented them in the previous chapters in order to make the comparisons more fluent.

86

Chapter 3

(1) a. Dopo tutto, Luca non è arrivato primo? (NRQ) after all, Luca EN is arrived first ‘After all, hasn’t Luca arrived first?’ = Luca arrived first / #Luca did not arrive first b. E Luca non mi è arrivato primo?! (Sneg) and Luca EN ED.to me is arrived first ‘Luca arrived first’ / ‘#Luca did not arrive first’ According to Sadock (1974) and Han (1998, 2002), RQs appear as questions but perform as assertions of opposite polarity: if they are negative in form, their meaning is affirmative; if they are affirmative in form, their meaning is negative. Such a “polarity inversion effect” clearly applies for Snegs too (1b). Interestingly, this similarity goes even deeper. In fact, NRQs belong to the strong class of EN exactly like Snegs (see Chapter 1). Just to be clear, I recall here some fundamental data comparing Negative Questions (NQs), NRQs and Snegs in order to show the similarity between the two EN structures as opposed to the standard one: - Both do not legitimate n-words: (2) a. Non ha telefonato nessuno per me? neg has called nobody for me ‘Anyone called me?’ b. *Dopo tutto, non ha telefonato nessuno per me? after all, EN has called nobody for me c. *E non mi ha telefonato nessuno per me?! and EN Cl.to me has called nobody for me

(NQ) (*NRQ) (*Sneg)

- Both do not allow not-also conjunctions (3) a. Maria non ha aiutato Gianni e neanche Luca? (NQ) Maria neg has helped Gianni and neither Luca ‘Has not Maria helped either Gianni or Maria?’ b. *Dopo tutto, Maria non ha aiutato Gianni e neanche Luca? (*NRQ) After all Maria EN has helped Gianni and neither Luca c. *E non mi è scesa Maria dal treno e neanche Luca?! (*Sneg) and EN Cl is got-off Maria to-the train and neither Luca - Both do not license either strong or weak NPIs98: 98

For similar data, see Vann Rooy (2004) and Guerzoni (2004).

What Snegs are not

87

(4) a. Luca non ha capito un tubo della lezione? (NQ) Luca neg has understood a tube of-the class ‘Has not Luca understood a tube about the class?’ a’. Luca non ha alzato un dito per aiutare Gianni? Luca neg has raised a finger to help Gianni ‘Has not Luca raised a finger to help Gianni?’ b. *Dopo tutto, Luca non ha capito un tubo della lezione? (*NRQ) after all Luca EN has understood a tube of-the class b’. *Dopo tutto, Luca non ha alzato un dito per aiutare Gianni? after all Luca EN has raised a finger to help Gianni c. *E Luca non mi ha capito un tubo della lezione?! (*Sneg) and Luca EN ED has understood a tube of-the class c. *E Luca non mi ha alzato un dito per aiutare Gianni?! and Luca EN ED has raised a finger to help Gianni - Both legitimate PPIs (5) a. *Luca non ha già finito i compiti? (*NQ) Luca neg has already finished the homework b. Dopo tutto, Luca non ha già finito i compiti? (NRQ) after all Luca EN has already finished the homework ‘After all, hasn’t Luca finished the homework?’ c. E Luca non mi ha già finito i compiti?! (Sneg) and Luca EN ED has already finished the homework ‘Luca has already finished the homework!’ The sentences (2-5) show that NRQs and Snegs belong to the same EN class. Moreover, they also share the same behavior with the ED and the Expletive E: (6) a. E dopo tutto, Luca non mi è arrivato primo alla gara? (NRQ) EE after all Luca EN ED is come first to-the race ‘After all, hasn’t Luca come in first in the race?’ b. E Luca non mi è arrivato primo alla gara?! (Sneg) EE Luca EN ED.to me is come first to the race ‘Luca has come first in the race!’ In the previous chapter we saw that ED and EE are strongly connected to Sneg structures. NRQs seem to be equally sensitive to them, and such a relation further reinforces the hypothesis that Snegs are instances of NRQs.

88

Chapter 3

Unfortunately, this is not the case because they differ in several other aspects, making their assimilation impossible. First at all, Snegs are correct answers to questions, whereas NRQs are not: (7) A: Sembri sconvolto. Cos’è successo? ‘You look upset. What’s happened?’ B: Non mi è scesa dal treno Maria?! ‘Maria got off the train!’ B’: *Dopo tutto, non è scesa dal treno Maria? after all EN is got off the train Maria

(Sneg) (*NRQ)

Along the same line, NRQs usually allow hearers to answer back either for confirmation or for refutation, whilst Snegs do not: (8) A: Dopo tutto, Luca non è arrivato primo come avevo detto? after all Luca EN is come first as had said ‘After all, hasn’t Luca come in first as I said?’ B: E si, hai sempre ragione tu! ‘Yes, he did! You always is right!’

(NRQ)

(9) A: E Luca non mi è arrivato primo?! EE Luca EN ED.to me is come first ‘Luca has come in first!’ B: *E si, hai sempre ragione tu! EE yes have.2ndsing.Pres always right you

(Sneg)

Interestingly, Snegs do not also display some stylistic elements uniquely associated with the RQs. For example, according to Han (1998, 2002), RQs can be introduced by the expressions dopo tutto (after all) and comunque (anyway), whereas Snegs completely reject it: (10) a. Dopo tutto, Luca non è arrivato primo? (NRQ) after all Luca EN is come first ‘After all, hasn’t Luca come in first?’ a’. Comunque, Luca non è arrivato primo? anyway Luca EN is come first ‘Anyway, hasn’t Luca come in first?’ b. *Dopo tutto, Luca non mi è arrivato primo?! (*Sneg) after all Luca EN ED.to me is come first b’. *E comunque, Luca non mi è arrivato primo?! EE anyway Luca EN ED.to me is come first

What Snegs are not

89

Furthermore, from the pragmatic point of view, NRQs differ from Snegs because the former do not necessarily carry a surprise effect, whereas the latter do. As a proof, NRQs can occur with some expressions usually associated with the repetition of an action or with some habits (ex. come al solito), Snegs cannot (see §2.2) because such expressions lack unexpectedness: (11) a. Come al solito, Luca non è arrivato primo? (NRQ) as to the usual Luca EN is come first ‘As usual, hasn’t Luca come in first?’ b. *E come al solito, Luca non mi è arrivato primo!? (Sneg) EE as to the usual Luca EN ED is come first Finally, Snegs and NRQs also differ in allowing wh-elements, in fact, only NRQs allow them, Snegs do not99: (12) a. Dopo tutto, che cosa non ha fatto Maria per Gianni? (NRQ) after all what thing EN has done Maria for Gianni ‘After all, what hasn’t Maria done for Gianni?’ b. *E che cosa non mi ha fatto Maria per Gianni?! (*Snegs) and what EN ED has done Maria for Gianni To conclude, on the one hand Snegs and NRQs are similar because both display strong EN, ED and EE; on the other hand, they behave differently in respect to some pragmatic and semantic aspects: only NRQs display whelements and allow elements such as dopo tutto, comunque and come al solito, whilst Snegs do not; on the contrary, only Snegs can be used as correct answers to questions, unlike NRQs. As a consequence, we have to discard the hypothesis that Snegs belong to the class of rhetorical questions. Since Snegs are not NRQs, a coherent alternative is to consider them an instance of exclamatives.

3.2 Snegs are not Expletive Negative Exclamatives We saw above that Snegs’ intonational pattern looks like the interrogative one, because both display a rising pattern at the end of the utterance. However, if we look at the beginning of the utterance, we find that the F0 in 99

The Sneg sentence in (12b) could appear grammatical for some Italian speakers because it can be easily confused with an interrogative. A correct Sneg intonation is necessary to see the ungrammaticality of (12b).

90

Chapter 3

Snegs looks like that in exclamatives (see. §2.2). Starting from this fact, I wonder–as in the case of rhetorical questions–whether the common intonational pattern suggests a common categorical membership between Snegs and exclamatives. Before going into details, we need to identify a specific type of negative exclamative in order to make the comparison with Snegs more accurate. In fact, under the label “exclamatives” there are several different elements (see Sorianello 2011) that can potentially confuse the analysis. To begin, I restrict the use of “exclamatives” to only the exclamative statements; leaving out the other clausal types with an exclamatory force (Huddleston 1984; Collins 2005). To be clearer, I will not consider sentences such as “What the hell are you doing?” as exclamatives because they are questions, even though they have the illocutionary force of exclamatives. Moreover, moving on to negative exclamatives, we find that such sentences are often semantically ambiguous between two readings: one in which the clausal meaning is negative, and one in which it is affirmative. Consider, for example, the following sentence: (13) Che cosa non ha mangiato Maria! what neg has eaten Maria a. ‘What Maria has not eaten!’ (=Maria has not eaten something) b. ‘What Maria has eaten!’ (=Maria has eaten everything100) In the first reading, the speaker highlights that Maria has not eaten something and the clausal meaning is still negative; in the second, the speaker communicates that Maria has eaten everything, displaying an expletive negation. From these facts, I will call the first type “Negative Exclamative” (NE) and the second “Expletive Negative Exclamative” (ENE)101. Interestingly, the two types of structures differ grammatically. According to Grimshaw (1979) and Zanuttini and Portner (2003), exclamatives are factives and, therefore, they can only be embedded under factive predicates (14a), rejecting the non-factive ones (14b): (14) a. E’ incredibile [che cosa ha visto Gianni]! is incredible what has seen Gianni ‘It is incredible what Gianni has seen!’ 100 101

See Chapter 1 for the discussion on the semantics of exclamatives. To the best of my knowledge, nobody has made such a partition before.

What Snegs are not

91

b. *Luca crede [che cosa ha visto Gianni]! Luca believes what has seen Gianni Like other exclamatives, NEs and ENEs follow this pattern: (15) a. E’ incredibile [che cosa non ha mangiato Luca]! is incredible what neg has eaten Luca . ‘It is incredible what Luca has not eaten’ (Luca has not eaten a particular food and this fact is unbelievable) (NE) . ‘It is incredible what Luca has eaten’ (Luca has eaten everything and this fact is unbelievable) (ENE) b. *Luca crede [che cosa non abbia mangiato Luca]! Luca believes what neg had.3rdsing.Subj eaten Luca However, focusing on a specific class of factive predicates, to know-verbs, we find that only the NEs can be embedded; ENEs cannot: (16) Luca sa [che cosa non ha mangiato Maria]! Luca knows what neg has eaten Maria a. ‘Luca knows what Maria has not eaten’ b. ‘*Luca knows that Maria has eaten everything’

(EN) (*ENE)

The expletive reading is completely ruled out in (16), whereas the standard one is preserved102. From this fact it follows that we must consider NEs and ENEs as different structures. Coming back to the comparison with Snegs, ENEs seem to be the closest exclamative type to them. In fact, both structures share, at least, the occurrence of EN, whereas NEs do not. Like NRQs, ENEs have the illocutionary force of an assertion, exactly as Snegs do103. (17) a. Che cosa non ha mangiato Maria! (ENE) ‘What Maria has eaten!’ (=Maria has eaten everything) b. E non mi è scesa dal treno Maria!? (Sneg) ‘Maria got off the train’

102

Finding an explanation for such a phenomenon is beyond the aim of this book. For the moment, I only use the descriptive attestation. 103 I repeat here the sentence in (13) in order to be clearer.

92

Chapter 3

Interestingly, they also share the same strong type of EN. In fact, ENEs license PPIs (18) but not NPIs (19)–both weak and strong–and they do not allow n-words (20) and not-also conjunctions (21)104: (18) Che cosa non ha già mangiato Maria! what EN has already eaten Maria ‘What Maria has already eaten’ (19) a. *Chi non ha alzato un dito per Gianni! who EN has lifted a finger for Gianni b. *Che cosa non ha affatto regalato Gianni! what EN has at all given Gianni (20) *Che cosa non ha fatto nessuno per nessuno! what EN has done nobody for nobody (21) *Che cosa non ha fatto Gianni e neanche Luca! what EN has done Gianni and neither Luca Moreover, both Snegs and ENEs carry a surprise evaluation for the event described by the sentence105. The speaker attitudinal component in ENEs is marked by a sense of unexpectedness much like the one seen for Snegs. As a proof, both Snegs and ENEs cannot occur with the expression come al solito (as usual): (22) a. *Come al solito, che cosa non ha mangiato Maria! (ENE) as to the usual, what neg has eaten Maria b. *E come al solito, non mi è scesa dal treno Maria?! (Sneg) and as to the usual neg ED is got off-the train Maria Finally, ENEs also legitimize ED and EE: (23) Maria è uscita a cena ieri sera. E che cosa non Maria is gone out to supper yesterday evening. EE what EN non mi ha mangiato! Del caviale! EN ED.to me has eaten of the caviar ‘Maria went out for a dinner yesterday night. What she has eaten! Caviar!’ 104

We have already seen these data in Chapter 1, I repeat here for the sake of clearness. 105 For the semantic value of exclamatives see, above all, Michaelis and Knud (1996), Zanuttini and Portner (2003), Collins (2005) and the references therein.

What Snegs are not

93

Since ED and EE are specifically associated with Snegs, it is reasonable to interpret Snegs and ENEs as common members of the same grammatical category. Once again, this is not the case. First of all, Snegs cannot be embedded under both to know-verbs (24), and any other factive predicates (25), contrary to ENEs: (24) *Luca sa [che Maria non mi è scesa dal treno]?! Luca knows that Maria EN ED is got off-the train (25) a. E’ incredibile [che cosa non abbia mangiato Luca]! is incredible what EN had.3rdsing.Subj eaten Luca ‘It is incredible what Luca has eaten’ (ENE) b. *E’ incredibile che Maria non mi sia scesa is incredible that Maria EN ED.to me was.3rdpres.Subj got dal treno!? (*Snegs) off-the train Such a difference may alone distinguish Snegs from exclamatives. However, other elements can be considered in order to reinforce this distinction. For example, Snegs can be answers to questions; ENEs cannot. More specifically, according to Grimshaw (1979) and Zanuttini & Portner (2003), the unavailability to be used as answers is a crucial property of exclamatives. Since Snegs are perfectly grammatical in these cases, this further establishes a distinction between them: (26) A: Sembri sconvolto. Cos’è successo? ‘You look upset. What’s happened?’ B: Laura non mi ha rubato un orologio a Marco?! Laura EN ED.to me has stolen a clock to Marco ‘Laura has stolen a clock to Marco!’ B: *Che orologio non ha rubato Laura a Marco! what clock EN has stolen Laura to Marco

(Sneg) (*ENE)

Another fact is that Snegs and ENEs differ in allowing wh-elements. Only ENEs allow them, Snegs do not106: (27) a. Che cosa non ha fatto Maria per Gianni! what EN has done Maria for Gianni ‘What has Maria done for Gianni!’ 106

(ENE)

Some Italian speakers may be tempted to interpret the Sneg sentence in (27) as grammatical. See the note 12.

94

Chapter 3

b. *E che cosa non mi ha fatto Maria per Gianni?! and what EN ED has done Maria for Gianni

(*Sneg)

In summary, even though Snegs and ENEs show the same kind of EN and both host ED and EE, they differ in some other respects: ENEs, but not Snegs, can be embedded under factive predicates and display wh-elements; on the contrary, only Snegs can be used as correct answers to a question. Such elements exclude that Snegs belong to the same grammatical category of ENEs 107. Since I had already discarded the hypothesis that Snegs are NRQ, I am tempted to assume that any assimilations of Snegs into other grammatical structures is impossible. However, before pursuing this path, one last alternative has to be tested. More specifically, I’m going to compare Snegs with a structure displaying both expletive and standard negation.

3.3 Snegs are not Bi-clausal Structures According to Poletto and Zanuttini (2013), Italian displays a type of emphatic sentence pointing out the speakers’ high level of agreement, or disagreement, with a previous assertion or question. For example, in the following conversation, the answer B strongly confirms the expectation presupposed by the question; the answer B’, instead, strongly contradicts it: (28) A: È poi arrivato Gianni? (Poletto and Zanuttini 2003:124) is then arrived Gianni ‘Did Gianni arrive in the end?’ B: Sì che è arrivato. yes that is arrived ‘Of course he arrived!’ / ‘Absolutely!’ B’: No che non è arrivato. no that neg is arrived ‘He did not!’ / ‘Not at all!’ 107

Interestingly, Italian does not display a polar version of ENEs. In fact, a possible correspondent polar form for the ENE sentence “Che cosa non ha mangiato Laura!” is unavailable: (i) Laura non ha mangiato caviale! Laura neg has eaten caviar a. Laura has not eaten caviar (NE) b. *Laura has eaten caviar (*ENE) Such unavailability possibly reinforces the different status of Snegs, since they only realized a polar form (see Ch. 5).

What Snegs are not

95

The Italian polar particles si (yes) and no (not) emphasize the content of the sentence, which is introduced by the declarative complementizer che (that). Starting from the sequence of words polar particle - complementizer, Zanuttini and Poletto call such sentences si che/no che sentences. Interestingly, the negative form in (28B’) displays two negative elements setting up a unique instance of negation. I will, in the first instance, consider the first no as expletive. Before investigating the relation between no-che sentences and Snegs, we need to know something more about such structures. In Zanuttini and Poletto’s analysis, si che/no che sentences are biclausal structures and, according to the cartographic approach developed by Rizzi (1997) and Benincà and Poletto (2004), they can be captured by the following representation: (29) The polar particles are merged in a higher clause and, from there, they bind an operator in the lower one introduced by the complementizer che; moreover, the higher clause displays in hanging topic position (HTP) a copy of the uttered sentence which is deleted before being pronounced.108 Interestingly, no-che sentences display some features shared by Snegs and, therefore, I wonder whether they are bi-clausal structures as well. Theoretically, nothing forbids EN in Snegs from belonging to another clause, which in turn selects the uttered sentence. First of all, (28) shows that no-che structures are correct answers to questions, exactly like Snegs. Moreover, both Snegs and no-che clauses do not legitimize wh-elements (30) and fronting focused elements (31): (30) a. *Che cosa no che non fai? (*No che-clause) what not that neg does b. *E che cosa non mi ha fatto Maria per Gianni?! (*Snegs) and what EN ED has done Maria for Gianni (31) A: Immagino che voi abbiate già visto Avatar e imagine that you have.2ndplu.Subj already seen Avatar and anche The Artist (P&Z, p. 128) also The Artist ‘I figure you have already seen Avatar and also The Artist’ 108

See Zanuttini and Poletto (2013) for a detailed analysis.

96

Chapter 3

B: *AVATAR no che non abbiamo visto (ma l’altro Avatar no that neg have.1stplu. seen (but the-other film si) (*no che sentence) movie yes) B’: *AVATAR non abbiamo visto (ma l’altro film si)?! (*Sneg) Avatar EN have seen (but the-other movie yes) Again, both the structures can be naturally introduced by the EE: (32) a. E no che non sono andato al cinema (No che-clause) EE no that neg am went to the movie theatre ‘No, I did NOT go to the cinema!’ Finally, no-che structures cannot be embedded under any predicates, exactly like Snegs: (33) a. *Mi ha detto che/di no che non viene Luca109 to me has said that/of no that neg comes Luca b. *Mi spiace che no che non viene Luca to me am sorry that no that neg comes Luca Nevertheless, the similarities in (30)–(33) are not sufficient. In fact–exactly as in the case of NRQs and ENEs–there are several other elements distinguishing the two structures. The most significant one is the different nature of the negative marker that is involved: non is a head, whereas no is a maximal projection (see Chapter 1 for a detailed discussion). Beside this syntactic difference, there is another one: no-che structures are always negative in meaning, whilst Snegs are affirmative: (34) a. No che Maria non è scesa dal treno no that Maria neg is got off-the train ‘Maria did NOT get off the train!’ b. E Maria non mi è scesa dal treno?! ‘Maria got off the train’

(No che-clause) (Sneg)

Crucially, Z&P (2013) shows that no cannot ever occur in affirmative sentences:

109

See Sanuttini and Poletto (2013), p. 138.

What Snegs are not

(35) a. No che non mi hanno invitato a casa loro! no that neg me have invited to home their ‘They did NOT invite me to their home’ b. *No che mi hanno invitato a casa loro. no that me have invited to home their

97

(Z&P 2013:136)

From such a fact, Z&P propose that the polarity particle in the higher clause binds an operator in the lower one, which, in turn, selects the polarity of the clause: (36)

It is clear that the negative polarity of the sentences is related to the negative polarity of the particle no, which, therefore, cannot be expletive at all. In fact, coherently with the real status of negation, no-che sentences allow weak and strong NPIs, N-words and not-also conjunctions: (37) a. No che Luca non ha alzato un dito per Gianni no that Luca neg has lifted a finger for Gianni ‘Luca did NOT lift a finger to help Gianni!’ b. No che Luca non ha capito un tubo no that Luca neg has understood a tube ‘Luca did NOT understand anything!’ c. No che Luca non ha visto nessuno no that Luca neg has seen nobody ‘Luca did NOT see anybody!’ b. No che Gianni non ha studiato e neanche Luca no that Gianni neg has studied and neither Luca ‘Gianni did NOT study and Luca either!’ The negativity of no-che sentences definitively distinguishes them from Snegs. Moreover, the impossible mismatch between the polarity of the higher clause and the polarity of the lower one (see 35), rules out the chance that the negative marker in Snegs is dislocated in a different clause from the uttered one. In fact, Snegs are affirmative regardless of the negative morphology of the negative marker non; it means that the negative marker does not bind any polar operator and, therefore, it is interpreted as an expletive.

98

Chapter 3

To sum up, in this chapter we have seen that Snegs cannot be assimilated to other grammatical structures, even though they share with them several properties. I have excluded that Snegs are either rhetorical questions or exclamatives. We have also seen that Snegs are not bi-clausal structures like some negative Italian sentences. Therefore, we have just one last possible alternative: Snegs display their own syntactic structure. I will dedicate Chapter 5 to pursuing this hypothesis. Before doing so, it can be useful to recall some specific properties of Snegs. We saw that Snegs belong to the strong class of EN (they do not allow weak and strong NPIs, n-words, not-also conjunctions, etc.) and that they require the negative marker non. From the pragmatic point of view, we saw that the whole sentence in Snegs carries new information and, therefore, Snegs can be congruent answers to what I called “Propositional Questions”. The speaker attitude marks the unexpectedness of the fact described by Snegs, and from this follows the surprise perception of the hearer. Finally, I have shown that some elements are specifically associated with Snegs: fronting structures, ED, EE and the specific intonational pattern melting together some exclamative and interrogative acoustic features. Interestingly enough, all these structures share the peculiarity of interacting with the CP-layer of the sentence. This is the reason why we need to shed light on the structure of the left periphery before presenting a possible syntactic analysis for Snegs,.

