Negation raising : logical form and linguistic variation 9781443825061, 1443825069

Citation preview

Negation Raising

Negation Raising: Logical Form and Linguistic Variation

By

Vincenzo Moscati

Negation Raising: Logical Form and Linguistic Variation, by Vincenzo Moscati This book first published 2010 Cambridge Scholars Publishing 12 Back Chapman Street, Newcastle upon Tyne, NE6 2XX, UK British Library Cataloguing in Publication Data A catalogue record for this book is available from the British Library Copyright © 2010 by Vincenzo Moscati 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): 1-4438-2506-9, ISBN (13): 978-1-4438-2506-1

TABLE OF CONTENTS

Preface ....................................................................................................... vii List of Abbreviations .................................................................................. xi Chapter One................................................................................................. 1 Parameters and Cross-linguistic Variations in the Syntax of Negative Markers Chapter Two .............................................................................................. 19 Negative Complementizers Chapter Three ............................................................................................ 39 Negation Raising and Scope Widening Chapter Four .............................................................................................. 63 Clause Typing and Negative Chains Chapter Five .............................................................................................. 77 Negative Arguments Chapter Six .............................................................................................. 107 Domain Widening and Strict Negative Concord References ............................................................................................... 131

PREFACE

Some years ago, I planned to work on language acquisition and my attention was captured by an early construction which is sometime reported in the speech of two-year-old children: it has been sporadically noticed that children initially use negation in the first clausal position. I thought that this option could belong to set of UG possibilities and the next step was to observe the range of variation of clausal negation in Language. This was literary the end of the original language acquisition project and the beginning of this book. From the observation of a sufficiently large linguistic sample, it is evident that the expression of sentential negation is variable. Some languages, such as Irish, have negative complementizers, while others, like French, double the negative maker and still others, such as some Romance dialects, present an adverb at the lower edge of the VP. Given this state of affairs, it is largely unclear whether these differences affect the interpretive properties of the negative markers. There are two possibilities: either the link between Logical Form and Phonological Form is isomorphic, or it is not, and a mapping operation is required. In the first case, which I consider to be the null hypothesis, we expect that logic scope will be determined by the surface position occupied by the negative maker. Throughout the course of this book, I will attempt to falsify this hypothesis. I suggest that an indirect link exists between LF and PF and the two levels of representation are mediated by a syntactic operation which I call “LF Negation Raising”. The first step is to define where, exactly, is the position of negative makers, disentangling the many factors which complicate the analysis. In chapter one, I will try to isolate the parameters of variation which are directly related to the clausal position of NegP. In the second chapter, I will extend the analysis from the inflectional system to the complementizer layer. On the basis of the data presented in the first two chapters, I will draw a cartography of sentential negative markers. By including non-Romance varieties and the complementizer’s layer, this book will integrate and extend the work of Zanuttini (1997). The second essential step toward the falsification of the “isomorphic hypothesis” is to determine the scope of the different negative markers.

viii

Preface

Cinque (1999) empirically showed that the structural position of modality is fixed. In the same way, it seems that also its scope is invariable and tied to its surface structure. Given this fixity, we have a privileged point of observation for the scope of negation. It is possible for us to observe the scope interaction between the sentential negative markers and modal verbs: if PF-LF mapping is direct, low negative markers will allow only narrow scope readings under modality. In chapter three, I will consider the scope of negative markers in this light, showing that inverse scope is relatively easy to obtain and that the scope of negation is not tied to its surface syntactic position. An operation as LF Negation Raising has to be allowed. In chapter four I will propose the existence of a link with the complementizer’s system. The final issue which will be examined in the last two chapters concerns negation in the nominal domain and some consequences on the treatment of negative concord readings. The main contribution of chapter five will be to present a set of peculiar phenomena related to negative quantifiers in post verbal positions. The asymmetries between preverbal and postverbal positions of negative quantifiers in non concord languages shares relevant similarities with n-words in asymmetric negative concord languages. I will suggest that the distribution of n-words and the behavior of VP internal negative quantifiers can be reduced to the same constraint, which is only differently modulated in asymmetric concord languages. The final chapter closes the book by considering several issues related to NPIs and Negation Raising. Some apparent problems can be solved by adopting a semantic analysis of NPI along the lines suggested by Kadmon and Landman (1993) and further refined in Chierchia (2004). This monograph is based on my Ph.D. dissertation and it extends that work in many directions, although the core idea remains the same. Parts of this book have been previously presented elsewhere. The operation of LF-Negation Raising was presented at the 33rd Incontro di Grammatica Generativa (Bologna, 2007) and the main data on Germanic languages were discussed at the 34th Linguistiktagung (2006, Klagenfurt). A seminal work on downward entailing particles in the Italian complementizer was presented at the 31st conference of the Società Italiana di Glottologia (2005, Scuola Normale Superiore, Pisa). During these moments, I received many insightful comments and my gratitude goes out to the audiences in attendance. However, the most interesting discussions related to this work took place in more informal occasions, especially with the members of the CISCL at the University of Siena during winter seminars and spring

Negation Raising

ix

lunches. Thanks to Adriana Belletti, Valentina Bianchi and Cecilia Poletto for discussing with me some of the ideas in this book. I am very indebted, for his helpful comments, to Luigi Rizzi, who followed this work from the earliest stages and to my PhD thesis committee: thanks to Liliane Haegeman, Carlo Cecchetto, Günther Grewendorf and Ian Roberts. The improvements from the 2006 version are mainly due to their revisions.

LIST OF ABBREVIATIONS

1 2 3 Acc Agr Asp Aux Cl Comp Cont D Decl Deon DN Ep Foc Fut I Inf Irr Man Mod

first person second person third person accusative case agreement aspect auxiliary clitic pronoun complementizer continuative determiner declarative deontic double negation Epistemic focus future interpretable non-finite irrealis manner modality

N Neg Nom Non.x O.cl Obj OP P Past PPast Pres Q Real S S.cl Su Subj T Top Val Vol

noun negation nominative case lack of the x feature object clitic object operator plural past past participle present question realis singular subject clitic subjunctive subject tense topic valued volitional

CHAPTER ONE PARAMETERS AND CROSS-LINGUISTIC VARIATIONS IN THE SYNTAX OF NEGATIVE MARKERS

Introduction: parameters and variations As in any aspect of cross-linguistic comparison, also the syntax of negation presents a constellation of different constructions. Many of these variations, however, can be considered to be only the reflexes of other syntactic factors. If we assume here that a great, but limited, number of parameters is responsible for language variation, a preliminary issue concerns the nature of the parameters which directly relate to the syntax of negation. Since parameters interact between them like different genes interact to create different phenotypes, it is not easy to individuate and isolate specific parameters as it is not easy to detect the role of single genes. In abstract terms, an ideal parameter will be a syntactic key feature, possibly with a binary value, which directly affects only a specific component of grammar, even if cascade effects have to be expected in other places as well. Being firmly aware of this difficulty, let me begin by considering a first possible classification of the languages of the world in relation to the means they have to convey a negative meaning. A first, basic difference is in the particle used to signal negation. Early typological studies as Dahl (1979) and Payne (1985) classified languages by distinguishing three main strategies in accordance to the kind of negative element adopted. This can be an affix, a free particle or a dedicated verbal form as a negative auxiliary. Although this is a fundamental distinction, things are more complicate. The choice and the position of negative morphemes, for example, clearly obey to the interactive dynamic of inflection. In a more recent survey, Miestamo (2005) makes further distinctions and he considers the effects that the presence of a negative operator has on the expression of categories as Tense, Aspect and Mood. As it is easy to

2

Chapter One

imagine, what emerges is an intricate scenario of complex interactions. At this point, a first possibility is to descriptively account for the differences postulating a specific parameter for each of them. However, some are clearly the byproduct of more general syntactic features, not directly related to the expression of negation. In order to show how two different parameters might interact, consider for example the formation of negative sentence in English: (1) a. I (*not) read (*not) b. I don’t read c. I can’t read Example (1) shows that a lexical verb cannot support negation (1a) and that the presence of an auxiliary is obligatory (1b-c) even in simple tenses. Consider now a language as Italian. Here the negative affix might be hosted both by lexical verbs in simple tenses (2a) and by auxiliaries in complex tenses as passato prossimo (2b) (2) a. Io non leggo I neg read-1s-pres ‘I don’t read’ b. Io non ho letto I neg aux-1s read-pastpart ‘I didn’t read’ The difference between (1) and (2) shows that in Italian the auxiliary is not required in negative clauses. We could account for this in two ways: the first is to focus on the syntax of negation, and say that the two languages differentiate for the capacity that the negative affix has to incorporate on lexical verbs. The second is to relate the difference to the properties of Italian lexical verbs. If we choose the first hypothesis, we could assume that the syntax of negative affixes must include a parameter which specifies the kind of host. Something like [+/- lexical verb] is encoded. However, there are good reasons to believe that this distinction is epiphenomenal: a mere consequence of some deeper variation. We have indication of that if we consider a diachronic change occurred in the passage from Middle English to Modern English around the middle of the 16th century. In fact, Middle English was similar to contemporary Italian in that the negative affix ne could appear directly on the lexical verb (van Kemenade 1999)

Parameters and Cross-linguistic Variations in the Syntax of Negative Markers

(3) He ne andwyrde ðam wife æt fruman ‘He didn't answer the woman at first’

3

ÆcHom.ii.110.33

Successively, at least two important syntactic changes happened. The first was the lost of V-to-I movement, occurred around the middle of the 16th century (Roberts 2007; Kroch 1989). The second was a gradual change in the morpho-syntactic status of negation, which through different stages, became the particle not. At around the same time when English lost V-to-I movement (van Kemenade 2000), not was a syntactic head in a functional projection NegP immediately dominating the Tense Phrase (4)

NegP Neg’ not loss of V-to-I

TP VP

V

At this stage, what happened was that once the verb lost his ability to raise higher in the structure, not was still needing an host. A possible rescue strategy to satisfy the need of the negative affixal head was the insertion of an auxiliary, able to reach NegP in the course of the derivation. This is exactly what happened in the transition from Middle English to Modern English and do support began to be consistently attested in questions, negative declaratives and negative imperatives. The gradual emergence of do in these environments is shown in the figure below.

4

Chapter One

Fig.1. Percent of do forms in various sentence types (from Ellegård 1953)

In the course of the 16th century, do-support was almost obligatory in negative questions, when the verb was required to move up to the complementizer, soon followed by negative declarative. The fragment in (5), reported in van Kemenade (2000) shows that at an earlier stage not could move together with the auxiliary dyd to C, confirming the fact that it was a syntactic head which moved together with the auxiliary (5) Dyd not I send unto yow one Mowntayne that ...?

(Mowntayne.210)

At this stage, both lexical verbs and auxiliaries could transit trough NegP. However, when lexical verbs definitively lose V-to-I movement, negation always required the presence of an auxiliary, as in present day English. Once we took this snapshot of the events changing the English grammar in the 16th century, it is evident that the parameter ‘host of the negative affix’ [+/- lexical verb] can be decomposed into two different parameters: a parameter regulating the morpho-syntactic status of the negative marker and a different parameter regulating the movement of lexical verbs.

Parameters and Cross-linguistic Variations in the Syntax of Negative Markers

5

1.2. NegP and the X-bar schema in diachronic change The examples given in (1) and (2) from Italian and English show that, even in the case of languages whose grammar is well-known and documented, it is hard to isolate grammatical parameters which might be directly related to the syntax of negation. For this reason, we must be guided by some theoretical assumption on the status of parameters. If we intend a parameter as a variation related to functional projections via lexical items (Borer 1984, Kayne 2005), then a good candidate is the morpho-syntactic status of the sentential negative marker. This element will be marked as [+/- affixal], fitting the distinction between negative affixes and negative adverbials already made in Dahl (1979) and Payne (1985). We may look again at some changes occurred in the syntax of English, in order to demonstrate that variations along this dimension are not related to other general changes. If we consider the period of around 1000 years, from the 7th to the 17th century, we observe an interesting oscillation. By observing the cyclic grammaticalizations of negative markers, which swung between free morpheme/affix, Otto Jespersen proposed the existence of a syntactic cycle which affects the morpho-syntax of negation. In the heroic poem Beowulf, there is trace of a sentence initial negative marker, not incorporated to the verb, as shown by the following examples1: (6) No he wiht fram me flodyþum feor fleotan meahte, hraþor on holme neg he thing from me on waves far swim could, quicker in water ‘In no way could he swim far from me on the waves of the flood’ (7) No ic fram him wolde neg I from him wanted ‘I would not consent to leave him’ At this stage, the free particle no was not the sole negative marker and it probably coexisted with the affix ne. This latter element soon replaced no, becoming the only negative marker in the Old English prose of 9th and 10th century:

1

The cycle has been well documented in van Kemenade (2000).

6

Chapter One

(8) Ne sende se deofol ða fyr of heofenum, þeah þe hit ufan come neg sent the devil then fire from heaven though that it from-above came ‘The devil sent not fire from heaven, though it came from above' ÆCHom.i.6.13 In the course of two centuries, English underwent to a gradual change in the form of its sentential negative marker: from a sentence initial free particle to a verbal affix. The cycle went on and in the passage from Old English to Middle English, the affix ne was coupled with a second negator na, appearing in a lower clausal position: (9) þonne ne miht þu na þæt mot ut ateon of ðæs mannes eagan then neg could you not the speck out draw of man's eye ‘then you could not draw the speck out of man’s eye’ ÆHomP.XIII.153 in (9) the particle na is separated from the verb, confirming that this element was an adverbial particle. The successive development in the Middle English period is primarily characterized by the ongoing weakening of ne and the transformation of the reinforcing negator na to some spelling variant of not (10) I ne may nat denye it I neg may neg deny it ‘I may not deny it’

CMBOETH,435.C1.262

(11) ac of hem ne speke ic noht but of them neg spoke I neg ‘but I did not speak of them’

CMTRINIT,95.1271

The passage from an affix to an adverbial sentential negative marker was finally completed after the 14th century, when not or some of its allomorphs became the only sentential negative marker: (12) He yaf nat of that text a pulled hen He gave neg of that text a pulled hen ‘He didn’t give a thing about that text’

CMROLLTR,43.880

Parameters and Cross-linguistic Variations in the Syntax of Negative Markers

(13) I know nat the cause I know neg the cause ‘I do not know the cause’

7

CMMALORY,627.3550

All these successive stages are well captured through a quantitative analysis of the Middle English corpora given in Wallage (2008) and summarized in Table 1: Table 1. Distribution of ne, ne...not, not. From 1150AD to 1500AD. Period 1150-1250 1250-1350 1350-1420 1420-1500 Total

ne 436 166 43 14 660

% 60.5 22.9 1.9 0.8 16.3

Ne...not 277 490 236 18 1021

% 38.5 67.7 10.5 1.0 20.7

not 7 68 1959 1842 3876

% 1.0 9.4 87.5 98.2 63.0

Although it is impossible to draw a precise border between the different stages, we can clearly individuate at least three passages between the middle of the 12th century and the end of the 16th century. First, in the beginning of this period, ne was the only negator and it appeared as an affix on the verb. Second, in the following two centuries ne, gradually weakened and a second negative element as noht/nat appeared. Third, in the final stage around the 15th century, this second element became the sole negative marker, substituting the morpheme ne. This cycle (Jespersen 1917) can be found in many other languages (Dutch, Zeijlstra 2004; German, Jäger 2008; French, Roberts 2007) and it might capture variations in both directions: from affixes to free morphemes and vice-versa: (14)

affix

Æ

free morpheme + affix

Æ

free morpheme

This cyclic change supports the idea that all the variations are placed on a single dimension and that only one parameter is responsible for a long cycle of fluctuations. A way to express this parameter is to assume that a NegP may host both an adverbial specifier and an affixal head. The idea of a dedicated functional projection goes back to Pollock (1989) and the cycle can be captured by the three structural changes presented in (15):

8

Chapter One

(15) Jespersen’s cycle in Middle English Stage 1: 1150-1250 a.c. NegP

Stage 2: 1250-1350 a.c. NegP

Neg’ ne

Stage 3: 1350-1500 a.c. NegP Neg’

not nat noht

ne

Neg’ not

The variation here is restricted within a single functional head, and the lexical elements only minimally vary with regard to the dimension [+/head]. In addition, given the existence of the Stage 2, we need to say something else for the simultaneous presence of two negative markers. Assuming that parameters only have binary values, we can account for the Jespersen’s cycle with the following two parameters: (16) Negation is signaled by an overt adverbial in the specifier or a bound morpheme in the head of NegP. (17) Both the specifier and the head positions of NegP may be phonologically realized. These two parameters capture the three different main stages of the Jespersen’s cycle and, as I will try to show in the next section, have a great empirical coverage also in relation to synchronic variations.

1.3. Parameters and typology How good is the cross-linguistic coverage of the two parameters given in (16) and (17)? We could try to assess it by looking at typological data. I’ll make first some further assumptions with regard to NegP, assuming that functional projections are unique and cannot be duplicated2. This is the same as saying that each functional projection may express one and only one feature. We have than a very strong prediction: if there is only 2 This is compatible with a cartographic approach to clausal representation (Cinque and Rizzi 2008).

Parameters and Cross-linguistic Variations in the Syntax of Negative Markers

9

one NegP in the clause expressing a plain negative meaning, we could find at most one negative element of the same kind, being it an affix or a functional head. If we consider the diachronic variations described through the Jespersen’s cycle, they can be well captured by assuming only one functional NegP. The question is whether we can find two elements with the same meaning and of the same kind in other languages. In this case, we must assume the presence of more than one NegP. In order to determine this, we must look at a relevant sample of the world languages. The following discussion will be based on the extensive typological survey in Miestamo (2003, 2005). He considered a sample of 297 languages, whose majority expresses sentential negation using only one element, being it a free morpheme or an affix (258 languages on 297). This means that the parameter in (16) covers the 86.9% of the sample. What remains is a group of 39 languages, which adopt composite negation. Here sentential negation is expressed by more than one element. This is not unexpected since NegP might host both an adverbial and a functional element, as in the Stage 2 of the Jespersen’s Cycle. Almost all the languages of this group uses two elements of a different kind, being one a free morpheme and the other an affix. Such cases, which can be captured by the parameter in (17) can be observed also in some languages of Europe, as French and West Flemish (Haegeman 1995): (18) Jean ne vient pas Jean neg come.3s neg ‘Jean does not come’ (19) Da ze nie ketent me euren kado en-was that she neg contented with her present neg-was ‘That she was not pleased with her present’

(French)

(West Flemish)

Languages belonging to this group are similar to 13th century Middle English, with a composite negation formed by an affix (ne in French and en in West Flemish) plus a free morpheme (pas in French and nie in West Flemish). Given that affixes arguably occupy the head of NegP and free morphemes its specifier, this kind of composite negation might be captured by assuming a single NegP, compatibly with (17). It seems that by combining the two parameters in (16) and (17), we can capture a wide range of languages by assuming only a single Negative Projection in the clause structure and the following generalization largely holds:

10

Chapter One

(20) In the clause, only one NegP is phonologically realized. However, there are a few recalcitrant cases and I’ll briefly discuss some of them. In Bafut, a Niger-Congo language, negation is duplicated by a two particles, KƗƗ and sì: (21) KƗƗ mbìƾ sì lǀò neg rain neg fall ‘It has not rained’

(Bafut, Chumbow & Tamanji 1994)

both KƗƗ and sì have to be considered as free particles, apparently forcing us to assume the presence of a second NegP. However, an important fact is that KƗƗ is optional, a fact which suggests that it can be considered to be a reinforcer or an element carrying presuppositional meaning (Horn 1989). Moreover, the duplication of negative markers can be generated by complex interactions with other functional categories. Seiler (1985) showed that in Imonda what apparently looks as a second negative element is instead the expression of an irrealis feature. If we look at sentences as (22), the two elements së and –m could be considered, at first sight, as two negative particles: (22) Ehe së eg-l-uagl-f-me 3 neg follow-obj-p-go-pres-neg/Q ‘He does not follow them’

Imonda (Seiler 1985)

However, once we have more data at hand, this intuition turns to be wrong. In fact, if we consider also interrogative clauses, we find the same suffix –m (23) Ne uagl-f-me 2 go-pres-neg/Q ‘Are you going?’ Under a closer scrutiny then, the apparent composite negation in Imonda can be decomposed in a combination of +irrealis and +negation, two different features arguably hosted by two different functional heads3. 3

The only cases which appear to be really problematic in the set of languages discussed in Miestamo (2003, 2005) come from Maricopa and Maranungku. Here two negative elements, two affixes in Maricopa and two free morphemes in Maranungku, seem to express exactly the same meaning and they are both

Parameters and Cross-linguistic Variations in the Syntax of Negative Markers 11

A word also deserves to be spent on languages with more than two negative markers. It is not impossible to find languages where an extremely articulated form of negation is expressed by multiple elements. Zanuttini (1997) reports the following example taken from Parry (1997), relative to the Italian dialect of Cairo Montenotte, spoken in Liguria: (24) Dy’menika u *(n)e pa ‘vnynu Sunday s.cl neg-is neg come-neg ‘He didn’t come on Sunday’

(Cairese)

Sentence in (24) appears extremely repetitive and we find three negative elements: the affix on the copula n-, the adverbial pa and the suffix –nu. However, this redundancy is only apparent. In fact Parry suggests that the three elements absolve to different functions: while the first –n is obligatory, the other two are optional and enrich the meaning of the sentence in different ways. The element pa carries a presuppositional meaning while –nu is used as a reinforcer. On the basis of this discussion, it seems that the generalization in (20) is substantially valid and that the two parameters in (16) and (17) are adequate not only to capture major syntactic changes in a diachronic dimension, but also to cover a wide range of synchronic differences. From the examples discusses up to now it emerges another important difference, and it regards the structural position where the negative element appears: clearly, the distinction between affixes and free particles is not sufficient to account for the clausal distribution of the sentential negative marker. In order to deal with this, we must consider the relation between NegP and the other functional projections. obligatory.Maricopa is a native American language spoken in Arizona, which allows the presence of two negative affixes, waly and ma: (1) waly-puy-ma-k (Maricopa, Gordon 1986) neg-die-neg-R ‘He didn't die’ In Maranungku instead, a Norhtern Australian language, more than one negative adverb might be present at the same time: (2) piya ka-nga-ni way (Maranungku, Tryon 1970) neg non.fut-1s-go neg ‘I am not going’ The presence of languages of this kind seems to argue in favor of the possibility to have more than one NegP in the clause structure, both expressing the same meaning. We should be aware of the existence of these languages, but their rarity (2 on 297 languages, less than 1%) and the relative paucity of the available data do not seems sufficient, for the moment, to abandon the generalization in (20).