CHAPTER 4 A PRELIMINARY STEP: A VIEW OF THE LEFT PERIPHERY

Finding the syntactic representation of a sentence has always been at the heart of generative grammar. Several attempts have been made, starting from the early works of Chomsky in the 1950s110, and the debate is still open. Recapitulating this discussion is beyond the aim of this book, and I have chosen to focus on a very restricted issue: the left periphery. As we saw in Chapter 2, Snegs specifically interact with some elements involving the left periphery of the sentence, the CP-fields. Rizzi (1997) was the first to extensively investigate the complex nature of CPs. Before facing the CP layers, let me briefly introduce Rizzi’s idea on the structure of the left periphery. A basic assumption in generative grammar is that a sentence involves lexical categories (ex. nouns, verbs, adjectives, etc.) and, beside them, some other categories called “functional”.111The functional categories are usually more abstract than the lexical ones, belonging to closed classes of elements, and are phonologically and morphologically dependent. For example, the determiners, the agreement morphemes and the verbal tense are usually associated with a functional head selecting some other projections as its argument. 112 Chomsky (1986) proposes that the functional projections display a X-bar schema, just like the lexical ones (Chomsky 1970), and this has been the common assumption in the theory until the minimalist approach.113 110

See Chomsky (1995) and Graffi (2001) for an accurate historical and critical description of the generative program. 111 See Abney (1987), Graffi (2001) and Cinque (2002) for a definition of a “functional category” as opposed to the lexical one. 112 The determiner, for example, selects a noun as its argument; see below. 113 We take the X-bar theory as the theoretical substrate of this chapter. Even though it has been modified during the evolution of the generative enterprise, most of the works presented here used to deal with the X-bar generalization.

100

Chapter 4

Chomsky (1986) suggests at least, two functional projections setting up a sentence: the Complementizer Phrase (CP) and the inflectional one I(NFL)P. Roughly speaking, CP defines the type of the clause (interrogative, declarative, exclamative, etc.) and host discourse-related elements (such as topics and foci) and operator-like elements (like interrogatives, relatives, etc.); IP settles the inflectional derivation of the verb. For example, in (1), the morpheme –av- marks the past tense of the verb, and the morpheme –a marks the grammatical number: (1) Luca mangi-av-a Luca eat.Pass.3rdsing ‘Luca ate’ Interestingly, during the evolution of the generative enterprise, the single Xbar projection for such functional categories turned out to be insufficient to capture the complexity of a language114. As we saw in Chapter 1, IP has been therefore split into 2 different projections (Moro 1988, Pollock 1989 and Belletti 1990): Agreement Phrase (AgrP) and Tense Phrase (TP). The former settles the value 1st, 2nd and 3rd singular and plural person; the latter settles the value +/- finite. Before this step, TP was considered as a single projection; after this step, it was considered a layer with as many projections as the number of features carried by the verb. For example, Kayne (1985, 1989a) argues for the existence of another AgrP regulating the relation between verbs and direct objects115 (to distinguish the agreement between subjects and verbs from that between verbs and objects, this has been labeled AgresP for the former and AgroP for the latter). In Italian, such a relation is evident in CLRD structures: (2) a. Io l’ho mangiat-a la ciambella-a I Cl.it.have.1stsing eaten.3rdsing.fem the doughnut.sing.fem ‘I ate a doughnut’ b. Io le ho mangiat-e le ciambell-e I Cl.them have.1stsing eaten.3rdplu.fem the doughnuts.plu.fem ‘I ate the doughnuts 114

The theoretical background is the binary branching hypothesis. According to (Kayne 1984), the number of specifiers and complements is, at most, one (see also Kayne 1994). Such a constraint had led to the introduction of new XPs in order to capture the complexity of the linguistic data instead of freely expanding a unique XP. 115 See also Belletti (1990) and Chomsky (1993). For a critique of this proposal see D’Alessandro and Roberts (2008).

A preliminary step

101

The bound-morphemes –a and –e point out the grammatical number (respectively, singular and plural) which is shared by both the direct object and the past participle, signaling the grammatical agreement between them. Larson (1988) proposes a similar “split” operation for VP. He works on double object constructions and on oblique dative ones: (3) a. John sent Mary a letter b. John sent a letter to Mary Notably, in such structures the order of the direct and indirect objects is not free, and it raises asymmetries when they host elements in which the structural relations play a central role. For example, the anaphora in (4) is correctly valued only when it is c-commanded by the antecedent Mary; the opposite order is ungrammatical: (4) a I showed Mary herself a’. *I showed herself Mary b. I showed Mary to herself b’. *I showed herself to Mary To take into account the pattern in (4), Larson follows Chomsky’s (1955/1975) proposal on oblique dative structures. More specifically, Chomsky argues that the basic structure of (3b) is (5): (5) John sent a letter to Mary

The structure is built compositionally: the V0 send takes to Mary as its argument, forming the small predicate V’ send to Mary, which, in turn, is predicated of the direct object a letter forming VP a letter send to Mary; such a predicate is taken as argument of an higher VP (the landing site for the verb movement) which is predicated of the subject John. The

102

Chapter 4

configurational position of the direct object in the lowest VP is like that of the subject in a transitive sentence ([Spec, IP]); in the same way, the relationship between verbs and PPs looks like the one between verbs and direct objects. This is the reason why the lowest VP is called “clause-like structure”. Interestingly, only the object in (5) c-commands PP. Such a fact captures the asymmetry in (4b-b’): the antecedent is the direct object ccommanding the anaphoric element. Larson extends Chomsky’s analysis to double object constructions, with the only difference that the direct object is taken as an adjunct (exactly as the subject in the passive clauses): (6) John sent Mary a letter

The verb send directly selects the indirect object Mary as its argument; the complex send-Mary is predicated of the direct object a letter. The inner NP raises in the structure in order to get the Case from VP116. The final result is that only the inner object Mary c-commands the adjunct NP a letter, explaining the asymmetry in (4ab). Anyway, the crucial point is that Larson proposes the splitting of VP in two projections for multiple-argumental verbs117. A single VP has been transformed in a VP-layers set up by more than a phrase (exactly as IP seen above). Until now, we saw that the “splitting” operation firstly proposed for IP also affects the VP-structure. Along the same line, it has been proposed that noun phrases (NP) display a multiple-phrasal organization too.

116

See Larson (1988) for a detailed explanation. See also Hale and Keyser (1993). The VP-splitting operation also was the background for the distinction proposed by Chomsky (1995) between vP and VP.

117

A preliminary step

103

According to Abney (1987), some nominal structures display a sentencelike physiognomy. For example, the English “Poss-ing” Gerund seems to combine together some nominal and verbal properties: (7) John’s building a spaceship

(Abney 1987: 13)

From the distributional point of view, Poss-ing Gerund behaves as a NP and, in fact, can occur in subject and object positions (ex. “Did [John’s building a spaceship] upset you?” and “I told you about [John’s building a spaceship]”). However, at the same time, the morphological morpheme -ing constitutes the verbal form for gerundives. To take into account their double nature, Abney proposes a complex structure in which NP and VP are generated in the Specifier and Complement positions of a Determiner Head respectively (DP-analysis): (8) John’s building a spaceship

DP also has an agreement function from that the noun John gets the genitive case 118 . From this point of view, DPs look like inflectional elements connecting two maximal projections. By assumption, all noun phrases are headed by the functional element D0 in a similar way to which an Agr0 heads a sentence. Crucially, as in the case of VP and IP, the single XP projecting the noun phrase has been split in a multiple-XPs layer119. To sum up, a “split-operation” originally thought for IP has been exported for VP and NP. They were originally analyzed as single X-bar projections, then they were split in multiple-projections constituting a layer. Every projection has a head with a proper X-bar skeleton. IP has been split in AgrP and TP (and, subsequently, into even more elements); VP in two VPs and NP in DP and NPs. Such extension allows capturing of the major complexity 118 119

See Abney (1987) for a detailed discussion on the agreement property of D0. See also Cinque (1994).

104

Chapter 4

of the data, giving a more fine-grained view of the syntax120. Interestingly, a similar fate also affected the CP structure.

4.1 The syntactic status of the left periphery-CP field The tendency towards microscopic analysis of linguistic structure has also influenced the approach to Complementizer phrases (CP). CP has been traditionally considered as the functional projection displaying the elements associated with the type of the clause (interrogative, declarative, exclamative, etc.), the finiteness of the IP embedded in it (finite/non-finite), the operator-like elements (such as wh-elements, relatives, etc.) and the discourse-related features (ex. topic and focus). We shall shortly see some details of each of those elements. The first property of CP is to point out the type of the clause. The expression of the clausal type, also known as the Force, interacts with the selection criteria of a potential high-clausal verb; for example, the verb say selects a declarative introduced by the complementizer that, and the verb wonder selects an interrogative introduced by whether: (9) a. I say that Luca will come b. I wonder whether Luca will come As Rizzi (1997) points out, from this point of view, CP looks outward at the border of its clause. At the same time, CP also looks inward to its clause, and it specifies the finiteness of the IP embedded in it. For example, in (10a), the complementizer that selects a +finite value of the verb to rob giving the (past) finite tense robbed; on the contrary, the complementizer to selects a –finite value of the verb to eat giving the non-finite tense eat (10b). Such a phenomenon is a sort of agreement between C and I. (10) a. She says that Luca robbed Giorgio b. She wants to eat a pizza

120

An important step in the syntactic analysis of functional heads is the Mirror Principle’s hypothesis. See Baker (1985, 1988) and Harley (2013) among others.

A preliminary step

105

Finally, CP also hosts some operator-like elements, such as the interrogative what and the relative who, and discourse-related elements, like topics and foci121: (11) a. What booki did you buy ti? b. The man, who has broken the arm, is here. c. Your book, you should give ti to Paul not to Bill (Rizzi 1997:285) d. YOUR BOOK, you should give ti to Paul (not mine) Topics and Foci differ in several aspects. Above all122, topics pragmatically express old information already present in the discourse context, whereas focus expresses new information. In (11c), the phrase your book is the information already shared by the speaker and the hearer; the rest of the utterance “you should give to Paul not to Bill” is a comment (or rheme) to the topic. In (11d), the phrase your book is instead new information given by the speaker. Clearly, the material that is not focalized is old information, and is considered as a presupposition. In some languages, like Italian and English, focalized structures are used to correct some wrong beliefs, or knowledge, already presented in the context of speech (therefore, they are labeled “contrastive”). (11d) is one of those cases, and the “corrected belief” is that the book that the hearer should give to Paul belongs to the speaker instead of to the hearer. Topics and Foci display different prosodies123 as well, and they also select different syntactic structures. For example, even though already discussed in Ch. 2, it is useful to recall that Italian marks the topical nature of a fronting element through Clitic (left) dislocation structures (Cinque 1990) displaying a resuptive clitic coreferential with the topic124 (see below): (12) a. Il ragazzoi loi ha salutato the guy him has greeted ‘The guy s/he has greeted him’ 121

(Topic)

In the rest of this Chapter I will focus on focalized and topicalized structures because this is the part of the CP-field that most concerns the analysis of Snegs. 122 See Rizzi (1997) and Cruschina (2012) for a complete discussion on the grammatical difference between topics and foci. 123 Cfr. Bocci (2007, 2008) and Frascarelli (2000). Since focus phases are marked with a rising of the tonal function (L+H* in ToBi transcription), they are usually signaled with the uppercase. 124 Rizzi (1997) further distinguishes topics and foci on the basis of some other syntactic observations. He argues that only foci raise WCO effects, allow bare quantifiers and are incompatible with wh-phrases.

106

Chapter 4

b. *IL RAGAZZOi loi ha salutato the guy him has greened

(Focus)

In summary, CP displays, at least, four different functions: it expresses the force of the clause, it specifies the IP finiteness and it hosts operator-like and discourse-related elements (like foci and topics). Such complexity has led scientists to hypostasize a complex structure. Two points of view, among all, have been debated during the last 20 years: the Cartographic Project and the Minimalism Program. I will sketch them in the following sessions, starting from the former.

4.1.1 The Complementizer phrase and the Cartographic Approach The fact that CP hosts heterogenic elements has made it suitable for the splitting-approach seen above for IP, NP and VP. In fact, a single CPprojection was felt to be insufficient to take into account such a complex situation, and it has been proposed that CP is set up by an array of functional projections, each separately headed. According to Rizzi (1997, 2001), the CP-field displays, at least, the following heads125: (13) CP…Force°… (TopP°)… Int°… (TopP°)… Foc°… (Top°)… Fin°…IP The order of the functional heads is based on distributional evidences. Force° and Fin° host, respectively, elements such as what and to (as we saw above) 126 and they “sandwich” the CP-field. Italian correspondent complementizers, che (‘that’) and di (‘to’), clearly show this order when they are considered in relation to topicalized phrases (ex. il tuo libro, En. your book): che must precede the TopP (14b-b’), whereas di must follow it (14c-c’). (14) a *Maria crede, [TopP il tuo libro, [ che lo potrà leggere]] ‘Maria believes, your book, that (she) will be able to read it’

125

Other scientists before Rizzi (1997) proposed two or more projections forming CP (cfr. Cruschina 2012 and the reference therein). I refer to Rizzi (1997) as the fundamental work because he is the first to have extensively investigated the CPstructure and his work has mostly influenced the successive research on this subject. 126 Rizzi (1997) proposes that relative pronouns are in Force as well, but they are generated in Specifier positions instead of in head positions like complementizers.

A preliminary step

107

a’. Maria crede che [TopPil tuo libro, [ lo potrà leggere]] ‘Maria believes that, your book, she will be able to read it’ b. Maria crede, [TopPil tuo libro, [ di poterlo leggere]] ‘Maria believes, your book, of to be able to read it’ b’. *Maria crede di, [TopPil tuo libro, [ poterlo leggere]] ‘Maria believes of, your book, to be able to read it’ The pattern in (14) also suggests that Top° is between Force° and Fin° and, in fact, it cannot precede che (14b) or follow di (14c’). The same goes for Foc°: (15) a. *Maria crede, [FocPIL TUO LIBRO[, che potrà leggere]] (non il mio) ‘Maria believes, YOUR BOOK, that (she) will be able to read it’ (not mine) a’. Maria crede che [FocPIL TUO LIBRO [, potrà leggere]] (non il mio) ‘Maria believes that, YOUR BOOK, she will be able to read it’ (not mine) b. Maria crede, [FocPIL TUO LIBRO [, di poter leggere]] (non il mio) ‘Maria believes, YOUR BOOK, of to be able to read it’ (not mine) b’. *Maria crede di, [FocPIL TUO LIBRO, [potere leggere]] (non il mio) ‘Maria believes of, YOUR BOOK, to be able to read it’ (not mine) If we put together the patterns in (14-15), it generates the sequence of heads in (13). Interestingly, Italian simultaneously shows the realization of multiple elements in the CP-domain: (16) Credo [CP [Force che [TopP ieri [FocP QUESTOk, [TopP a Giannii, think.1stsing.Pres that yesterday this to Gianni [TP i tuoi amici avrebbero dovuto dir-glii tk]]]]]] (non altro) the your friends have.3rdplu.Cond. had to tell.him (not other) ‘I think that yesterday your friends would have had to say to Gianni THIS (not other)’ The complementizer che expresses the declarative nature of the subordinate clause and makes the sentence suitable for the selecting requirement of the higher verb credere (to think); ieri (yesterday) and a Gianni (to Gianni) express the topicalized function and surround the focalized element

108

Chapter 4

QUESTO (this). In this example, it is evident how the complementizer and the inflectional material sandwich topicalized and focalized elements127. According to Rizzi (1997), a single head can host different types of elements. For example, interrogative pronouns in the main questions and focalized elements compete for the same structural position (i.e. Spec of Foc°) and, therefore, they cannot co-occur in the same sentence: (17) a. [FocP Con chii [ha mangiato Luca ti]]? ‘With whom did Luca eat?’ b. [FocP LA PIZZAi [ha mangiato ti Luca]] (non la pasta) ‘PIZZA, Luca has eaten (non pasta)’ c. *[FocP Con chi LA PIZZA [ha mangiato Luca]? (non la pasta) ‘With whom has Luca eaten the pizza?’ The impossible co-occurrence between wh-phrases and discourse-related elements disappears when wh-phrases occur with topicalized elements, (18) since they do not compete for the same structural position: (18) [FocP Chi [TopP la pizza [l’ha mangiata a casa]]]? ‘Who ate the pizza at home?’ The schema in (13) takes into account the multiple-nature of CP: for each function (Force, Finiteness, Topicalization and Focalization/operator-like elements) there is a proper projection128. According to the original proposal, each head projects its own X-bar schema with specific interpretative features. For example, the Top° projects a TopP assigning the topical interpretation to its Specifier, and the comment interpretation to its complement (see 19a); the Foc° projects a FocP assigning the focal interpretation to its Specifier and the presupposition interpretation to its complement (see 19b).

127

According to Cecchetto (1999) and Belletti (2004), there is another topic position right above VP. See below for a more detailed discussion. 128 Rizzi (2001) proposes an additional projection for some interrogative elements like the Italian perchè (why) that he called Int(errogative). See Rizzi (2001) for the detailed proposal.

A preliminary step

(19) a.

109

b.

Interestingly, the topic is recursive (thus the multiple occurrences of Top° in 13), unlike the focus: (20) a. Al ragazzo, i libri, le maestre glie-li to the guy the books the teachers Cl.to him.Cl.them hanno regalati have.3rdplu given for free ‘To the guy, the teachers gave the books for free’ b. *AL RAGAZZO, I LIBRI, le maestre hanno to the guy the books the teachers have.3rdplu regalato (non al fratello i quaderni) given for free not to the brother the notebooks According to Rizzi (1997), the non-recursive nature of the focus is related to its specific interpretative properties indicated by the X-bar schema in (19b). More specifically, if the FocP were recursive, it would mean that the Foc° selects another FocP as its argument, which would be presupposed and, that at the same time, it would contain a focal constituent giving an interpretative crash. On the contrary, nothing excludes that the comment interpretation in the complement of the Top° displays, in turn, a TopicComment structure129. Topicalization and focalization, as well as wh-structures, are criterial (Brody 1990, Rizzi 1997, 2000, 2006, Frascarelli 2000, Cinque and Rizzi 2010, Cardinaletti 2009, Haegeman 2012) and their associated constituents are endowed with a proper uninterpretable feature that has to be checked in a Spec-Head configuration with the Top° and Foc° by movement (“lastresort” operation, cfr. Chomsky 1995, 2000). More specifically, according to Rizzi (2006), topicalized and focalized elements are generated in a semantic position (s-position), where they satisfy the thematic properties, 129

See Rizzi (1997) and Cinque and Rizzi (2010).

110

Chapter 4

then they move to a criterial position (A’-position) in the Spec of some head in the CP-field130, which is dedicated to scope-discourse semantics131. (21) [FocP / Criterial position Which book] should you read [s-position t]? [TopP /Criterial position This book] you should read [s-position t] [FocP /Criterial position YOUR BOOK] you should read [s-position t] (not mine) Topicalized and focalized elements land in Specifier positions respectively of Top° and Foc° which, in languages like Italian, are phonologically unrealized. Crucially, some languages display a morphological representation of the focus-topic heads as a proof of their syntactic codification. According to Aboh (2004, 2010), Gungbe is one of them132: (22)

(Aboh 2004:1)

The topicalized element Setu, and the focalized Marì, are dislocated in the left periphery of the sentence and they are marked, respectively, by the topic head yà and the focus head wè. Coherently with the schema in (13), TopP and FocP follow the complementizer but precede the IP-internal subject pronoun é. A stronger version of Rizzi’s proposal is found in Benincà and Poletto (2004) 133 . They suggest a “one-to-one” mapping between the syntactic positions in the CP-field and the pragmatic interpretations (functions). From this point of view, the discourse-related features are read off the syntactic structures. As a consequence, they reject the recursive analysis of TopP, 130

In literature there are some alternative proposals. For example, Samek-Lodovici (2006) rejects Rizzi’s hypothesis and proposes that the FocP position is the rightmost position in Italian clauses. Frascarelli (2004) proposes that some topicalized phrases are directly merged in the criterial position. I do not go into depth on such alternatives because this is beyond the aim of my book. 131 In Chomsky (1995, 2004)’s system, s-positions are represented by the external merge, and criterial ones by the internal merge (the previous “move” operation). I do not keep track of the evolution of the theory, leaving this task to Cinque and Rizzi (2010) and Cruschina (2012). 132 See Cruschina (2009) and the references therein for some other cases. 133 See also Cruschina (2009, 2012) and Brunetti (2004) for a detailed review and a rationalization of the Foc-Top labels.

A preliminary step

111

arguing that each TopP is unique and obtains its own interpretative feature from a specific X-bar projection. More specifically, moving from comparative data in Romance languages, they propose four heads for the four Top° in (13)134: Hanging Topic°, Scene Setting°, Left Dislocation° and List Interpretation°. Each head carries its own interpretative feature. On the basis of the same distributional data, they also propose the existence of more than one focalized structure; more specifically, two contrastive FocPs and one informational FocP. From this point of view, they individuate two (sub)fields splitting the CP one: the Topic field, in which there are non-operator elements, and the Focus-fields, in which there are operator-like elements. The Topic-field is structurally higher than the focus one, excluding the possibility that a TopP follows a FocP (as 13 admits). As a proof, Italian topicalized elements can only precede the focalized one and cannot follow them135: (23) a. *A GIANNI, un libro di poesie, lo regalerete (B&P2004: 54) TO GIANNI a book of poems Cl.it give.2ndplu.Fut b. Un libro di poesie, A GIANNI, lo regalerete. a book of poems TO GIANNI, Cl.it give.2ndplu.Fut ‘You will give a book of poems to Gianni’ The final result is the schema in (24): (24) [Force° [Hanging Topic° [Scene setting° [Left Dislocation° [List Interpretation° [Contrastive Focus1° [Contrastive Focus2° [Informational Focus° [Fin° [TP]]]]]]]]] The array of projections forming the CP-field has been increased compared to Rizzi’s first proposal. Identical objects, like Topics, are not more present and each head projects its own X-bar schema. To sum up, we saw that CP has also been affected by the split-operation previously reserved for IP, VP and NP. The result is a growing number of functional projections encoding the morphological and interpretative properties displayed by a language. Such a tendency to map each functional projection of a sentence has spawned what is called “The cartographic project” (cfr. Cinque 2002, 2006, Rizzi 2004; Belletti 2004b, Cinque and

134

See next paragraph for the discussion of the data. I do not agree with the grammaticality judgements in (23a), but it may depend on the different sub-standard type of Italian I speak. 135

112

Chapter 4

Rizzi 2010, Shlonsky 2010, Cruschina 2012). The core mission of the cartographic project is “the attempt to draw maps as precise and detailed as possible of syntactic configurations” (Cinque and Rizzi 2010: 51). Crucially, the cartographic project relies on two theoretical principles that need to be briefly discussed: the Linear Correspondence Axiom (Kayne 1994) and the Uniformity Principle (Chomsky 2001). The Linear Correspondence Axiom (LCA) maps hierarchical structures into a linear order, which, in turn, is accessible to the human articulatory system. According to this axiom, if the word a precedes the word b, then the node A dominating the word a asymmetrically c-commands the node B dominating the word b. This relation between the linear word order and the hierarchical structure of the syntactic projections is the basis of the schema in (13). The cartographic approach also suggests that the sequence of functional projections dominating VP, NP, IP, CP etc., is universal, even if it is differently realized in different languages. The Uniformity Principle supports such an implication: Uniformity Principle: In the absence of compelling evidence to the contrary, assume languages to be uniform, with variety restricted to easily detectable properties of utterances (Chomsky 2001:2) Crucially, Cinque (1999) argues for the existence of a universal hierarchy of functional projections based on cross-linguistic data. More specifically, put in Cinque’s own words, “the same number, type and order (hierarchy) of functional projections holds across languages and clause types, despite apparent counterevidence” (p.v). He starts showing that optional elements like adverbs, usually associated to adjuncts, are rigidly fixed in the succession of syntactic projections and, therefore, in the succession of the words order (by LCA). For example, three Italian adverbs like solitamente (usually), mica (neg) and già (already) are in the specific ordered relationship solitamente > mica > già (Cinque 1999: 4): (25) a. Alle due Gianni non ha solitamente mica mangiato, ancora at two, Gianni neg has usually neg eaten yet ‘At two, Gianni has usually not eaten yet’ a’. *Alle due Gianni non ha mica solitamente mangiato, ancora at two, Gianni neg has neg usually eaten yet b. Gianni non ha mica già mangiato Gianni neg has neg yet eaten ‘Gianni has not eaten yet’ b’. *Gianni non ha già mica mangiato Gianni neg has yet neg eaten

A preliminary step

113

The sentence in (25a) shows that solitamente (usually) can only precede mica (neg); the sentence in (25b) shows that mica can only precede già (already). For transitivity, solitamente precedes già as well. Building on this kind of data and expanding the observations to some other comparative materials, Cinque proposes that a fixed order of AdvPs is invariant across languages (even though a language does not morphologically realize some of these) and it replays the following schema: (26) [frankly Moodspeech act [fortunately Moodevaluative [ allegedly Moodevidential [ probably Modepistemlc [ once T(Past) [ then T(Future) [ perhaps Moodirrealis [ necessarily Modnecessity [ possibly Modpossibility [ usually Asphabitual [ again Asprepetitive(I) [ often Aspfrequentative(I) [ intentionally Modvolitional [ quickly Aspcelerative(1) [ already T(Anterior) [ no longer Aspterminative [ still Aspcontinuative [ always AsPperfect [just Aspretrospective [ soon Aspproximative [ briefly Aspdurative [ characteristically(?) Aspgeneric/progressive [ almost Aspprospective [ completely AspSgCompletive(I) [ tutto AspPlCompletlve [ well Voice [fast/early Aspcelerative(II) [ again Asprepetitive(II) [ often Aspfrequentative(II) [ completely AspSgCompletive(II) ]]]]]]]]]]]]]]]]]]]]]]]]](Cinque 1999: 106) Such a rich functional make-up affects the whole sentence. As for the CPfield, the basic assumption is that each adverb is in a Spec-Head relation with a different functional head carrying its own semantic feature (mood, modality, tense, aspect, and voice). Putting together the array of functional categories proposed for CP, IP, VP, and for NP, there emerges a very complex structure undergoing a sentence. I recall that the basic assumption in the cartographic approach is that every linguistic function is properly projected in a X-bar schema and that such an array of functional projections is invariant across languages and across the sentences of each language (regardless of the phonological realization of them or their optionality). The cartographic project, therefore, consists in “the view that syntactic structures are uniform, locally simple and both necessary and sufficient to structurally represent the grammatical or functional information relevant for semantic/pragmatic interpretation” (Shlonsky 2010: 418). Among other things, such an approach has the great advantage of pursuing a microscopic analysis of the grammatical components of a language and, regarding the CP-field, of proposing a syntactic codification for the discourse-related features. Both these points will be useful for my analysis of Snegs.