12

Chapter One

1.4. Negation and the inflectional system We know that languages express negation in different clausal position and this is certainly a key difference. Even languages belonging to the same family and spoken in the same area, might vary in the position where the sentential negative marker is realized. This point has been well illustrated for a small group of Italian dialects reported in Zanuttini (1997). Under the assumption that adverbs do not move unless topicalized, she considered the relative order between negation and other adverbs expressing aspect, habituality and other event modifications. Limiting the discussion to the three adverbials already, no more and always in Standard Italian and in two Northern Italian dialects as Piedmontese and Milanese, we have the following liner order (see also Cinque 1999) (25) Italian: Piedmontese: Milanese:

già > più > sempre gia > pi nen > sempre gemò > pü > semper ‘already’ ‘no more’ ‘always’

thus in Italian già ‘already’ precedes più ‘no more’ and this, in turn, precedes sempre ‘always’. This order, which can be easily determined by observing the position that the three adverbs occupy when they appear within the same clause, is respected also in the two other varieties4. Now, consider the position of the negative marker in Piedmontese and Milanese in relation to the adverbials in (25). In Piedmontese the sentential negative marker nen is an adverbial and it occupies a position between gia ‘always’ (26) and sempre ‘always’ (27) (26) A l’ avia gia nen salutami cul di la (Piedmontese) s.cl s.cl’aux already neg greeted-me that day there ‘Already on that day he had not greeted me’ (27) A l’ ha nen dine sempre tut s.cl scl’aux neg told-us always everything ‘He hasn’t always told us everything’

(Piedmontese)

The example (27) also shows that nen precedes not only the adverbial sempre, but also the past participle dine. If we recast the linear order in

4

The inverse order is forbidden or prosodically marked.

Parameters and Cross-linguistic Variations in the Syntax of Negative Markers 13

terms of structural dominance, we have an indication that in Piedmontese the following order holds: NegP > Past-Part P > Habitual P. This order, however, is partially reversed in Milanese. Negation is expressed by the postverbal adverbial no5 and this particle occupies a position immediately above the VP, linearly to the right of the past participle: (28) El l’ ha scrivuu no s.cl. s.cl’aux written neg ‘He hasn’t written’

(Milanese)

Consider that Past Participle in Milanese occupies a very low structural position and it is arguably lower than in Piedmontese, since it follows the habitual adverb semper: (29) L’ ha semper di tüscòs s.cl’aux always said all ‘He has always said everything’

(Milanese)

In this vaiety, negation seems to be lower both than aspectual adverbs and past participle. This is confirmed by the following example where no follows the two elements: (30) L’ a semper pagà no i tas s.cl’aux always paid neg the taxes ‘It’s always been the case that he hasn’t paid taxes’

(Milanese)

On the basis of these considerations, we have two different orders of functional projections: (31) Piedmontese: NegP > Past-Part P > Habitual P Milanese: Habitual P > NegP > Past-Part P The differences between Milanese and Piedmontese shows that while it is possible to find a strict correspondence in the order of adverbials in different languages (Cinque 1999), the position of negative adverbs resist to a cross-linguistic generalization.

5 In Milanese there is also another negative marker, the adverbial minga which has a presuppositional meaning (see Zanuttini 1997).

14

Chapter One

This conclusion is also supported if we consider not only the ordering of adverbs, but also the ordering of affixes. Ouhalla (1990) reports the following difference between Turkish and Berber, two rich agglutinating languages: (32) Jan elmarlar-i ser-me-di-∅ John apples-acc like-neg-past-agr ‘John didn’t like apples’

(Turkish)

(33) Ur-ad-y-xdel Mohand dudsha neg-fut-agr-arrive Mohand tomorrow ‘Mohand won’t arrive tomorrow’

(Berber)

While in Turkish (32) the particle –me is closer to the verbal root than the tense morpheme –di, the inverse happens in Berber. In fact, in the Berber example (33), the tense particle –ad is closer to the root than the negative marker ur-. If we assume (Baker 1985) that the order of morphemes reflects the ordering of functional heads, the examples in (32) and (33) argue in favour of a different ordering between NegP and TP in the two languages: (34) Turkish: AgrP > TenseP > NegP > VP Berber: NegP > TenseP > AgrP > VP The problem is that the comparison between (32) and (33) may be misleading for various reasons. First, it is not obvious that past and future tenses are hosted by the same functional projection. Second, while the Turkish verbal morphology illustrated in (32) shows a series of suffixes, Berber instead uses a sequence of prefixes. Although these considerations may weaken his conclusions, I believe that the argument presented by Ouhalla has a substantial validity. We only have to find better examples. Again, a typological survey might help us to find languages with the appropriate characteristics. Miestamo (2003) reports some data on Malayalam, a Dravidian language which is one of the official languages of India, spoken in the state of Kerala. Malayalam has two different suffixes for negation and past, with the past morpheme closer to the verbal root (35) Avan pa|hi-c-illa he study-past-neg ‘He did not study’

(Malayalam)

Parameters and Cross-linguistic Variations in the Syntax of Negative Markers 15

This order of suffixation may be captured if we assume that the morpheme expressing past is hosted in a functional head closer to the VP while -illa is hosted in an higher NegP. Malayalam then show the order NegP > PastP which is reversed from Turkish: (36) Turkish: Malayalam:

AgrP > PastP > NegP > VP NegP > PastP > VP

In this case we compared two series of suffixes and we also considered the order of two morphemes expressing the same tense, overcoming the drawbacks of Ouhalla original comparison. We may push the comparison even further, and consider other tense morphemes as well. If we consider present tense, the inverse order between negation and present tense morphology is shown by two SouthCentral American languages, Barasano and Kuna. Barasano is spoken in Southern Colombia while Kuna in the northern part of the country, at the border with Panama. Consider the order of suffixed in (37) and (38) (37) Wa-be-a-ha yu move-neg-pres-non.3p 1s ‘I am not going’

(Barasano)

(38) an pinsa-e-suli 1s remember-pres-neg ‘I don't remember’

(Kuna)

the order of negative suffixes in Barasano and Kuna, with the negative affix more internal in the first case, suggests again that the position of NegP is variable: (39) Barasano: Kuna:

PersonP > PresP > NegP > VP NegP > PresP > VP

Finally, let me also consider the relation between future and negative morphemes. Miestamo reports data from Inanwatan (de Vries 1996) and Tonkawa (Hoijer 1993), two languages with a sequence of distinct suffixes for negation and future tense. Inanwatan, spoken in Indonesia and North West Papua, negates the clause with the suffix –aigo, which can be optionally preceded by the particle náwo. Example (40) shows that the morpheme –aigo follows the future morpheme

16

Chapter One

(40) (náwo) né-se-sa-aigo neg 1s-walk-fut-neg ‘I will not walk’

(Inanwatan)

The opposite ordering is instead found in the Native North American language Tonkawa, documented by Henry Hoijer: (41) Yalon-ab-ado-no-’c kill-neg-fut-cont-1s ‘I will not kill him’

(Tonkawa)

Again, by looking at the sequence of affixes, we have another crosslinguistic difference for what concerns the ordering of projections: (42) Inanwatan: Tonkawa:

NegP > FutP > VP PersonP > Continuative P > FutP > NegP

In conclusion, the comparison between all the previous couple of examples supports Ouhalla’s original idea that the position of the NegP is variable across languages. We have then converging evidence both from the order of adverbials and affixal morphology showing that there is no fixed position for NegP crosslinguisitcally. This is a remarkable exception, since other inflectional projections are instead rigidly ordered in accordance to the articulated hierarchy in (43): (43) [franklyMood-speech act [fortunately Mood-evaluative [allegedly Moodevidential [probably Mod-epistemic [once T(Past) [thenT(Future) [perhaps Mood-irrealis [necessarily Mod-necessity [possibly Modpossibility [usually Asp-habitual [again Asp-repetitive (I) [often Aspfrequentative(I) [intentionally Mod-volitional [quickly Aspcelerative(I) [already T(Anterior) [no longer Asp-terminative [still Aspcontinuative [always Asp-perfect [just Asp-retrospective [soon Aspproximative [briefly Asp-durative [characteristically Aspgeneric/progressive [almost Asp-prospective [completely AspSgCompletive(I) [tutto Asp-PlCompletive [well Voice [fast/early Aspcelerative(II) [again Asprepetitive(II) [often Asp-frequentative (II) [completely Asp-SgCompletive(II) (Cinque 1999) As we saw in the pairs Turkish/Malayalam (36), Barasano/Kuna (39) and Inanwatan/Tonkawa (42), NegP occupies a variable position cross

Parameters and Cross-linguistic Variations in the Syntax of Negative Markers 17

linguistically. In principle, it seems that it may appear between any of the projections in (43). In order to express the fact that language differentiates with respect to the position where they realize the NegP, we may assume that this functional projection varies within the functional space above the VP (44) NegP c-commands XP or is c-commanded by XP where XP is any of the functional projections immediately dominating VP The parameter in (44), as stated, only indicates the lower limit (VP) where NegP could be realized. I take this move since we are not yet in a position allowing us to determine the upper bound of NegP, having not considered the complementizer layer. In fact, additional functional projections, related to the complementizer, can be individuated on the top of the hierarchy in (43) and nothing excludes that negation can be instantiated by complementizer’s particles. In the next chapter, I’ll try to extend the range of cross-linguistic variation in order to include also languages with negative complementizers.

CHAPTER TWO NEGATIVE COMPLEMENTIZERS

We saw that, although many different factors indirectly influence the way negation is expressed, there are at least three parameters which are directly responsible for the variations illustrated in the previous chapter. These three parameters, which I report again below, may capture the differences in the morphology of the negative marker, the reduplication in cases of composite negation and the position of the negative marker with respect to other constituents: PARAMETERS (45) Negation is signaled by an overt adverbial in the specifier or a bound morpheme in the head of NegP. (46) Both the specifier and the head positions of NegP may be phonologically realized. (47) NegP c-commands XP or is c-commanded by XP, where XP is any of the functional projections immediately dominating VP. Let me briefly recapitulate: the parameter in (45) casts the morphological differences into the X-bar schema and it is needed to capture the distinction between free negative particles and bound morphemes; for what concerns (46), it only says that the two possibilities given by (45) might be simultaneously realized, in order to capture cases as French or West Flemish. The parameter in (47) deserves instead more attention: in chapter one, we saw only negative markers which are hosted in the inflectional portion of the clause, overlooking cases where negation surfaces in higher structural positions. In fact, negation in the complementizer is a possibility well-attested in many languages of the world. Some of them have a specific complementizer morpheme, or sometimes even a series of morphemes, used to introduce a negative clause. This means that it is

20

Chapter Two

reductive to assume that negation can be realized only in the range of IP positions6. In the rest of this chapter I’ll present several languages showing negative complementizers, distinguishing them on the basis of their finiteness, interpretability and the possibility to occur in matrix clauses.

2.1. Irish A well documented case of a language with a rich system of complementizers is Irish (McCloskey 1979, 2001; Duffield 1995). It presents an articulated system of complementizer particles, also including several negative ones. The first particle I wish to consider is the element nach, which is used to introduce an embedded negative clause (or a negative question). The analysis of nach as a CP element is uncontroversial, also assuming a position such as that of Duffield (1995), who restricts the analysis of complementizer particles only to the elements showing the following properties7 altogether: eclipsis (or nasalization) on the verb, selection of the dependent verbal form and the selection of the past tense marker –r. As example (48) shows, both the positive complementizers go and nach trigger nasalization on the first segment of the immediately following verb (48) Eclipsis a. Creidim go gcuirfidh sí isteach air. I-believe comp put-fut she in on the job ‘I believe that she'll apply for it’

6 A caveat has to be introduced here. Individuating a NegP in the complementizer might be slightly misguiding. In fact, NegP as intended within the inflectional system has no other role than hosting a negative feature. When no such a feature is present in the clause, NegP is inert. In the case of negative complementizers, instead, the same position might be filled either by a negative feature or by a declarative or interrogative feature. In this respect, the host of negation within the CP doesn’t seem to be a specialized projection and some care is needed. The formulation in (47) has to be maintained assuming that negative features might be realized up to the complementizer, even if the label NegP can be properly used only for inflectional projections. 7 McCloskey (2001) considers these features to be a requisite sufficient but not necessary for individuating complementizer particles.

Negative Complementizers

21

b. Creidim nach gcuirfidh sí isteach air. I-believe comp-neg put-fut she in on it ‘I believe that she won't apply for it’ Another property of Irish complementizers is the selection of the –r ending for past tense. In (49a) we see that in matrix clause the past-tense marker is d(o). In embedded clauses (McCloskey, 2001), instead, both the negative complementizer nach (49c) and the positive one g- (49b) triggers the past tense –r: (49) a. D-ól siad an t-uisce past-drink they the water ‘They drank the water’ b. Deir siad gu-r ól siad an t-uisce say they comp-past drank they the water ‘They say that they drank the water’ c. Deir siad ná-r ól siad an t-uisce say they comp-neg-past drank they the water ‘They say that they didn't drink the water’ These properties, plus the fact that nach is in complementary distribution with the positive complementizer go, support the analysis of this particle as a true negative complementizer. For what concerns the meaning of this particle, it is clear that nach has a negative meaning on its own, being it the only overt negative element in the sentence. Another element of great interest is the particle cha, found in the Ulster variety of Irish (McCloskey 2001). This particle behaves like nach and go with respect to eclipsis8, dependent forms and past tense markers. The important difference is that it appears in clause initial position of matrix sentences (50) Cha-r ól tú a gcuid uisce comp-neg-past drank you their portion water ‘You didn't drink their water’

8

McCloskey (2001, 2003) notes that there is a sub-dialectal variation regarding lenition and eclipsis.

22

Chapter Two

Example (50) shows that cha does not need any other overt negative element and it suffices to make sentence (50) negative by itself. Irish has, then, at least two negative particles that carry an interpretable negative feature into the CP system: one in the embedded and one in the Matrix CP. Table 2. Negative complementizers triggering i. eclipsis, ii.dependent verbal forms, ii. alternative past-tense markers.

nach cha (Ulster)

Matrix +

Embedded + -

Interpretability + +

In Table 2, I report the syntactic distribution of the two particles (matrix/embedded) and their interpretability. They are both marked as interpretable, since they convey sentential negation on their own. In addition to these two particles, other elements in Irish can also be considered to be negative complementizers. As McCloskey notes, the morpho-syntactic tests previously seen only indicate the properties of some prototypical complementizer. It is not at all clear that one element lacking one of these features cannot be considered to be in CP. There are, in fact, also other relevant factors to be considered, like the distribution at the left edge of the sentence and the complementary distribution with positive complementizers. The preverbal negative marker ní, for example, always appears in clause initial position, it is in complementary distribution with other complementizers, it triggers the choice of the verbal dependent form (fhaca, in (51a)) and it forms the past tense by means of the suffix –r in (51b)9 (51) a. Ní fhaca tú mo nighean comp-neg see-past you my daughter ‘You didn't see my daughter’ b. Níor ól tú a gcuid uisce comp-neg-past drank you their portion water ‘You didn't drink their water’

9

Duffield has proposed that nì is not a complementizer, lacking eclipsis. McCloskey (2001) instead considers the absence of eclipsis as an idiosyncratic property of this special element.

Negative Complementizers

23

Following McCloskey (2001), we can assume that there are two classes of comp elements, one with the cluster of morpho-syntactic properties of go and nach and another which presents only a subset of such properties. Besides nì, members of this second class are mura(r) and gan. The first is the negative counterpart of má, the equivalent of English if, that is the negative form for conditionals (52) Mura gcuireann tú isteach ar an phost if-neg put-pres you in on the job ‘If you don't apply for the job’ The other element gan10 introduces negative non-finite clauses and it can also be considered in CP (53) Ba mhaith liom gan e' a bheith a'balta e'iri' were good at.me comp-neg him to be able get.up ‘I would like him not to be able to get up’ The case of gan is not isolated, and also in Hebrew (see section 2.4 in this chapter) it is possible to find a similar element. The variegated paradigm of Irish complementizer’s particles is summarized in Tab.3. below. Table 3. Irish negative complementizers. Matrix nach mura gan ní cha (Ulster)

+ +

Embedded + + +

Finiteness + + + +

Interpretability + + + + +

As it is possible to see by looking at the last column of the table above, all these particles are best analyzed as carrying an interpretable [+neg] feature, given that they do not require any other negative element to convey a negative meaning. The diversity of the elements in Table 3 covers almost all the possible cases of negative complementizers. The only option which Irish seems not to realize is the possibility to have uninterpretable complementizer 10

Example (53) is from Svenonius (1994).

24

Chapter Two

particles dependent from another negative element in the embedding matrix.

2.2. Latin A language whose negative complementizers are well known from old texts is Latin. Traditional grammars describe ne and quin as negative complementizers. Examples involving these constructions are easy to find in classical authors, as in (54) and (55) (54) a. Dionysius tyrannus [ne tonsori collum committeret,] tondere suas filias docuit ‘the tyrant Dyonisius, [in order to not commit the neck to a barber], taught his daughter to shave’ (Cicero, Tuscolanae Disputationes) b. Timeo ne Verres fecerit ‘I fear that Verres has done’ (Cicero, Actionis in C. Verrem Secundae) (55) numquam tam male est Siculis, [quin aliquid facete dicant] ‘Sicilians are never so bad, not to say something witty’ (Cicero, Actionis in C. Verrem Secundae) Examples in (54) introduce the element ne, which is analyzed in traditional Latin grammars as equivalent to ut+non 11 and which is usually used in embedded negative final clauses as in (54a) or in sentential complements of verbs of fearing (54b). In both situations, this complementizer is the only overt negative element. Let me briefly consider how ne replaces ut in embedded clauses. Example (56) illustrates a plain negative matrix sentence with the sentential negative marker non (56) Hominem non video man-acc neg see-1s ‘I don't see a/the man’

11

Liliane Haegeman (p.c.) pointed out that ne resembles the meaning of the English complementizer lest. Also English seems to admit the presence of overt negative features in its complementizer system.

Negative Complementizers

25

In (57) the affirmative counterpart of (56) is embedded in a final clause introduced by ut (57) Veni ut hominem videam I-came comp man-acc see-1s-su ‘I came to see the man’ Now let us consider negative final clauses. If we want to embed sentence (56) with its negative meaning, the negative marker non and the complementizer ut are collapsed together and they are replaced by ne (58), which carries both the features [+final] and [+neg]: (58)

Veni ne hominem videam came-1s comp-neg man-acc see-1s-su ‘I came in order not to see the man’

Let us now consider quin, the other negative complementizer presented in (55). A difference between ne and quin is that this last element introduces a negative clause but it is dependent from negation overtly expressed in the matrix. In (55) it is dependent from numquam ‘never’ in the embedding clause. For this reason, we cannot conclude that quin expresses a negative meaning on its own. Table 4 illustrates the fact that Latin had at least two negative complementizers, which were used to introduce negative embedded clauses. A salient difference between the two is that while ne has in its meaning an inherent negative content and conveys negation alone, quin depends instead on a c-commanding negative operator. Table 4. Latin negative complementizers. Matrix ut quin

Embedded + +

Finiteness + +

Interpretability + -

2.3. Basque Basque presents an interesting alternation between the two particles (e)nik and (e)la, as documented in Laka (1990). A few words on Basque are necessary to understand the role of these particles. Laka assumes a head-

26

Chapter Two

final analysis for Basque, with the auxiliary appearing to the right of the lexical verb (59) a . Etxea erori da house fallen aux ‘The house fell down’ b. *Etxea da erori c. *Erori etxea da The same order, with finite verbs or auxiliaries to the right, is found in embedded clauses. Here the complementizer appears on the finite verb which moves to the head (final) CP position (60)

CP1

[

CP2 IP

[ [Galapagoak muskerrez beterik] [daudela]] diote ] Galapagos lizards-of full are-that say-they ‘They say that the Galapagos are full of lizards’

Now consider the following examples, showing that the negative complementizer enik replaces its positive counterpart ela if the sentence is embedded under a negative matrix (61) a. CP1[ Amaiak CP2[ IP[inork gorrotoa] [dionik ]] ukatu du] Amaia anyone hated has-comp-neg denied has] ‘Amaia denied that anybody hated her’ ]] esan] b. CP1[Ez du Zuriñek CP2[ IP[inor etorriko] [denik no has Zuriñek anyone come will- comp-neg said 'Zuriñek has not said that anybody will come' When the embedded clause is selected by a positive declarative matrix sentence (60), the complementizer ela must be chosen. On the other hand, the negative complementizer enik must be obligatorily selected when the matrix is negated by means of an adversative predicate (61a) or the negative marker ez (61b). Enik is thus selected by a c-commanding negative element and cannot be considered as carrying an interpretable negative feature by itself. In this respect, it is similar to the Latin particle quin.

Negative Complementizers

27

2.4. Hebrew Landau (2002) discusses at length the property of the Hebrew particle me12, which he considers to be homophonous between a non-finite complementizer and a preposition. However, once island violations are taken into account, it is possible to detect the properties of the complementizer. Consider sentence (62) (62) *[Im mi] Gil hitragez me-ledaber? with whom Gil became-angry from-to-talk ‘To whom did Gil become angry from talking?’ this example shows that, when me- is a preposition, extraction of a whconstituent is impossible given that this will violate strong islands constraints. The structure of (62) is given below, with im mi ‘to whom’ being extracted from a PP: (62)’ * CP[ [Im mi]

IP

[Gil hitragez

IP

[

me-ledaber

PP

[ t]]]]

Interestingly, there are certain cases where constituents can be extracted, in an apparent violation of islandhood (63)

[Im mi] Gil nizhar me-ledaber? with whom Gil was-careful from-to-talk ‘to whom was Gil careful not to talk?’

Why extraction is possible in (63) but forbidden in (62)? The pair minimally differs only in the matrix verb. According to Landau there is a special class of verbs with a negative meaning which introduce a nonfinite embedded clause. Nizhar belongs to this class and the particle -me is not a preposition but instead a non finite complementizer. With respect to the class of verbs able to licence the complementizer me-, we find some verbs as nimna 'refrain' which obligatorily selects it (64) and other verbs which only optionally selected it as nizhar (65)

12

Landau mentioned also ilmale, (i)lule, pen as possible negative complementizers, as suggested by Aldo Sevi in a reply to a query appeared on LingList (http://linguistlist.org/issues/9/9-1331.html#1). At any rate, examples with these particles are not clear (Ur Shlonsky, p.c.), therefore they won’t be considered here.

28

Chapter Two

(64) Gil nimna me-leha’aliv et Rina Gil refrained from to-insult ACC Rina ‘Gil refrained from insulting Rina’ (65) a. Gil nizhar me-leha’aliv et Rina Gil was-careful from-neg to-insult acc Rina ‘Gil was careful not to insult Rina’ b. Gil nizhar lo leha’aliv et Rina Gil was-careful not to-insult acc Rina ‘Gil was careful not to insult Rina’ This last kind of verbs is particularly telling since they always demand the presence of a negative element in the embedded clause: either the negative complementizer me- (65a) or the sentential negative marker lo (65b). Notice that the complementary distribution of me and lo indicates that also me has an interpretable feature like the sentential negative marker lo. Since we may consider selectional requirements as being satisfied in the CP layer of the embedded clause, an intuitive way to state the dependence between the matrix verb and its negative clausal complement is to say that a CP with the feature [+neg] is required. What (65) shows is that this feature can be provided in two ways: either by the sentential negative marker lo or by the non-finite negative complementizer me-, base-generated in the CP. This means that a [+neg] feature in CP can be provided through in situ lexical insertion or through agreement with the SNM (65)a. CP1[ Gil nizhar

CP2

[me+neg -leha’aliv [IP [VP et Rina ]]]

(65)b. CP1[ Gil nizhar CP2[+neg [IP lo+neg leha’aliv [VP et Rina]]]  The representation in (65b) presents a distance relation between two structural positions, suggesting that lexical insertion is not the only way to provide a negative feature to the complementizer but that a mechanism as Agreement might be involved as well. I’ll come back to this point later (see chapter four), in relation to the more general discussion on clause typing. 

Negative Complementizers

29

2.5. Features and positions within the extended complementizer Let me take stock here and summarize the properties of the various kinds of negative complementizers encountered so far. Although the constellation is likely to enlarge by broadening the empirical survey, it is already possible to give a first taxonomy with the elements at hand. They differentiates along three main oppositions: interpretability (+/- i[neg] ), embedding (+/- matrix) and finiteness (+/- [fin]). Regarding the first opposition +/- i[neg], we saw that some particles are obligatorily selected by a negative matrix. In this case there is no evidence that such particles express an interpretable negative feature. In such a slot we find the Basque negative complementizer enik and Latin quin, in opposition to the Irish nach that independently signals a negative meaning. With respect to the second opposition, embedding (+/- matrix), this is illustrated by considering elements as Basque enik and Ulster Irish cha. While enik is found exclusively in embedded clauses, the particle cha surface in sentence initial position of main clauses. The last element of variation +/- [fin] has to do with the finiteness of the clause introduced by the particle. Hebrew me and Irish gan differ from all other particles exactly along this dimension. With these oppositions in mind, table 5 summarizes the characteristics of all the negative complementizers examined. Table 5. Taxonomy of Negative Complementizers Matrix nach mura ne quin enik gan me ní cha

+ +

Embedded + + + + + + +

Finiteness + + + + + + +

Interpretability + + + + + +

At this point, we have to assign a structural representation to these elements, possibly accounting for their different feature specification. In

30

Chapter Two

doing that, I assume an articulated CP structure as in Rizzi (1997, 2001) and reported in (66) (66)

ForceP

(TopP*) IntP (TopP*) FocusP (TopP) FinP

Positions as TopicP and FocusP are dedicated to host constituents made prominent in the discourse and are not directly relevant to our discussion. Instead the two external positions ForceP and FinP are the ones where notion as finiteness and typing (Cheng 1991) are encoded and are arguably responsible for hosting the particles previously analyzed. According to (66), complementizers introducing non-finite clauses mark the lower edge of the CP layer: elements as the Irish gan and the Hebrew me- sit in this position, respectively in the specifier and in the head of FinP. The other complementizers in table 5 introduce instead finite clauses, either matrix or embedded, and are hosted in ForceP. On the basis of the structure (66), it is possible to trace a cartography of the positions occupied by the elements in table 5: (67) ForceP

CP - matrix

cha (Ulster Irish) ni (Irish)

ForceP CP - embedded ne (Latin) FinP mura (Irish) nach (Irish) gan (Irish) me (Hebrew)

In (67), only complementizers which are interpretable as negative operators are reported, being them the sole elements which can unambiguously be considered as negative. For what concerns instead the Basque enik and the Latin quin, it is less clear whether they have to be treated as negative. In fact, they cannot freely appear in embedded clauses and need another negative element to license them. This means that a mechanism is required to capture their

Negative Complementizers

31

dependency from another c-commanding negative element. Tentatively, we could assume that such complementizers are a kind of polarity items. Under this view, they are not equipped with a negative feature and do not belong to the set of negative elements13.

2.6. Morpho-syntactic reflexes Once recognized the fact that the CP system might host negative features, we might look back at the languages which are usually assumed to typically express negation only in the IP. Some traces of an involvement of comp in negative sentences have been sporadically noticed in English, but these phenomena have never received an organic treatment. For example it is well known, at least since the seminal work of Edward Klima (1964), that English has another way of expressing sentential negation besides the use of negative markers as n’t and not. This strategy consists in fronting a negative constituent of various kind14, being it an adverbial, a negative quantifier or a negated DP or PP: (68) a. On no account will I read emails b. Nothing have I seen that could rival the pyramids c. No race could Lewis win d. With no job would Mary be happy Sentences in (68) (see Haegeman 2000b and references therein) can be considered as true negative sentences for their ability to pass several diagnostics to reveal sentential negation. They are the capacity to host positive tags (see also section §5.1.), to be conjoined with neither and to licence NPIs:

13 Hebrew me- need to be treated differently, since it has to provide the embedded CP with a negative feature in order to satisfy the s-selectional requirement of the matrix verb. A solution is to consider me- as having only a formal negative feature. Such a feature is uninterpretable and it won’t be passed to the interpretive interface. 14 Minimal value adverbials can also be fronted, triggering Subject-Auxiliary Inversion: a. Rarely has so much been done for so many by so few b. Scarcely had the bell rung when John commenced to speak c. Seldom do I see him nowadays

32

Chapter Two

(69) a. Not often does jack attend parties, does he? b. Not often does Jack attend parties and neither does John c. With no job would she ever be happy Klima describes these constructions saying that the initial expansion node of a sentence may optionally include negative constituents. This is equivalent to say, in our terms, that the CP is a clausal category able to attract negative material15. Another interesting construction is discussed in Progovac (1994). She considered sentences embedded under a matrix adversative predicate as doubt. Consider the following sentences and the distribution of NPIs: (70) a. I doubt that you saw anybody b. I doubt that anybody will come c. *I doubt of anybody In (70a-b) the NPI anybody is within the scope of doubt which might be considered as its licensor. However, doubt c-commands the NPI also in (70)c and nevertheless the sentence results ungrammatical. The contrast between (70a-b) and (70c) follows instead if we assume that verbs as doubt cannot directly take an NPI as a complement but instead have to select a CP with a negative feature (70)a I doubt [ForceP Øi[+neg] that [IP you saw anybody]] Progovac’s analysis ultimately provides us with another indication for a negative feature hosted in the English complementizer system16 and

15

Haegeman (2000b) considerd fronted negated constituents in Focus P. This possibility was actually the standard way of expressing negation in Old English around the 8th century (see van Kemenade 2000). We find several examples with the sentential negative marker nǀ in clause-initial position, as shown in the examples below from Beowulf: (1) Nǀ hƝ wiht fram mƝ flǀdyþum feor fleotan meahte, hraþor on holme not he thing from me waves-DAT.PL far swim could, quicker in water ‘In no way could he swim far from me on the waves of the flood, more quickly on the sea’ (2) Nǀ ic fram him wolde. not I from him wanted ‘I would not consent to leave him’ The sentential negative marker nǀ appears to the left of the subject, in a position that it is best analyzed as being inside the CP (Christensen, 2004). Notice that nǀ 16

Negative Complementizers

33

sentence (70c) can be ultimately analyzed along the lines proposed for Basque sentences, whose embedded clauses is introduced by the negative complementizer enik. The only difference is that in English the complementizer remains phonetically null. English seems not to be an isolated case and I wish to consider Italian next, where even less attention has been paid to the emergence of negation within its CP system.

2.7. Italian Left-Edge Licensors Italian typically signals negative sentences with the negative marker non, considered to be an X° element base generated in a position between AgrsP and TP (Belletti 1990). With regard to the CP layer, no dedicated complementizer for negation can be found. Italian thus seems to be able to express negative meaning only by means of its inflectional system. However, to a closer scrutiny, it seems that there are indeed some elements, equipped with a negative feature (or at least with a DE operator), which appear in a position above the IP. Let us first consider the shape of the left edge of the clause in Italian. Italian marks its higher structural position with the finite complementizer che (Rizzi 1997), which precedes all the remaining clausal material. In the example below we see that both topicalized and focused constituents must obligatorily follow che: (71) Credo ForceP[che TopP[Gianni FocP[LA PARTITATopP[ieri IP [abbia visto] believe-1s [that Gianni the match yesterday has seen ‘I believe that Giovanni, THE MATCH, yesterday, saw’ Under this organization, we expect that the finite complementizer che can never be preceded by any other element. But this is not entirely true, since members of a given class of words may indeed precede the complementizer che (Moscati 2006a). Among these elements, some of them such as mai ‘never’, prima ‘before’, senza ‘without’, and the

appears in the left-periphery of the clause, but the verb stays lower in the structure, excluding the possibility that nǀ has moved there together with the finite verb.

34

Chapter Two

negative marker non itself trigger downward entailing contexts17. Consider the examples given below: (72) a. Prima che tu riesca a vedere Gianni before that you manage.su to see Gianni ‘Before you manage to see Gianni’ b. Prima che mi vedesse nessuno before that me see.su anybody ‘Before anybody saw me’ c. Prima che nessuno riuscisse a fermare Gianni before that anybody manage.su to stop Gianni ‘Before anybody manage to stop Gianni’ (73) a. Senza che tu riesca a fermare Gianni without that you manage.su to stop Gianni ‘Without you managing to stop Gianni’ b. Senza che ci vedesse nessuno without that us see.su anybody ‘Without anybody seeing us’ c. Senza che nessuno riuscisse a fermare Gianni without that anybody manage.su to stop Gianni ‘Without anybody managing to stop Gianni’

17 There are other words which can also appear before che. These words, though having an intuitive negative flavour, are unable to licence NPIs themselves and the presence of another negative maker is required (1) No che non ho visto nessuno neg that neg have seen anybody ‘I deny that I saw anybody’ (2) Tranne che non arrivi nessuno, staccherò presto dal lavoro unless that neg comes anybody, quit early from work ‘unless somebody comes, I’ll finish early to work’ (3) A meno che non venga nessuno, staccherò presto dal lavoro unless that neg comes anybody, quit early from work ‘unless somebody comes, I’ll finish early to work’

Negative Complementizers

35

(74) a. Mai che tu riesca a fermare Gianni never that you manage.su to stop Gianni ‘You never manage to stop Gianni’ b. Mai che ci vedesse nessuno never that us see.su anybody ‘Nobody ever saw us’ c. Mai che nessuno riesca a fermare Gianni never that anybody manage.su to stop Gianni ‘Nobody ever managed to stop Gianni’ (75) a. Non che tu riesca a fermare Gianni not that you manage.su to stop Gianni ‘Not that you manage to stop Gianni’ b. Non che io abbia visto nessuno not that I Aux.su seen anybody ‘Not that I saw anybody’ c. Non che nessuno riesca a fermare Gianni not that anybody manage.su to stop Gianni ‘Not that anybody manages to stop Gianni’ The a. examples show that all these particles may precede the finite complementizer che at the left edge of the sentence. With regard to their semantic status, they introduce a Downward Entailing operator since they licence n-words in their scope18, both in object (b. examples) and in subject (a. examples) position. For their ability to appear to the left of che and to licence polarity items, I descriptively labelled these elements as Left-Edge Licensors (LEL, Moscati 2006c). These elements precede che but no material can intervene between them. For this reason, I consider LELs to form a unit with the complementizer in the specifier of ForceP19, as in (76) 18

The status of niente and nessuno is controversial (see Acquaviva 1999, Haegeman 1995, Longobardi 1991 and Zanuttini 1991). The issue will be considered in chapter five. Notice that in all the examples from (72) to (75) the licensor of n-words is represented by the complementizer’s particles. 19 When in an embedded clause, senza and prima can be also preceded by the finite complementizer che:

36

(76)

Chapter Two

ForceP

[prima che/senza che/mai che Force’[... TopP[...]]]

Conclusions In this chapter we saw that negative complementizers are easy to find in the languages of the world and even in a single language as Irish we may find more than one negative complementizer. Sometimes, pre-sentential negation has been considered as a special kind of metalinguistic or presuppositional negation (Horn 1989). This notion is clearly not adequate to capture the properties of negative complementizers illustrated in the previous sections. In fact, these particles carry a plain negative meaning that directly corresponds to a negative operator (or to an interpretable negative feature, using a different terminology). In this respect, negative complementizers might work exactly as sentential negative markers. Now, I want to combine the observations made in chapter one with the typology of complementizers drawn in this second chapter and trace a map of the different positions where negative features are inserted. The representation in (77) expresses the different surface positions of the negative elements encountered:

(1) Io credo che prima che tu veda Gianni, dovresti parlare con Ada I believe that before that you see Gianni, should-2s talk with Ada ‘I believe that, before you see Gianni, you should talk to Ada’ (2) Io credo che senza che Gianni se ne accorga, tu ruberai l’anello I believe that without that Gianni himself of-it notice you steal-fut the ring ‘I believe that, without Gianni noticing, you could steal the necklace’ However, these sentences have a parenthetical intonation. Sentence (1), for example, can be represented as follows: (1)’ Io credo che [ForcePprima che tu veda Gianni], dovresti parlare con Ada This suggests that prima and sempre really mark the left-edge of the clause.

Negative Complementizers

37

(77) ForceP (matrix) NegP1 cha (Ulster Irish) ni (Irish) TP -illa (Malayalam) -suli (Kuna)

NegP2

nen (Piedmontese)

AspP NegP3

no (Milanese)

VP ForceP (embedded)

prima che (Italian) ne (Latin) FinP Ø (English) mura (Irish) nach (Irish) gan (Irish) me (Hebrew) In (77), I report only negative markers which are not dependent from another negative element. Thus cases of complex negation (as in Standard French and Inanwatan) and negative embedding complementizer introduced by a negative matrix (as in Basque) have not being considered. Moreover, markers which express a presuppositional meaning have been also excluded. This makes safe to conclude that sentential negation may range from the border of VP in languages as Milanese, to the topmost position of the CP. Looking at (77), a natural question is whether the position of negation in these different languages has any consequence on its logic scope. I’ll try to address this issue in the next chapter.