114

Chapter 4

4.1.2 The Complementizer phrase and The Minimalist Program The main alternative to the cartographic approach is the Minimalist Program (MP) inaugurated by Chomsky (1995). The MP proposes a contrary view to the cartographic one: it pursues a rationalization of the computational operations and, therefore, of the number of projections (Chomsky 1995, 2001, 2008, 2016). The basic assumption of the minimalist framework is that languages are optimal objects regulated by economical and simple computational operations such as “Merge”. Merge is a syntactic operation that takes two objects, X and Y, and forms the new one Z 136 . “Merge” can be either “internal” or “external” depending on which kind of element it takes: if X is distinct from Y, it is an “external merge”; if one of them has already been generated, it is an “internal merge” (cfr. Chomsky 2004). The “internal merge” is a feature-driven operation. The syntactic objects created by Merge are set up before the “spell-out” operation, as the schema in (27) shows: (27)

After the Spell-out, the syntactic objects receive their “interpretative” representations traditionally associated with the Phonetic Form and the Logical form (which, in turn, are respectively associated with the sensorimotor system for externalization and with the conceptual-intentional system for reflection, interpretation, inference, etc.). When objects receive interpretation at one interface, the representation satisfies the Full interpretation condition. After the Spell-out, the liner word order, the prosody and the detailed phonetic properties are elaborated in the Phonetic Form interface. Crucially for our discussion, the early version of the minimalist framework (Chomsky 1995) interprets discourse-related elements, like topics and foci, as features visible only at the interface levels. Roughly speaking, discourse-related elements depend on the “interpretive 136 See Moro (2000, 2009) and Narita (2011) for a detailed discussion on the labelling of the new objects derived from the Merge operation.

A preliminary step

115

requirements that are externally imposed by our systems of thought” (Chomsky 1998 in Moro 2000). Under this approach, topic and focus are taken as “surface” phenomena affecting the linearity imposed by the phonological form. They are excluded from the narrow syntactic elaboration before the spell-out point. In a successive version of the MP, the theoretical position about discourserelated elements changes in favor of their syntactic codification. More specifically, Chomsky (2008) proposes specific edge-features hosted by the phase heads C and v*. Such edge-features encode scope and discourserelated functions and, moreover, they trigger the A’-movement of an element from its “generated” position (external merge) to the fronting one (internal merge). To quote Chomsky, (28) it will suffice to define an A'-position as one that is attracted by an edge-feature of a phase head; hence typically in SPEC-C or outer SPEC of v* [Chomsky 2008: 150]. Edge-features affect syntactic derivation and, exactly like other syntactic objects, they are associated with the interpretative effects of the outcomes (the edge-effects). Crucially, focus and topic are associated with edgeeffects. From this point of view, the minimalist and the cartographic approaches do not differ excessively since both interpret discourse-related elements as syntactically encoded. However, Chomsky (2008) proposes that only one head, C or v*, hosts all the functions traditionally associated with the complementizer field (force, topic, focus, etc.), whereas the cartographic approach assumes that there are as many heads as functions. Such a difference is related to a basic assumption that the minimalist framework, but not the cartographic one, supports: the MP adopts the bare phrase structure theory in which both the classical X-bar schema and the LCA are ruled out. In fact, the minimalist approach argues that “merge” creates symmetric binary objects and that linearization is required only in PF. In such a configuration, nothing prevents a head from having multiplespecifiers each hosting different features (this is ruled out in X-bar theory by LCA; cfr. Kayne 1994). However, as Cruschina (2009, 2012) states, a single head for all complementizer features is unable to capture the complexity of some data, particularly the cases in which a language phonologically overtly realizes the functional heads for topics and foci (as Gungbe in 22 repeated here as 29).

Chapter 4

116

(29)

As Cruschina (2012) states, a model with only one complementizer head is unable to take into account cases in which multiple-heads realize different edge-features. I will follow the proposal supported by both views that the complementizer area is a complex object hosting multi-elements, even though deciding which of the two approaches is more accurate is beyond the aim of this book137. I will follow Rizzi’s original proposal for the theoretical reasons seen above (I repeat here 13 as 30) 138: (30) CP…Force°… (TopP°)… Int°… (TopP°)… Foc°… (Top°)… Fin°…IP Before going any further, I will show some examples of focalized and topicalized structures in Italian. We saw some theoretical discussions in the last two paragraphs; I move now on to the investigation of their applications.

4.2 An overview of Topicalized structures In this paragraph I am going to investigate the structure of some Italian topicalized constructions. I already discussed single cases of topics in the previous chapters. I want now to collect and discuss that data. Before going into detail, let me recall the theoretical basis of topics. According to Stalnaker (1978, 2002), the set of propositions usually presupposed in the domain of discourse formed the “common ground” (CG). The CG contains the referents of discourse either newly introduced 137

Many studies have proposed a sort of conciliation between the MP and the cartographic approach; see, among others, Frascarelli (2000), Belletti (2004b), Cinque and Rizzi (2010). 138 Theoretically, I could also follow either the Belletti & Poletto (2004) version or Rizzi’s hypothesis, or other similar proposals (eg. .Frascarelli and Hinterhölzl 2007); however, this would not change the essence of my proposal. Moreover, since the discussion on the structure of the CP-field is still open (cfr. Abels 2012), its resolution is beyond the aim of this book.

A preliminary step

117

or already present. The former are traditionally associated with the focus, the latter with the topic139. As we saw above, in the early stages of the cartographic approach, topics were considered as recursive functions allowing simultaneous realizations. In the successive phases a mapping of one-to-one between specific, different projections has been argued for (Benincà and Poletto 2004, Frascarelli and Hinterhölzl 2007, Bianchi and Frascarelli 2010). This is the reason why different labels have been proposed for different topics. We are going to see two different proposals140. The first one is found in Benincà and Poletto (2004). They propose that there are, at least, four typologies of topics: Hanging Topic, Left Dislocation, Scene setting and List interpretation. I will start by showing the differences between those structures. Compare, for example, the following sentences141: (31) a. Mario, gli amici, gli hanno fatto un brutto Mario the friends to-him have.3rdplu done a bad scherzo (Hanging Topic) joke 'Mario, his friends played a bad joke on him' b. A Mario, gli amici, gli hanno fatto un brutto to Mario the friends to-him have.3rdplu done a bad scherzo (Left Disl.) joke 'To Mario, his friends played a bad joke on him' c. Domani Gianni lo vedo. (Scene Setting) tomorrow Gianni Cl.him see.1stsing.Pres 'Tomorrow I will see Gianni’ d. La frutta *(la) regaliamo, la verdure *(la) the fruit Cl.it give for free.3rdplu the vegetables it vendiamo (List interp.) sell.3rdplu.pres 'We give fruit for free, while we sell the vegetables' 139

For a more fine analysis see Lambrecht (1994) and Brunetti (2009). Even though I follow Rizzi’s proposal on the CP-field, in this paragraph I will refer to Benincà and Poletto (2004) and Frascarelli and Hinterhölzl (2007) because they deeply investigated the different realizations of topics. I choose their system of labeling because their classification is syntactically led (coherently with the spirit of this book). For a different approach, among others, see Bianchi and Frascarelli (2010). 141 The data is taken from Benincà and Poletto (2004). 140

118

Chapter 4

In (31a-d), the italic constituency represents the topic of the sentence. Interestingly, the syntactic physiognomies of the four topics are different even though they belong to the same interpretative class. For example, the hanging topics (HTs) and left dislocation sentences (LDs) differ in at least four aspects (Cinque 1982, Benincà and Poletto 2004): i) topics in LDs maintain the Case (as is shown by the particle a in 31b) whereas the HT do not (consequently, they can only be bare DPs); ii) HT is unique (32a) and LDs can be multiple (32b); iii) the clitic resuptive pronoun is obligatory in HT (33a) and it agrees with the topicalized phrase in gender and number, but not in Case (33c); in LD the resuptive pronoun is required only when it refers to the direct or partitive object of the sentence (33b) and it agrees with the topicalized constituent in gender, number and in Case; iv) LD structures can be embedded in relative clauses, whereas the HT ones cannot (34). (32) a. *Gianni, questo libro, non ne hanno parlato a lui (HT) Gianni this book neg Cl.of-it have.3rdplu talked to him b. A Gianni, di questo libro, non gliene hanno mai to Gianni of this book neg Cl.to him.of-it have.3rdplu never parlato (LD) talked 'They did not ever talk to Gianni about this book' (33) a. Marioi, gli amici, *(gli) hanno fatto un brutto scherzo Mario the friends to-him have done a bad joke 'Mario, his friends played a bad joke on him.' b. A Giannii, questo librok, non (glie)i*(lok) hanno dato to Gianni this book neg Cl.to him.it have.3rdplu given 'To Gianni, they did not give this book’ c. Mario, non ne parla più nessuno Mario not Cl.of-him talks anymore nobody 'Mario, nobody talks of him anymore' (34) a. Luca è una persona che, di questo segreto non ne Luca is a person who of this secret neg Cl.of it parlerà mai talk.3rdsing.Fut never ‘Luca is a person who will never talk about this secret’ b. *Luca è una persona che, questo segreto non ne Luca is a person who this secret neg Cl.of it parlerà mai talk.3rdsing.Fut never

(LD)

(HT)

A preliminary step

119

The Scene Setting structure (SS) in (31c) displays its own syntactic identity too. According to Benincà and Poletto (2004), in languages such as Italian there is no reason to distinguish it from HTs or LDs; however, it can be done taking into consideration other Romance languages such as RhaetoRomance. In fact, the SS in Rhaeto-Romance displays specific syntactic behaviors, such as the impossibility of being embedded (contrary to LDs) and the absence of resuptive clitics (contrary to HT): (35) a. Duman va-al a Venezia tomorrow goes.he to Venice ‘Tomorrow he will go to Venice’ b. *Al m a dit c duman va-al a Venezia. he me has told that tomorrow goes-he to Venice The temporal adverb duman is clearly in a topic position in (35) because it anticipates the clitic subject al (he). Assuming that every language universally shares the hierarchy of functional structures, it follows that Italian realizes such an element as well. As a clue, Benincà and Poletto (2004) show that SS (italic character in 36) displays a rigid order when it co-occurs with other topicalized elements like HT (underlined): (36) a. Mario, nel 1999, gli hanno dato il premio Nobel Mario in the 1999 C.to-him have.3rdplu given the Prize Nobel ‘To Mario, they gave the Nobel Prize in the 1999’ b.*Sul giornale Mario ne hanno parlato malissimo on the newspaper Mario of him have.3rdplu spoken very badly Finally, the List Interpretation (LI) syntactically differs from other topics since, for example, it requires the presence of a resuptive clitic such as la in (31d) (contrary to the SS) and it can be embedded in relative clauses (contrary to HT): (37) E’ bella la decisione [che la frutta la regaliamo, e la is nice the decision that the fruit Cl.it give for free.3rdplu and the verdura la vendiamo] vegetable Cl.it shell.3rdplu ‘It’s a nice decision giving fruit for free and selling vegetable’ An alternative system is proposed by Frascarelli and Hinterhölzl (2007). Like Benincà and Poletto (2004), they propose that topics are not recursive

120

Chapter 4

but they fall into three classes recognisable both syntactically and phonologically: the aboutness topic, contrastive topic and familiar topic. The aboutness topic describes the point of discussion of a sentence; the contrastive topic is “an element that induces alternatives which have no impact on the focus value and creates oppositional pairs with respect to other topics”; the familiar topic is a “d-linked constituent, which is typically destressed and realized in a pronominal form” (p. 88). Let me show an example taken from the original Frascarelli and Hinterhölzl (2007) work: (38) a. La situazione è questa: l'insegnante come ho detto ai ragazzi è in maternità, ha una gravidanza difficile e sta usufruendo di quella legge particolare della maternità anticipata per ora ha avuto un mese io penso che non tornerà però lei m'ha detto ah di non dirlo ancora ai ragazzi perché per motivi suoi-comunque io signora penso di chiudere l'anno [...] questo comunque io ai ragazzi non l'ho detto direttamente. ‘This is the situation: the teacher, as I told students, is pregnant, she’s having a difficult pregnancy and she is now benefitting from a specific law that allows for an early maternity-leave. So far, I was given one month of teaching supply. I don’t think she is coming back this year, however she told me not to tell students, because- well, she has her reasons. However, I think I will keep the class till the end of the year [...] anyway I did not tell this fact to students directly.’ b. Questo, io ai ragazzi non l’ ho detto direttamente this I to.the boys not Cl.it.have.1stsing told directly ‘I did not tell that fact to my students directly’ The story is about a teacher who is speaking with a mother about a difficult situation at school. In (38b) I have focused on the sentence in which all the three types of topics appear: the familiar topic is ai ragazzi (to the boys). In fact, it has the resuptive pronoun and is repeated several times in the sentence; the subject io (I) is the contrastive topic and it stresses the fact that the speaker–the teacher–is keeping a secret; the demonstrative pronoun questo (that) is the object under discussion in the sentence or, put in other words, it is the “what about” of the sentence. From a syntactic point of view, the different classes of topics occupy different positions in the C-system and, therefore, they follow a rigid linear word order when they occur both in multiple topic structures (as in 38b or 40) and in focalized clauses (39 and 41):

A preliminary step

121

(39) a. Quello ANCHE SUL QUESTIONARIO l’ho that also on-the questionnaire cl.it have.1stsing scritto (About.>Foc) written ‘I wrote that also ON THE QUESTIONNAIRE’ (40) Io francamente questo- questa attività in particolare I frankly this this activity in particular non me la ricordo (Contrastive>familiar)142 neg Cl.me Cl.it remember.1st sing.pre. ‘Frankly, I don’t remember that particular activity.’ (41) L’ha sempre tirato fuori LUI, il discorso143 (Foc>famil. topic) Cl.it has always took out his the topic ‘It is HIM who always took this topic.’ The aboutness topic precedes both the focalized phrase (39) and the contrastive topic (40). The contrastive topic precedes the familiar one (40), which, in turn, follows the focus (41). From this data, the authors propose that each kind of topic occupies a different syntactic position in the CP-field, giving the following structure: (42) [Aboutness Topic [Contrastive topic [Focus [Familiar Focus* …144 Interestingly, starting from Cinque (1990), Frascarelli (2000, 2004) also proposes that topics are base generated in the C-domain when they show a resuptive clitic, to which they are connected by binding a pro145. Clearly, some structures discussed in Benincà and Poletto (2004) and Frascarelli and Hinterhölzl (2007) are the same but are labeled in a different manner. I do not deepen this issue here and I will refer to the topicalized structures for the common features shared by all: the displacement in the 142

As Rizzi (1997) points out, only one contrastive element can occur to each sentence. As a consequence of this general fact, the co-occurrence of a contrastive topic with focus is impossible. This is the reason why I have to use a topical element and not a focused one in (40), in order to show the liner order of the contrastive topic. 143 Unfortunately, the original data discussed in Frascarelli and Hinterhölzl (2007) present the focus in situ instead of in a fronting position. 144 They propose that Familiar Topic is recursive since a sentence can display more than one. It seems to me that this fact contrasts their basic assumption that there is not recursion for topic. 145 See the original works for more details.

122

Chapter 4

frontal position of the sentence and the presence of resuptive pronouns. In particular, I refer to the clitic resuptive pronoun as a diagnostic for topicalized structures both in right and left dislocation structures (cfr. Cinque 1990, Kayne 1994, Cecchetto 1997). Actually, the two structures are not identical. For example, the resuptive clitic is obligatory in CLLDs but not in CLRDs: (43) a. Il gelato io l’ho già mangiato the ice-cream I Cl.it have.1st already eaten ‘I have already eaten the ice-cream’ a’. *Il gelato io ho già mangiato the ice-cream I have.1st already eaten b. Io l’ho già mangiato, il gelato I Cl.it have.1st already eaten the ice-cream ‘I have already eaten the ice-cream’ b’. Io ho già mangiato, il gelato I have.1st already eaten the ice-cream ‘I have already eaten the ice-cream’

(CLLD)

(CLRD)

The contrast in (43a’-b’) has led some scholars (cfr. Benincà 1988, Cecchetto 1997, 1999, Cardinaletti 2002) to propose that such an asymmetry is due to the non-assimilation of (43b’) to the CLRD structures. In fact, they argue that (43b’) is a case of marginalization, a different syntactic structure. According to this view, I will not use marginalization as a case of prototypical topicalized structure, and I will always refer to the topics as indicated by the resuptive clitic pronoun.

4.3 An overview of Focalized structures As I outlined in §4.1, the sentential focus corresponds to the pieces of new information introduced in the context of speech (cfr. Calabrese 1982, Brunetti 2004, 2009). From this point of view, focus is complementary to the topic of the sentence, which refers to the “old information”. As a consequence, sentences are composed by both focus and topic (Vallduvì 1992): (44) A: Why did you call the police? B: [I called the police]Topic [because I lost my wallet]Focus! The informative part of the answer is the content of the reason clause because I lost my wallet; it represents the focus of the sentence and I will

A preliminary step

123

call it informational focus (cfr. E’ Kiss 1998) Crucially, the label “focus” is also used to indicate discourse-related features, like the case of contrastive focus seen above146. To refresh your memory, see the dialogue (45): (45) A: Luca stole a phone B: A WALLET, Luca stole (not a phone) The speaker B rejects the information that what Luca has stolen is a phone, and corrects it by asserting that it was a wallet. Intuitively, the contrastive focus A WALLET also carries new information, exactly like the informative focus in (44). However, according to E’ Kiss (1998) and Belletti (2004), informational and contrastive focus describe different syntactic structures: contrastive focus expresses a quantificational-like operation, it is optional and it requires the movement of the focalized element in a fronting position (Rizzi 1997); informational focus conveys non-presupposition information, it is always realized in a sentence because it express the informational newness, and it does not require any kind of movement147. Compare, for example, the following sentences: (46) A: What did you eat yesterday? B: I ate [pizza]Informational Focus (47) A: Luca lost the wallet B: [The phone]Contrastive Focus Luca lost ti In (46B), the direct object realizes the informational focus of the sentence; it carries the new (non-presupposed) information and it does not move from its basic position. On the contrary, in (47B), the direct object realizes the contrastive focus of the sentence; it denies a part of the previous information given by A (the fact that it is the wallet that Luca lost) and replaces it introducing a new reference (the phone) as the object of the verb; moreover, its position is not standard, but has been moved there (internal merge position).

146

Such a double meaning of the label “focus” has been deeply investigated by Cruschina (2012). 147 The two types of foci also display different prosodic features. Cfr. Belletti (2004) and Donati and Nespor (2003).

124

Chapter 4

From the distributional point of view, the informational and contrastive focus may be argument-focus, predicate-focus or sentence-focus, depending on which part of the sentence falls under the focal scope (cfr. Lambrecht 1994, Cruschina 2012). The three types describe the focalization of a verbal argument (48a–49a), of the whole predicate (48b–49b) and of the entire sentence (48c–49c), respectively: (48) a. (Context: What did you break down?) I broke down [my car]argument-focus b. (Context: What happened to your car?) My car [broke down]predicate-focus c. (Context: What happened?) [My car broke down]sentence-focus (49) Context: Laura dice che i turisti pensano che Milano sia piovosa ‘Laura says that tourists think Milano is rainy’ a. Laura dice che [TRENTOi]Contr.Focus i turisti pensano Laura says that Trento the tourists think.3rdplu.sing che ti sia piovosa (non Milano) that is.3rdsing.Subj rainy (not Milano) ‘Laura says that the tourists think TRENTO is rainy (not Milano)’ b. Laura dice che i turisti pensano che [SIA Laura says that the tourists think that is.3rdsing.Subj SOLEGGIATAi]Contr.Foc Milano ti (non piovosa) sunny Milano (not rainy) ‘Laura says that the tourists think Milano IS SUNNY (not rainy)’ c. Laura dice che [ROMA SIA SOLEGGIATAi]Contr.Foc i Laura says that Roma is.3rdsing.Subj sunny the turisti pensano ti (non che Milano sia piovosa) tourists think (not that Milano is rainy) ‘Laura says that the tourists think ROMA IS SUNNY (not that Milano is rainy)’ Crucially, informational and contrastive foci are in complementary distribution in languages like Italian148. Taking the displaced position as the diagnostic for the contrastive focus, the dialogue in (50) shows that only the informational focus can be an answer to questions; on the contrary, (51) shows that only a contrastive focus can be a rejection to a previous assertion (cfr. Belletti 2004). 148

For a different view see Benincà and Poletto (2004) and Cruschina (2012).