CHAPTER THREE NEGATION RAISING AND SCOPE WIDENING

Introduction Having considered the possible surface variations of negative markers, I want to dwell here on the effect that these variations have on the scope of the negative operator. I take the move from the fact, already noticed in Jespersen (1917), that negation might be interpreted in a position different from the surface one. Jespersen pointed out that in sentences with a matrix psych-verb, negation can refer to the event expressed in the embedded clause, although it surfaces in the matrix. For example, consider the following sentence: (78) I don’t think that he has come in (78) we observe that the negative marker n’t is used not to deny the belief expressed in the matrix, but instead the event of coming. For some reason, the negative operator seems to be reconstructed in a lower position. The most appropriate paraphrases of (78) is (78a) and not (78b): (78) a. I think that he hasn’t come (78) b. # It is not the case that I think that he has come. This kind of reconstruction is visible also within the same clause and what seems prima facie a sentential negation, must be instead interpreted with narrower scope, as noted by Stockwell, Schachter and Partee (1973) in relation to sentences as (79): (79)

I don’t cut my salami with a hacksaw

Again, the negative marker appears on the auxiliary, but it actually has scope only on the constituent with a hacksaw. Thus the sentence in (79) is logically equivalent to (79a):

40

Chapter Three

(79) a. I cut my salami, but not with a hacksaw I take these facts, referred in Jespersen (1917) as attraction of negative20, to indicate that some mechanism can separate the LF interpretation of negation from its PF realization. Notice that the attraction of negative cannot be easily reduced to other kinds of movement. It is immediately evident that it cannot be A or A’ movement. But it cannot be simple head movement either, as in (78a) the element n’t crosses clause boundaries. This is forbidden, unless we are in some kind of restructuring domain (Rizzi, 1982). This is clearly not the case for the English sentence (78a). However, even in languages as Italian, attraction of negative does not behave as we might expect in environments where restructuring is allowed. We know that modal verbs create monoclausal effects allowing clitic climbing from embedded to matrix clauses. This is possible with modals (80)a but not with psych verbs as pensare ‘to think’(80b): (80) a. Gianni lo vuole vedere Gianni o.cl want to see o.cl ‘Gianni wants to see him’ b. *Gianni lo pensa di vedere Gianni o.cl think comp-to see o.cl ‘Gianni believes he will see him’ While clitic climbing is excluded in (80b), attraction of negative is acceptable in the same context: (81) Gianni non pensa di vedere Carlo Gianni not think comp-to neg see Carlo ‘Gianni thinks that he won’t see Carlo’ The contrast between (80b) and (81) shows the limit of an analysis which would consider attraction of negative as a particular kind of head movement, on a par with clitic climbing. The phenomenon seems to be more general and, although it is still poorly understood, it shows that the surface position of negation doesn’t tie its scope. 20

Other authors as Horn (1989) changed the terminology, using the term Negative Raising to describe sentences where a matrix psych verb attracts the negative marker. I wish here to maintain the original label attraction of negative, since the term raising, in the current framework, is generally used to denote LF operations.

Negation Raising and Scope Widening

41

This phenomenon clearly indicates that the interpretation of negation is not bound from its surface manifestation and that a negative chain links the overt sentential negative marker with a lower structural position. Now, I will consider at length the opposite situation, showing that LF displacement works in both directions: while in the case of attraction of negative we observed a scope narrowing, there are many cases where the reverse situation holds and the scope of negation is widened.

3.1. Scope widening The question I want to address here is whether languages which express negation in a structurally low position have means to widen its logic scope at LF. As we saw in the case of the attraction of negative, it seems that the relation between LF and PF is not isomorphic. This is true in at least one direction: scope can be narrowed, as in the examples (78) and (79). However, this is not sufficient to admit free LF displacement. It is possible that, while scope narrowing (or reconstruction) is possible, scope widening is not an available option. Differently stated, it is perfectly possible that negation may receive its maximal scope in the position where it is expressed at PF. If this hypothesis is on the right track, we expect that the scope of negation in Milanese should be narrower than in Ulster Irish. Are things really like that? A way to check this prediction is to consider the scope of negation in relation to other scope bearing elements. A possible testing ground is given by the interactions between negation and modality. In cases where a modal c-commands negation at PF, the prediction is that negation will receive only narrow scope if negation is interpreted in situ. Let us go back to Milanese and see what happens once a modal is inserted. We showed in chapter one that the sentential negative marker no must follow both past-participles and low aspectual adverbs, as in (30) repeated here as (82) (82) L’ a semper pagà no i tas s.cl’aux always paid neg the taxes ‘It’s always been the case that he hasn’t paid taxes’

(Milanese)

Unsurprisingly, no also follows infinitive forms selected by the quasimodal verb gà, similar in meaning to “have to” in English:

42

(83)

Chapter Three

El gà de studià no s.cl must of to-study neg

A representation for (83)21 is the one below (83)’

AgrP

ModP

[Elj' gak

[OPƑ tk

TP

[de studiàv

NegP

[OP¬ no

VP

[ tj tv ]]]]]

This representation only allows narrow scope of negation under modality and the sentence could be uniquely interpreted as a prohibition against the event of studying. However, according to many native speakers, sentence (83) can be felicitously uttered to express not only the meaning in (83a), but also the one in (83b): (83) El gà de studià no s.cl must of to-study neg a. ‘He is required not to study’ b. ‘He is not required to study’

Ƒ¬p ¬Ƒp

The existence of this reading in (83b) shows that the scope of the negative operator is not constrained by the position where it surfaces and that its scope can be widened in order to derive scope-shifted interpretations. At this point, it is clear that we need a mechanism able to derive the alternative representation (83b). As in the case of quantifiers, a possible solution is to derive (83b) by applying some further operation to (83)’. If we assume that this process applies in syntax, we may raise the negative operator in a position above ModP: (84)

XP

[OP¬

AgrP

[Elj' gak

ModP

[tk OPƑ

TP

[de studiàv

NegP

[no VP[ tj tv ]]]]]

This operation is not new in the literature and a similar proposal is presented also in Haegeman (1995), Haegeman & Zanuttini (1991), Acquaviva (1999) and Watanabe (2004). However, there is at least an important difference between the operation I’m proposing here and the one they suggested. The difference is that Haegeman & Zanuttini (1991), for example, focused on cases where 21

I adopt the view in Cinque (2006) who considers modal as functional heads. Nothing changes, for our purposes, if we adopt instead a restructuring process (Rizzi 1982).

Negation Raising and Scope Widening

43

raising is necessary in order to have scope over the event variable. Consider sentence (85): (85)

John sees nothing, does he?

if this sentence is used to express the fact that the event of seeing doesn’t take place, the negative operator on the quantifiers has to move in order to scope over the event variable. The sentential scope of negation in (85) is confirmed by the well-known Klima’s test: positive tags are licensed only by negative clauses22. This operation is more similar to the obligatory LFmovement which has been proposed by Heim & Kratzer (1998) for object quantifiers. The operation I propose in (84) is, instead, of a different kind. Negation already has sentential scope and the representation is well-formed before the application of Negation Raising. What is the reason for this movement? Clearly it has to be related to scope-shift and it is of the same kind than the operation required to shift the scope between two quantifiers, as in the example (86): (86) A doctor will interview every new patient a. [IP A doctor will [VP [every new patient]1 [VP interview t1]]] b. [XP [every new patient]1 [IP A doctor will [VP t’1 [VPinterview t1]]]] In (86), raising of the universal quantifier [every new patient] is by no means necessary from the point of view of derivational convergence. It is only required to generate the alternative meaning (86b). For this reason, it is optional in syntax23. I believe that this is a better way to capture the operation in (84) and it can be referred to as ‘Optional Negation Raising’: (87) Optional Negation Raising is possible when: Į c-command ȕ at PF but ȕ scopes over Į. where: Į is a logic operator and ȕ is the logic operator ‘¬’ In the following section, I’ll consider other cases where the scope-shifting operation described in (87) is required. Once we direct our attention towards this kind of constructions, many other instances can be found. I’ll 22

VP-internal negative quantifiers also allow positive tags. This indicates that negation in direct object position only optionally types the clause as negative (see chapter 5, section 5.1) 23 According to Fox (2000) and Reinhart (2006) it is also costly.

44

Chapter Three

consider next the case of West Germanic languages, where inverse scope readings over modality can be found not only with the sentential negative markers but also with negative quantifiers.

3.2.1. Inverse scope in West Germanic languages West Germanic languages as German, Dutch and West Flemish24 mark sentential negation with a negative adverbial, hosted in a position immediately above the VP (Laenzlinger 2004, Haegeman 2002). According to this analysis, the structural position and the syntactic status of the sentential negative marker in West Germanic languages shares relevant similarities with the Milanese no. In what follows I will show that the similarity is real and that analogous inverse scope readings can be found. In order to do that, it is necessary to control at least two relevant factors: the first is the clausal architecture with respect to the precise order of functional projections; the second concerns the polarity restrictions active on the modal paradigm. Let me consider the latter point first. Many scholars (van der Auwera 2001; de Haan 1997) pointed out that the choice of the modal in German and Dutch is sensitive to polarity. Different forms with the same strength can be employed to explicitate their scope relation with negation. However, even if it is certainly true that scope can be disambiguated through the appropriate lexical choice of the modal, it is also true that this mechanism is not perfect and some modals present a certain grade of ambiguity. Consider the German modal mussen ‘must’ in its deontic use: (88) Du musst nicht soviel rauchen you must neg so much smoke a. ‘You don’t have to smoke that much’ b. ‘You must not smoke that much’

¬Ƒp Ƒ¬p

According to Büring (1997), this sentence can be uttered to express either the meaning (88a) or (88b). What is especially interesting for our purposes is the interpretation (88a). This reading is perfectly licit also in absence of a special intonation (rising pitch accent on the modal, required for (88b)) and the same intuition has been confirmed with some native speakers for sentences (89) and (90), who accepted interpretations (89a)-(90a): 24 West Flemish, in addition to the negative adverbial nie also has the negative head en-, which marks the clausal interpretation of negation.

Negation Raising and Scope Widening

45

(89) Hans muss Julia nicht sehen Hans must Julia neg to-see a. ‘Hans is not required to see Julia’ b. ‘Hans is required not to see Julia’

¬Ƒp Ƒ¬p

(90) ...dass Hans Julia nicht sehen muss ...that Hans Julia neg see-inf must a. ‘... that Hans is not required to see Julia’ b. ‘... that Hans is required not to see Julia’

¬Ƒp Ƒ¬p

At this point, mussen can be considered to have, if any, only a ‘mild’ polarity and in any case, it can be felicitously interpreted within the scope of negation. Now, if NegP is projected immediately above the VP, we can imagine that Negation Raising applies to derive the interpretations in (88a), (89a) and (90a). However, some care is needed here since this conclusion relies on the assumptions we make with respect to the position of the modal verb. In what follows, I’ll consider in more detail what is the proper status of the modal mussen in German and what is its relation with the sentential negative marker nicht. For what concern the functional status of mussen, this analysis presents several empirical advantages over a VP internal analysis of modal verbs, as discussed in Wurmbrand (2004). She extended Cinque’s (2001, 1999) proposal to German modals, including them in the set of verbal forms which have to be considered as functional heads. Her analysis is based on the fact that modals do not behave as lexical verbs in at least two important respects: first, they are not able to assign theta roles; second, they show rigid ordering restrictions which can be captured by assuming an ordered hierarchy of functional projections25. With regard to the absence of theta role assignment, it is possible to demonstrate that modals do not select an external argument26 and that they behave as raising verbs. A first argument comes from the fact that they

25

The form sehen muss (90) has been treated as a composite head (von Stechow, 1984) to account for the fact that it cannot be interrupted. However Haider (2003) shows that these complex forms can be separated and that smaller constituents may be topicalized, as in the following example: (1) [Erklären ]VP müßte man das können tVP ‘one ought to be able to explain that’ 26 It is harder to show that modals do not assign thematic roles to internal arguments, since they select a VP which can absorb the thematic role. For this issue, see Wurmbrand (2004, 2007) and Cinque (2004,2006).

46

Chapter Three

licence non-thematic expletives in subject position, as in (91) where the expletive ‘es’ corresponds to a wheater-it subject27 (91) Es muß morgen schneien it must tomorrow snow ‘It must snow tomorrow' Another argument in favour of the idea that modals do not assign theta roles to external arguments is confirmed by the impossibility to form passives. In fact, if we embed an unaccusative verb under a modal, none of them is able to assign a theta role to the external argument. Consider first the contrast between (92a-b): (92) a. unergative Es wurde einen Abend lang getanzt it was an evening long danced ‘They danced for an evening’ b. unaccusative *Es wurde am Flughafen angekommen it was at-the airport arrived ‘They arrived at the airport’ the different grammatical status of (92a-b) shows that only unergative verbs can be passivized, while unaccusatives cannot be used to form passives. The generalization is that passives are only possible if the predicate involves an underlying external argument. Wurmbrand shows that, if we embed an unaccusative under a verb which is able to assign a theta role to the subject (as ‘to try’), passives are possible. The reason is that it is the embedding verb which saves the structure by theta role assignment. This, however, is not possible under modals as mussen28: 27 Not all the verbs selecting an infinitival complement can be considered to be functional. For example, modals contrast with other verbs as plante ‘planned’ which cannot have subject expletives: (1) *Es plante zu schneien it planned to snow ‘It planned to snow’ 28 The ungrammaticality of modal passives does not seem to be limited to German, but appears to be a more general property of modals (Aissen and Perlmutter 1983, Burzio 1986). Also in Italian passivization has a degraded status, as shown in the following sentences:

Negation Raising and Scope Widening

47

(93) *Der Wagen wurde (zu) reparieren gemusst/müssen the car-nom was (to) repair must-part/inf ‘They had to repair the car’ this again supports the idea that modals are inert with regard to theta-roles assignment. The second kind of arguments in support of the functional analysis of modals comes from the ordering restrictions active on their sequence. The pair in (94) shows that dürfen and mussen need to respect a fixed order, with the first modal being higher in the structure and able to reach the V2 position: (94)

a. Er dürfte zu Hause sein müssen he might at home be must ‘He might have to be at home’ b. *Er muß wieder singen dürften he must again sing might ‘It must be the case that he might sing again’

Some considerations are in order here. In (94) dürfen has an epistemic meaning while mussen has a deontic one. It could be possible that semantic restrictions determine the c-command relations. However this doesn’t seem to weaken the argument, since the epistemic modal have to be merged higher than the deontic modal in order to be successively able to move, in syntax, to V2. This means that syntax has to insert the epistemic modal above the deontic projection, in conformity with the ordering ModEpistemicP > ModDeonticP. All these effects (ordering restrictions, lack of theta-role assignment, raising verb properties) can be easily accounted for by assuming that modals are not lexical verbs taking a VP complement but are instead directly generated within the inflectional system. Wurmbrand labels this kind of structures as functional restructuring, to combine the early (1) *L'esercizio è stato dovuto riscrivere the exercise is been required to rewrite (2) *Quel tramonto non fu più potuto rivedere that sunset not was anymore can see-again Again the ungrammaticality of (i-ii) might be related to the purely functional role expressed by modals. Cinque (2004) suggests that the ungrammaticality of (2) is due to the fact that modals are base-generated outside the VP, above the functional projection VoiceP responsible for passivization.

48

Chapter Three

restructuring properties noted by Rizzi (1982) with the more recent proposal of a primitive monoclausal structure (Cinque 2006). Modal projections are then part of the inflectional system of German, sandwiched between other functional projections and dominating the VP: (95)

functional restructuring FP2 ModP modal

FP1 vP VP v

obj

At this point we can try to determine the PF-structural relation between negation and modality. This can now be restated in functional terms as the attempt to determine the position of NegP and ModP(s). We can provisionally assume that functional heads in West-Germanic languages are ordered as follows: (96)

ModP > NegP > VP

This ordering, with ModP above NegP can derive the correct word order either assuming a head final (HF) or a remnant movement(RM) analysis. Sentence (97) has the following representation under the HF account (for RM, see Moscati 2007, 2008): (97)

...dass Hans Julia nicht sehen muss

(97)’ [...[ModP [AgrOPJuliaObj [NegPnicht [VP tSub tObj sehen]]] muss]] Even if (97)’ might capture the right word order, it does not provide evidence in favour of a structure ModP > NegP. In fact, modals (in embedded clause without V-to-C movement) are always expected to surface to the right of nicht. But the same correct word order also obtains if we invert the order of projections:

Negation Raising and Scope Widening

(98)

a. ModP > NegP

49

b. NegP > ModP

ModP

NegP nicht

NegP

mogen

ModP

nicht mogen We clearly need to integrate the picture with more elements to decide between (98a) and (98b). Given that nicht is an adverbial in German, significant data come from the ordering of nicht with respect to other adverbials. Laenzlinger (2004) presents a rich series of examples regarding the ordering of adverbials in German, tracing a fine-grained topography of the syntactic space (mittelfeld) between the left periphery and the upper bound of the VP. For what concerns modality, he considered the adverb wahrscheinlich ‘probably’ to be the specifier of a modal projection expressing epistemic modality. Such an adverb precedes to the left other modal adverbs, with freiwillig ‘spontaneously’ being the lowest one: (99)

...weil Hans wahrscheinlich freiwillig einen Apfel oft gegessen hat ...because Hans probably spontaneously an apple often eaten has ‘...because frankly Hans probably spontaneously often ate an apple’

What is especially interesting is the fact that low IP adverbials as freiwillig must appear to the left of the sentential negative marker nicht29, as sentence (100), adapted from Laenzlinger, shows: (100) ... weil der Mann freiwillig diese Sonate nicht gespielt hat ...because the man spontaneously this sonata neg played has ‘...because the man didn’t spontaneously played the sonata’ 29

Nicht may be followed by other adverbials as oft and gut which express, respectively, an aspectual and a manner meaning. Such adverbs might occur to the right or to the left of negation. In this former case, it is possible that these adverbs can be replicated inside the VP (as suggested in Cinque 1999) or that NegP is higher than aspect and manner.

50

Chapter Three

On the basis of the adverbial ordering, Laenzlinger proposes the following organization of the German inflectional system: (101) SubjP > MoodPspeech.act > ModPep > ModPvol > ObjP > NegP > AspP > MannP At this point, there is sufficient evidence to support the claim that, also in German, modality is expressed in the inflectional system in a position above NegP. Given this situation, let us go back to sentence (90), repeated here as (102), and to its possible interpretations: (102)

...dass Hans Julia nicht sehen muss ...that Hans Julia neg see-inf must a. ‘...that Hans is not required to see Julia’ b. ‘...that Hans is required not to see Julia’

¬Ƒp Ƒ¬p

Given that ModP dominates NegP at PF, we need a way to capture the inverse scope reading in (102a). At this point, the same mechanism proposed for scope-shift reading in Milanese can be applied also here, raising the negative operator above ModP: (103) ... NegP2[OP¬ModP[AgrOP[JuliaiNegP[nichtVP[tHanstiV°[sehen]]]]Mod°[muss]]]

3.2.2. Double modal constructions I will now consider the possibility that the wide scope reading of negation is obtained in the opposite way: by lowering the scope of the modal operator. This could be a viable hypothesis if modals are base generated VP internally, as lexical verbs. However, we saw that they resist to such an analysis, not being able to have an external argument and showing rigid ordering restrictions. A further argument against reconstruction comes from sentences with more than one modal verb, where reconstruction is blocked. Consider for example the following sentences from Italian: (104)

Gianni deve poter parlare Gianni must can speak.inf a. ‘It is necessary that Gianni can talk’ b. *‘It is possible that Gianni must talk’

¡p *¡ p

Negation Raising and Scope Widening

(105)

Gianni può dover parlare Gianni can must speak.inf a. *‘It is necessary that Gianni can talk’ b. ‘It is possible that Gianni must talk’

51

* ¡p ¡ p

In both sentences (104) and (105), the only possible interpretation is the surface scope reading: in (104) the modal deve ‘must’ takes wide scope over the modal potere ‘to be able to’, while the inverse relation holds in (105). This phenomenon can be accounted for in different ways. A first possibility is that the higher of the two modals expresses Epistemic Modality and that it is generated in a higher functional head. A second possibility is that the inverse scope reading is excluded by derivational principles. In this case, reconstruction is blocked by the Minimal Link Condition (Chomsky, 1995) or Relativized Minimality (Rizzi, 1990): since reconstruction of the highest modal needs to create a link over the lowest modal, an intervention effect occurs and reconstruction is forbidden. This observation is not restricted to Italian, but it extends also to German. Consider sentence (106), with the two modals können and mussen: (106)

...dass ich einschlafen können muss ...that I fall.asleep can must a. ‘...that I must be able to fall asleep’ b. *‘...that I can necessarily sleep‘

¡p *¡ p

this sentence is unambiguous and only reading (106a) is possible. Again, the only interpretation is the one consistent with surface scope. The conclusion is that whenever two modal verbs are present within the same sentence, their scope is rigidly fixed by their surface position. This fact is relevant in connection with the interpretation of negative sentences. If we assume that inverse scope is derived through reconstruction of the modal, we expect that such inverse scope readings will be excluded whenever reconstruction is forbidden, as in the double modal construction in (106). However the following example shows that, although the scope of the highest modal is frozen by the presence of the second modal, negation is able to scope over it:

52

(107)

Chapter Three

a. Karl muss nicht schwimmen können Karl must neg swim can ‘It is not necessary that Karl is able to swim’

¬ ¡p

b. ... dass Karl nicht schwimmen können muss ... that Karl neg to swim able must ‘... that it is not necessary that Karl is able to swim’

¬ ¡p

In the matrix (107a), the scope of the modals können and mussen is fixed and only surface scope is possible. Now that reconstruction is blocked, the only way to derive the paraphrased meaning is to raise the negative operator above the highest modal mussen, and the same applies to the embedded sentence (107b). (108) XP Op ¬

ModP1

ModP2

muss reconstruction

NegP

können

nicht VP This kind of sentences shows that the desired scope-shift cannot be obtained by lowering the scope of the modal below negation, since reconstruction is blocked by an intervention effect. The inverse scope reading can be instead generated widening the logic scope of negation.