A preliminary step

125

(50) A: Che cosa hai rotto? (What did you break down?) B: Ho rotto [la macchina]Inf. Focus B’: *[LA MACCHINA]Cont. Focus ho rotto ti (51) A: Luca ha perso il portafoglio (Luca has lost the wallet) B: #Luca ha perso [il cellulare]149Inf. Focus B’: [IL CELLULARE]Cont. Focus Luca ha perso ti In order to take into account the data in (46-51), I will follow E’ Kiss (1998) and Belletti (2004)’s distinction between informational and contrastive focus150. Moreover, I will refer to contrastive focus as the discourse-related phenomenon requiring a specific syntactic representation in the CP-field (as Rizzi argues). Starting from these bases, I now want to investigate some other Italian focalized structures as they have been presented in literature. First of all, Italian displays focalized elements in situ as well. Consider, for example, the following dialogue: (52) A: Luca ha vinto una medaglia Luca has won a medal ‘Luca won a medal’ B: Luca ha vinto UNA MAGLIETTA (non una medaglia) Luca has won a t-shirt (not a medal) ‘A T-SHIRT, Luca won (not a medal)’ The phrase una maglietta carries the same contrastive semantics and the same intonational patter (Avesani and Vayra 2003) as the fronting focus. According to Rizzi (1997), when the focalized phrases stay in situ there is still a movement in the [Spec, FocP] but it is covert (or, in other words, it happens in LF). From this movement, focalized elements get their semantics and prosody. Like other focalized structures, the in situ forms may be argument-focus, predicate-focus or sentence-focus:

149

Italian native speakers could judge the sentence in (51) perfectly fine if the object il cellulare is interpreted as an in situ focalized element. In fact, Italian also displays this strategy to focalize a phrase (see below). In such a case, the focalized element requires a marked intonation as well. 150 Cfr. Brunetti (2004) for an opposite view.

126

Chapter 4

(53) Context: I turisti pensano che Milano sia piovosa ‘Tourists think Milano is rainy’ a. I turisti pensano che ROMA sia piovosa (non Milano) ‘The tourists think Roma is rainy (not Milano)’ b. I turisti pensano che Milano SIA SOLEGGIATA (non che sia piovosa) ‘The tourists think Milano is sunny (non that it is rainy)’ c. I turisti pensano CHE ROMA SIA SOLEGGIATA (non che Milano sia piovosa) ‘The tourists think that Roma is sunny (non that Milano is rainy)’ Moreover, the focalized phrases are often connected to some particles (see Brunetti 2004) such as anche (also), solo (only) and nemmeno (not even), both in fronting and in situ positions: (54) a. Anche LA PIZZAi ha mangiato Paolo ti also the pizza has eaten Paolo ‘Paolo also ate pizza’ a’. Paolo ha mangiato anche LA PIZZA Paolo has eaten also the pizza ‘Paolo also ate pizza’ b. Nemmeno LA PIZZAi ha mangiato Paolo ti neither the pizza has eaten Paolo ‘Paolo did even not eat pizza’ b’. Paolo non ha mangiato nemmeno LA PIZZA Paolo neg has eaten neither the pizza ‘Paolo did even not eat pizza’ c. Solo LA PIZZAi ha mangiato Paolo ti only the pizza has eaten Paolo ‘Paolo only ate pizza’ c’. Paolo ha mangiato solo LA PIZZA Paolo has eaten only the pizza ‘Paolo only ate pizza’ These particles are called “focalizing” because they carry a contrastive-like semantics. More specifically, also contrasts the belief that Paolo did not eat something more than the expected set of foods; not even, on the contrary, contrasts the belief that Paolo ate something more than the expected set of foods; solo contrasts the idea that Paolo ate something more than one

A preliminary step

127

element among the set of foods151. Interestingly, in a restrictive set of cases Italian assigns other semantic values to the fronting focalized element as well as the contrastive one seen above. See, for example, the following sentence: (55) Ora ricordo, una sciarpa rossai mi ha regalato ti now remember.1stsing.pres a scarf red cl.to-me has given Luigi per Natale (Brunetti 2004: 121) Luigi for Christmas ‘Now I remember: it was the gloves that Luigi gave me as a present for Christmas.’ The phrase una sciarpa rossa is moved into a non-argumental position in the left periphery of the sentence. This position cannot be TopP because the construction does not allow resuptive clitics as the topicalization usually does. (56) *Ora ricordo, una sciarpa rossai me l’ha now remember.1stsing.pres a scarf red cl.to-me cl.it.has regalata ti Luigi per Natale (Brunetti 2004: 121) given Luigi for Christmas According to Lonzi (2004), Brunetti (2004, 2008) and Cruschina (2012), (55) is a case of focalization. However, the semantics of the focalized element in this case is not contrastive (even though it displays an emphatic marked intonation) since it misses a previous referent against which the new information contrasts. Quoting Cruschina (2012), I will refer to such a structure as a case of “Mirative fronting Focus”152. Mirative focus sentences are usually uttered out-of-the-blue and carry an effect of surprise because the information associated with them is beyond the speaker’s expectations. In (55), the surprising event is that the speaker received a red scarf for Christmas. Beside the Mirative focus, there are, at least, two more structures in which the contrastive semantics of the Italian focus seems to lack or, more 151

The connection between these particles and the focalized structures has led some scholaras to propose that they represent the heads of the FocP, or the heads of CP in the minimalism framework. See Brunetti (2004) and the reference therein. 152 For different labeling, see Lonzi (2004), Brunetti (2004). I will not refer here to the debate about the proper syntactic characterization of the Mirative fronting focalization because it is beyond the aim of this book. See Cruschina (2012) for a more extensive discussion.

128

Chapter 4

precisely, to be dilute: Quantifier Fronting (QP-fronting) (Benincà 1988; Quer 2002; Benincà and Poletto 2004; Cruschina 2012) and the AnaphoricAnteposition. Consider the following examples: (57) Nientei ha concluso ti Luca stando qui (QP-fronting) nothing has accomplished Luca being here ‘Luca did not accomplish anything by being here’ (58) A: Gianni è andato al bar invece di andare a scuola (Anaph. Ant.) ‘Gianni went to a bar instead going to school’ B: E questoi ha fatto ti anche Lorenzo and this has done also Lorenzo ‘Lorenzo did the same’ The bare quantifier niente in the QP-fronting construction in (57) has moved from its base generated position to the left periphery. Interestingly, QPsfronting constructions do not display clitic resumptions and they cannot cooccur with wh-elements: (59) a. *Niente lo ha concluso ti Luca stando qui nothing it has accomplished Luca being here b. *Niente chi ha concluso stando qui? nothing who has accomplished being here The patterns in (57-59) reproduce the one associated with focalization (cfr. Rizzi 1997), even though the QP-fronting does not show a particular marked intonation or a strong contrastive semantics153. According to Quer (2002) and Benincà and Poletto (2004), QPs-fronting lands in FocP position–as the other focalized elements–and this is the reason why fronting focalized structures and QP-fronting ones share the same syntactic behaviors154. In (58), the demonstrative questo lays in a fronting position and, since it strongly refers to a referent already present in a previous assertion, such a structure has been called Anaphoric Anteposition. Even though the 153

Actually, the structure maintains a weak contrastive semantics. For example, in (57) the quantifier niente is used to correct the previous expectation that Luca should accomplish something by being in a certain place. 154 Cruschina (2012) proposes that the landing site of QPs-fronting is an Informational Focus Phrase (IFocP) located in the left periphery. I do not deepen such a proposal because it does not change the nature of the syntactic movement displayed by the QP-fronting structures.

A preliminary step

129

reference to a previous constituent makes it similar to the topicalized function, anaphoric anteposition constructions cannot allow clitic resumptions and wh-elements, exactly like QPs-fronting: (60) a. *La maestra mi ha detto di portare un libro e un libro the teacher Cl.to me has said of bring a book and a book lo ho portato155 st Cl.it have.1 . brought b. *Questo chi ha fatto? this who has done According to Benincà and Poletto (2004), anaphoric anteposition constructions are focus-like structures without the particular marked intonation and the contrastive semantics usually associated with the focus156. Since some fronting focalized structures (Mirative focus, QP and anaphoric anteposition constructions) do not show strong contrastive semantics and marked intonations, I will just take the unavailability with the resuptive clitic pronoun and the impossible co-occurrence with the wh-elements as the diagnostic for the fronting focus.

4.4 The left periphery in other XPs Finally, the focus and topic projections originally discussed for CP have also been proposed for the periphery of both vPs (Belletti 2001, 2004; Rizzi 2004; Belletti, Bennati, Sorace 2007) and NPs (Giusti 1996). Let me start with the former case. Italian, as many other pro-drop languages, can display a subject in postverbal positions as well as in preverbal positions: (61) a. Luca ha telefonato Luca has called ‘Luca called’ 155

According to Benincà and Poletto (2004), the repetition of an element in the fronting position of a coordinate clause is a case of anaphoric anteposition structure. 156 Like the previous case, the anaphoric anteposition construction maintains a sort of contrastive semantics even though it is weak. In (58) the demonstrative questo corrects the wrong belief that Lorenzo would not behave like Gianni by going to a bar instead of going to school.

Chapter 4

130

b. Ha telefonato Luca has called Luca ‘Luca called’ The availability of a post-verbal subject is commonly called “free subject inversion”. Subjects in post-verbal position usually carry new information, as the following dialogue shows: (62) A: Chi ha rotto la macchina? ‘Who broke down the car?’ B: L’ha rotta Luca ‘Luca broke it’ The post-verbal subject Luca represents the new information introduced by the answer157. Comparative data suggests that the position of the subject in the linear word order is not free, and must follow the low adverbs of Cinque’s (1999) hierarchy (see the schema in 26). Consider, for example, the adverb tutto (everything): (63) a. ?Spiegherà tutto Maria explain.3rdsing.Fut everything Maria al direttore (Belletti 2004a: 19) to the director ‘Maria will explain everything to the director’ b. *Spiegherà Maria tutto al direttore explain.3rdsing.Fut Maria everything to the director The subject cannot precede the adverb, and this also suggests that it appears in a lower structural position. Starting from the observations in (61-63), Belletti (2004a) proposes that the post-verbal subject moves from its base position (i.e. Spec, vP) to a focalized one immediately above the argumental part of the structure (vPleft periphery). There, it takes its own interpretative feature conveying the new informational part of the sentence. From this point of view, Belletti proposes a sort of left periphery in the low part of the sentence above vP:

157

I have already discussed this fact in relation to the informational focus. See the paragraph above.

A preliminary step

131

(64)

(Belletti 2004a, p. 25) A sentence such as (61b) (Ha telefonato Luca) is associated with the representation in (65): (65) [CP ......[TP pro...ha telefonato ... [TopP [FocP Gianni [TopP [VP.......]]]]] Theoretically, an alternative proposal to (65) says that the post-verbal subject is still in the left periphery (CP-field), and IP passes it by remnant movement (in order to give the correct word order). According to Belletti, this is not the case. A possible proof that the post-verbal subject has raised in the low (vP) focus position is given by negative quantifier subjects. As I discussed in Chapter 2, n-words must be c-commanded by a negative operator when they occur in a post-verbal position, otherwise it raises ungrammaticality. This is also true when the n-word realizes the subject of the sentence. Compare, for example, the following sentences: (66) a. Non ha mangiato nessuno neg has eaten nobody ‘Nobody ate’ b. *Nessuno non ha mangiato nobody neg has eaten Since the negative operator is in a IP-internal position, we have to exclude that the subject nessuno raises in the left periphery and IP passes it by remnant movement. The only available possibility is that the subject raises in the vP-Foc position, remaining in the government domain of negation. In Belletti’s proposal (64) there are also topicalized projections in the array of functional heads forming the left periphery of vP. Such an assumption takes into account other types of structures in which the post-verbal subject does not seem to be in the focalized position. For example, we know that wh-questions do not allow focalized structures (cfr. Rizzi 1997) and, at the same time, we also know that focus, but not topic, raises weak-crossover violations. In a sentence such as (61), such constrains are violated but the sentence is perfectly fine:

132

Chapter 4

(67) Attualmente, in quale suoi appartamento vive Giannii? at present in which his apartment lives Gianni? ‘In which of his apartments does Gianni live at the moment?’ (Belletti 2004a: 41) If the post-verbal subject Gianni were in the focalized position, we would expect that (67) was ungrammatical. Belletti (2004a) proposes that in such a case the post-verbal subject raises in the topic position instead of in the focus one. This fact is also supported by other contexts in which the postverbal subject clearly carries a topicalized function: (68) A. Che cosa ha poi fatto Gianni? ‘What has Gianni finally done?’ B. Ha parlato Gianni. ‘He has spoken, Gianni.’ In (68B), Gianni does not carry any new information but, on the contrary, refers to the old one. To sum up, the introduction of some functional projections in the periphery of vP can easily take into account some very productive Italian structures such as the free-subject inversion158. Giusti (1996, 2005) made a similar proposal for NPs. Compare the following sentences: (69) a. [I suoi capelli bianchi] caddero a terra the her/his hair white fell to ground b. [I suoi bianchi capelli] caddero a terra the her/his white hair fell to ground c. [I bianchi suoi capelli] caddero a terra the white her/his hair fell to ground ‘Her/his hair fell down to the ground’ The basic word order of the elements is given by (69a): the possessive suoi precedes the noun capelli, which, in turn, precedes the adjective bianchi. In (69b) and (69c), the adjective has moved before the noun and before the possessive respectively. As Giusti points out, when the adjective is in a 158

As Rizzi (2004) points out, the proposal that the verbal phrase is also endowed by topicalized and focalized phrases (as with the CP) matches with the idea of the edge of phases in Chomsky (2001). Such a parallelism could be a good overlap between the minimalist and the cartographic approach.

A preliminary step

133

prenominal position, it refers to some shared information already present in the context of speech (put in other words, it expresses the topicalized function). In order to explain the data in (69b-c), Giusti suggests that the DPstructure displays an array of functional projections with specific discourserelated features exactly like the left periphery and the vP-periphery. More specifically, she proposes that DPs are associated with the following representation: (70) [DP … [D .[TopP TopP [FocP FocP [… N]]]]] According to Giusti (1996), only the topic positions allow movement of the adjective (69a-c). In fact, Italian does not permit focalization in DPs: (71) a. *[I suoi BIANCHI capelli] caddero (non i neri) the her/his white hair fell not the black b. *[I BIANCHI suoi capelli] caddero (non i neri) the white her/his hair fell not the black However, the structure in (70) maintains the FocP position because some other languages, such as Serbo-Croatian, realize it. In conclusion, in this chapter we have seen that the split-operation, originally proposed for IP, has been progressively adopted for all the other functional projections (CP, DP, ext.). I have focused my attention on the structure of CP because Snegs interact with it. We saw that CP is a complex head displaying different functions: it realizes the type of the clause and the finiteness of the IP embedded in it; it hosts operator-like elements and discourse-related features. Such complexity has led scholars to decompose CP in an array of functional heads (cartographic approach) or in a head with multiple-specifiers (minimalist approach). Since Snegs only allow topicalized structures and not focalized ones, I have investigated the syntactic differences between them. I also sketched some focalized and topicalized Italian clauses as examples of the high productivity of these constructions in Italian. We have now all we need to investigate the structures of Snegs.

CHAPTER 5 SNEGS AND THE LEFT PERIPHERY: TOWARD A SYNTACTIC HYPOTHESIS

I started this book wondering whether and how a peculiar instance of an expletive negation sentence, Snegs, were syntactically encoded. I have already answered the “whether” question in the previous chapters, showing that Snegs interact with other well-known grammatical structures (eg. ethical dative). I classified Snegs as strong expletive negation clauses and I pointed out their semantic and pragmatic features. In Chapter 3, I attempted to consider Snegs as instances of some other grammatical clausal types–like exclamatives or interrogatives–because of some common intonational and distributional patterns. Unfortunately I have excluded such a possibility and, therefore, the “how” part of my question remains unresolved. The alternative that I want to pursue is to consider Snegs as clauses with their own syntactic configurations. This is precisely what I am going to show in this chapter. In Chapters 3 and 4, I argued that the relationship between Snegs and the left periphery is crucial for the correct representation of Snegs. I started from the observation that Snegs cannot co-occur with fronting focalized phrases, but only with topicalized ones. In the brief discussion that followed, we saw that several Italian clauses engage in discourse-related phenomena and, I focused on them. Now I want to investigate more thoroughly the relationship between Snegs and all those structures in order to shed light on the nature of the Sneg-CP interaction. After that, I will elaborate a syntactic hypothesis for Snegs from which it is possible to derive their grammatical features.

5.1 A crucial observation In Chapter 2, we saw that Snegs cannot co-occur with focalized phrases, but only with topicalized ones, both in fronting (1) and in situ (2) positions:

Snegs and the Left Periphery

135

(1) a. *E GIANNI non mi è sceso dal treno (non Maria)?! and Gianni EN ED.to me is got off-the train (not Maria) b. E il libro Gianni non me lo ha dato a Luca?! and the book Gianni neg ED.to me Cl.it has given to Luca ‘The book, Gianni gave it to Luca!’ (2) a. *E non mi è sceso dal treno GIANNI (non Maria)?! and EN ED.to me is got off-the train Gianni (not Maria) b. E Gianni non me lo ha dato a Luca il libro?! and Gianni EN ED.to me Cl.it has given to Luca the book ‘The book, Gianni gave it to Luca!’ In Chapter 4, we also saw that different focalized and topicalized structures conveyed different semantic and pragmatic interpretations. A reasonable question is whether the pattern in (1-2) only concerns one typology of discourse-related element, or if it works for every instance of them. More specifically, according to Rizzi (1997), the Italian focalization in (1-2) carries the semantic feature of contrast. It may be possible that Snegs are semantically incompatible with such a feature but not with the focalization in itself. We know that the contrastive flavor is absent, or at least extremely weak, in some other focalized structures (eg. the anaphoric-anteposition). If Snegs are banned even in these cases, we could have a strong clue that it is the syntactic structure involved in focalization that plays a central role in the incompatibility with Snegs and not (only) the semantic characterization. A similar discussion also goes for topicalized structures since, as we saw, they display multiple semantics. Let me start with this. The crucial point in the relation between Snegs and topics is that they can always co-occur in the same sentence regardless of the typology of the topic. Even though we consider the different cases proposed in Benincà and Poletto (2004), in Frascarelli and Hinterhölzl (2007) and in other works (such as Benincà 1988, Cecchetto 1997, 1999, Cardinaletti 2002), the pattern does not change: Snegs are always allowed. (3) a. E Mario, gli amici, non gli hanno fatto un brutto EE Mario the friends Sneg to-him have.3rdplu done a bad scherzo?! (Hanging Topic) joke 'Mario, his friends played a bad joke on him!'

136

Chapter 5

b. E a Mario, gli amici, non gli hanno fatto un brutto E to Mario the friends Sneg to-him have.3rdplu done a bad scherzo?! (Left Dislocation ) joke 'To Mario, his friends played a bad joke on him!’ c. E domani Gianni non mi viene a trovare?! (Scene Setting) EE tomorrow Gianni Sneg Cl comes to see 'Tomorrow Gianni is coming to see me!’ d. E la frutta non la regaliamo e la verdure EE the fruit Sneg Cl.it give for free.3rdplu.pres and the vegetables non la vendiamo?! (List Interpretation) Sneg Cl.it sell.3rdplu.pres 'We give fruit for free, while we sell the vegetables!' e. E questo, io ai ragazzi non l’ ho detto EE this I to.the boys Sneg Cl.it.have.1stsing told ‘I told that fact to my students direttamente?! (Aboutness, contr. & familiar topic) directly!’ f. E Luca non ha già mangiato, il gelato?! (Marginalization) EE Luca Sneg has already eaten the ice-cream ‘Luca has already eaten the ice-cream!’ g. E Luca non l’ha già mangiato, il gelato?! (CLRD) EE Luca Sneg Cl.it has already eaten the ice-cream ‘Luca has already eaten the ice-cream!’ In (3a-g), different types of topicalized structure co-occur with Snegs. According to the original proposal sketched in Chapter 4, they occupy different positions in the CP-field. Since there is no limitation of topics in Snegs, all their syntactic projections are available. Let us now consider the focalized structures. As (1) and (2) show, Snegs cannot co-occur with contrastive focalized phrases. This fact is consistent regardless of the typology of the focalized phrase, i.e. an argument, a predicate or a whole sentence, and regardless of the fronting (4) or in situ (5) position: (4) Context: E Laura non mi dice che i turisti pensano che Milano sia piovosa?! (‘Laura says that tourists think Milano is rainy!’) a. *E Laura non mi dice che [TRENTOi]Contr.Focus i turisti EE Laura Sneg ED says that Trento the tourists pensano che ti sia piovosa?! (non Milano) think.3rdplu.sing that is.3rdsing.Subj rainy (not Milano)

Snegs and the Left Periphery

137

b. *E Laura non mi dice che i turisti pensano che EE Laura Sneg ED says that the tourists think that [SIA SOLEGGIATAi]Contr.Foc Milano ti ?! (non piovosa) is.3rdsing.Subj sunny Milano (not rainy) c. *E Laura non mi dice che [ROMA SIA SOLEGGIATAi]Contr.Foc EE Laura Sneg ED says that Roma is.3rdsing.Subj sunny i turisti pensano ti ?! (non che Milano sia piovosa) the tourists think (not that Milano is rainy)’ (5) A: E Luca non mi ha vinto una medaglia alla gara di corsa?! EE Luca Sneg ED has won a medal to the race of running ‘Luca won a medal at the running race!’ B: *E Luca non mi ha vinto UNA MAGLIETTA alla gara di EE Luca Sneg ED has won a t-shirt to the race of corsa?! (non una medaglia) running (not a medal) B’: *E Luca non mi ha PERSO alla gara di corsa?! (non EE Luca Sneg ED. has lost to the race of running (not ha vinto una medaglia) has won a medal) According to Rizzi (1997), focalized phrases in situ also display a (covert) movement to the left periphery. As a consequence, cases such as (4) and (5) do not clarify the nature of the incompatibility between Snegs and focalized structures; in fact, both the syntactic movement and the semantic interpretation may be the cause of such an incompatibility. Fortunately, there are, at least, two cases helping us in discerning the nature of such incompatibility: i) some structures display fronting focalized elements without a contrastive semantics (i.e. Mirative fronting focus (6), QPfronting (7) and Anaphoric-Anteposition(8)); ii) some structures display a contrastive-like focal semantics (ex. the focalized particles(9)) without the movement to CP. Let us examine those cases. (6) *Ora ricordo. E una sciarpa rossai non mi ha now remember.1stsing.pres EE a scarf red Sneg cl.to-me has regalato ti Luigi per Natale?! (Mirative fronting focus) given Luigi for Christmas (7) *E qualcosai non mi ha concluso ti Luca stando EE something Sneg ED has accomplished Luca being qui?! (QP-fronting) here

138

Chapter 5

(8) A: Gianni è andato al bar invece di andare a scuola (Anaph. Ant.) ‘Gianni went to a bar instead going to school’ B: *E questoi non mi ha fatto ti Lorenzo?! EE this Sneg ED has done Lorenzo (9) a. ?/*E anche la pizzai non mi ha mangiato Paolo ti?! EE also the pizza Sneg ED has eaten Paolo a’. E non mi ha mangiato anche la pizza Paolo?! EE Sneg ED has eaten also the pizza Paolo ‘Paolo also ate pizza!’ b. *E solo la pizzai non mi ha mangiato Paolo ti ?! EE only the pizza Sneg ED has eaten Paolo b’. E non mi ha mangiato solo la pizza Paolo?! EE Sneg ED has eaten only the pizza Paolo ‘Paolo only ate pizza!’ In (6), the Mirative fronting focus una sciarpa rossa (a red scarf) is moved to the left periphery and the Sneg sentence is completely ungrammatical. According to Lonzi (2004), Brunetti (2004, 2008) and Cruschina (2012), such a structure displays fronting focus without contrastive semantics. Consequently, the ungrammaticality with Snegs has to be related to the syntactic movement of the focalized constituent into the CP-field. The same is true for (7) and (8): the contrastive semantics does not play a central role in such structures and, therefore, the ungrammaticality with Snegs is related to their syntactic configuration. This consequence is further highlighted by (9). All the sentences in (9) display a contrastive semantics given by the focalized particles. However, in (9a-b) the focalized phrase moves to the CP-field whereas it does not in (9a’-b’). In both cases, the particle gives a contrastive semantics to the focalized phrase, but Snegs are ruled out only when the focalized element is fronted. Since the contrastive semantics is always present, the incompatibility with Snegs has to be related to the syntactic movement to the [Spec, FocP]. The sentences (6-9) are strong clues concerning the central role of focus projection in the CP-field in defining the structure of Snegs. To sum up, we saw that Snegs allow topicalized structures but not focalized ones (regardless of the strength of the contrastive semantics). The discriminant fact seems to be the syntactic configuration of those structures: Snegs do not allow elements to move into the focus position of the CP-field, but allow elements moving in the topic positions. This conclusion partially matches the observation in Rizzi (2004a) on the interaction between some

Snegs and the Left Periphery

139

adverbs and Italian negation. He notes that adverbs like rapidamente (rapidly) cannot be focalized in a fronting position when they occur in a negative sentence (10b) while, at the same time, they can be topicalized (10c): (10) a. RAPIDAMENTE i tecnici hanno probabilmente risolto il ‘RAPIDLY the technicians have probably solved the problema (non lentamente) (Rizzi 2004a: 235) problem (not slowly)’ b. *RAPIDAMENTE i tecnici non hanno risolto il rapidly the technicians not have.3rdplu solved the problema problem c. Speravo proprio che potessero sbarazzarsi rapidamente di questo problema, ma devo dire che, rapidamente, non lo hanno risolto. ‘I really hoped that they could rapidly get rid of this problem, but I must say that, rapidly, they didn’t solve it.’ The data in (10) led Rizzi to propose that negation and focalization display the same kind of licensing features (i.e. quantificational), making (10b) ungrammatical by a relativized minimality violation. In any case, what I want to emphasize here is the particular interaction between focus and negation. Such an observation is crucial for Snegs since they do not have a correspondent affirmative form (as we saw in Ch. 2): (11) a. E non mi è scesa dal treno Maria?! ‘Maria got off the train’ b. *E mi è scesa dal treno Maria?! EE ED is got off to the train Maria It is clear that, on the one hand, negation is the crucial element characterizing Snegs and, on the other hand, the interaction with the left periphery strongly affects their distribution (topicalized structures are allowed, unlike focalized ones). I will coherently start my research on the syntactic configuration of Snegs by investigating the relationship between negation and the CP-field.

5.2 CP and neg features At the beginning of this book we saw that every language displays one or more elements representing sentential negation. In such a rich set of

140

Chapter 5

possibilities, some of them are particularly interesting for the relation between negation and the CP-field, i.e. the negative complementizers. Negative complementizers are present in many languages and they melt the function of both linguistic negation and the complementizer, making negative the clause that they select. For example, Irish displays a complex system of negative complementizers and, according to McCloskey (2001) and Duffield (1995), two of them are widespread: i) nach, which introduces both negative embedded clauses (12a) and interrogatives (12b); ii) cha, which introduces matrix clauses (13)159. (12) a. Creidim nach gcuirfidh sì isteach ar an phost. I-believe NEG- COMP put [FUT] she in on the job ‘I believe that she won’t apply for the job’ů (McCloskey 2001: 75) b. Nach gcuirfidh sì isteach ar an phost? NEG- COMP put [fut] she in on the job ‘Won’t she apply for the job?’ (13) Cha cheangaileann sé é NEG-COMP tie [PRES] he it ‘He doesn’t tie it.’

(McCloskey 2001: 80)

The negative meaning of nach and cha is clear in (12) and in (13) because they are the only negative element in the clause160. Some other examples of negative complementizers are the Modern Greek min (14) (Makri 2013), the Latin ne and quin (15) (Moscati 2010), the Basque enik and ez (16) (Laka 1990): (14) Kitakse min to eleghe charitologhontas look NEG it.CL said jokingů ‘Check, if he said (that) joking.’

(Makri 2013:15)

(15) a. Dionysius tyrannus [ne tonsori collum committeret,] tondere suas filias docuit (Cic., Tuscolanae Disputationes in Moscati 2010:24) ‘The tyrant Dyonisius, [in order to not commit the neck to a barber], taught his daughter to shave’ 159

McCloskey (2001) discusses some other cases of negative complementizers. I voluntary omit them because a deeper discussion on the Irish complementizer system is beyond the aim of this book. 160 See Moscati (2010) and the references therein for a detailed discussion on the complementizer nature of such elements.

Snegs and the Left Periphery

141

b. Non possumus, quin alii a nobis dissentiant, NEG can that-NEG others ABL. us disagree recusare (Cic. Ac. 2, 3, 7 in Makri 2013: 14) refuse ‘We can’t refuse others disagree with us’ (16) a. Amaiak inork gorrotoa dionik ukatu du (Laka 1990:210) Amaia anyone hatred has-comp-neg denied has ‘Amaia denied that anybody hated her’ b. [Erori ez-den] etxea (Laka 1990: 44) fallen no has-that house-the ‘The house that didn’t fall-down’ According to Moscati (2010), negative features are realized as negative morphemes in the specifier position of either Force° or Fin°–heads usually associated with the complementizer elements (cfr. Rizzi 1997): (17) [forceP negative compl. [Force° (TopP*) IntP (TopP*) FocusP (TopP) [FinP negative compl [Fin° … ]] Once again, the presence of negative complementizers reinforces the idea that the relation between CP and negation is particularly strong, exactly like in Snegs. Unfortunately, Italian does not display such negative complementizers, and the evidence of the relation between negation and CP seems to be less strong. However, some other phenomena suggest an opposite conclusion. Let me start with an observation in Progovac (1994)161 about the doubt verb in English. Verbs such as doubt are adversative predicates usually introducing a subordinate clause or a PP: (18) a. ‘I doubt [that you did it]’ b. ‘I doubt [of you]’ Interestingly, subordinate clauses introduced by doubt can trigger NPIs by themselves: (19) I doubt [that you saw anybody]

161

See also cfr. Moscati (2010).

142

Chapter 5

The verb to doubt clearly triggers the NPI anybody; if we consider other verbs, such as to say, this possibility is ruled out: (20) *I say [that you saw anybody] The pattern in (19 -20) can be correctly predicted if we hypothesize that verbs like to doubt select a CP with an active negative feature: (21) I doubt [ForceP Øi[+neg] that [IP you saw anybody] (Moscati 2010: 32) The negative feature in the CP-domain triggers the NPI anybody. As a proof, such a possibility is ruled out when the dubitative predicate selects a PP, which lacks CP: (22) *I doubt of anybody Moscati (2010) observes that the derivation in (21) is not different from what I have discussed about negative complementizers in Basque, Latin, etc. In fact, in both cases a negative feature establishes the “denying” part of the selected sentence. Such a feature requirement is not ad hoc proposal for to doubt. As is well known (cfr. Grimshaw 1979 and Pesetsky 1991), verbs such as to say and to wonder display specific requirements on the selected subordinate clause when they are in a matrix clause: (23) a. I wonder [whether John comes] b. *I wonder [that John comes] (24) a. *I say [whether John comes] b’. I say [that John comes] Verbs like to wonder only select interrogative complementizers such as whether; on the contrary, verbs like to say only select declarative complementizers such as that. This fact has been explained by selecting requirement: to wonder selects a [+Question] feature carried by whether and to say a [+Declarative] one carried by that. The violation of this selecting requirement gives the ungrammaticality in (23b) and (24a). Mutatis mutandi, this is exactly what happens with to doubt and the [+negative] feature: the verb in the matrix clause selects and imposes a negative feature on CP of the embedded clause (even though it is not phonologically represented).

Snegs and the Left Periphery

143

Crucially for my research, Italian displays a negative feature in CP too. Two cases, above all, need to be considered: dubitative verbs and the withoutclause. The Italian dubitative verb dubitare behaves exactly like the English correspondent form to doubt: it selects either a CP (25a) or a PP (25b,) and it can trigger a NPI only when the matrix verb selects a CP (compare 26a and 26b). (25) a. Dubito [che Maria abbia capito] Doubt.1stpres that Maria have.3rdsing.conj understood ‘I doubt that Maria understood b. Dubito [di Luca] Doubt.1stpres of Luca ‘I doubt of Luca’ (26) a. Dubito [che James possa spifferare niente ai russi ... I-doubt that James could-su reveal anything to-the Russians è la nostra spia migliore]! (Moscati 2010: 92) is the our spy best ‘I doubt that James could reveal anything to the Russians... he is our best’ spy!’ b. *Dubito [di niente] Doubt.1stsing.pres of anything Like the English cases, only a negative operator can trigger the NPI niente when it occurs in subordinate clauses. In fact, this possibility is ruled out when the matrix verb does not select a CP with an active negative feature. Consider, for example, a verb such as pensare (to think), the affirmative form of dubitare: (27) a. *Penso [che James possa spifferare niente ai think.1stsing.pres that James could reveal anything to-the russi… è la nostra spia migliore]! .. Russians is the our spy best (28) b. Penso [che James possa non spifferare niente ai think.1stsing.pres that James could neg reveal anything to-the russi è la nostra spia migliore]! Russians is the our spy best ‘I do not think that James could reveal anything to the Russians...he is our best’

144

Chapter 5

The sentence in (27a) is the correspondent form of (26a), but with the verb pensare it results as ungrammatical. The only way to rescue it is to add the negative marker non to the predicate162. Such asymmetry between dubitare and pensare can be easily taken into account if we assume that only the former verb selects a CP with a negative operator, whereas the latter does not (cfr. Moscati 2010). A further proof that the selected CP hosts a negative feature is that it also allows strong NPIs like un tubo, which can never be triggered by downward entailment contexts (as in the case of niente), but only by negation (cfr. Chapter 1): (29) Dubito [+Neg che Luca abbia capito un tubo doubt.1stsing.pres that Luca have.3rdsing.subj understood a tube del piano] of the plan ‘I doubt that Luca has understood anything of the plan’ The idea of a negative feature is also supported by without-clauses. Like the verb dubitare, such clauses allow n-words (30a) and strong NPIs (30b) (cfr. Alonso-Ovalle and Guerzoni 2004): (30) a. Maria è uscita dal negozio senza [+neg aver salutato nessuno] Maria is left to the shop without to have greet anybody ‘Maria left the shop without saying goodbye to nobody’ b. Maria è uscita senza [+neg aver capito un tubo della lezione] Maria is left without to have understood a tube of-the class ‘Maria left the class without having understood anything’ Dubitative verbs and without-clauses further reinforce my observation about the interaction between negation and CP-field. Even though Italian does not 162

The sentence in (27b) can sound a bit weird for an Italian native speaker because the negative marker usually raises in the matrix clause for negative raising (see chapter 1): (i) Non penso [che James possa spifferare niente ai Russi… neg think.1stsing.pres that James could reveal anything to-the Russians è la nostra spia migliore]! is the our spy best ‘I do not think that James could reveal anything to the Russians... he is our best spy’

Snegs and the Left Periphery

145

display explicit negative complementizers (as some other languages), we have seen that it may display a covert negative feature hosted by CP. We can now add this observation to the one seen above, i.e. that discourserelated structures like focalization and topicalization play a central role in the relation with negation in Snegs. What I am going to argue in the next session is that the couple “negation-CP” is the core element for deriving all the features seen until now when characterizing Snegs. We are arriving at the core of my thesis.

5.3 A syntactic proposal for Snegs In this journey we have faced up to the syntactic structure of Snegs. It is important to recall their features briefly: i) ii) iii) iv) v) vi) vii) viii) ix) x) xi) xii)

Snegs can host any kind of predicate, regardless of their valency and their diathesis; They allow the subject to be in a pre-verbal and post-verbal position (even though it carries a topicalized flavor when it occurs in the preverbal position) both in nominative and dative cases; They are instances of strong expletive negations (affirmative in meaning, they allow p-words but not strong/weak NPIs, n-words and not-also conjunctions); They do not allow al-words and elements requiring downwardcontexts; They are only realized by the negative marker non, rejecting other denying elements like n-words in a preverbal position; Pragmatically, they carry new information with an unexpected value; They can be coherent answers to propositional questions; They have an intonational pattern melting together interrogative and exclamative prosody; Their interpretation is strongly related to the occurrence of the Ethical Dative and of the expletive E; They interact with the left periphery allowing topicalized structures and rejecting focalized ones (both in fronted and in situ positions). They cannot be identified with negative exclamatives, rhetorical questions and bi-clausal structures even though they share some features with them. They cannot be embedded under any kind of predicate.

146

Chapter 5

A perfect proposal for a syntactical representation of Snegs would be able to correctly predict all the elements in i-xii or, at least, most of them. Cleary, the behavior of Snegs with some of those elements could be related to independent reasons and, therefore, we need to verify them case by case. Let me now introduce my hypothesis. What I would like to propose is that in Snegs the Italian negative marker non is externally merged in the left periphery of the sentence. It displays a head status and selects a FocP as its argument, which is compulsorily filled by a TP that has been moved there (internal merge). The structure in (31) shows this proposal graphically: (31) [CP … [X° non ] … [TP Foc° [… tTP …]] Let us consider step by step the schema in (31), starting from the first piece: the negative marker non. As we saw in Ch. 1, the Italian negative marker non displays a head status (Kayne 1989b; Zanuttini 1996, 2001) realizing the head of NegP. I do not depart from this assumption because there are no theoretical and empirical reasons to do so. In fact, as we know, Italian also displays a negative form with the maximal projection nature, no, but Snegs refuse to occur with it: (32) *E no mi è scesa dal treno Maria?! EE neg ED is got off-the train Maria We also do not need to change the idea that the specifier position of NegP is filled by a negative operator (contra Espinal 1992, 2000 and van der Wurff 1999) as in the case of standard negation. The Snegs negative marker non is exactly the same as the one in standard negation. The original part of my proposal is that in Snegs, non is merged directly in the CP-field instead of in the TP-area163.

163

Clearly, a reasonable question is what non denies, since the meaning of Snegs is affirmative whilst the negative marker is still negative; I do not face this issue now, but I will do later. In Ch. 1 we saw that a similar proposal is also sketched in Makri (2013), Abels (2002, 2005) and Delfitto and Fiorin (2014). I depart from Abels (2002, 2005) in that I propose that the negative marker in Snegs is externally and not internally merged in the CP.

Snegs and the Left Periphery

147

From this point of view, it is extremely economical to consider expletive and standard negation as the same morphological and semantic phenomenon with a different syntactic representation: when the negative marker non is externally merged in the TP area (cfr. Belletti 1990; Laka 1990; Zanuttini 1996, 1997, 2001), it gives the standard negation reading; when it is externally merged in a higher position, the CP-field, it gives the expletive negation reading (at least, in the Snegs’ case)164. For the same reason, I also refuse to consider non as the morphological realization of two different elements, i.e. EN and standard negation (as in Portner and Zanuttini 2000 and Zanuttini and Portner 2003), but I assume that Italian displays unique negation, with different interpretative rules depending on the place in which it is merged. Referring to other proposals seen in Chapter 1, we have no evidence that non is a member of a negative concord phenomenon as well (cfr. Fischer 1992; van Der Wouden 1994a, 1994b, 1997), since it is not licensed by any other negative element. We are also forced to reject the idea that non is an evaluative mood marker (cfr. Yoon 2011) since Snegs do not accept the subjunctive mood usually required by those kinds of elements: (33) *E non mi sia scesa dal treno Maria?! EE EN ED be.3rdsing.subj got off to the train Maria In my proposal, the negative marker non is always the expression of Italian linguistic negation. Coherently, Snegs display the same restrictions as the other Italian negative sentences seen in Chapter 1. For example, it must be adjacent to verbs because of its clitic-like nature and, therefore, it cannot occur in TP-lacking structures, such as past participle clauses (34a), Absolute Constructions165 (34b), and Imperative clauses166 (34c): (34) a. *E non raggiunta Maria, siamo andati al bar?! EE EN reached Maria are.1st.plu went to the bar b. *E non arrivata in ritardo, Maria mi è stata sgridata EE EN came in late Maria ED is been reprimanded dalla maestra?! by the teacher

164

See below for a discussion on the reason why the negative marker is not able to deny the standard meaning of the sentence when it is merged in the CP. 165 Above all, cfr. Zanuttini (1996), Belletti (1992, 2006), Kayne (1989b); 166 Cfr. Kayne (1992); Riviero (1994); Riviero and Terzi (1995); Zanuttini (1994, 1997) and Zeijlstra (2004).

148

Chapter 5

c.*E tu non mi telefona a Gianni! EE you EN ED call to Gianni For the same reason, Snegs cannot accept elements located between non and the verb (35a), unless they were already adjuncts to the verb like pronominal clitics (35b). (35) a. *E Maria non ieri mi ha telefonato?! EE Maria ED yesterday ED has phoned b. E Maria non gli ha telefonato nel cuore della notte?! EE Maria EN Cl.to him has phoned in the heart of the night ‘Maria phoned to him in the dead of the night!’ Verbal adjacency is also the reason why Snegs cannot display the subject between negation and the verb: (36) *E non Maria mi è scesa dal treno?! EE EN Maria ED is got off to the train In fact, according to the schema in (31), the word order non>subject>verb is the expected one in Snegs since the negative marker dominates the whole TP which hosts the subject in its specifier position (between negation and the verb). This order is correctly ruled out in my proposal for the basic requirement that non has to be adjacent to the verb because of its “standard” nature. Just because non behaves like standard negation in Snegs, one could conclude that there are no reasons to assume that it lays in the CP-layer instead of at the edge of the TP area. Alternatively, it is possible to follow a mixed-path, i.e. non is externally merged in the TP-area and then it moves to the CP-field. However, we have some distributional reasons to reject such alternatives. In fact, when negation is generated above TP and either stays there or moves to CP by internal merge, it displays all the negative-scope phenomena absented in Snegs: it triggers n-words, strong and weak NPIs and not-also conjunctions. Consider, for example, the Aux-to-Comp phenomenon (Rizzi 1981, 1982; Cinque 1999). As I discussed in Chapter 1, in these structures the negative marker non carries along verbs moving in the CP-field. Crucially, the “negative” force of negation is still present:

Snegs and the Left Periphery

149

(37) a. Non essendo arrivato nessuno, siamo venuti da soli (n-word) neg be.gerund arrived nobody be.1stplu come by alone ‘Not being arrived anyone, we have come by ourselves’ b. Non avendo alzato un dito per aiutare Luca, Laura è stata neg having lifted a finger to help Luca Laura is being sgridata (weak NPI) reprimanded ‘Not having lifted a finger to help Luca, Laura has been reprimanded’ c. Non avendo capito un tubo, Laura è stata bocciata neg having understood a tube Laura is being rejected all’esame (strong NPI) to the exam ‘Not having understood anything, Laura has been rejected to the exam’ d. Non essendo arrivato Gianni e neanche Luca, siamo venuti neg being arrived Gianni and neither Luca be.1stplu come con Maria (not-also conjunction) with Maria ‘Not being arrived Gianni and Luca either, we have come with Maria Even though negation is in the CP domain at the end of the derivation, it has already triggered the n-words, NPIs and other negative elements. As we know, Snegs reject all the elements in (37a-d). This fact is easily seen if we assume the different derivation of the negative marker in Snegs: it is externally merged in the CP domain when the v*P phase has already closed167. In order to better understand this step, I have to devote a few words to the theory of phases as formulated in Chomsky (2000, 2001, 2004 and 2008). Chomsky proposes that the derivation of a sentence is cyclically structured in different phases in order to minimize its computational load. In fact, the computational system does not select a unique Lexical Array (LA) for the whole sentence, but rather several sub-arrays which have to be exhausted one-by-one in a sequential order: the computation first exhausts the first sub-array LAi forming a syntactic object, and then it passes to the second 167

This fact also suggests that negation does not impose any requirements on the selected predicate, which, therefore, can host any kind of verb regardless of its valency and its diathesis. This is exactly what happens.

150

Chapter 5

one LAi+1, etc. Each syntactic object created by exhausting each sub-array is a phase. Chomsky (2001, 2004) considers, at least, two phases: the v*P phase and the CP phase. The v*P phase is the argumental one with both the internal and the external arguments (as in transitive verbs); the CP phase is the “minimal construction that includes tense and event structure and (at the matrix, at least) force” (Chomsky 2004: 124). Every phase is inaccessible to the subsequent one, given that they are “semantically and phonologically coherent and independent”. When a phase is closed, it is mapped to the Sensorimotor and to the Conceptual-Intentional Systems by Transfer. The relation between two or more phases is led by the Phase Impenetrability Condition proposed by Chomsky (2000, 2001)168: (38) Phase Impenetrability Condition (PIC): The domain of H is not accessible to operations at ZP; only H and its edge are accessible to such operations. (Assumed structure: [ZP Z…[HP Į [H YP ]]]) (Chomsky 2001: 14) The domain of the head phase H is its complement (in this case the YP). PIC proposes that the head phase Z cannot see into YP when the phase H is closed. According to (38), it still possible to extract some material from a phase only passing through its head and its edge (the specifier). Since the C° and the v° are the head phases, it means that it is possible to extract by them passing through the [Spec, CP] and the [Spec, vP]. As has been emphasized (cfr. Chomsky 2001; Richards 2011; Citko 2014 and the references therein), a phase is closed only when the head of the subsequent one has been merged: the complement of the v is inaccessible only when the head of C is merged; the complement of the (embedded) C is inaccessible only when the matrix-v° is merged. Consider, for example, the following couple of sentences: (39) a. [CP2 Whati do you [vP think [CP1 i Maria gave i to Gianni]? b. *[CP2 To whomi do you [vP think [CP1 whatj Maria gave j i]?

168

The first formulation is found in Chomsky (2000). Chomsky (2001) changes the original proposal, introducing new elements (cfr. Citko 2014 for a detailed discussion).