Negation Raising and Scope Widening

53

3.3. Split scope readings We have considered the case where the negative operator was introduced within the inflectional system by the sentential negative marker. However, sentential negative markers are not the only lexical elements able to introduce negative operators. Another interesting piece of evidence in favour of LF-Negation Raising comes from the scope of negative quantifiers interacting with modality. This kind of constructions has been discussed in Jacobs (1991) and Rullmann (1995). For Dutch, Rullmann (1995) showed that certain readings cannot be derived by covertly moving the quantifier as a whole. Consider the following sentences from Dutch: (109) Ze mogen geen eenhoorn zoeken they are allowed no unicorn seek a. ‘there is no unicorn that they are allowed to seek x’ b. ‘they are allowed to seek no unicorn’ c. ‘they are not allowed to seek a unicorn’

¬‫¡>׌‬ ¡ > ¬‫׌‬ ¬>¡>‫׌‬

(110) Ze hoeven geen verpleegkundigen te ontslaan they need no nurses to fire a. ‘For no nurse x it is the case that it is necessary to fire x’ ¬‫>׌‬ b. #‘It is necessary that they fire no nurses’ > ¬‫׌‬ c. ‘It is not necessary for them to fire a nurse’ ¬ > >‫׌‬ Notice that here we don’t have negative concord and negative quantifier can negate a sentence as the only negative elements. In both sentences (109) and (110) the negative operator introduced by the quantifiers creates complex interactions with the modal and three different interpretations are possible. If we consider geen eenhoorn ‘no unicorn’ in sentence (109), it might be interpreted taking either wide or narrow scope over the modal mogen. This results in the interpretations (109a-b). (109a) describes a situation where there exists no unicorn, such as it is possible to search for it. This interpretation is given by the narrow scope of the modal below the negative quantifier. Interpretation (109b) is true instead in the situation where it is permitted to search for unicorns, although they do not exist. In this case the negative quantifier has to be interpreted within the scope of modality. The last interpretation is the most natural and the most interesting for our purposes. According to reading (109c), negation takes scope over the modal and the sentence expresses a prohibition to search

54

Chapter Three

for unicorns which may not exists. Here the quantifier doesn’t take an existential interpretation while negation is interpreted with the widest possible scope. This reading shows that in ‘geen eenhoorn’ quantification and negation have independent scope. Here the negative feature might be raised in isolation to take scope over the modal. This reading becomes even more prominent if we use a modal as ‘hoeven’, which takes narrow scope under negation. In this case, the negative operator must obligatory raise (with the quantifier (110a) or without it (110c)) to take scope over the modal. The interpretations in (109c)-(110c) are referred to as the split scope readings and the solution proposed by Jacobs (1991) and Rullmann (1995) is to consider quantifiers as ‘geen’ as complex elements which include a quantifier plus a negative operator30. If this analysis is correct, the mechanism proposed for deriving the inverse scope reading can straightforwardly apply, raising the negative operator in a position dominating modality: (111)

XP Op ¬ ModP AgrOP

mogen

geen t¬ ∃ The same kind of data can be replied for German, as reported in Zeijlstra and Penka (2005). If Negation Raising is possible in German, we expect that also in this language the negative feature may be separated from the object quantifier and interpreted with wide scope over modality. Sentence (112) confirms this expectation:

30

For a different analysis, see de Swart (2000).

Negation Raising and Scope Widening

(112)

Du musst keine Krawatte anziehen. You must no tie wear a. ‘It is not required that you wear a tie’ b. ‘There is no tie that you are required to wear’ c. ‘It is required that you don't wear a tie’

55

¬> >∃ ¬∃ > > ¬∃

Among the different readings available for this sentence, the most salient is the one given in (112a), which corresponds again to the split-scope interpretation. Once more, the interpretation (112a) shows us that also the negative quantifier keine is composed by two elements: a negative operator and a quantifier. Given this analysis, reading (112a) can be derived by raising the negative feature in isolation above ModP.

3.4. A refinement of the notion of modality In order to account for inverse the wide scope readings, the solution I propose is to admit a covert process able to widen the scope of the negative operator by raising it in a higher structural position above modality. At this point, we could ask where this position is. Anticipating the conclusions, a precise projection cannot be individuated, since the evidence based on scope interaction is insufficient. However, this doesn’t mean that the analysis cannot be refined. I will consider next different kinds of modality, in order to detect if there exists an upper bound to the scope of the negative operator.

3.4.1. Epistemic, Alethic and Deontic modality In the previous section we considered modality as a whole and all the examples discussed were involving deontic modality. It is time now to consider also different kinds of modality, in order to check if one of them blocks Negation Raising. In this case, we could individuate with a certain accuracy the highest structural position triggered by Negation Raising. In what follows, I’ll consider three different kinds of modality. Let me briefly introduce their semantics. Modality in general refers to notions such as ‘possible’ and ‘necessary’ truth, but the domain of the possible worlds where the truth conditions are calculated is variable. In this respect, some distinctions between different kinds of modality have been proposed. I will limit the discussion here to alethic, epistemic and deontic modality (Lyons 1977; Palmer 1986). For what concerns their different semantics, alethic modality is considered to quantify over all possible worlds, regardless to the speaker’s

56

Chapter Three

beliefs. In this respect, alethic necessity/possibility is similar to logic necessity/possibility. Conversely, epistemic modality is dependent to the speaker’s model of the world, expressing “the status of the proposition in terms of the speaker’s commitment to it” (Palmer, 1986: 54-5). The last kind of modality, the one we have been implicitly using so far, is usually referred to as deontic modality, and it is used to indicate necessity and possibility with reference to moral beliefs, obligations or norms in general. Differences in meaning might be also syntactically encoded and become visible once different modal forms are employed. For example, Danish marks the distinction between epistemic and alethic modality using two distinct forms: kunne for alethic modality and skulle for epistemic modality (Vikner 1988). These two modals occupy different structural positions, with kunne appearing lower than skulle. Also the distinction between epistemic modality and deontic modality might be expressed. English, for example, encodes this distinction by using the two different forms may and can. As we may expect, there are also languages which, istead of using modal auxiliaries, uses instead affixes on the verb. Also in these cases, we find different morphemes hosted in dedicated functional projections. An example is Una, an Irian Jaya language of New Guinea, which shows two different suffixes for epistemic and deontic modality. Consider the following pair from Louwerse (1988): (113) a. Er bin-kwan-de-darib she go-fut-3s-ep.mod b. Ni buk-ti-nyi I sit-deon.mod-pres Una shows a difference in the linear order between suffixes expressing deontic and epistemic modality with regard to suffixes expressing tense. In (113a), the epistemic particle –darib results external to the tense particle kwan. In (113a) instead, the root modal particle –ti appears internally to the present tense suffix –nyi. On the basis of the ordering of modals and affixes, Cinque (1999) traces a topography of the modal projections in which epistemic modality is the topmost category: (114) Epistemic Mod > Alethic Mod > Root/Deontic Mod

Negation Raising and Scope Widening

57

At this point, what we want to know is whether one of these different modalities blocks wide scope of negation. In the case of deontic modality, it is less controversial that negation may easily take wide scope over it (Cinque 1999, Drubig 2001). In fact, all the modal sentences presented in previous sections are best interpreted as instances of deontic modality. Consider, for example, the two following sentences from Italian, with the modal dovere ‘must’ and potere ‘can’: (115) tu non devi prendere l’ aspirina (...se non vuoi) ¬ P you neg must take the aspirin ‘It is not necessary that you take the aspirin. (If you don’t want to)’ (116) tu non puoi prendere l’ aspirina (...se sei allergico) ¬¸p you neg can take the aspirin ‘It is not possible that you take the aspirin. (if you have intolerance)’ In sentences (115) and (116), the modal expresses prescription or permission, two cases falling into the definition of deonticity. In (115) the modal dovere falls within the scope of negation, and the sentence expresses the lack of the necessity of p. In (116), where the modal indicates permission, again negation takes wide scope resulting in a prohibition. Similar examples might be easily construed for many languages. This shows that deontic modality doesn’t oppose any resistance to be interpreted under negation. This means that there is no reason to believe that the highest interpretable site for negation has to be in a NegP below Deontic ModP. Consider now alethic modality in the example (117): (117) Il numero dei pianeti del nostro sistema solare non deve essere per forza dispari ‘the number of planets of our solar system needs not to be odd’ ¬Ƒp This sentence can be considered ambiguous between various readings, however one of the possible interpretations is surely the one reported in (117). It says that there is no logic necessity on the parity of planets and it is perfectly plausible that new planets will be discovered. In sentence (117) negation can be interpreted (and actually this is the favoured interpretation) with wide scope over alethic necessity. Again, no logic constraint seems to exist on the wide-scope of negation over alethic modality. It is clear that the negative operator may scope both on Alethic

58

Chapter Three

and Deontic modality. This means that a NegP, able to host a negative operator, might exist on the top of these two modal projections: (118) EpistemicModP>NegP2>AlethicModP>NegP1>Root/DeontModP. In (118) only epistemic modality it is not roofed by a negative projection. Here things are slightly more complicate and I will consider epistemic modality at length in the next paragraph.

3.4.2. Epistemic modality Authors like Cinque (1999) and Drubig (2001) claim that epistemic modality cannot fall under the scope of negation. Drubig in particular takes the following examples as indications that scope readings between epistemic modals and negation are unidirectional and only epistemic modality can scope over negation: (119) a. John must not be at home. b. John may not be at home.

ƑEP ¬p ; * ¬ƑEPp ¸EP¬p ; * ¬¸EPp

However, his argument appears to be weak, given that the inverse scope readings may be ruled out for different reasons. In fact, as noted by many authors (Rullmann 1995, Hoeksema 2000 for Dutch and German, Stowell 2004, von Fintel and Iatridou 2003), modal verbs may be sensitive to polarity. And in fact, may and must can be considered as positive polarity modals, given that they, when within the scope of negation, have to be replaced by can and need. Consider the following examples ((120) from von Fintel and Iatridou (vF&I) and (121) from Stowell (2004)): (120) a. John need not be home (...he might be at work) b. John can’t be at home (...he must be at work).

¬ƑEP p ¬¸EP p

(121) That can’t be a Dodo bird: they are extinct

¬¸EP p

Modals as need and can are the equivalent of must and may in negative contexts and are forced to be interpreted with narrow scope. The situation now reverses and the problematic examples in (119), if carefully considered, can be taken instead as a confirmation that negation is indeed able to scope over epistemic modality, given that even specialized forms exist.

Negation Raising and Scope Widening

59

This conclusion is further supported by different kinds of evidence. For example, quasi-modal verbs in English are able to convey epistemic modality and they are perfectly fine under the scope of negation (Papafragou 2006): (122) a. John does not have to be the prime suspect. b. John is not necessarily the prime suspect.

¬ƑEPp ¬ƑEPp

Moreover, even in languages where the modal paradigm is poorer than English, it is possible to see that epistemic modality can be interpreted with narrow scope. In Italian, for example, there is no dedicated form for epistemic modality. Nevertheless, it is possible to recognize when a modal is used with an epistemic meaning. Zagona (1990) noticed that deontic modals may trigger a future shifted interpretation of the eventuality time. Intuitively, this is due to the fact that permissions can be asked only for actions not started yet, as in the following two sentences: (123) a. John can run b. John can stay in class

FUTURE-SHIFTED FUTURE-SHIFTED

In sentences (123) the event time results t+1 with respect to the modality time t. This means that if we utter (123a) in order to gives John the permission to run, the act of running can only be successive to the time t. Instead, in the case of epistemic modals, we have another possibility. Here a simultaneous reading, in which the event time coincides with the modality time, is possible: (124) John could be in class today

SIMULTANEOUS OR FUT-SHIFTED

This fact indicates that when a simultaneous reading is obtained, the modal has to be interpreted with an epistemic value. At this point we have a diagnostic to reveal the kind of modality also in Italian. Consider the examples in (125) and (126): (125) a. Giovanni può andare al mare PERMISSIVE MOD. ÆFUT-SHIFTED Giovanni can-3s go to-the beach ‘Giovanni can go to the beach’

60

Chapter Three

b. Elvis può essere vivo Elvis may-3s be.inf alive ‘Elvis may be alive’

EPISTEMIC MOD.

Æ SIMULTANEOUS

(126) a. Giovanni deve andare al lavoro OBLIGATION MOD.ÆFUT-SHIFTED Giovanni must-3s go to-the work ‘Giovanni must go to work’ b. Elvis deve essere vivo EPISTEMIC MOD.Æ SIMULTANEOUS Elvis must-3s be.inf alive ‘Elvis must be alive’ In the b. examples above we have the possibility to interpret the sentence without a future shift. This means that the proposition can be true in some possible world at the time of utterance,. We have then an interpretation of the modal which is epistemic. Even the same sentence may show an epistemic/root modality flipflop, resulting in a time shift. Consider the same sentence in the two contexts (127) and (128) (127) SITUATION A. Permissive. Future shifted We are in a music school. Due to financial problems, there are not enough music rooms for the proper training of all the students. To face this shortage, a music teacher proposes the following solution: I pianisti più talentuosi potrebbero esercitarsi di meno ‘the most talented pianists could practice less’ (in the next future) (128)

SITUATION B. Epistemic. Simultaneous We are in a conversation between critics. They are discussing why some pianists are better than others. One of them thinks that excessive training impairs personal creativity and utters: I pianisti più talentuosi potrebbero esercitarsi di meno ‘the most talented pianists could be so because they practice less’ (in the present)

When the context favours a permissive reading of the modal verb, we have the future-shift effect, shown in (127)-(128). When an epistemic

Negation Raising and Scope Widening

61

interpretation of the modal is instead favoured, the event time is anchored to the modal time (128). With this distinction in mind, let us go back to negative sentences and consider a situation in which you enter into the house and you try to discover if the sauce you left on the stove is ready. We are clearly in an epistemic situation. Since there is no smell, you may utter (129): (129) Il sugo non può essere pronto: non ne sento l'odore the sauce neg can-3s aux ready neg of-it smell the scent ‘The sauce can‘t be ready yet: I don’t smell it’ As expected, this sentence can be interpreted only simultaneously, with modal evaluation coinciding with eventuality time. This confirms that the modal is used epistemically. What about the interpretation of negation? Can it receive wide scope? This is actually the only interpretation, reported in (129a): (129) Il sugo non può essere pronto: non ne sento l'odore a. ‘It is not possible that the sauce is ready’ b. ‘It is possible that the sauce is not ready’

¬ ¸EPp # ¬ ¸EPp

It seems than that also epistemic modality can fall within the scope of the negative operator and that an interpretive site for negation exists also on the top of the functional projection hosting epistemic modals: (130) NegP3 > Epistemic ModP > NegP2 > Alethic ModP > NegP1 > Root/Deontic ModP.

Conclusions In this chapter we consider the problem of syntactic variation from a different angle, looking not only at the clausal position of the sentential negative marker but also at the scope of the negative operator. These two facets have to be kept distinct on the basis of the fact that logic scope doesn’t coincide with the surface position of negation. The scope of negation can be narrower and reconstructed in a lower structural position, as in the cases of attraction of negative or widened through Negative Raising. We ultimately need a link between the surface structure and the logical form and this link cannot be direct and isomorphic. For this reason, I suggest an operation able to raise the negative operator. Such an operation

62

Chapter Three

has a welcome consequence: we don’t need to assume that negation have different scope in languages using different surface positions. This means that we have a way to account for the fact that in Milanese (low) and Irish (high) the negative operator might have exactly the same scope. A second issue concerns the highest scope site available for interpreting negation. A way to reveal it is to consider different kinds of scope bearing elements. For this reason, I refined the notion of modality in order to distinguish between epistemic, alethic and root/deontic modality. We saw that, in many cases, negation might take wide scope even on epistemic modality, which is considered to be hosted in the highest positions of the inflectional system (Cinque 1999). This means that a NegP might be projected above the whole IP. We said that the operation of Negation Raising allows us to consider the scope of negation in Milanese as wide as in Irish. At this point, remember that there are many languages showing negation in the CP. Given this state of affairs, we cannot exclude that the negative operator is interpreted in a position as high as the topmost clausal projection. In the next chapter I’ll follow this idea and I’ll propose that the CP is always involved in negative sentences.

CHAPTER FOUR CLAUSE TYPING AND NEGATIVE CHAINS

A way to deal with the scope ambiguities presented in the previous chapter is to admit a covert operation able to provide an additional interpretive site for negation. In order to distinguish it from other kinds of covert movements not related to scope-shift, the operation has been presented as follows: (9) Optional Negation Raising is possible when: Į c-command ȕ at PF but ȕ scopes over Į. where: Į is a logic operator and ȕ is the logic operator ‘¬’ This formulation characterizes negation raising as a last resort operation admitted only when an alternative reading has to be generated. It ultimately permits to account for the different readings of sentences such as (102), repeated in (131) below: (131)

...dass Hans Julia nicht sehen muss ...that Hans Julia neg see-inf must a. ‘...that Hans is not required to see Julia’ b. ‘...that Hans is required not to see Julia’

¬Ƒp Ƒ¬p

An open question is whether it is possible to identify a structural position where the negative operator is displaced. In order to try to answer this question, we could look at the relative scope of negation and other operators, adopting a logic similar to the one in Beghelli and Stowell (1997) for quantifiers. In the previous chapter I considered the interaction between negation and different kinds of modal operators, showing that negation might take wide scope over each of them. This suggests that Negation Raising might trigger a position in the topmost portion of the IP. However, it is also possible that even higher positions within the complementizer could be

64

Chapter Four

involved. In this chapter I’ll explore this possibility, making the strong claim that a link between NegP and the CP is always created. Let me briefly recapitulate some of the conclusions from previous chapters. Starting from the observation that inverse scope readings are allowed, I proposed the application of Negation Raising in order to derive scope-shift. This operation widens the scope of the negative operator, with the welcome result that its scope is not tied to cross-linguistic variations. The second conclusion is that scope can be widened over high (in the structural sense of Cinque 1999) IP operators which occupy a fixed position. This is shown by considering inverse scope over different kinds of modality. The third is a typological observation about the surface realization of sentential negative markers, which can be realized within the complementizer in different languages. In this chapter I will try to organically account for these three main conclusions by suggesting a mechanism which capitalizes on the notion of clausal typing proposed by Cheng (1991). This move will differentiate the current proposal from previous ones, given its emphasis on the role of the complementizer system.

4.1. Clausal Typing Let me take the move from the contrasts in (132) and (133) (132) a. Mary believes that he will come b. *Mary believes whether he will come (133) a. Mary wonders whether he will come b. *Mary wonders that he will come these pairs show that lexical matrix verbs as believe or wonder select a sentential complement introduced by the appropriate complementizer. In (132) the verb believes is not capable to select an indirect interrogative clause and the complementizer that is chosen while wether is excluded. The situation reverses in (133), where the matrix verb wonder demands for a complementizer able to introduce an indirect question. These examples show that different complementizers are chosen in accordance with the selection requirements (Grimshaw 1979; Pesetsky 1991) of the matrix verb. This can be expressed by saying that verbal forms such as say and wonder respectively select for a complementizer carrying the feature [+question] or [+declarative]:

Clause Typing and Negative Chains

65

 (134) a. say Æ b. wonder Æ

CP

[[+decl] ... CP [[+Q] ...

In this respect, [+decl] and [+Q] are s-selected and can be considered as typing features. Also in the case of negation, there are reasons to believe that [+neg] is indeed a typing feature that may (and sometimes must) be s-selected by the main clause. Let us consider again the verb doubt and the distribution of NPI in the embedded clause. The examples (70b-c), repeated here in (135), show that doubt cannot directly licence NPI objects, but it must first select a clausal complement (135) a. I doubt that anybody will come b. *I doubt of anybody In this case the licensor for the NPI anybody is not directly the matrix verb itself but instead the complementizer of the embedded equipped with a negative feature: (135)’ I doubt Force[that [+neg] IP[ anybody will come]] This analysis is confirmed by the existence of languages such as Basque, which overtly use a special negative complementizer when the subordinate clause is introduced by verbs as ‘deny’. Look again at the examples (60) and (61a) repeated here in (136): (136) a.

CP1

[

CP2 IP

[ [Galapagoak muskerrez beterik] [daudela]] diote ] Galapagos lizards-of full are-that say-they ‘They say that the Galapagos are full of lizards’

b. CP1[ Amaiak CP2[ IP[inork gorrotoa] [dionik ]] ukatu du] Amaia anyone hated has-comp-neg denied has ‘Amaia denied that anybody hated her’ In (136a) the verb diote ‘they say’ selects for a declarative embedded clause, while in (136b) ukatu ‘denied’ selects instead for a negative one. This is overtly expressed by the choice of two different complementizers, enik[+neg] and ela[-neg]. Such examples show that [+neg] is actually a typing feature that can be hosted in the topmost position of the clause.

66

Chapter Four

Now this observation can be related to the proposal in Cheng (1991). She hypothesizes that the presence of a typing feature is a requirement in force on every clause and this is expressed in the statement (137): (137) Clause Typing Hypothesis Every clause needs to be typed. this hypothesis is aimed at capturing the intuition that syntax possesses formal means to indicate the type of clause, it being a question, a declarative, an imperative, a negative sentence and so on. This idea results as less controversial, but has actually never been wholly developed for negative clauses. I’ll try to do that by reconsider the general definition in (137) to the light of a ‘Split CP’ model. In a layered structure as in Rizzi (1997), the topmost position ForceP is the projection devoted to signal the type of the clause. Consider again the different projections which articulate the complementizer: (138) FORCE

(TOP*) INT (TOP*) FOC (TOP*) FIN

given the structure in (138), the clause typing hypothesis can be restated in the following terms: (139) Clause Typing Hypothesis (Split CP) ForceP must host an appropriate typing feature Assuming that a negative feature can be a typing feature, we have to determine how ForceP can be provided with a negative value. The case of Basque shows that lexical insertion of an overt complementizer can fulfil this requirement. However this cannot be considered as the only way and other mechanisms can do the job as well, with some important consequences on the syntax of negation in general. I’ll consider these in section §4.3.

4.2. Lexical Insertion One way to satisfy the requirement in (139) is through the direct insertion of a typing feature, carried by a lexical element. We already saw that certain complementizers are equipped with different typing features, as in the case of English whether[+Q] or Basque enik[+neg], and actually all the cases in chapter three can be considered as examples of a lexical insertion of a negative feature in CP.

Clause Typing and Negative Chains

67

 Notice that the choice of the complementizer is not always determined by selectional requirements. In fact, negative complementizers may be inserted in matrix and embedded clauses without being s-selected by the matrix verb, as in the case of the Irish example reported again below (140) Creidim nach gcuirfidh sí isteach air I-believe neg-comp put-fut she in on it ‘I believe that she won't apply for it’ in (140) nach is embedded under a psych verb, which doesn’t require a negative feature in the complementizer of the embedded clause. This example shows that s-selection is not a necessary requirement for having negative complementizers. In fact, negative complementizers might type also matrix clauses. We saw that Irish types them as well through the insertion of negative complementizers: (141) nìor (char) ól tú a gcuid uisce comp-neg-past drank you their portion water ‘You didn't drink their water’

(McCloskey 2001)

Irish examples as (140) and (141) give us very clear indications on typing through lexical insertion. However, also languages lacking negative complementizer might insert a negative feature in ForceP. We saw in chapter two that also languages as Italian, which do not have negative complementizers of the Irish kind, have means to mark the complementizer with a negative feature. In fact, elements as mai ‘never’ might be inserted preceding the complementizer che: (142) Mai che ci vedesse nessuno never that us see.su anybody ‘Nobody ever saw us’ (142)’

ForceP

[mai[neg] che IP[ci vedesse VP[ nessuno ]]]

here mai appears on the left of the complementizer che31, and from this position takes scope over all the remaining sentential material. 31

The presence of a higher projection ȈP (Laka 1990) on the top of ForceP is another possibility. This is suggested by positive elements which can also appear to the left of che:

68

Chapter Four

Summarizing the discussion held so far, it is clear that the typing requirement can be fulfilled by inserting an appropriate lexical element able to type the clause, being it a matrix or a subordinate and that Clause Typing in (139) is not a principle uniquely dependent form S-selection. The On the basis of the existence of negative complementizers or lexical elements with a negative feature in the topmost projection ForceP, negation can be considered to belong to the set of appropriate typing features. But what happens when no lexical element appears in typing position? One possibility is to have a null-element as, for example, in sentences introduced by doubt. Another possibility is that a negative feature (and typing features in general) is copied in ForceP, satisfying clause typing by a mechanism similar to Agree.

4.3. Typing through Agree If typing is a general syntactic requirement, lexical insertion of overt material doesn’t suffice to cover all the possible cases. Many languages, in fact, lack negative complementizers and do not show any morphological manifestation of a negative feature in Comp. Should we assume that the Clause Typing Hypothesis stated in (139) doesn't apply in these cases? Interrogative sentences can give us some indication in this respect. Let me leave aside, for the moment, the case of negation and turn our attention to sentences where Force is typed by a question feature [+Q]. To detect the typing feature, we can subordinate a clause under a matrix verb which introduces an indirect question. Consider for example the Italian pair in (143): (143)

a. Io penso che tu abbia visto il film I believe that you have seen the movie ‘I believe that you have seen the movie’ b. *io penso se tu abbia visto il film I think whether you have seen the movie

(1) Sì che Gianni verrà Yes/Sure that Gianni will-come ‘Gianni will come for sure’ (2) Certo che Gianni verrà certainly that Gianni will-come ‘Gianni will certainly come’

Clause Typing and Negative Chains

69

 Example (143a) shows that the complementizer se ‘whether’ cannot be used in the embedded clause if the matrix verb does not select an indirect question. This is due to the fact that se carries a [+Q] feature, incompatible with the declarative feature required by the matrix verb. If we change the matrix verb, sentence (143a) becomes perfectly acceptable (144) Io mi chiedo se tu abbia visto il film I myself wonder whether you have seen the movie ‘I wonder whether you have seen the movie’ here the verb chiedere, which in its reflexive use is similar in meaning to ‘wonder’, selects for an embedded clause marked with [+Q]. For this reason, the interrogative complementizer se is allowed. Interestingly, such interrogative feature does not need to be inserted only by an interrogative complementizer, but a wh- element might do the job as well: (145) Io mi chiedo cosa tu abbia visto I myself wonder what you have.su seen ‘I wonder what you have seen’ A point worth noticing is that, in both (144) and (145), the feature [+Q] is not directly inserted in the topmost position ForceP. It can easily be shown that se ‘whether’ and cosa ‘what’ are lower, given the possibility to have a Topic to their left: (146) a. Io mi

chiedo, A GIANNI, se dovremmo dargli quel regalo

I myself wonder to GianniTOP whether should-we give-him that present

‘I wonder if we should give that present TO GIANNI’

b. Io mi chiedo, A GIANNI, cosa dovemmo dargli I myself wonder to GianniTOP what should-se give-him ‘I wonder what we should give TO GIANNI’ In both the examples in (146), the topmost visible position is TopP and the [+Q] feature is inserted in lower position: IntP32 in (146a) or FocusP in (146b). However, TopP cannot be the highest structural node, since the matrix verb c-selects for a clausal complement and not for a Topic. 32

Following Rizzi (2001) I will assume that se occupies a projection IntP between Force and Focus.

70

Chapter Four

At this point we have on one side the need to satisfy the requirement of the matrix verb in the embedded ForceP and on the other side the presence of a suitable [+Q] element in a lower position. A way to solve the problem is to permit ForceP33 to probe its search domain looking for a typing feature: (147) a.

b.

ForceP Typing [+Q]

TopicP a Gianni

IntP se [+Q]

FocusP cosa [+Q]

in (147a) Force needs to be typed and searches within its domain for a suitable feature. In the case of example (146a) it is satisfied by the feature [+Q] inserted with the interrogative complementizer. In the case of (146b), the typing requirement is instead satisfied by the [+Q] feature of the Whelement. In both cases, a typing feature is copied in ForceP as in (147b)34. Let us now turn to negative features. We saw that matrix verbs can select negative clauses introduced by a negative complementizer. This forces us to include [+neg] in the set of the typing features. Now the mechanism in (147) can be straightforwardly applied to embedded clauses. We said that negative markers35 typically carry a [+neg] feature which might be used to satisfy clause typing, and also negative constituents can provide a typing feature, similarly to wh- constituents. In Italian, for example, clause typing could be satisfied by the negative feature carried by sentential negative marker (148) Gianni non vede Piero Gianni neg sees Piero ‘Gianni doesn’t see Piero’ 33

Or FinP in non finite sentences. The fact that only one [+Q] feature needs to be checked in ForceP may also account for the impossibility to have cosa and se within the same clause, as shown below: cosa dovremmo regalare (1) *io mi chiedo, A GIANNI, se I myself wonder to Gianni whether what should-se give-him 35 The clause typing hypothesis states that each clause has to be typed and this should be intended as “typed by at least one appropriate feature”. In fact nothing prevents to have interrogative negative clauses. 34

Clause Typing and Negative Chains

71

 (148)’

ForceP ... Typing [+neg ]

NegP non [+neg]

or by a negative quantifier, in this case an n-word in preverbal position (see chapter five) (149) Nessuno vede Piero nobody sees Piero ‘nobody sees Piero’ (149)’

ForceP ... Typing [+neg ]

AgrP Nessuno [+neg]

This will explain how typing obtains for languages which manifest a sentential negative marker inside the IP layer, such as Italian, English, German and many others. If an operation such as that in (147) is allowed, we have duplication in ForceP of the negative feature visible inside the IP. In this way, requirements imposed by clause typing are satisfied. At this point, we have sketched the contours of a general mechanism which creates a link between ForceP and a lower NegP, with two instances of the same negative feature. Could we relate this mechanism to negation raising? A way to do so is to reduce scope shift to the problem of determining which one of the two instances of the negative feature is interpreted. Consider again the Milanese example (83), repeated here as (150), and its possible interpretations: (150) El gà de studià no s.cl must of to-study neg a. ‘he is required not to study’ b. ‘he is not required to study’

Ƒ¬p ¬Ƒp

Sentence (150) has the following surface structure, with the negative marker no hosted in NegP just above the VP (150)’ ForceP[ AgrP[Elj' gakModP [OPƑ tk TP[de studiàv NegP[no[+neg]VP[ tj tv ]]]]]] Being [+neg] a typing feature, it can met the probe send by ForceP and its value is copied there:

72

Chapter Four

(150)’’ ForceP[[+neg]AgrP[Elj'gakModP[OPƑtk TP[de studiàvNegP[no[+neg]VP[tj tv ]]]]]] This representation is passed to the interface and either the negative feature in ForceP or the one in NegP could be interpreted. This means that, whenever scope-shift readings have to be generated as in (150)b, the higher negative feature is interpreted at the interface. Negation Raising might then be reformulated in terms of chains, where the chain is created for clause typing purposes: (151)

Covert links of a negative chain are interpretable only if they generate alternative readings

Given (151), we capture the fact that scope-widening is allowed only for scope shifting purposes. Negation Raising is then reduced to the matter of determining which link of the chain is interpretable. This means that the operation can be considered, in a certain sense, as parasitic on the formation of the negative chain.36 Appling the filter in (151), we have a way to decide which one of the negative features presented in (150)’’ is interpreted. Without any other scope bearing element, it is the overt negative feature in NegP to be interpreted, with the highest one used only for typing. This is also what happens for the narrow scope reading of negation in (150a): (150) a.ForceP[[+neg]AgrP[Elj'gakModP[OPƑtk TP[de studiàvNegP[noi[+neg]VP[tj tv ]]]]] However, when a different reading must be generated, it is possible to interpret also cover instances of the negative feature in order to have scope-shift. In this case, the topmost link of the chain receives interpretability at LF and we have the alternative reading in (150b): (150) b.ForceP[i[+neg]AgrP[Elj'gakModP[OPƑtk TP[de studiàvNegP[no[+neg]VP[tj tv ]]]]]

4.4. Valuation of Typing Features I wish to conclude this chapter by considering a particular way to build feature chains which is slightly different from the feature copying mechanism presented in Chomsky (2001). In the previous paragraph, I presented the possibility to cast negation raising in terms of an LF choice 36

A mechanism in this spirit has been proposed in Hornstein (1995) for QR.

Clause Typing and Negative Chains

73

 between possible interpretable features. However, this is not the only option available and it exists also an alternative in order to avoid the generation of ambiguous structures at LF. According to this view, different representations will be generated in syntax: it is possible that the syntactic engine creates alternative but unambiguous logic representations. If we adopt this view, we must clean the syntactic output from duplication of interpretable features. This means that chains of features are created with only a single interpretable link. This is the proposal in Pesetsky and Torrego (2004), which can be also exploited in order to formalize negative chains. The most important innovation in the system of Pesetsky and Torrego is the enrichment of the feature typology presented in Chomsky (2001). According to this last view, semantic Interpretability and syntactic Valuation go together and interpretable features are also valued in syntax, resulting in the two possibilities given below: (152) Typology of feature in Chomsky (2001) a. F1 i[val] b. F1 u[] In (152a) the feature F1 is ‘complete’ and it does not start any syntactic process by itself. It can only be linked to other instances of F1 able to probe in need of interpretability and valuation, as the one in (152b). Basically, interpretability and valuation go togheter in the typology in (152). However, Pesetsky and Torrego (2004) pointed out that while interpretability is a notion which mainly concerns the semantic interface, valuation is instead a syntactic notion. For this reason, there is no necessity to assume that the two things go together. It is possible, for example, to have interpretable features which need to be syntactically valued. To me, this is equivalent to have a null-operator which needs syntactic licensing. On the other side, it is also possible to have uninterpretable features which are required only by the syntactic component. According to this intuition, the possibilities in (153) must be added to the ones in (152) (153) Extended Typology in Pesetsky and Torrego (2004) a. F1 i[] b. F1 u[val]

74

Chapter Four

Derivationally, we have three kinds of features which probe, looking either for interpretability (153b), valuation (153a) or both (152b)37. We can now look at how this mechanism, considering how it could be concretely implemented in the case at interest. Let us consider again sentence (150), repeated as (154) (154) El gà de studià no s.cl must of to-study neg a. ‘he is not required to study’ b. ‘he is required not to study’

Ƒ¬p ¬Ƒp

The wide scope reading (154b) can be accounted assuming that an i[] feature is inserted in ForceP, probing for a valuation: (154) a. ForceP[i[]AgrP[Elj'gakModP[OPƑ tk TP[de studiàvNegP[no u[val]VP[tj tv ]]]]] once the evaluated but uninterpretable instance of the negative feature introduced by the sentential negative marker is encountered, its value is copied in ForceP, satisfying clause typing and providing an interpretable site for negation with wide scope over modality: (154) a’.ForceP[i[val]AgrP[Elj'gakModP[OPƑ tk TP[de studiàvNegP[no u[val]VP[tj tv ]]]]] This is basically a mechanism able to introduce a negative operator by transforming a formal feature into an interpretable one. As we said, this option must be allowed only to derive scope-shift. Otherwise, the system privileges the insertion of features interpretable and valued, in order to have surface scope as the default. This is the case of the surface scope reading in (154b), which corresponds to the derivation below: 37

All the possible outcomes of Agree are the ones below: (1) a. F1 i[val] Å F1 i[val] b. F1u[val] Å F1 u[val] c. F1 i[val] Å F1 u[val] d. F1u[val] Å F1 i[val] Configurations (1a) and (1b) are problematic at the semantic interface, since (1a) is ambiguous and (1b) is vacuous. In fact no interpretable feature is visible in the chain (1b) and two interpretable features are instead visible in (1a). These configurations have to be excluded by a filter on chains, as proposal in Brody (1997) based on the thesis of radical interpretability. See also Moscati (2006a, 2008).

Clause Typing and Negative Chains

75

 (154) b. ForceP[u[]AgrP[Elj'gakModP[OPƑ tk TP[de studiàvNegP[no i[val]VP[tj tv ]]]]] (154) b’. ForceP[u[val]AgrP[Elj'gakModP[OPƑ tk TP[de studiàvNegP[noi[val]VP[tj tv ]]]]] The representations in (154a’) and (154b’) ultimately show the two possible feature-chains. Each chain has only one interpretable feature and each of them corresponds to one of the two possible readings for sentence (154).

Conclusions In this chapter I have implemented the idea that wide scope readings of negation can be derived using an independently created syntactic chain. In doing so, I tentatively reformulate Negation Raising in terms of interpretable links of a single negative chain. The question is what motivates the creation of such a chain. I proposed that this can be due to the satisfaction of the requirement proposed by Cheng (1991), according to which every clause must be specified in relation to its type. This requires the copy of a typing feature in the topmost position of the clause. At this point, our system only needs to decide which link is interpreted at LF. Here we have two possibilities. The first is to resolve this ambiguity at LF. This means that syntax creates ambiguous representations which are later disambiguated by the interpretive component. An alternative view is instead to assume that only one interpretable feature arrives at LF and that two representations are created in syntax. This last option is the one which I explored in the ending section, using an extended feature typology as in Pesetsky & Torrego (2004). Having tried to propose a technical implementation of Negative Raising, I’ll turn next to other elements which arguably carry a negative feature. After having considered negative complementizers and inflectional negation within NegP, the last category remained to be examined is the one of negative DPs, which will be the topic of the remaining two chapters.