Snegs and the Left Periphery

151

In (39a), CP1 phase is closed when the matrix vP is merged and its complement (TP) is transferred to the Sensorimotor and to the ConceptualIntentional System. The wh-element what can escape out CP1 because it is in its edge area, the [Spec, CP1]. In (38b), the derivation crushes because CP1 has already been Transferred when the wh-element to whom escapes out and this is not allowed by PIC. Phase-theory gives several advantages to the study of language (cfr. Lòpez 2009; Gallego 2012 and the references therein; see Chesi 2007 for a critical view), and it is also useful for the analysis of Snegs. More specifically, I assume that the negative marker non is externally merged in the CP-phase when the vP-one is already closed and transferred to the interfaces. Schematically, we can see it in (40): (40) [CP … [X° non ] … [FocP TP [Foc° [TP [vP [… v°… ]]]i]]] The underlined elements represent the phase of the derivation169. When non is merged as a head in the CP-field, the domain of vP is impenetrable and, therefore, negation cannot see inside it. From this fact, it follows that negation in Snegs is unavailable to trigger negative-scope elements like nwords, weak/strong NPIs and not-also conjunctions 170 . For the sake of clarity, I write the examples seen in Chapter 2 emphasizing the different phases in which different elements occur: (41) a. *[CP E non [vP mi è sceso dal treno nessuno]]?! EE EN ED.to me is got off-the train nobody b. *[CP E Luca non [vP mi ha la più pallida idea]]?! EE Luca EN ED has the most faintest idea c. *[CP E Maria non [vP mi ha alzato un dito per aiutare Luca]]?! EE Maria EN ED has lifted a finger to help Luca d. *[CP E non [vP mi è scesa dal treno Maria e EE EN ED.to me is got off-the train Maria and neanche Gianni]]?! neither Gianni 169

I did not point out other phases like the DP, PP, etc. I was only interested in propositional ones, as has been referred to in Chomsky (2000). For other types of phases, see Boškoviü (2002) and Chomsky (2008). 170 According to Giannakidou (1997) and Zeijlstra (2004), a negative operator (realized by the negative marker non in the Italian case) binds all free variables in the vP domain, and this allows NPIs, NC, etc. If there is not an available negative operator, or if vP is already closed when it is merged (like in the case of Snegs), then such a binding fails and all operator-variable structures are excluded.

152

Chapter 5

Coherently, Snegs allow PPIs such as già because these elements require to be in an affirmative domain: (41) [CP E non [vP mi è già scesa dal treno Maria]]?! and EN ED.to me is already got off-the train Maria ‘Maria already got off the train!’ Crucially, when the negative marker is first merged in the TP area, it can see inside vP because TP belongs to the v*P phase. In fact, according to Chomsky (2001), TP does not count as a head phase (only CP does). From this fact follows the pattern seen in the Aux-to-Comp structures in (37). A theoretical issue may arise at this point of the discussion. In Chapter 4 I assumed Rizzi’s split-CP hypothesis as the correct description of the CPfield. We have also seen that this assumption departs from the minimalism approach inasmuch as it proposes an array of functional heads forming CP, instead of one head with multiple specifiers (as proposed by Chomsky 2008). I want now to consider which head in the CP-field exactly is the phase head triggering the Transfer of vP. Theoretically speaking, each of the heads in the CP-field, reported here in (42), could be a good candidate. (42) [CP…Force°…(TopP°)…Int°…(TopP°)…Neg°…Foc°…(Top°)… Fin°…IP171 The solution of this issue is beyond the aim of this book, and also in literature it is far from been resolved. However, many proposals have been advanced and only some of them give good predictions for my hypothesis. A reasonable possibility is to consider Force° as the head of the CP phase since it closes up the sequence of functional heads in (42) (cfr. Rizzi 2006). Unfortunately, in this way my proposal would turn out to be wrong because the vP-complement would still be open at the point in which negation is merged in CP. As predicted by PIC, only the head of the phase triggers the Transfer of the previous phase and, consequently, the negative marker non would still interact with TP before the Force° is merged in the structure. An alternative solution is to consider the other “extreme” head of the CP-field, the Fin°, as the head of the phase. This has been proposed, among 171

Remember that I chose not to take a clear position on the different proposals present in the literature about the correct sequence of functional heads after the first proposal in Rizzi (1997). See Chapter 4 for a more detailed discussion on some alternatives.

Snegs and the Left Periphery

153

others, by Barrier (2007), Lopez (2009) and Robert (2012). According to this view, in Snegs, the negative marker is merged when the vP-complement is already closed, because non follows the Fin° in derivation and Fin° caused the Transfer of the vP-phase. Another alternative is proposed by Totsuka (2013). He states that both Force° and Top° are head phases. If it is true, my derivation of Snegs would be unproblematic because, as in the previous case, non is merged in the structure after the Top°, which already closed the vP-phase. Crucially, it has also been proposed that Foc° plays a central role in defining a phase. Shlonsky (2010) notes that the [Spec, Foc°] is the landing site of wh-elements (according to the original proposal in Rizzi 1997) and he wonders whether Foc° is the head phase of CP. Kiss (2006) supports this conclusion, arguing that the Foc° is the phase head in languages such as Hungarian. Again, Poletto (2006) emphasizes that in Old Italian, the “features of a functional head as Focus are parametrized as phase independent properties” (both in vP and CP edges). The hypothesis that Foc° is the head of the CP-phase is coherent with my analysis as well: negation dominates it and, therefore, the vP-domain is already closed at the time in which non is merged. We can also apply some tests of phasehood elaborated in the literature in order to investigate such a hypothesis (see Boškoviü 2014). According to Citko (2014), a XP can be isolated (for example, being an answer for a question), moved or eliciting ellipsis because of its head is a phase head; crucially, all these phenomena occur in Italian focus sentences. Consider, for example, the following dialogue: (43) A: Che cosa pensa Luca? ‘What does Luca think?’ B: Che Roma sia sporca that Roma be.3rdsing.subj dirty ‘Roma is dirty’ Extending Brunetti’s (2003, 2004) analysis172, this kind of structure displays the pronounced part in [Spec, FocP] with an ellipsis of TP: (44) [FocP Che Roma sia sporca [Foc° [TP pensa Luca]]] Moreover, according to Gallego (2010), a head phase is a source of uninterpretable features and we know that Foc° displays such properties. 172

See below for a more detailed discussion.

154

Chapter 5

By the way, the fact that non in Snegs belongs to a different phase to v*P correctly derives the impossible co-occurrence with negative-scope elements (as we saw above) and also takes into account some other phenomena. More specifically, negation in Snegs differs from the standard one in at least three more crucial aspects: it does not display the NEG-raising phenomenon; it does not interact with high adverbs; it does not legitimize the negative concord phenomenon. All those limitations are easily explained assuming that the negative marker in Snegs belongs to a different phase than the vP one. Let me start with the NEG raising phenomenon. As it is well-known (cfr. Jespersne 1917; Fillmor 1963; Horn 1989) that when verbs such as to think and to want take negative clauses as their complements, negation appears in the matrix clause (45a) instead of in the subordinate one. Consider, for example, the following sentence: (45) a. I do noti think Luca ti is smart b. ‘I think Luca is not smart’ According to Collins and Postal (2014), the sentence in (45) involves the raising of negation from the embedded clause (where actually it is interpreted) to the matrix one173. Interestingly, Snegs cannot co-occur in such structures (46): (46) Luca noni pensa che ti sia scesa dal treno M. (SN) Luca neg thinks that be.3rdsing.subj got off-the train Maria ‘Luca thinks that Maria did not get off the train’ E Luca non mi pensa che sia scesa dal EE Luca EN ED thinks that be.3rdsing.subj got off-the treno Maria?! (Sneg) train Maria . ‘#Luca thinks that Maria did not get off the train!’ . ‘Luca thinks that Maria got off the train!’ The content of Luca’s belief in the assertive clause (46) is negative because negation is externally merged above the embedded TP and, from there, it 173

See the original work for the arguments in support of the syntactic approach to the NEG raising. Collins and Postal (2014) discusses at least three reasons for which the syntactic approach is preferable: it explains why NEG raising obeys an island constraint; it predicts the properties of a peculiar NEG raising structure (the Horn clause); it explains the interaction with parenthetical structures.

Snegs and the Left Periphery

155

denies the subordinate predicate (as predicted by the NEG raising). In the Sneg sentence (46) the negative reading is ruled out. The only available reading is affirmative and the content of Luca’s belief is not denied. My hypothesis correctly predicts this pattern. In Snegs the negative marker is not externally merged above TP, but directly in the CP-field. From there it cannot see inside the subordinate predicate because the vP-phase is already closed. The NEG raising is, therefore, incompatible with Snegs. Let me now face up to the interaction of Snegs and various adverbs. As already discussed in Chapter 4, Cinque (1999, 2002) proposes that the order of AdvPs is fixed across languages, and follows a specific schema. In Chapter 1, we also saw that negation interacts with them giving some unexpected scope phenomena. For example, when negation follows some adverbs in the linear word order, it can either scope over the adverb or not, depending on the position in which adverbs are generated174: (47) Luca non ha deliberatamente fatto cadere la penna Luca neg has deliberately made drop the pen a. ‘Luca deliberately has not dropped the pen’ (deliberately…not) b. ‘Luca has not-deliberately dropped the pen’ (not…deliberately) If the adverb is generated above negation, we have the reading in (47a); if negation is generated above the adverb, we have the reading in (47b). Therefore, as Cinque states, “what is crucial for determining the scope of sentence negation is not its "surface" position (the one at "Spell-out"), but its "base" position” (Cinque 1999: 124). Surprisingly, both the readings in (47) are ruled out in the correspondent Sneg sentence: (48) E Luca non mi ha deliberatamente fatto cadere la penna?! EE Luca EN ED has deliberately made drop the pen a.‘*Luca deliberately has not dropped the pen!’(*deliberately…not) b.‘*Luca has not deliberately dropped the pen!’(*not…deliberately) c.‘Luca has deliberately dropped the pen!’ In (48), the negative marker does not interact with the adverb; neither does it scope over the adverb, nor does the adverb scope over it. According to Mizuno (2010), Edelstein (2012) and Giorgi (2016), adverbs such as deliberatamente are base generated in the vP-phase. Also in this case my hypothesis on the syntactic structure of Snegs takes into account the data in 174

See Chapter 1 for the syntactic explanation.

156

Chapter 5

(50): negation and adverbs do not interact because they belong to different phases and, therefore, they cannot see each other. Finally, there is another aspect that is useful to consider for my analysis: Snegs cannot license the negative concord (NC) phenomenon175. Compare the following couple of sentences: (49) a. Non è sceso dal treno nessuno neg is got off-the train nobody ‘Nobody got off the train’ a’. *E non mi è sceso dal treno nessuno?! EE EN ED is got off-the train nobody b. Luca non ha mangiato niente Luca neg has eaten nothing ‘Luca ate nothing’ b’. *E Luca non mi ha mangiato niente?! EE Luca EN ED has eaten nothing

(SN) (*Sneg) (SN) (*Sneg)

The NC is only allowed in standard negation structures; Snegs reject it. Much has been written on the syntactic and semantic derivation of the NC, and many theories have been discussed. Leaving aside most parts of this debate, I want to focus on a basic observation: the NC needs a minimal syntactic domain in order to be instantiated and it is not applicable across the board of a phase. (50) *Luca non ha detto [CP che Maria ha mangiato niente] Luca neg has said that Maria has eaten nothing Negation cannot interact with the argument of the v*P-phase because it is merged in the matrix clause when the embedded vP is already impenetrable: niente cannot stand alone without negative elements legitimizing it (as we saw in Ch. 1). According to my proposal, the same goes for Snegs in (49a’b’): the negative marker cannot stay in a NC relation with niente because non is merged in CP when the vP-phase has already been Transferred, and niente lacks a trigger176. To sum up, I have proposed that the nature of the negative marker in Snegs is the same as the one in standard negation: it is the head of NegP and it 175 176

See Chapter 1. See Tubau (2008) for an overview on the NC and the Phase-theory.

Snegs and the Left Periphery

157

must satisfy all the requirements usually associated with it (such as the phonetic proximity to the verb). I hypothesized that the first notable difference in Snegs is that non is externally merged in the CP-field when the v*P-phase has been exhausted by merging the CP-phase head. In such a syntactic configuration, I have proposed that non is unable to see inside the argument of vP banning the NC phenomenon, the occurrence with NPIs, nwords, NEG raising, etc. We can now focus on the second part of my proposal in (31): the head non selects a FocP as its argument requiring focalization on the whole TP177. The second crucial aspect of my analysis is the focalization of the whole TP. I have already discussed in Chapter 4, the criterial nature of the focalization and the mechanism of feature-checking lead to phrasal movement, or internal merge, in the sentence fronting position. I want now to propose that in Snegs, the negative marker non imposes on its argument, the FocP, the selection of the whole TP. The reasons for such an assumption are (for this moment) descriptive and predictive of what happens in Snegs. However, there are also some theoretical arguments supporting this hypothesis such as for example, the fact that negation in Snegs is propositional and, therefore, it still affects the whole flexed predicate (TP). What I would like to propose is that in Snegs, the whole TP is focalized. From this point of view, I can write the original example as in (51): (51) [CP E non [FocP [TP mi è scesa dal treno Maria] [ Foc°] [… tTP …]]]?! TP is internally merged in the [Spec, FocP] because of an uninterpretable [+Foc] feature carried by it. This assumption has a series of welcome consequences. First of all, the negative marker non does not select TP as its complement, as it usually does in standard negation sentences, but it selects the FocP. I want to suggest that this is the reason why surprise negation does not reverse the polarity of the sentence, since it plays a role on the presuppositional layer required by a focus. Such a hypothesis goes in the same direction of Benincà (1996)178. As a consequence, negation in Snegs would maintain its standard nature, losing its capability to deny a sentence because of its syntactic configuration. 177

I did not discuss the reason why (31) excludes the Snegs introduced by n-words. I could add more than one reason: it is because Snegs require non by definition; or because only non has the required head status; etc. 178 See Ch. 1.

158

Chapter 5

Secondly, since the [Spec, FocP] in Snegs is occupied by TP, the structure cannot display other focalizations (at least, with regard to the preposing position). Such a proposal easily predicts the asymmetry between the topicalized and focalized elements seen at the beginning of this chapter: only the former are allowed in Snegs, the latter are excluded. Compare, for example, this couple of sentences (I write here the sentences in 1 as 52): (52) a. *E GIANNIi non [FocP mi è sceso dal treno ti [Foc°] [tTP]]] and Gianni EN ED is got off-the train (non Maria)?! (not Maria) b. E [TopP il libro [Top°] Gianni non [FocP me lo ha dato and the book Gianni EN ED Cl.it has given a Luca [Foc°] [tTP]]]?! to Luca ‘The book, Gianni gave it to Luca!’ In (52a), the name Gianni cannot be focalized in a fronting position because TP has already satisfied the foc-criterion occupying the [Spec, FocP], leaving no more space for other foci. In (52b), such an impediment is ruled out because the fronting phrase has been moved forward of the TopP and not forward of the FocP. As I argued above, the impossible co-occurrence of Snegs and focalized elements is syntactically guided and my hypothesis seems to be on the right track. On the same line, my hypothesis also predicts that Snegs cannot display in situ focalizations. According to Rizzi (1997) and successive works179, there is only one (contrastive) focus phrase in a sentence180 and, therefore, only one focalized element (regardless of the position in which it is pronounced). Since TP is focalized, it follows that no other focalized elements can occur in Snegs even though they are in situ181. This is exactly what happens in sentences like (53):

179

See Chapter 4 for a detailed analysis. It has been suggested that Rizzi’s proposal is true only for contrastive focus and not for the other types such as the informational one (cfr. Cruschina 2012). In any case, such a theoretical background would not affect the central idea of my hypothesis. I continue to consider Snegs as a case of TP (contrastive) focalization. See Chapter 4 for a more detailed discussion. 181 According to Rizzi (1997), in situ focalized elements nevertheless display a (cover) movement (as I discussed in Ch. 4). If it is true, I have another reason to exclude that more than one focused element can occur in the same Sneg sentence. 180

Snegs and the Left Periphery

159

(53) a. *E non [FocP1 mi è sceso dal treno [FocP2 GIANNI] EE EN ED is got off-the train Gianni [Foc°] [tTP]]] (non Maria)?! (not Maria) b. E. G. non [FocP me lo ha dato a Luca [TopP il libro]] [Foc°]?! EE G. EN ED Cl.it has given to Luca the book ‘The book, Gianni gave it to Luca!’ The co-occurrence of two foci in the same sentence is banned by definition and, therefore, (53a) is ungrammatical. Conversely, (53b) is perfectly fine because there is only one focalized element (TP). Moreover, there is one more fact that the focalization of TP predicts: WH-elements cannot co-occur in Snegs because they compete for the [Spec, FocP] position (cfr. Rizzi 1997)182. (54) a.*E [FocP2 da quale treno]i non [FocP1 ti è scesa Maria ti]?!183 EE from which train EN ED is got-off Maria b. *E [FocP2 chi]i non [FocP1 mi è sceso dal treno ti]?! EE who EN ED is got off-the train c. *E [FocP2 che cosa]i non [FocP1 mi ha mangiato ti Luca]?! EE what EN ED has eaten Luca d. *E [FocP2 con chi]i non [FocP1 mi è arrivato Gianni ti]?! EE with who EN ED is arrivated Gianni Again, given that the [Spec, FocP] is already filled, there is not space for the wh-element (thus the ungrammaticality of 54a-d). The fact that TP lies in the [Spec, FocP] also suggests another remarkable consequence, i.e. that the whole predicate is the new information introduced in the context of the speech. In Ch. 2, I stated that this specific property of Snegs makes them suitable answers for what I called propositional questions. (55) A: Sembri sconvolto, cos’è successo? ‘You seem shocked, what did you happen?’ B: [CP Non [FocP [TP mi è scesa dal treno Maria] [Foc°] [tTP]]]?! ‘Maria got off the train!’ 182

For a detailed discussion on the relation Focus/Wh-elements, see Cruschina (2012). 183 For some native speakers, the sentences in (54) seem to be grammatical. This is because they interpret them as rhetorical negative questions.

160

Chapter 5

The whole predicate is the new information introduced by (55B). According to Ovalle and Guerzoni (2004) and Brunetti (2004), the part carrying new information in an answer lies in the focus domain, and this perfectly matches with my idea. Let me briefly show their proposal. They discuss the well-known fact that Italian displays two different forms for answering questions: one “extended” in which the speaker repeats part of the question under discussion (the topic/background) and adds new information (like in 56B); one “short”, in which the speaker only pronounces the part carrying new information (like in 56B’). The former form is usually judged redundant even if it is correct from the pragmatic and semantic points of views; the latter is more natural and preferred. (56) A: Che cosa ha vinto Gianni? ‘What has Gianni won?’ B: [Una maglietta]NEW ha vinto Gianni a shirt has won Gianni ‘Gianni has won a shirt’ B’: [Una maglietta]NEW a shirt ‘(Gianni has won) a shirt’ Intuitively, the “short” form presents the ellipsis of a part of the sentence (the background). According to Ovalle and Guerzoni (2004) and Brunetti (2004), the ellipsis regards TP and the part pronounced has been moved in the FocP escaping the deletion184: (57) [Foc [DP La maglietta]j [IP ha [VP vinto[DP Gianni] tj]] The pronounced part in (57) is the new information, and the idea that it lies in the FocP also follows the traditional distinction between the topic and the focus as the old and the new information constituting an utterance185. 184

See the original works and the references therein for a more detailed discussion. Interestingly, Ovalle and Guerzoni (2004) discuss cases in which a singular nword represents the short answer to the question. Consider, for example, the following dialogue: 185

(ii) A: Chi è venuto? ‘Who came? B: Nessuno ‘Nobody (came)’

Snegs and the Left Periphery

161

Crucially, my proposal on Snegs’ syntactic structure is perfectly coherent with the Ovalle and Guerzoni (2004) and Brunetti (2004)’s hypotheses. I just add a caveat. In my idea, the whole TP is focalized and, therefore, there is not evidence of ellipsis. For the rest, I present the same outcomes: Snegs are suitable answers to propositional questions because TP is focalized and, therefore, it carries new information. To sum up this last part, we have seen that the focalization of TP predicts several aspects of Snegs: they cannot host focalized elements in fronting or in situ positions; they also reject wh-elements; they are suitable answers for prepositional questions and the whole TP carries new information. Beside those syntactic considerations, I can also speculate that the interaction with the focus causes the particular prosody of Snegs. I do not deepen this consideration because the focus of this book is the syntactic representation of Snegs, but I can assume that the interaction with the left periphery also affects their intonational pattern. We know that a syntactic object forming a phase is transferred to the phonetic interface after the spellout. Since focus always has its own syntactic and intonational representation (cfr. Bocci 2008 and the references therein), it could also affect the intonational of Snegs. This is only a suggestion and we need further investigations in order to verify it. Leaving aside this latest part, we saw in this paragraph that the schema proposed in (31) and repeated here in (58) is able to take into account several aspects of Snegs. (58) [CP … [X° non ] … [TP Foc° [… tTP …]] Crucially, (58) also predicts a series of other phenomena that further reinforce the predictive power of my hypothesis.

In order to explain the syntactic configuration of B, they propose a FocP/TP-ellipsis structure like the one seen above. They also propose that Foc° can host a negative feature semantically active but phonologically empty. This would be the reason why n-words occur in pre-verbal positions even when no negative markers trigger them, like in (iii): (iii) [FOCP Nessunoj [FOC Ø [+neg] [IP tj [ è venuto]]] I report this hypothesis because this association between focus and negation captures an aspect also presented in my proposal.

162

Chapter 5

5.4 Advantages and consequences of the proposed analysis The first notable fact due to the syntactic proposal in (58) is that Snegs cannot be embedded in subordinate clauses. As we saw in Chapter 3, Snegs are a root phenomenon and they cannot be embedded under any kind of predicates. Among others, Snegs cannot realize the argument of indirect questions (IQ), of that-clauses (TC), of factive predicates (FP) or, more generally, of bridge verbs (VB)186: (59) a. *Mi chiedo [se non mi sia scesa dal treno Cl.to me wonder if EN ED be.3rdsing.subj got off-the train Maria]?! (IQ) Maria b. *Luca sa [che non mi è scesa dal treno Maria]?! (TC) Luca knows that EN ED is got off-the train Maria c. Mi spiace [che non mi sia scesa dal Cl.to me sorry that EN ED be.3rdsing.subj got off-the treno Maria]?! (FP) train Maria d. Luca dice che [non mi è scesa dal treno Maria]?! (BV) Luca says that EN ED is got off-the train Maria All sentences in (59) do not accept the Snegs reading, but only allow the negative standard one (as the English translations show). Let me discuss this fact. It has been proposed (cfr. Grewendorf 2002; Haegeman 2004, 2012) that some predicates, like factive verbs, select a reduced CP, i.e. a CP lacking some functional phrases. Crucially, such a CP lacks the FocP187. An easy way to take into account their ungrammaticality with Snegs is to assume my hypothesis: since Snegs display the whole predicate in the [Spec, FocP], they cannot occur in FocP-lacking clauses because there is no way to satisfy the focus requirement. Unfortunately, this explanation is good for factive predicates but not for the other ones. More specifically, we know that bridge verbs such as to say select a full-fledge CP (cfr. Benincà and 186 Prima facie, Italian speakers may consider the following sentences as grammatical, but this is true only when negation is interpreted as standard and not as expletive. 187 See Grewendorf (2002) and Haegeman (2004) for the theoretical and distributional arguments in favor of this hypothesis.