CHAPTER FIVE NEGATIVE ARGUMENTS

In previous chapters, we considered only negative elements inserted in functional positions above the VP. For this reason, negative nominal elements have been only marginally mentioned. It is time, now, to look at negative nominal syntagms in argument and adjunct positions and these last two chapters will be entirely dedicated to negative DPs. Again, linguistic variation will be our starting point. Since it is impossible to consider negative nominal elements without taking into account the phenomenon of concord, we can begin the discussion by introducing the basic distinction between negative concord and non negative concord languages. In languages belonging to the last group, negative DPs can express negation by themselves. In Standard English, for example, negative quantifiers as nobody and nothing might be the sole negative elements in the clause. In languages showing negative concord, instead, the presence of the sentential negative marker is required to license negative arguments. This results in a reduplication of negative morphology which is usually referred as negative concord. The picture is further complicated by the fact that negative concord languages do not form a homogeneous class, since some languages show pre-post verbal asymmetries while others don’t. For this reason, we can use the label asymmetric concord to indicate asymmetries in the interpretation of negative constituents and strict concord in cases where no pre/post verbal asymmetry exists. The differences between non concord, concord and asymmetric concord are illustrated below with examples taken from languages belonging to the different groups: (155)

a. I saw nobody b. Nobody saw me c. I didn’t see nobody

(English)

78

Chapter Five

(156) a. *Ja videl ni-kogo I saw nobody

(Russian)

b. *Ni-kto zvonil nobody called c. Ja ne videl ni-kogo I neg saw nobody ‘I saw nobody’ (157) a. *Ho visto nessuno aux seen nobody

(Italian)

b. Nessuno ha visto me nobody aux seen me ‘Nobody saw me’ c. Non ho visto nessuno neg aux seen nobody ‘I saw nobody’ In the English examples (155a-b) the negative quantifier nobody is the sole negative element in the sentence and it suffices to make the sentence negative without the negative marker not or n’t. When not is introduced, as in (155c), the sentence has a double negation meaning similar to the paraphrases ‘it is not true that I saw nobody’, where the two negations cancel each other out. The situation is instead different in Russian, which belongs to the group of strict negative concord languages. In these languages, in fact, negative elements always need the presence of the sentential negative marker and double negation is impossible. This is illustrated in the examples (156) from Russian (Brown 1999). Examples (156a-b) show that negative constituents as nikogo are always ungrammatical without the sentential negative marker ne. When this last element is present as in (156c), the combination of nikogo with the negative marker results in a single negation and a double negation reading is impossible. The last group of examples in (157) illustrates the mixed case of asymmetric negative concord. Here it is necessary to distinguish between preverbal and post-verbal positions: if appearing in post-verbal position (157a), negative elements behave as in negative concord languages and they require a negative licensor; in pre-verbal position, instead, negative

Negative Arguments

79

 constituents do not need the presence of the sentential negative marker (157b) and behave as English negative quantifiers. The main difference between negative constituents in the three groups regards their dependency from the negative marker and the possibility to generate double negation readings. In addition, in order to capture asymmetric concord, preverbal and post-verbal positions have to be kept distinct. Given the properties just illustrated, only negative DP in non concord languages might be unambiguously considered as equipped with truly negative interpretable features. The other two groups, the ones showing concord, are instead more problematic. In fact the relation between the negative DP and the other negative element might be analyzed in different ways. For this reason, the neuter term n-words (Laka 1990) is usually adopted. The main properties of negative nominal constituents are summarized in table 6 below: Table 6. Properties of DP with negative morphology in concord, asymmetric concord and non concord languages. negative marker required 1. Non Concord 2. Asymmetric 3. Concord

preverbal position × × ¥

postverb. position × ¥ ¥

double negation ¥ ¥ ×

For what concerns n-words in concord languages, it is possible to analyze them in at least two ways. A first possibility (Zanuttini 1991, Haegeman and Zanuttini 1991, Haegeman 1995) is to consider them as true negative quantifiers entering in a concord relation through a syntactic mechanism. However, this analysis is not uncontroversial and an alternative analysis assumes that n-words are simply a special kind of polarity items(Laka 1990). This will explain the impossibility to generate double negation. I’ll adopt and develop this view for strict negative concord languages in this and in the next chapter, arguing that concord readings might be reduced to NPI licensing. In this chapter I’ll limit the discussion to asymmetric and non concord languages, emphasizing the relatively unnoticed fact that even no concord languages present pre/post verbal asymmetries.

80

Chapter Five

5.1. Negative quantifiers in non concord languages The principal characteristic of negative constituents in non concord languages is that they respect semantic compositionality: each of them contributes to the sentential meaning with an interpretable negative feature. This is confirmed by the capacity to convey negation alone and by the possibility to generate a double negation reading in combination with another negative element. In accordance with these diagnostics, a nourished set of languages can be considered to be non negative concord. For example German, Dutch (Zeijlstra 2004) and Norwegian (Christensen 2005) belong to this group. Below I report some examples showing that negative constituents might appear without negative markers (a. examples) and that double negation readings are possible (b. examples): (158) a. Ich habe niemand gesehen I have nobody seen ‘I have seen nobody’

(German)

b. Hans sieht nicht Nichts Hans sees neg nothing ‘Hans doesn’t see nothing’ (159)

a. Jan ziet niets Jan sees nothing ‘Jan sees nothing’

(Standard Dutch)

b … dat Jan niet niemand ziet … that Jan neg nobody sees ‘… that Jan doesn’t see nobody’ (160) a. Jeg har ingen bøger læst I have no books read ‘I read no book’

(Norwegian)

b. Ole sier ikke ingenting Ole says neg nothing ‘Ole doesn’t say nothing’ the examples in (158a), (159a) and (160a) show that negative DPs do not require an additional negative element, while (158b), (159b) and (160b)

Negative Arguments

81

 illustrate the availability of the double negation readings. There is less doubt, in these cases, that such DPs have a negative feature. Now consider how this feature behaves in relation to clause typing. In English we have a good diagnostic to detect whether a clause has been typed as negative. This test has been first proposed by Klima (1964), who noticed that English tag-questions behave like polarity reversing expressions, sensitive to the positive/negative Force of the antecedent. Consider the following pair: (161) a. John likes mushrooms, doesn’t he? / *does he? b. John doesn’t like mushrooms, does he? / *doesn’t he? a positive sentence as the one in (161a) only allows for a negative tag38, while a negative sentence (161b) requires instead a positive one. In order to make this test telling, we should analyze the structure of tags. Following a fairly standard analysis, I’ll consider tags as a repetition of the preceding sentence with VP ellipsis. Let me illustrate this point by dividing (161) into two different sentences, with the complete structure of the tag question given in (162b) (162)

a. John likes mushrooms, b. doesn’t he [likes mushrooms]?

According to (162b), the tag is a complete sentence with the elided VP. The overt sentential negative marker types the sentence as a negative (interrogative) sentence, imposing a reversion of polarity with respect to the antecedent: if the antecedent is positive, the tag is negative and vice versa, as shown in (161). This can be expressed by a rule able to operate on and reverse the typing value in ForceP. Let us call the antecedent α and the tag β, then the tag rule is the following: (163)

TAG QUESTIONS - TYPE INVERSION RULE ForceP β must be typed by the opposite value of ForceP α with respect to +/- [neg]

The tag rule here formulated is active on ForceP and can be used to detect its features. 38

Liliane Haegeman (p.c.) pointed out that in certain echoic contexts when the speaker challenges what the interlocutor has just said, a positive tag is allowed: (1) John likes exercises, does he? (Well, I'll show him what exercising means)

82

Chapter Five

Let us go back to negative quantifiers, and see what kind of tag they select. Consider sentence (164): (164)

Nobody came, did he/ *didn’t he?

the example shows that a negative quantifier in preverbal position triggers a negative tag, just like the sentential negative marker. This means that the type inversion rule is fed by a negative feature in the antecedent and gives back a positive value for the tag Antecedent: ForceP[[+neg] Tag:

ForceP

(164)’

[[-neg] did

IP

[Nobody[+neg] came VP[...]]

IP

[he VP[came]]

ForceP

[[+neg] IP[Nobody[+neg] came VP[...]],ForceP[[-neg] did IP[he VP[came]]

This means that, in analogy with sentential negative markers, a negative quantifier in non concord languages bears a negative feature able to enter in a typing relation with the topmost clausal projection. The conclusion is that negative constituents in non concord languages contribute with a negative feature of exactly the same kind of the one carried by sentential negative markers. After this preliminary discussion on tag questions, I would like to consider next whether the position of negative quantifiers has consequences on typing.

5.2. Pre/post verbal asymmetries in non concord languages We saw that the tag test might be used to detect the polarity of ForceP and that negative quantifiers are visible to the tag inversion rule. However, it seems that the position of negative quantifiers is not irrelevant. In fact, when they are in object position, the tag inversion rule only optionally applies. In (165) I report an original example from Ross (1973), who pointed out that sometimes the tag inversion rule fails to apply: (165)

John saw nobody, didn’t he?

Negative Arguments

83

 In (165) nobody in object position allows for a negative tag39 and a possible explanation is that the antecedent clause has not been typed as negative. The same effect is found not only in object position, but also in adjunct position, as in the example (166) from Jackendoff (1972) (166)

Kim looks good in no clothes, doesn’t he?

If we consider negative tags to indicate that ForceP has received its default positive value, it seems that something happens when the negative element is in post-verbal position. The contrast might be captured in various ways40, but the edge of the VP seems to play a key role. If we assume a theory of syntactic phases as in Chomsky (1998), the Phase Impenetrability Condition (PIC) might capture the fact that the link between ForceP and VP-internally negative quantifiers cannot be established. According to the PIC, VP internal material left in situ is not visible for further syntactic operation, including Agree. From this it follows that ForceP might look for an appropriate typing feature only up to to the edge of the vP phase, failing to see the negative feature in internal argument position. In this situation, the negative feature in post-verbal position is invisible and the default value must be chosen. If VP internal negative features are inert, due to the opacity created by some general principle as the PIC, we expect that the subject-object asymmetries found in English are not an isolated case. And in fact, other non concord languages reveal to be less ‘symmetric’ than previously thought. A complication is due to the fact that asymmetries cannot be easy to detect due to other syntactic factors. In fact, many Germanic languages which do not have negative concord are SOV. If we follow the fairly 39

Huddleston (1984) reports the possibility to have also positive tags, as in the following sentence: it solves nothing, does it? However, notice that even if we believe that there is a degree of optionality here, this is found only for VP internal negative constituents and that positive tags are always required for subject (and VP external) quantifiers. This confirms the existence of a subject/object asymmetry. 40 An alternative to the syntactic account based on the notion of phases is to consider the event variable and the scope of the negative quantifier. If the event variable e is inserted on the top of the VP, negation fails to take scope over the event, resulting in a constituent negation reading. The account which I’m going to propose can be reformulated also in these terms, by saying that a negative chains must be created in order to create a link which have an interpretable negative feature c-commanding e.

84

Chapter Five

standard assumption (Koster 1994, 2000; Zwart 1993, 1997, 2001; Pearson 2001; Haegeman 1998a, 1998b, 2000a, 2001) that the unmarked word order is obtained from an underling SVO order and that the object has been moved outside the VP, we must look for these particular cases where the direct object fails to move. Let us consider German first. In this language (Hamann 1993), a negative quantifier in object position normally appears to the left of adjectives as stolz, see (167) below: (167) ... dass Peter auf niemanden stolz ist …that Peter of nobody proud is ‘... that Peter isn’t proud of anybody’

(German)

However, Hamann noticed that, in certain cases, it is also possible to leave the quantifier inside the VP. In these cases, the negative quantifier niemanden will appear to the right of stolz: (168)

...dass Peter stolz auf niemanden ist …dass Peter proud of nobody is ‘...that Peter is proud of nobody’

In (168), if the vP really creates an opaque domain for Agree, we expect that niemanden should be invisible and that it cannot enter into a negative chain. We predict that Negation Raising will be impossible and that only narrow scope is allowed. We can check this prediction in the usual way: by adding another scope bearing element. Sentence (169) constitutes a minimal pair with (167), the only difference being the insertion of the modal mussen ‘must’. Here negation might easily take wide scope over modality: (169)

…dass Peter auf niemanden stolz sein muss …that Peter of nobody proud be must ‘... that Peter need not be proud of anybody’

¬Ƒp

Now let us build another minimal pair, but this time with sentence (168), where the negative quantifier is inside the VP: (170)

… dass Peter stolz auf niemanden sein muss …that Peter proud of nobody be must ‘... that Peter need not be proud of anybody’

*¬Ƒp

Negative Arguments

85

 interestingly, in (170) the inverse wide scope reading is impossible and negation cannot take scope over the modal. The only licit interpretation is the one where negation has narrow scope. An appropriate paraphrases will be ‘...that Peter must not be proud of anybody’, in accordance with the scope must > not. This is readily explained if we consider, as in the case of English, negative quantifiers inside the VP as unable to enter into negative chains. This situation seems to be a general one and it can be retrieved in many other languages, supporting the idea that VP internal/external asymmetries are found also in languages without negative concord. Norvegian shows a similar asymmetry with respect to the negative determiner ingen ‘no’ (Svenonius 2002), which conveys sentential negation only when it is moved out of the VP. The usual situation in finite clauses without auxiliaries is the one presented in (171), where the VP is empty: the finite verb goes in its V2 position and the direct object appears in its object-shift landing site (171)

Vi vant ingen konkurranse we won no competition ‘We didn’t win any competition’

However, there are cases in which the object must be frozen in its base position, as in the configurations excluded by Holmberg generalization41. For example, when the lexical verb is not moved in its V2 position due to the presence of an auxiliary, ingen konkurranse cannot move outside the VP: (172)

*Vi kunne ingen konkurranse vinne we could no competition won

Importantly, it cannot even stay in its base position (173a), contrary to non-negative constituents (173b): (173) a. *Vi kunne vinne ingen konkurranse we could win no competition

41

It states that movement is impossible when overt material appear VP internally (Holmberg, 1999).

86

Chapter Five

b. Vi kunne ikke vinne konkurransen we could not win the.competition ‘we could not win the competition’ (173) shows there is some of constraint blocking negative constituents in VP-internal position. The pattern which emerges from Svenonius’ argumentation is very clear, and we again have a contrast between VP internal/external positions42. This is reminiscent of the well-known asymmetry found in asymmetric concord languages and it is possible that the differences between the two groups might be partially reduced to a different modulation of the constraint in force on asymmetric concord languages: in fact, while the constraint seems to be mainly interpretive in the non concord languages group, we will see that in asymmetric concord languages it is also syntactic. In the next section, I’ll consider the case of Romance varieties as Italian and Spanish.

5.3. Asymmetric concord languages Negative quantifiers in non concord languages have usually been considered to have properties which remain constant, regardless to their 42

Negative constituents forced to remain in situ might be saved only if they receive narrow scope reading (see section 5.4). In this case, negation has to be interpreted as constituent negation (1)a or as a minimal quantity expression (1)b: (1) a. Vesna ser fin ut i ingen klær Vesna looks fine out in no clothes ‘Vesna looks good naked’ b. Han flirer av ingenting he laughs of nothing ‘he is laughing at nothing’ (‘for no reason’) The negative constituents in (1) do not have a minimal quantity meaning per se, since once moved outside the VP, they regain their sentential interpretation, supporting positive tags and licencing PIs: (2) a. I ingen klær ser Kim fin ut, gjør han vel? in no clothes looks Kim nice out does he well ‘In no clothes does Kim look good, does he?’ b. I ingen klær ser Kim fin ut i det hele tatt in no clothes looks Kim nice out in the whole taken ‘In no clothes does Kim look good at all’ It seems that in English, German and Norwegian negative elements in post verbal position cannot take sentential scope, but are limited to a constituent negation interpretation.

Negative Arguments

87

 clausal position. However this is true only for some basic features, like the capacity to generate double negation and to be independently licensed without sentential negative markers. In fact, to a closer scrutiny, they reveal certain asymmetries related to their clausal positions which are reminiscent to the well-known asymmetries shown by n-words in Romance languages (Laka 1990, Moscati 2006a). Many scholars, starting from Zanuttini (1991), have tried to analyze nwords as negative quantifiers and my proposal will be partially in this vein, limited to preverbal positions. A difference with previous accounts is that I consider also non concord languages as being ‘asymmetric’ and I’ll try to derive the impossibility to have post-verbal negative quantifiers in Romance from the same mechanism blocking the formation of negative chains. Let me first introduce the basic facts on negative concord in Romance. The first observation is that a series of nominal constituents with negative morphology shows a mixed behavior, depending on their clausal position. For this reason, we need to distinguish between pre/post verbal positions. When they precede the verb, Romance n-words typically show the properties of negative quantifiers. In in all the examples below43, negative constituents in pre-verbal position express sentential negation alone and sentences are grammatical even in absence of the sentential negative marker: (174) a. Nessuno ha telefonato nobody aux telephoned ‘Nobody telephoned’

43

(Italian)

b. Ninguém lhe telefonou nobody him called ‘Nobody called him’

(European Portoguese)

c. Nada funciona nothing works ‘Nothing works’

(Spanish)

Sardinian examples from Jones (1993); Catalan and Spanish from Valduvì (1994); European Portuguese from Matos (1999).

88

Chapter Five

d. A NINGÚ ha vist nobody aux seen ‘He has seen nobody’

(Catalan)

e. Neune at mai peccatu nobody aux never sinned ‘Nobody has ever sinned’

(Sardinian)

in this respect, they look very similar to negative quantifiers in non concord languages. The parallel is also maintained if we look at double negation. In fact, double negation readings are possible when negative constituents appears pre-verbally in combination with the sentential negative marker. This is illustrated by the following Italian example (175)

Nessuno non ha capito il problema nobody neg has understood the problem ‘Nobody didn’t understand the problem’

(DN)

in (175) nessuno appears in pre-verbal subject position together with the negative marker non. This sentence, pronounced with a neutral intonation, is naturally interpreted as a double negation. This effect is not only found with subjects, but also with other constituents dislocated to the left of the verb. Consider for example sentence (176b), uttered in the given context: (176) Speaker A: a. Dicono che i suoi ospiti non abbiano mangiato tutte le portate ‘They say that his guests didn’t eat all the dishes’ Speaker B: b. Macchè! NIENTE non hanno mangiato! ‘Quite the contrary! NOTHING, they didn’t eat!’ Æ they eat everything

(DN)

In sentence (176b) the n-word is in contrastive focus position and it can be interpreted with a double negation reading. What sentences (175) and (176b) show is that, although rare in the everyday speech, double negation readings are relatively easy to obtain once the appropriate context is given.

Negative Arguments

89

 The same is true for other asymmetric concord languages44. In (177) and (178) I report two examples, respectively from Spanish and Sardinian: (177) Nadie no ha venido nobody neg aux come ‘Nobody hasn’t come.’ (178) Neune no’ at mai peccatu nobody neg aux never sinned ‘Nobody has never sinned’

(Spanish) DN (Sardinian) DN

The examples from (174) to (177) confirm that preverbal n-words have the same interpretation of negative quantifiers. On the contrary, when they follow the verb, their behavior dramatically changes. In this position, they cannot appear alone and they do not generate double negation readings: (179) a. *(Non) vidi niente neg saw-1s anything ‘I did not see anything’

(Italian)

b. *(Não) vi ninguém neg saw-1s anyone ‘I did not see anyone’

(Europena Portuguese)

c. *(No) vino nadie neg came-3sg anyone ‘Nobody came’

(Spanish)

d. *(No) he dit res neg have-1s said anything ‘I didn’t say anything’

(Catalan)

e. *(No’) est vennitu neune neg is came anybody ‘Nobody came’

(Sardinian)

In all the examples in (179), the post-verbal n-word requires the presence of the sentential negative marker and the only possible interpretation is the 44

Among Romance, only Catalan seems to be exceptional in not allowing double negation readings (see Espinal 2000). For EU Portuguese, see Martins (2000) and Matos (1999).

90

Chapter Five

concord reading. Descriptively, the the pre/post verbal alternation can be captured by the following generalization45, proposed by Ladusaw (1992): (180) In a NC language it is impossible to express the negation only in VP This generalization assumes the edge of the VP as the discriminating factor for the asymmetric behaviour of n-words in Romance. But we saw that the edge of VP also conditions some properties of negative quantifiers in non concord language. In what follows, I suggest that the difference between the two groups of languages might be extremely subtle and that it derives from the same constraints blocking sentential interpretation of VPinternal negative quantifiers in non concord languages.

5.4. VP-internal negative concord is NPI licensing I believe that the similarity with non concord languages can be pushed up to the point where concord can be avoided altogether in Romance. Under this view, all the asymmetries previously illustrated can be reduced to the opposition VP internal/external. Many accounts dealt with the properties of n-words, fluctuating between the negative quantifiers (Zanuttini 1991, Haegemann & Zanuttini 1991) or the NPI analysis (Giannakidou 1998, Zeijlstra 2004). Analyses as in Zanuttini (1991) easily capture the behaviour of pre-verbal n-words, but problems come in when we consider the drastic change in post-verbal positions. Conversely, NPI accounts easily explain the behaviour of nwords only in post-verbal positions. In what follows, I’ll propose a mixed approach capitalizing on the VPinternal/external asymmetry found also in non concord languages. We saw that this asymmetry is more common than previously thought and that changes in the syntax/semantics of negative quantifiers can be detected also in non concord languages, where the VP border has several effects. We saw that it blocks clause typing (in English) and sentential scope (in Scandinavian and West Germanic languages). It is then possible that these effects are simply strengthened in Romance and that negative quantifiers are blocked VP internally. 45

Zanuttini (1991) proposed a similar account for the distribution of negative elements, but she assumed Tense as the relevant category. She proposed the following constraint on the assignment of sentential scope to negation: ‘negation can take sentential scope only if at s-structure it is in a position from which it ccommands both the Tense Phrase and the Agreement Phrase’.