Snegs and the Left Periphery

163

Poletto 2004; Featherston 2005), which expresses the whole array of the functional projections including the FocP. As a consequence, there are no reasons for the ungrammaticality in (59d). An alternative is to assume Grimshaw (1979) and Pesetsky (1991)’s perspective on the selection requirement of a subordinate CP188: since verbs such as to wonder (59a) and to say (59b) require a [+Question]/[+Declarative] feature in their selected CP, Snegs are ruled out because they do not express such features but, rather, a [+neg] one. This solution is appealing because in my proposal the negative marker non realizes a real instance of negation, introducing a [+neg] feature in the CP. Unfortunately, we have to reject this solution for both theoretical and distributional reasons. In fact, nothing prohibits the CP from realizing both the [+Question]/[+Declarative] and the [+Neg] features. Negative complementizers testify that it is possible (as we saw at the beginning of this chapter). Moreover, Snegs are also ruled out in structures that require a [+neg] feature in the embedded CP, such as the to doubt verb: (60) Dubito [che non sia scesa dal treno M.]?! doubt.1stpres that EN be.3rdsing.subj got off-the train Maria . ‘I doubt that Maria did not get off the train’ (SN) . ‘#I doubt that Maria got off the train’ (#Sneg) We have to reject the selection requirement as the cause of the pattern in (60). Crucially, there is a basic observation about all the predicates seen above, and it can help us: they allow the focalization of an internal TPelement but not of the entire TP. Consider the following dialogues: (61) A: Mi chiedo [CP1se i turisti pensino [CP2 che Cl.to me wonder whether the tourists think.3rdplu.subj that Firenze sia bella]] (IQ) Firenze be.3rdsing.subj beautiful ‘I wonder whether tourists think that Firenze is beautiful’ B: Mi chiedo [CP1 se [FocP [ROMA]i i turisti Cl.to me wonder whether ROMA the tourists pensino [CP2 che ti sia bella]] (non Firenze) think.3rdplu.subj that be.3rdsing.subj beautiful (not Firenze) ‘I wonder whether tourists think that ROMA is beautiful (not) Firenze) 188

See above.

164

Chapter 5

B’: *Mi chiedo [CP1 se [FocP [CHE ROMA SIA Cl.to me wonder whether that Roma be.3rdsing.subj PULITAi] i turisti pensino [CP2 [tTP]] (non che Firenze clean the tourists think.3rdplu.subj not that Firenze sia bella) be.3rdsing.subj beautiful (62) A: Mi spiace [CP1 che i turisti pensino [CP1 che Firenze Cl.to me sorry that the tourists think.3rdplu.subj that Firenze sia brutta]] (FC) be.3rdsing.subj ugly ‘I’m sorry that tourists think Firenze is ugly’ B: Mi spiace [CP1 che [FocP [ROMAi] i turisti pensino Cl.to me sorry that Roma the tourists think.3rdplu.subj [CP2 che ti sia brutta]] (non Firenze) that be.3rdsing.subj ugly not Firenze ‘I’m sorry that tourists think ROMA is ugly (not Firenze)’ B’: *Mi spiace [CP1 che [FocP [CHE ROMA SIA Cl.to me sorry that that Roma be.3rdsing.subj SPORCA] i turisti pensino [CP2 [tTP]] (non che dirty the tourists think.3rdplu.subj (not that Firenze sia brutta) (Firenze be.3rdsing.subj ugly) (63) A: Luca dice [CP1 che i turisti pensano [CP2 che Firenze Luca says that the tourists think.3rdplu.sing that Firenze sia bella]] (BV) be.3rdsing.subj beautiful ‘Luca says that the tourists think Firenze is beautiful’ B: Luca dice [CP1 che [FocP [ROMA]i i turisti pensano Luca says that Roma the tourists think.3rdplu.sing [CP2 che ti sia bella (non Firenze) that be.3rdsing.subj beautiful (not Firenze) ‘Luca says that the tourists think ROMA is beautiful (not Firenze)’ B’: *Luca dice [CP1 che [FocP [CHE ROMA SIA Luca says that that Roma be.3rdsing.subj PULITA]i i turisti pensano [CP2 [tTP]] (non che Firenze clean the tourists think.3rdplu.sing not that Firenze sia bella) be.3rdsing.subj beautiful

Snegs and the Left Periphery

165

All predicates allow the focalization of a TP-internal element (61B-62B63B), but, at the same time, they completely reject the focalization of the entire inflected-TP (61B’-62B’-63B’). To the best of my knowledge, this observation is new in literature and it deserves further insights. However, at the moment, it provides us with a way to take into account Snegs’ incompatibility with subordinate clauses: it is a consequence of an independent restriction, i.e. the fact that the clause does not allow the focalization of the entire inflected TP. Since I have proposed that the focalization of TP is just one of the core properties of Snegs, it follows that Snegs cannot occur in contexts where this operation is ruled out. Once again, my proposal on the syntactic characterization of Snegs allows us to predict the pattern in (59) that, otherwise, would be obscure. There is, at least, another fact confirming my hypothesis on the syntactic structure of Snegs. We already knew (see Ch. 1-2) that Italian displays the sentential subject both in in pre- and post-verbal position. Snegs do not depart from this pattern (I write here the sentences 100 of Ch. 1 as 64): (64) a. E Maria non mi è scesa dal treno?! Maria got off the train!’ b. E non mi è scesa dal treno Maria?! ‘Maria got off the train!’

(Pre-verbal subject) (Post-verbal subject)

Crucially, when the subject is in a pre-verbal position, it has a “topic” nuance disappearing when it occurs in the post verbal position. In fact, its topic interpretation can also be syntactically marked by co-referentiality with a resume pronoun (usually reserved for topicalized structures; cfr. Cecchetto 1999): (65) a. E Maria non mi è scesa dal treno, lei?! EE Maria EN ED is got off-the train she ‘Maria got off the train!’ We saw above that Snegs can be a coherent answer to prepositional questions, such as the one in (66A), because the whole TP is focalized. Crucially, this is true only when the subject occurs in the post-verbal position (like in 66B); when it occurs in the preverbal position, the sentence yields ungrammaticality (66B’). (66) A: Sembri sconvolto. Cos’è successo? ‘You are shocked, what’s happened?’

166

Chapter 5

B: Non ti ho incontrato in stazione Maria?! EN ED have.1stsing met in station Maria ‘I met Maria at the train station’ B’: *Maria non ti ho incontrato in stazione (lei)?! Maria EN ED have.1stsing met in station (her) I assume from now on that the preverbal subject lies in [Spec, TopP] instead of in the [Spec, TP] like it usually does. From there, it receives the topicalized interpretation and, as a consequence, it cannot occur in context only requiring new information, as in (66)189. Again, this is predicted by my hypothesis. The real negative status of the non in Snegs imposes that negation is phonologically close to the verb, excluding the interposition of the subject (see above). As a consequence, when the subject is pre-verbal, it can only precede the head non and the FocP in the sequence of the CPfunctional projections occupying the higher available position, i.e. [Spec, TopP]: (67) [CP …[ToP subjectk [X° non ] … [TPi Foc° [TP tk ]i] The subject in (67) has been internally merged in the [Spec, TopP] before TP is moved in [Spec, FocP]. This correctly predicts both the word order subject-non-verb and the topic interpretation associated with the pre-verbal subject in Snegs. As last consequence of my hypothesis, let us consider the relation between Snegs and the expletive conjunction E (and). I have diffusely associated the EE with the Sneg sentences, but I have said nothing about its nature (except its vacuous semantics). This is the reason why I am going to start with the analysis proposed in Poletto (2005). She discusses the Old Italian form of EE. Consider, for example, the following text: (68) e quando avea forbiti i piedi ed elli tornava fuori and when had cleaned the feet and he came back outside e rinfangavalisi vie piu` e tornava a ricalpitare and got mudded more and more and came back to step on il letto. E partisi e disse a Platone: . . . the bed and went away and said to Plato 189

For a deep discussion on the criterial interpretation of the preverbal Italian subject, see Bianchi and Chesi (2014).

Snegs and the Left Periphery

167

‘and when he had cleaned his feet went back outside, put mud on them, came back inside and went up onto the bed. He left and told Plato’ (FF, 124 in Poletto 2005, 227) The vacuous semantics of and is highlighted by its occurrence after the full stop, even though it is generally forbidden. Moreover, and introduces every clause in (68), even though it usually appears only before the last member of a list. For these reasons, and since it seems that the conjunction does not connect anything, it has been called Expletive E (EE). EE must occur at the beginning of the sentence that it introduces regardless of which functional projection in the CP-field is activated. This is visible with Fin° in absolute small clauses (69a), with focalized elements in Anaphoric Anteposition structures (70), with topicalized elements in CLLD structures (71) and with the Force° in relative clauses (72)190: (69) a. E [Fin° con] Gianni malato, è partita Maria! EE with Gianni sick is partita Maria ‘Maria has left even though Gianni was sick’ a’ *Con e Gianni malato, è partita Maria! with EE Gianni sick is partita Maria (70) A: Gianni voleva comprarsi un castello (Poletto 2005) ‘Gianni wanted to buy a castle’ B: E [FocP un castello [foc°] si è comprato] EE a casttle cl.to him self is bought ‘And a castle he bought’ B’: *Un castello e si è comprato a castle EE cl is bought (71) a. E [TopP i libri [Top°] me li porta Gianni]! EE the books Cl.to me cl.them brings Gianni ‘The books, Gianni brings them to me!’ b. *I libri e me li porta Gianni the books EE Cl.to me cl.them brings Gianni (72) a. E [Force° coloro] i quali sono arrivati ieri partiranno EE those the who are arrived yesterday leave.3rdplu.fut domani! tomorrow ‘Those who arrived yesterday will leave tomorrow!’ 190

I took the following sentences from Moro (2003).

Chapter 5

168

b. *Coloro e i quali sono arrivati ieri partiranno those EE the who are arrived yesterday leave.3rdplu.fut domani tomorrow Data in (69-72) shows that EE always occurs at the beginning of the sentence. We can wonder which elements it connects since the one on its left, (at least), is missing. In fact, according to Kayne (1994), a conjunction realizes the head of a conjunction phrase (&P), always selecting two elements, one as its specifier and one as its argument191: (73) [&P XP [&° e [XP]]] The fact that the missing element in (69a-70a-71a-72a) is the left one is coherent with the observation in Ross (1967): a conjunction forms a unit with the second conjunct but not with the first one 192 (that may be phonologically unrealized). As Kayne (1994) and Progovac (1998, 2003) highlight, in languages such as Italian, Ross’s observation suggests that e can only precede its complement and cannot follow it193: (74) Sono arrivati Anna [e Laura] (*e) are arrived Anna and Laura and ‘Anna, Roberto and Laura have arrived’ This is exactly what happens in (69-72), where the first conjunct is missing but not the second one. However, we do not know the identity of the elements that the conjunct coordinates since e can select different types of phrases. We only know that both the conjuncts must belong to the same 191

See also Munn (1993) and Johannessen (1998) for a general discussion on the syntactic status of the conjunction. 192 Ross (1967)’s analysis is based on the following contrasts: (iv) John left, and he didn’t even say good-bye. (v) John left. And he didn’t even say good-bye. (vi) *John left and. He didn’t even say good-bye. 193

The head status of the conjunction is confirmed by the fact that languages realizing the head on the right show the opposite order: complement> conjunction. This is exactly what happen with Japanese: (vii) [Robin-to Kim-to Terry-to]-ga [Robin-and Kim-and Terry-and-CASE]

(Progovac1998)

Snegs and the Left Periphery

169

syntactic category by the “law of coordination of likes” (Chomsky 1957; Williams 1978). Compare, for example, the following sentences: (75) a. Luca ha mangiato [DP una mela] e [DP una pera] Luca has eaten an apple and a pear ‘Luca has eaten an appeal and a pear’ b. Luca ha mangiato [AdvP voracemente] e [AdvP abbondantemente] Luca has eaten voraciously and abundantly ‘Luca has eaten voraciously and abundantly’ c. *Luca ha mangiato [DP una mela] e [AdvP abbondantemente] Luca has eaten an apple and abundantly In (75a-b) two elements belonging to the same syntactic category are coordinated: two DPs in (75a) and two AdvPs in (75b). The ungrammaticality in (75c) follows from the violation of the law of coordination of likes: the coordinated elements belong to two different syntactic categories. Unfortunately, according to Johannessen (1998) and Chomsky (1995), a &P takes its syntactic category features from its Specifier (via Spec-head Agreement) and this is exactly the element missing in (69-72). At this point, the question on which is the element filling the [Spec, &P] becomes even more important in order to understand the reason why EE is associated to Snegs. A possible clue is found in Poletto (2005): EE can only occur in front of different structures sharing the activation of the CP-layer. Consider, for example, canonical questions, rhetorical questions, non-echo wh–in situ questions, exclamative clauses, and Anaphoric Anteposition structures194: (76) a. E adesso? (Regular Question) ‘Now what?’ b. E cosa potrebbe fare in un frangente simile? (Rhetoric. questions) ‘And what could he do in such a case?’ c. E viene quando, allora? (Non-echo wh–in situ) ‘And come when, then?’ d. E che vestito che ti sei comprato! (Exclamative) ‘What a dress you bought!’ e. A: Gianni voleva comprarsi un castello (Anaphoric Anteposition) ‘Gianni wanted to buy a castle’ B: E un castello si è comprato] ‘And a castle he bought’ 194

All the following sentences are taken from Poletto (2005): 229-230.

170

Chapter 5

Crucially, all sentences in (76) share the activation of a specific functional head among the ones forming the CP-field, i.e. the FocP. According to Poletto (2005), EE in Modern Italian is, in fact, a Topic marker that is “parasitic an already existing CP layer” and, therefore, it precedes it. She proposes the following representation: (77) [TopicP (Null) TopP [Topic0 e [TopicP [CP]]] Poletto discusses in depth the reasons why the TopP is the element coordinated by e, but such a discussion is beyond the aim of this book. The crucial point is that EE requires an activation of the FocP in order to be elicited. Quoting her words, “It would be tempting to assume that e can only represent the continuation of a Topic if there is a Focus layer active” (p. 230). Actually, this is exactly what I have proposed for Snegs. EE selects the Snegs reading and rejects the standard one, because only in Sneg the FocP is involved: (78) [Top0 E [non [focP [ti incontro Maria che tornava a casa]TP [Foc°] … [tTP]]]]]?! Poletto’s analysis perfectly matches mine, and it sheds light on the close relation between EE and Snegs. Again, my proposal on the syntactic structure of Snegs predicts the fact that an element focus-sensitive can occur and interact with Snegs. To sum up, we have seen that assuming my proposal in (31) allows us to derive a series of different facts that otherwise would remain mysterious. We saw that the root status of Snegs lies in the fact that the matrix predicate does not accept the whole-TP focalization in their subordinate clause. Since Snegs display exactly such a configuration, it follows that they cannot be embedded. We have also seen that the topicalized flavor of the pre-verbal subject is a consequence of the real status of negation which, given that it is merged into CP and has to be close to the predicate, only allows the subject to stay in a TopP position. Finally, we have found in the EE analysis a further matter reinforcing my proposal: since EE requires a FocP activation, it can occur in Snegs and it selects the Snegs reading when it can be optionally present. Before deepening the experimental part of this book, I want to emphasize that what I have proposed for Snegs is not extravagant; it captures aspects already presented in some other Italian structures. I also want to add some

Snegs and the Left Periphery

171

comparative considerations that can confirm my hypothesis, albeit in an analogical way.

5.5 A comparative support in favor of the special status of expletive negation First of all, I want to point out that Italian displays, at least, another case of a high negation in the structure, i.e. the non che clause (Moscati 2010). Consider, for example, the following sentence: (79) Non che Gianni sia molto intelligente neg that Gianni be.3rdsing.subj very smart ‘Is not the case that Gianni is very smart’ The negative marker non precedes the complementizer che in the word order. Assuming that che realizes the head of the ForceP (cfr. Rizzi 1997), Moscati (2010) proposes that non therefore lies in the [Spec, ForceP]: (80) [ForceP Non [Force° che] Gianni sia molto intelligente We know that ForceP is the leftmost functional phrase in the CP-field and this means that the negative marker occurs very high in the structure. I do not face here the syntactic problems that such a proposal raises and I only refer to the distributional point of view: a negative marker can occur inside the CP-field. This suggestion matches my analysis of Snegs even though, in my case, there is not che (or similar elements) making the phenomenon visible. The second point that I want to emphasize is that Italian displays other structures with an element externally merged in the CP-field, i.e. the whyquestion (Rizzi 2001). According to Rizzi (2001), some Italian whquestions surprisingly do not require the inversion of the verb and the subject, as they usually do. Compare, for example, the following sentences: (81) a. Come ha fatto Gianni? how has done Gianni ‘How did Gianni do?’ b. *Come Gianni ha fatto? how Gianni has done

(Wh–Verb–Subject) (*Wh–Subject–Verb)

172

Chapter 5

(82) a. Perché è tornato a casa Gianni? (Why–Verb–Subject) why is come back to home Gianni ‘Why did Gianni come back home?’ b. Perché Gianni è tornato a casa? (Why–Subject–Verb) why Gianni is came back to home ‘Why did Gianni come back home?’ According to Rizzi, in (81b) the order Wh-element–subject–verb is ruled out because the wh-element and the verb movein [Spec, FocP] and [Foc°] respectively, in order to satisfy the Wh-criterion by Spec-head agreement. In such a configuration the subject cannot stay between them and, therefore, the sentence in (81b) is ungrammatical. In (82), such a requirement is not present and the order Wh-element–subject–verb is accepted. Moreover, even though we already know that wh-elements cannot co-occur with focalized phrases because both compete for the same [Spec, FocP] position, this observation does not apply to adverbs like why: (83) a *Come QUESTO avremmo dovuto dirgli, non how this have.1stplu.cond had sayd.Cl.to him not qualcos’altro? something other b. Perché QUESTO avremmo dovuto dirgli, non qualcos’altro? ‘Why THIS we should have said to him, not something else?’ In order to take into account the pattern in (81-82), Rizzi proposes that adverbs such as why are externally merged in the specifier position of an Interrogative Head intrinsically endowed with a Wh-feature. This configuration rules out the movement of the verb in C and, therefore, the inversion with the subject. At the same time, this also explains why the whyclause can host focalized elements: why and focalized elements do not compete for the same space (84). (84) FORCE° (TOP*) [IntP perché [INT° (TOP*) FOC° (TOP*) FIN° …] The crucial point for our discussion is that why is externally merged in the CP-field. This is exactly what I have proposed for the negative marker non in Snegs. Interestingly, the why-structures show the raising of TP to the CP domain in some Italian varieties spoken in the region of Marche. Consider, for example, the following couple of sentences:

Snegs and the Left Periphery

173

(85) a. [CP [IntP Perché [TP sei andato a casa]]]? why is.2ndsing gone to home b. [CP [Sei andato a casa]i [IntP perché [Int° …[TP ti]]]]? is.2ndsing gone to home why ‘Why did you go to home?’ In (85a), the adverb occurs in the fronted position preceding TP; in (85b), the order is the opposite. Moreover, the pragmatic interpretation of (85b) is different to the one in (85a): in the former, the information that the hearer went home is already present in the context of speech; in the latter it is not. Both the inverted order of the elements and the interpretation are taken into account if we assume that in (85b) the whole TP is internally merged in the topic position dominating the IntP (see 84). It follows that I can write the previous sentences in the following way: (86) a. [CP [IntP Perché [TP sei andato a casa]]]? why is.2ndsing gone to home b. [CP [TopP [Sei andato a casa]i [Top° [IntP perché [Int° [TP ti]]]]]? is.2ndsing gone to home why ‘Why did you go to home?’ The sentence in (86b) takes its topic flavor by occupying a Topic-position in the CP-field. Since this position is higher than the IntP (see 84), the final phrasal order is TP-why. I have proposed that a similar movement occurs in Snegs, with the crucial difference that the landing site for TP is the FocP; this leaves the non-TP order intact. In summary, why-clauses show at least two elements that affect Snegs as well: the external merge of an element in the CP-field, and the raising of TP. Interestingly, Italian displays at least one additional case in which an element marks the movement of TP in the CP-field: tipo-clauses 195 . Consider, for example, the following dialogue: (87) A: Ho speso tantissimo in spiaggia ‘I spent a lot of money at the beach’ B: Tipo [CP che ti hanno fatto pagare anche l’acqua kind that Cl.to you have.3rdplu make pay also the water del rubinetto]? of-the tap ‘Did they get you to pay also for the tap water?’ 195

I want to thank Andrea Moro for this central observation.

174

Chapter 5

The Italian noun kind selects as its argument a yes-no question introduced by the complementizer that. To the best of my knowledge there is not a syntactic analysis of this sentence in literature. My interest lies in one of their properties, i.e. they can display a movement of TP in a structurally higher position: (88) [CP [TPTi hanno fatto pagare anche l’acqua del Cl.to you have.3rdplu make pay also the water of the rubinetto]i tipo [ ti ]? tap kind ‘Did they get you pay also for the tap water?’ In (88), the movement of TP is evident because it precedes the noun kind instead of following it (like in 87B). I cannot deepen here the analysis of the syntactic structure of kind -clauses because it is beyond the aim of this book. I will assume that the movement of TP is towards the CP-field, without defining the correct position in which it lands. In fact, the central point is that the whole TP moves in (88) and this is similar to what I have proposed for Snegs. Remarkably, in such a clause TP-movement is marked by the occurrence of tipo at the end of the sentence. Both the pattern in why-clauses and in kind-clauses shows that my hypothesis on Snegs is not completely new within the panorama of Italian syntax. This is not a direct confirmation that I am right in my hypothesis, but, at least, it suggests that I am, plausibly, on the right track. Before concluding this chapter, I want to discuss some comparative data that can support my analysis. As I discussed in Chapter 1, it has been proposed that languages with two negative elements in the head and the maximal projection status only display the former in negative expletive structures. Consider, for example, modern French (Muller 1978; Makri 2013) and the Old English (Wim van der Wurff 1999)196: (89) a. Je ne nie pas que je Ș'aie ètè bien reçuů (Muller 1978) I neg deny neg that I neg have been well received ‘I do not deny that I was received well b. I drede not pat ne pe curs of God [...] wolde brynge me into a I doubt not that Neg the curse of God would bring me into a ful yitel eende if I contynuedepus very evil end if I continued thus ů 196

I repeat here as (89b) the sentence (53) in Ch.1.

Snegs and the Left Periphery

175

‘I do not doubt that God's curse would bring me to a very evil end if I continued like this’ (van der Wurff 1999) In (89a), the matrix clause hosts a standard negation displaying both the head ne and the specifier pas (see Chapter 1 for the syntactic characterization of French negation). Crucially, the subordinate clause only shows ne, the head of NegP, which realizes the expletive reading (as the English translations suggests). Similarly, the Old English sentence in (89b) displays two different negations with two different statuses (cfr. Wim van der Wurff 1999): the negative adverb not realizes the [Spec, NegP]; the negative marker ne realizes the Neg°. Again, only ne can be associated with the EN reading, as the subordinate clause shows. The fact that only elements with a head status can occur in EN structures is coherent with what I have proposed for Snegs: i.e., that the negative marker non is a head. Moreover, it also suggests that languages without negative elements with a head status cannot host Snegs. However, I require further research in order to verify this prediction. Finally, Latin displays some structures that are interesting for comparison with Snegs: Nonne and Numne. They are both rhetorical and interrogative elements that elicit an answer with a specific polarity (Cfr. Ernout and Thomas 1953): nonne requires an affirmative answer; numne, a negative one. Consider, for example, the following sentences: (90) a. Nonne concedas interdum ut excusatione summae stultitiae summae probitatis odium deprecetur? (Pro Caecina, Cicero) ‘Can you not sometimes permit one to avert the odium of the greatest wickedness by the excuse of the most abysmal folly?’ b. Numne, si Coriolanus habuit amicos, ferre contra patriam arma illi cum Coriolano debuerunt? (De Amicitia, Cicero) ‘If Coriolanus had friends, did they perhaps have to start a war against the homeland together with Coriolanus?’ Nonne and Numne are compounds formed either by the negative marker non or by the interrogative element num, plus the clitic –ne. I have already discussed the syntactic status of clitics (cfr. Ch. 1) and, therefore, I assume that –ne is a head, as the other interrogative elements nonne and numne lie in the CP-field. Interestingly, the clitic –ne can also be combined with a topicalized pronoun (cfr. Makri 2013) or with a V° displaying an opposite word order

176

Chapter 5

to that with nonne. Compare the following sentences (cfr. Ernout and Thomas 1953): (91) a. Non-ne Romam vidisti? Neg-ne Roma saw.2nd.sing ‘Didn’t you see Roma?’ b. Ego-ne tu interpellem? I-NEG you interrupt ‘What I interrupt you?’ c. Vidistii-ne ti Romam? saw.2ndsing-neg Roma ‘Did you see Roma?’

(Makri 2013: 80)

The order object-verb in (91a-b) is basic in Latin. In (91c), it is inverted, allowing the verb to precede the object. Crucially, in all cases, the clitic –ne shows up, arguably the head of a high CP-phrasal combined with the negative morpheme non or with the V0. I speculate that this head is the Foc° and that the non in Nonne is the equivalent of the non in Snegs. The movement of the verb is analogous to the movement of TP. Notably, ne is also a subordinate conjunction in negative final clauses in Latin (as we saw in Ch.1). To sum up in this paragraph, we have seen some comparative data arguably confirming my analysis of Snegs. I have discussed cases in which the negative marker non occurs high in the CP-field (non-che structures); we also saw that some Italian elements are externally merged in the CP-field, such as, for example, the adverb why in interrogative clauses (this is exactly what I propose for negation in Snegs). The why-structure displays a movement of TP in CP as well. I have pointed out the same for the kindclauses. In both why and kind cases the movement of TP is marked by the occurrence of those elements at the end of the sentence, whereas they usually occur at the beginning of it. Coming to some different languages, I have highlighted that the head status of a negative element is crucial in French and in Old English in order to obtain the expletive reading in a negative sentence. Finally, we saw that Latin displays a clitic –ne in the CPfield, and I have hypothesized that it is the realization of the Foc°. It also interacts with the negative marker non, reproducing the same neg-foc relation presented in Snegs.

Snegs and the Left Periphery

177

Crucially, in this chapter I have faced the core of the book, i.e. the syntactic analysis of Snegs. I have begun proposing an apparent simple schema like the following one: (92) [CP … [X° non ] … [TP Foc° [… tTP …]] I have assumed that the negative marker non is externally merged in the CPfield when the vP-phase is already closed. Non is the realization of Italian standard negation and it takes a FocP as its argument. Moreover, I have hypothesized that the FocP is compulsorily filled by TP and that it internally merges in its Spec position because of a [+Foc] feature. From (92), I have derived the main part of the properties concerning Snegs (I report here the list that I presented at the beginning of this chapter, adding the new elements discussed above): i) ii) iii) iv) v) vi) vii) viii) ix) x) xi) xii) xiii)

Snegs can host any kind of predicate regardless of their valency and their diathesis; They allow subjects in pre-verbal and post-verbal positions (even though it carries a topicalized flavor when it occurs in the preverbal position) both in nominative and dative cases; They are instances of strong expletive negations (affirmative in meaning, they allow p-words but not strong/weak NPIs, n-words and not-also conjunctions); They do not allow al-words and elements requiring downwardcontexts; They allow PPIs. They are only realized by the negative marker non, rejecting other denying elements like n-words in a preverbal position; Pragmatically they carry new information with an unexpected value; They can be coherent answers to propositional questions; They have an intonational pattern melting together interrogative and exclamative prosody; Their interpretation is strongly related to the occurrence of the Ethical Dative and of the expletive E; They interact with the left periphery allowing topicalized structures and rejecting focalized ones (both in fronted and in situ positions). They cannot co-occur with Wh-elements They cannot be identified with negative exclamatives, rhetorical questions and bi-clausal structures even though they share some features with them.

178

Chapter 5

xiv) They cannot be embedded under any kind of predicate. xv) They do not allow NEG-raising xvi) They do not interact with high adverbs I dedicated to each point a deep discussion and I have demonstrated that the schema in (31) (repeated here as 92) can predict them197. I also added some comparative considerations matching my hypothesis. As always, my proposal may not be true, but, in this case, a potential critic has the burden of proof to show that I’m wrong and, at the same time, has to propose an alternative hypothesis with the same predictive power. Finally, I am going now to show that some confirmations also come from the experimental part of my work. Let me open this discussion.

197

It is possible to extend my proposal to the other strong-EN structures, since they share several properties with Snegs. However, I will not face here this discussion because it is beyond the aim of this book. I will leave it for future works.

CHAPTER 6 AN EXPLORATORY PSYCHOLINGUISTIC APPROACH TO SNEGS

6.1 Introduction As discussed in Chapter 1, negation is one of the main elements characterizing the human language. From the semantic point of view, negation is a one-place operator reversing the truth-value conditions of the sentence in which it occurs (Bernini and Ramat 1996; Horn and Kato 2000; Speranza and Horn 2012). A sentence such as “Luca is not smart” is true if, and only if, the correspondent affirmative sentence “Luca is smart” is false, and vice versa. However, this logical mechanism underlying negation is not always employed. Consider, for example, the sentence “He is richer nor you’ll ever be” (Yoon 2012). In this case, negation does not reverse the truth-value conditions of the sentence, giving an example of what it is called expletive negation (cfr. Jespersen 1917; Horn 1989). In the present study I will face two essential questions raised by expletive negation: (i) How is negation processed when it does not deny a sentence, as it is supposed to do? (ii) Are expletive negative sentences elaborated either as affirmative clauses, according their semantic value, or as negative clauses, according to their morphological shape? Since different instances of expletive negative structures display important differences not allowing them to be treated as a unique phenomenon (cfr. Ch. 1), I will focus only on the Sneg case for this pioneering eye-movement study based on the visual-world paradigm. Processing of negative sentences Although there are no experimental studies in the psycho-linguistic literature on expletive negation, the works dedicated to standard negation can be useful in modeling the empirical approach to the Sneg case. One crucial piece of evidence on the elaboration of negative sentences is that they yield a longer processing time and higher error rates than non-

180

Chapter 6

negative ones, regardless of the experimental task involved (Wason 1961; Clark and Chase 1972; Carpenter and Just 1975; Trabasso, Rollins and Shaughnessy 1971; Carpenter et al. 1999; Hasegawa, Carpenter and Just 2002; Kaup, Lüdtke and Zwaan 2007). Moreover, it has also been shown that negation is incorporated into sentential interpretation processes only in a late stage, after some thousand of milliseconds from the end of the sentence in which it occurs (Hasson and Glucksberg 2006; Giora 2006; Kaup, Lüdtke and Zwaan 2005, 2006; Lüdtke, et al. 2008; Scappini et al. 2015). For example, Kaup, Lüdtke and Zwaan (2005) showed that negative sentences require two mental representations sequentially elaborated in order to be processed and that negation only affects the second one. They developed a sentence-picture verification task in which participants were presented with affirmative or negative sentences (e.g., The elephant is / is not above the giraffe) and with two cards either matching the current state of affairs (an elephant above a giraffe, in the affirmative case, or an elephant below a giraffe, in the negative case) or not (an elephant below a giraffe in the affirmative case, or an elephant above a giraffe in the negative case). With a variable latency of 750 msec and 1500 msec introduced between the sentences and the cards, they asked participants to evaluate whether the objects depicted in the cards had been mentioned in the sentence or not. Results show that after 750 msec, the affirmative sentences require shorter reaction time in the matching scenario than in the mismatching one, while negative sentences displayed the contrary pattern, pointing out that negation has not been elaborated yet at that early latency; after 1500 msec, participants valued both the affirmative and negative sentences with shorter reaction time in the matching scenario than in the mismatching one. According to the authors and subsequent works (Kaup, Lüdtke and Zwaan 2006, 2007), negative sentences require the elaboration of the contrafactual state of affairs (an elephant above the giraffe) and then the elaboration of the effective negative meaning (an elephant below the giraffe). Clearly, these two representations are not required by affirmative sentences, which just need one mental representation equivalent to their meaning. The longer processing time, the higher error rates, and the late integration have been taken in the literature as the consequences of two conflicting principles: negative sentences require higher processing costs compared to non-negative sentences (Carpenter and Just 1975; Carpenter et al. 1999; Hasegawa, Carpenter and Just 2002; Dale and Duran 2011); linguistic negation reduces the accessibility of the information taken under its scope (MacDonald and Just 1989; Kaup 2001; Tettamanti et al. 2008;

An exploratory psycholinguistic approach to Snegs

181

Liuzza, Candidi and Aglioti 2011; Bartoli et al. 2013). However, some intermediate positions have been advanced, proposing that specific linguistic and discourse contexts can reduce both the inhibitory power of negation (Kaup 1997; Giora et al. 2007) and its higher processing costs (Wason 1965; Villiers and Flusberg 1975, Glenberg et al. 1999; Lüdtke J. and Kaup B. 2006; Ferguson, Sanford and Leuthold 2008). Among others, Orenes, Beltrán and Santamaría (2014) show that an alternative term in the speech context facilitates the inhibitory power of negation. In a visual-world paradigm, negative and affirmative sentences (e.g., The figure is / is not red) were presented after a verbal context that was either binary (e.g., The figure could be red or green) or multiple (e.g., The figure could be red or green or yellow or blue), while four pictures with different colors were displayed on a screen. For the negative sentences ,eye-movement records showed that when the context was binary, participants shifted their attention toward the alternative color (e.g., green) with a delay of 1340 msec after the end of the sentence (confirming the late integration of negation), but when the context was multiple, they focused on the negated argument (e.g., red) with a delay of 400 msec after the end of the sentence, exactly as the affirmative target sentences do. In summary, the linguistic context in which negative sentences are evaluated affects the specific parameters usually associated with their elaborations (longer processing time, higher error rates and late integration) eventually making the processing of negative sentences similar to that of affirmative sentences. Based on the literature on standard negation, we can derive that the evaluation of Snegs will require a correct linguistic context in which the differences in the processing of negative and affirmative sentences will be clear (for example, a binary linguistic context in which the negated term cooccurs with an alternative one). Such a linguistic context will clarify whether expletive negative sentences are elaborated as affirmative clauses, according to their semantic value, or as negative clauses, according to their morphological shape.

6.2 The present study The goal of this study is to investigate how Snegs are interpreted. The prediction is that Snegs are elaborated as affirmative clauses, despite the occurrence of the negative marker non, since negation does not syntactically affect their predicative core (cfr. Ch. 5). Consequently, they would follow

182

Chapter 6

the same pattern in probability of fixation as affirmative sentences, differing from negative sentences. Based on Orenes, Beltrán and Santamaría (2014) and Orenes et al. (2016), I expect that in binary contexts (i) individuals will show an increased rate of fixation on the mentioned object in both affirmative and Sneg sentences but not in negative ones; (ii) individuals will show an increased rate of fixation on the alternative (not-mentioned) object after some thousands of milliseconds only in negative sentences due to the integration of negation. Moreover, based on studies in experimental pragmatics (Schumacher 2012; Masia et al. 2017), I expect that (iii) Snegs will induce some specific behavior due to their own properties in terms, for example, of higher processing costs given by the pragmatic integration of the surprising information in the discourse context. The theoretical assumption supporting this paradigm is that the comprehension of a sentence is mediated by its conceptual representation (Salverda, Brown, and Tanenhaus 2011), with the eyes moving toward the object denoted by a word when linguistic inputs match visual inputs (cfr. Cooper, 1974; Tanenhaus, Spivey-Knowlton, Eberhard, and Sedivy 1995, Allopenna, Magnuson, and Tanenhaus, 1998; Altmann and Kamide 2009; Kronmüller et al. 2017). According to Orenes, Beltrán and Santamaría (2014), this paradigm has the important advantage of leaving the subject free to run his/her attention to “whatever object is required for processing and representation”, since the pictures and words are presented at the same time (contrary to some other studies in which a forced delay was inserted between the visual and linguistic inputs; cfr. Hasson and Glucksberg 2006; Kaup et al. 2006; Lüdtke, et al. 2008). As a consequence of the conceptual mediation between the visual and the linguistic inputs, many works (Kamide, Altmann and Haywood 2003; Altmann 2004; Altmann and Kamide 2007) pointed out that the eyes move toward an object also before the subject hears the corresponding word (the anticipatory effect). Altmann and Kamide (1999) show that when a subject hears a sentence such as “The boy will eat the cake” presented with a screen with a cake and other inedible objects (toy train, toy car, etc.), the eyes begin to move toward the cake as soon as the verb eat appears. This anticipatory effect is the mark of the interaction between unfolding linguistic inputs and the visual inspection of a context. Unlike many other visual-world experiments (Altmann and Kamide 1999; Kamide, Altmann and Haywood 2003; Altmann 2004; Altmann and Kamide 2007; Orenes, Beltrán and Santamaría 2014) using stylized, drawing or white-black images, I used colored photos representing real-life

An exploratory psycholinguistic approach to Snegs

183

objects and animals in order to increase the ecological validity of the visual stimuli (Viggiano, Vannucci and Righi 2004) and their perceptual saliency (Brodie, Wallace and Sharrat 1991; Ostergaard and Davidoff 1985; Brodeur et al. 2010; Brodeur, Guérard and Bouras 2014). Even though the familiarity and the visual complexity of pictures are cross-linguistically stable variables (Cfr. Pompeia, Miranda and Bueno 2003), the name agreement is not (Sanfeliu and Fernandez 1996; Alario and Ferrand 1999; Pompeia, Miranda and Bueno 2003). This is the reason why all the photos used for the experiment needed to be evaluated by Italian speakers, even though most of them were extracted from rated corpora, such as BMB-corpus (Bates at al. 2000), COGPSY-corpus (Viggiano, Vannucci and Righi 2004) and BOSScorpus (Brodeur et al. 2010; Brodeur, Guérard and Bouras 2014). I also integrated the corpora with new non copyrighted photos taken from the Internet. All the pictures were rated in a naming and visual complexity task, separately conducted on subjects not participating in the main experiment. As we saw in Ch. 2, Snegs are limited to a specific context in which a speaker is struck by an unexpected fact. They have to be considered an oral phenomenon specifically affecting limited real-life situations. I tie the notion of expectedness to the more common notion of typicality (Rosch and Mervis 1975; Ashcraft 1978; Holmes and Ellis 2006; Maxfield, Stalder and Zelinsky 2014): the more typical an event is for a given situation or for a given semantic category, the more expected its realization is, and vice versa (Mccloskey 1980; Hampton and Gardiner 1983; Pèrez et al. 2015). For example, if a soup is considered a more atypical snack than an apple, then the sentence “The boy eats soup” will be considered more unexpected than the sentence “The boy eats an apple” in a context in which the hearer knows that that boy is eating a snack. This is the reason why, in this visual-world experiment, all sentences display atypical elements giving an unexpected value to the sentence, as Snegs require. The same subjects who rate the photos also judge the stimuli for typicality. For the visual-world paradigm, I presented short real-life stories (e.g., Laura invited some friends to her home. When they arrived, she showed them her domestic animal) introducing a specific semantic category (e.g., domestic animal) followed by a target sentence that was either affirmative (e.g, La ragazza ha mostrato un serpente translated as ‘the girl showed a snake’), negative (e.g., La ragazza non ha mostrato un serpente translated as ‘The girl did not show a snake’) or a Sneg (e.g., La ragazza non ha mostrato un serpente?! translated as ‘The girl showed a snake!’), presented via headphones while four pictures (e.g., snake, dog, backpack, air

Chapter 6

184

conditioning) were shown on the screen. Two of the objects presented on the screen belonged to the semantic category introduced by the story (e.g.,. the snake and the dog as domestic animal exemplars) and two did not (e.g., the backpack and the air-conditioning), giving a linguistic binary context in which the target term (e.g., the snake) co-occurs with an alternative one (e.g., the dog) belonging to the same semantic category. Such a context was required to distinguish the processing of affirmative and negative sentences (Orenes, Beltrán and Santamaría 2014). One of the two objects related to the story was a highly typical element for the category under discussion (e.g., the dog) and, therefore, highly expected; the other object related to the story was a highly atypical element (e.g., the snake) and, therefore, highly unexpected. Such a configuration allowed me to create a natural context in which a subject could correctly evaluate Snegs.

6.3 Methods Participants: Thirty-four native Italian speakers with a university education participated in the eye-movement experiment (thirteen males and twenty-one females; 24.43 mean age) without any payment. All of them had uncorrected vision or wore soft contact lenses. The Ethics Committee of the “Dipartimento di Scienze del Sistema Nervoso e del Comportamento” of the University of Pavia approved this study. Materials: Sixty experimental trials were used in the experiment. Each trial consisted of an introduction in which a real-life scenario was presented. The introduction could be followed by three types of target sentences representing the three experimental conditions: affirmative, negative and expletive negative sentences. Each trial was associated with a set of four pictures (two representing the plausible objects, either Expected or Unexpected, and two pictures were used as distractors). The introduction of the real-life story was composed by two parts roughly corresponding to two complex sentences: Part A: Lucia ha invitato delle amiche a giocare a casa sua (translated as Laura invited some friends to her home) Part B: Quando sono arrivate ha mostrato loro il suo animale domestico (translated as When they arrived, she showed her domestic animal to them).

An exploratory psycholinguistic approach to Snegs

185

Part A introduced the agent of the action and the first piece of the context. Part B introduced the most salient part of the context; it displayed the same past tense verb displayed by the target sentence (e.g., ha mostrato translated as showed) and the direct object corresponding to a specific semantic category (e.g. domestic animal). Following the introduction, each story ended with the target sentence that could be affirmative (e.g., La ragazza ha mostrato un serpente translated as ‘the girl showed a snake’), negative (e.g., La ragazza non ha mostrato un serpente translated as ‘The girl did not show a snake’) or expletive negative (e.g., La ragazza non ha mostrato un serpente?! translated as ‘The girl showed a snake!’). Each target sentence was built using a strict order of words: the subject introduced by an article (Il ragazzo, ‘the boy’, if context A displayed a male subject, or La ragazza, ‘the girl’, if it displayed a female one); the past tense verb resumed from the Part B of the story (eg. ha mostrato, translated as has showed); an article or a preposition introducing the direct object198; the direct object corresponding to an entity highly atypical for the semantic category introduced by context B (ex. ‘snake’). Such strict order was necessary to compare the different conditions. The negative and the expletive negative sentences displayed one extra word, the negative marker non, just before the verb. To compare all sentences regardless of the presence of the extra element non, I will focus my analysis on the eye movements after the verb region, where all the items had the same sentence structure. Each trial was associated with a set of 4 colored pictures (size = 450x450 pixels; resolution = 72 pixel/inch) which appeared on the screen in the four quadrants (Upper-Left, Upper-Right, Bottom-Left, Bottom-Right) (Fig. 1). Two of the pictures belonged to the semantic category introduced by the story (Related pictures; e.g., the snake and the dog) and two did not (Unrelated pictures; e.g., the backpack and the air-conditioning). One of the Related pictures represented the highly expected element of the category mentioned in the trial (eg. the dog as domestic animal); the other related picture represented the highly Unexpected element in the target sentence (eg. the snake). Such a configuration provides a binary visual context in which the participants could naturally interpret both affirmative and negative sentences. The Unrelated pictures were used as distractors, and were not exemplars of the semantic category introduced in the scenario. The pictures were displayed on the screen with a counterbalanced quadrant 198

In order to avoid any phonetic clue, both the unexpected and the expect objects associated with each story were introduced by the same article or preposition.

186

Chapter 6

rotation following these criteria: (i) across participants, the same photo occupied different positions with the same probability; (ii) within participants’ individual results, Related and Unrelated pictures occurred the same number of times in horizontal, vertical and diagonal arrangements independently of their position on the screen.

FIG 6-1: EXAMPLE OF SCENE ACCOMPANYING THE SENTENTIAL STIMULUS “THE GIRL SHOWED A SNAKE’.

A further 40 stories were used as fillers. These were created similarly to the experimental items, with the notable difference that the target sentences always referred to the Expected object. For example, after a short story (e.g., Oggi è il compleanno di Maria; quando è tornata a casa da scuola ha trovato un regalo per lei, translated as Today it is Maria’s birthday; she found a gift when she went back home after school) the participants could either listen to an affirmative sentence (e.g., La ragazza ha trovato un libro, translated as The girl found a book) or a negative one (La ragazza non ha trovato un libro, translated as The girl did not find a book) in which the theme was typical (book), contrary to the atypical one (moped) showed by the experimental items. The 30 affirmative and 10 negative fillers balanced the polarity of the experimental stimuli which were affirmative for one third and negative or expletive negative in the remainder, so that 50 affirmative (20 experimental items and 30 fillers) and 50 negative items (40 experimental items and 10 fillers) were part of the final set. I then created six different experimental lists according to a Latin square design: each list consisted of 60 experimental items (20 affirmative, 20 negative and 20 expletive negative) and 40 fillers (30 affirmative and 10 negative) for a total of 100 items. Trial presentation was pseudo-randomized in order to avoid the occurrence of two or more successive expletive

An exploratory psycholinguistic approach to Snegs

187

negative sentences. Linguistic materials were recorded by a native female Italian speaker at a sampling frequency of 96 kHz and a 32 bit. The auditory stimuli were presented via headphones. The onset and the offset time of each word forming the target sentence of each trial were marked using PRAAT software (Boersma 2001). Rating of the materials The rating of the material preceded the final experiment. There were two goals of the rating: (i) ensuring that participants correctly recognized pictures; (ii) verifying that the expectedness/unexpectedness value of the target sentences was consistently perceived. Procedure: I built two different on-line questioners using SurveyMonkey® software to rate pictures and sentences. Twenty-seven native Italian speakers voluntarily participated in the rating of the materials (ten males and seventy females; 26.17 mean age). To norm the pictures I asked participants to freely name them and to judge, on a five-point Likert scale (1 = very difficult; 5 = very easy), how easy it was to recognize the depicted items. Two hundred pictures (size = 450x450 pixels; resolution = 72 pixel/inch) corresponded to the Expected and Unexpected objects referred to in the target sentences. Each picture was presented one at time199. To norm the target sentences presented after the short real-life stories, in the second questionnaire I asked participants to judge on a five-point Likert scale how expected (1 = unexpected; 5 = very expected) the target sentence was, in the context provided by the short real-life stories. Real-life stories (e.g., Laura invited some friends to her home. When they arrived, she showed them her domestic animal) ended with a sentence that could either refer to the Expected object (e.g. The girl showed a dog) or the Unexpected one (e.g., The girl showed a snake). One hundred stories and two hundred target sentences were prepared. Participants were assigned to one of the two lists each displaying 100 target sentences and 100 pictures. Stimuli were organized in two lists according to a Latin square design, so that each participant was presented 199

I integrated the photos taken from the corpora with new ones taken from the internet. I chose pictures with a frontal point of view and I used an open-source graphic-editor software, GNU Image Manipulation Program, to make them comparable to the ones in the corpora: I cut out the object from the scene background and turned it a white color; I adjusted the lightness, the colors and the contrast; etc.

188

Chapter 6

once with each story (with equal numbers of typical and atypical target sentences) and each target sentence was judged by the same number of participants. After the rating of the materials, I excluded all the items that did not satisfy the following criteria: (i) cases in which more than 70% of participants did not agree on the name of the depicted object in the picture naming task (I considered correct all the narrow synonyms, such as croissant/brioche, computer/laptop, etc.); (ii) scores