Negative Arguments

91

 But, if such a block exists, why we find n-words in post verbal positions? The fact that they need the presence of the sentential negative marker and their negative polarity interpretation suggests that they might be elements of a different kind, homonymous with preverbal negative quantifiers. Let me illustrate this idea. In literary Italian, we have both a series of negative nominal constituents with negative morphology and a series of polarity items. The two distinct paradigms are reported in table 7. Table 7. Italian n- and al- series meaning person thing

n-series nessuno niente

al- series alcuno alcunchè

The al-series is used in the typical range of polarity licensing environments, as in negative contexts and in the scope of downward entailing operators in general: (181) Non ho detto alcunchè neg aux said anything ‘I did’t say anything’ (182) Ci penso sempre, prima di dire alcunché di sconveniente cl think always before to-say anything of impolite ‘I always reflect, before saying anything impolite’ Nowadays, however, the al-series has almost disappeared in the colloquial language and it is confined to literary style. But this does not mean that polarity items are disappearing in Italian. On the contrary, the n-series took over the al-series and it is now used in all the relevant syntactic environments, becoming the only really productive series. The n-series is used in many non-negative contexts as in questions (183a), indirect questions (183b), within the scope of ‘before’ (183c), in the complement of adversative predicates (183d), in comparatives (183e), superlatives (183f), protasis of conditionals (183g) and in combination with ‘too’ (183h):

92

Chapter Five

(183) a. Ha visto niente? aux see anything ‘did he see anything?’ b. Romeo si domanda se Giulietta ami nessun'altro Romeo o.cl wonders if Giulietta loves-su anyone else ‘Romeo asks himself if Giulietta loves anyone else’ c. Sparirono prima che Gianni potesse dir niente disappear-3p before that Gianni may-su say anything ‘They disappeared before Gianni could have said anything’ d. Dubito che James possa spifferare niente ai Russi… I-doubt that James could-su reveal anything to-the Russians ... è la nostra spia migliore! ...is the our spy best ‘I doubt that James could reveal anything to the Russians... ...he is our best spy!’ e. Il risultato è migliore di quanto nessuno si aspettasse the result is better than how anybody cl.3s expect-su ‘The result is better than anybody would have expected’ f. E’ l’ idea più stupida che sia venuta in mente a nessuno is.3sg the idea dumbest that be-su came in mind to anybody ‘It’s the dumbest idea that anybody has ever had’ g. Se dovesse venire nessuno, digli di ripassare dopo if may-su come anybody tell-to-him to come back later ‘If anybody comes, tell him to come back later’ h. Quel vecchio leone è troppo stanco per infastidire nessuno that old lion is too tired to bother anybody ‘That old lion is too old to bother anybody’ In all the examples above46, elements as nessuno and niente do not have a negative meaning and their interpretation is equivalent to anybody and anything in English. 46

There is a certain grade of variation in the acceptability of the examples (183) among speakers (see Acquaviva 1999). Although acceptability may vary, all Italian

Negative Arguments

93

 N-words follow the distribution of polarity items also in other asymmetric concord languages47, where they do not have any negative meaning. Below I report some examples from Catalan (Espinal 2000, Valduvì 1994) and Spanish (Laka 1990), showing the NPI interpretation of n-words: (184) a. QUESTIONS Que t’han dit res? ‘Did they told you anything?’

(Catalan)

b. INDIRECT QUESTIONS Si trobes cap paraigua com aquest, compra-me’l ‘If you find any umbrella like this one, buy it for me’ c. PROTASYS OF CONDITIONALS Si vol res ningù, aviseu-me if 3s-want anything anyone, 2p-imp-warn.me ‘If anyone wants anything, let me know’ (185) a. INDIRECT QUESTIONS (Spanish) Me preguntaron si nadie sabíe la respuesta me 3p-past-ask if anybody 3s-impf-know the answer ‘They asked me whether anybody knew the answer’ b. EVERY Todo aquel que tenga nada que dicir... all who that have.3sg anything that say ‘Everyone who has anything to say….’ c. ADVERSATIVEPREDICATES Dudo que venga nadie doubt that come anybody ‘I doubt that anybody comes’

speakers will accept an n-word in direct questions as in (183)a, clearly a non negative context. 47 In Portuguese the PIs seems to be a strong one, as shown by the ungrammaticality of the following example taken from Matos (1999): (1) *Telefonou ninguém? called.3sg n-person ‘Did anybody called?’

94

Chapter Five

Although there might be differences in acceptability among speakers, examples (184) and (185) illustrate the fact that Catalan and Spanish permit n-words as polarity items. Therefore, on the basis of the data in (183), (184) and (185), it seems to me unavoidable to admit a lexical entry for n-words as polarity items. Going back to Italian, the examples in (183) indicate that for many Italian speakers the distribution of this set of items is regulated by the same conditions in force on weak polarity items. This means that the nseries must have a lexical entry as in (186): (186)

niente/nessuno [+pol]

It is then unsurprising that n-words appear in post-verbal position within the c-command of negation, exactly as any in English. Given that postverbal n-words respect the usual licensing conditions, we could treat nwords in questions and in post-verbal environments in exactly the same way: (187) non vidi niente ‘I did not see anything.’ NegP

[OP¬ non vidi VP[niente[+pol]]]

(188)

ha visto niente? ‘did he see anything?’ CP

[OPQ IP[ha visto VP[niente[+pol]]]

In the examples above, niente appears within the scope of a negative operator (187) or a question operator (188), both suitable licensors for NPIs. Under this view, post-verbal n-words do not have anything special or “asymmetric”, but are simply polarity items subject to the usual licensing conditions. What remains to be explained is the negative interpretation of n-word in preverbal position, where the NPI analysis is problematic for many reasons. Remember that, in sentences as (189), n-words do not require a licensor and they might originate double negation: (189) Nessuno (DNnon) ha telefonato nobody neg has telephoned Nnobody telephoned’ (‘nobody didn’t telephone’)

Negative Arguments

95

 In this example nessuno has the same interpretation of a negative quantifier in non concord languages. Apparently, the simplest hypothesis is to treat pre-verbal n-words as negative quantifiers. But this is not straightforward, since we have to explain why negative quantifiers are licit only pre-verbally outside the VP. A possibility is to assume that negative quantifiers in Romance are subject to the same constraints forcing negative quantifiers in non concord languages to escape from the VP in order to have sentential scope. With the difference that in Romance the possibility to be left in situ is forbidden by an extra syntactic constraint. Imagine than that n-words have a negative feature which is interpretable only outside the VP. For this reason, they have to be moved out of it. This intuition is similar to the proposal in Haegeman & Zanuttini’s (1991) and can be casted in the feature system presented in chapter four without assuming any factorization process able to conflate multiple negation into a single operator. Let me illustrate how the system works starting from non concord languages. English negative quantifiers can be considered as self-standing elements, interpretable in situ without any additional negative marker. This can be expressed assuming that they have an interpretable and valued feature48: (190)

nobody/nothing i[val]

This insures that even in VP internal positions, they are licit and interpreted without sentential scope. In Italian, instead, negative quantifier cannot be interpreted VP internally and they need to enter in a chain with another interpretable negative feature. I consider Italian negative quantifiers as having a valued but uninterpretable negative feature: (191)

nessuno/ niente u[val]

The difference between (190) and (191) captures the variation between non concord and asymmetric concord languages. In asymmetric concord languages, the unevaluated feature must enter in relation with an additional interpretable negative feature, which can be hosted in ForceP or in NegP (as in the Neg-Criterion approach). Since VP internal positions 48

This does not exclude that whenever Negation Raising has to be applied for scope shifting purposes, they might alternatively be marked with an uninterpretable negative feature, linked to an interpretable one in ForceP.

96

Chapter Five

are invisible for the creation of negative chains, negative quantifiers are forced to move in order to be linked to an interpretable but unevalued negative i[] feature (which correspond to the null operator in the NegCriterion approach). Under this view, asymmetric concord languages are not asymmetric with regards to the licensing of their set of polarity items: the asymmetry concerns only negative quantifiers. Remember again that this asymmetry is reminiscent of a similar one found in some non concord languages. Consider again the Norwegian examples in (171b) repeated here as (192): (192) *Vi kunne vinne ingen konkurranse ‘We could win no competition’ here the negative constituent introduced by ingen cannot stay VP internally, even if this is possible for non-negative constituents. This shows that there is a constraint, also in Norwegian, against negative quantifiers in VP internal position. If we eliminate ingen, the sentence becomes perfectly acceptable: (193) Vi kunne ikke vinne konkurransen we could not win the-competition ‘We could not win the competition’ Sentences as (192) might marginally be acceptable only with very narrow scope constituent readings, as in the case of what Svenonius (2002) calls ‘trifling’ negation: (194) Han flirer av ingenting he laughs of nothing ‘He is laughing at nothing’ (‘for no reason’) It seems that even in some non concord languages, negative quantifiers are not interpretable in situ and they have an u[val] negative feature exactly as in Romance. If we consider ‘trifling’ negation in (194), once more the parallel with asymmetric concord languages looks very tight. In Italian, it is possible to have a negative quantifier in post-verbal position expressing very small quantities. Quite surprisingly, the sentential negative marker here is not required

Negative Arguments

97

 (195) a. Gianni ride per niente Gianni laugh-3s at nothing ‘Gianni laughs at nothing’ b. Gianni si accontenta di niente Gianni cl.3s satisfy of nothing ‘Gianni is satisfied with nothing’ This means that in certain particular cases, where negative constituents have a very specific meaning related to minimal quantities, they are interpretable inside the VP. In order to show how the whole system works, I wish to discuss next some key examples. Consider first a sentence with an n-word in postverbal position: (196)

*Ho visto nessuno aux-1s seen *nobody/*anybody

the ungrammaticality of (196) follows from the impossibility to have neither an NPI nessuno or a quantifier nessuno in this position: the NPI is excluded since there is no appropriate licensor and the negative quantifier cannot enter in a negative chains from inside the VP. Now consider what happens if we save (196) by inserting the sentential negative marker: (197)

Non ho visto nessuno not aux-1s seen *nobody/anybody ‘I didn’t see anybody’

the correct interpretation follows since the sentential negative marker non introduces an appropriate DE operator able to license the NPI nessuno. As in the previous example, the negative quantifier is excluded for the impossibility to receive interpretability inside the VP. For what concerns instead preverbal positions, it is easy to see that the opposite situation holds. In absence of an appropriate licensor, NPIs cannot appear in subject position (see next chapter for deriving this constraint from general semantic properties of NPI) and only the negative quantifier nessuno is allowed, as in sentence (198)

98

Chapter Five

(198) Nessuno è venuto nobody/*anybody is come ‘Nobody came’ Examples (196), (197) and (198) illustrate how the proposal works with respect to the core set of data needed to account for asymmetric concord languages. The system I’m proposing here covers also a wider range of phenomena which can receive an organic treatment. In the following section I’ll discuss some additional empirical facts, as concord readings in preverbal environments, negative spread and fragment answers.

5.4.1. Concord readings in pre-verbal position We expect that whenever the conditions relative to NPIs licensing are met, we can have ‘concord readings’. This means that even in preverbal position, if the licensing conditions for NPIs are satisfied, we can have single negation. Consider, for example, embedded clauses selected by a negative matrix: (199)

I don’t believe that anybody will come

in (199) negation in the matrix licences a subject NPI in the embedded clause. Under the current analysis, we also expect the NPI nessuno/niente will be possible in the same context, without generating double negation. The following example shows that this is exactly what happens in Italian, where sentence (200) below is interpreted without double negation49 (200) non voglio che nessuno mi disturbi not want that anybody me annoys a. # ‘I do not want nobody to annoy me’ b. ‘I do not want anybody to annoy me’

DN NC

This example suggests that whenever an NPI is licit, a ‘concord reading’ is possible. We can check this prediction in a wider set of contexts. For 49

Rizzi (1982) reports a double negation reading for the following sentence: (1) Non pretendo che nessuno ti arresti not demand that n-person you arrests ‘I do not require that nobody arrests you’ DN The contrast with (200) can be related to the degree of opacity created by the matrix verb or by the possibility to have different prosodic contours.

Negative Arguments

99

 example, weak NPIs in subject position can be licensed in the scope of before, in the second term of comparatives, in the scope of adversative predicates and in indirect questions. Again, concord is found in all these environments: (201)

a. BEFORE Scappammo prima che nessuno se ne accorgesse ‘We ran away before anybody noticed’ b. COMPARATIVES Quel libro era più interessante di quanto nessuno si aspettasse ‘That book was more interesting than anybody would expect’ c. ADVERSATIVE PREDICATES Dubito che nessuno di loro possa fare la spia ‘I doubt that anyone of them could reveal our secrets’ d. INDIRECT QUESTIONS La povertà di quel pittore mi induce a chiedermi se nessuno l'abbia davvero compreso ‘The poverty of that painter makes me wonder whether anybody did really appreciate him’

The examples in (200) and (201) suggest that negative concord in preverbal position is possible in exactly the same contexts licensing an NPI. We have thus additional evidence supporting the idea that negative concord can be reduced to NPI licensing.

5.4.2. Negative spread There is also another set of sentences that can be easily explained under the current account. I’ll follow Den Besten (1986) in referring to sentences with multiple n-words as ‘negative spread’. These sentences are possible in all the asymmetric negative concord languages, where multiple n-words can be simultaneously present in the clause, even without the sentential negative marker: (202) a. Nessuno ha letto niente n-person aux read n-thing ‘Nobody read anything’

(Italian)

100

Chapter Five

b. Nadie dijo nada n-person said-3s n-thing ‘Nobody said anything’

(Spanish)

c. Ningú va dir res n-person aux-3s say n-thing ‘Nobody said anything’

(Catalan)

d. Ninguem viu nada n-person saw-3s n-thing ‘Nobody saw anything’

(Portuguese)

An important characteristic in common to sentences in (202) is that at least one n-word must appear in preverbal position. This is easily explained if we consider preverbal n-words as negative quantifiers, able to licence post-verbal NPIs. As expected, negative spread will be impossible if all the n-words are in post verbal position: (203)

*Gianni ha letto niente a nessuno Gianni aux read anything to anybody ‘Gianni didn’t read anything to anybody’

the ungrammaticality of (203) follows if we consider the post-verbal nwords as NPIs. Here no licensor is present and NPI niente and nessuno are expected to be forbidden inside the VP. However, if a negative quantifier appears outside the VP, it is a proper licensor able to license NPIs, as in the examples in (202).

5.4.3. Fragments Another characteristic of asymmetric concord languages is the possibility to have n-words in fragment answers. Consider the following exchanges in different Romance languages: (204) Q: Chi hai visto ? who have-2s seen? ‘Who did you see?’ A: Nessuno ‘Nobody’

(Italian)

Negative Arguments

101

 (205)

Q: Ele telefonou a alguém? he phoned to anybody ‘Did he call anybody?’

(EU Portuguese)

A: A ninguém. ‘Nobody’ (206) Q: Qui t’ ha vist? who o.cl aux seen ‘Who saw you?’

(Catalan)

A: Ningú ‘Nobody’ (207) Q: ¿Desea algo más? ‘Would you like anything else?’

(Spanish)

A: Nada ‘Nothing’ in these examples n-words can be used in isolation, as negative quantifiers in non concord languages. This is a remarkable difference from NPIs, which are forbidden in fragments: (208)

Q: Who did you see? A: Nobody / *Anybody

The parallel between n-words and negative quantifiers is not surprising, if we consider n-words in fragments as negative quantifiers moved outside the VP. Some scholars consider fragments simply as nominal constituents (Ginzburg and Sag 2000) while others analyzed them as full clauses with an elided part (Hankamer 1979, Morgan 1973). The two representations are schematically given in (209): (209)

Chi hai visto? a. DP [nessuno] b. CP [non ho visto nessuno]

If we assume a representation such as (209a), it is unclear how the meaning of a whole proposition arises from a simple DP. The problem for (209b) is, instead, that phonological deletion affects a non-constituent

102

Chapter Five

portion of the sentence. This last problem can be solved by considering fragments as special cases of sluicing, a process originally investigated in Ross (1969) to account for sentences such as (210), where the whole IP is elided, leaving the wh- as the only overt element: (210)

Jack bought something, but I don’t know what Jack bought

The same proposal can also be extended to fragment answers, adopting the analysis in Merchant (2001). N-words, as wh-elements, are first moved in FocusP and successively the IP undergoes to phonological deletion. The representation for (208) will be the following: (211)

ForceP FocP

[

[nessuno IP [... ho visto VP [...]]]]

At this point, the fact that n-words behave as nominal quantifiers is not surprising since nessuno is outside the VP and has the expected pre-verbal properties. Given the representation in (211), we should also consider the issue of interpretability. In fact, (211) says nothing on the interpretive site of negation. However, given our assumptions on clause typing and on the uninterpretability of the negative feature of Italian n-words, it is necessary to create a chain between FocusP and ForceP. The chain links the uninterpretable feature of the n-word with an interpretable instance, insuring the satisfaction of clause typing: (212)

ForceP

[i[val] FocP [nessunou[val] IP [... ho visto VP [...]]]]

According to the representation in (212), the negative operator is in the topmost clausal position. From this it follows that negation should take wide scope if another operator is introduced. This is exactly what happens in fragments. Again, we can use modals to reveal the logic scope of the negative operator. Consider first the following declarative sentence with the negative quantifier nessuno and the modal dovere: (213)

Nessuno deve continuare a correre nobody must-3s to keep on-inf to run a. ‘it is not required that everybody keeps on running’ b. ‘it is required that nobody keeps on running’

¬ p ¬p

Negative Arguments

103

 as the paraphrases show, this sentence can be interpreted either with wide (213a) or narrow scope (213b)50 of negation. Consider now a fragment used as an answer to a question containing the same modal: (214)

Q: Chi deve continuare a correre? who must-3s keep on to run ‘Who must keep on running?’ A. NESSUNO [deve continuare a correre] ‘nobody’ must-3s keep on to run a. ‘It is not required that everybody keeps on running’ b. ‘It is required that nobody keeps on running’

¬ p * ¬p

in (214) only the wide scope reading is possible and reading (213)a disappears. But why negation raising is forced in these cases and we have a missing reading? A possible answer comes if we assume the mechanism of ellipsis resolution as in Merchant (2001). According to Merchant, the elided IP must respect the notion of e-GIVENESS which says that the antecedent and the elided portion should be in a reciprocal entailment relation51. This basically means that the operators and their scope in the elided part must parallel the meaning of the antecedent. In the case of (214), there is no negative feature in the antecedent question and the condition of e-GIVENESS imposes that also in the elided portion no negative operator has to be present. The only possibility, in order to respect the conditions on ellipsis resolution, is to have a negative operator outside the ellipsis. This means that negation must be raised and take scope outside the IP:

50

The narrow scope reading can be derived assuming that interpretation for nessuno is in NegP, by means of a negative feature interpretable and unevaluated. The negative constituent transits here to reach AgrSP and checks its uninterpretable negative feature . The derivational steps are indicated below: a. ForceP[ u[]neg AgrP[ij ModP[deve NegP[OPi[]neg VP[nessunoij,u[val]neg continuare CP [a correre]]]]]] b. ForceP[u[]neg AgrP[ij ModP[deve NegP[nessunoij,u[val]neg OPi[val]neg VP[t continuare CP [a correre]]]]]] ForceP AgrP ModP NegP VP [ u[]neg [nessunoij [deve [OPi[val] neg [ t continuare c. CP [a correre]]]]]] ForceP AgrP ModP NegP VP d. [ u[val] neg [ nessunoij [deve [ OPi[val] neg [t continuare CP [a correre]]]]]] 51 See Merchant (2001) for the issue of ‫׌‬-type shifting.

104

Chapter Five

(214) a. ForceP [i[val] FocP[ nessunok u[val] ]] ELIDED ModP { [deve VP[tk continuare CP[a correre]]]}]] The alternative representation with narrow scope is given in (214b) (214) b. ForceP[u[val] FocP[nessunok u[val] ELIDED ModP { [deve NegP[ OPi[val] VP[ tk continuare CP [a correre]]]]}]] but this representation is excluded since the negative interpretable feature appears within the ellipsis, violating the eGIVENESS condition. For this reason only (214a) is possible and this accounts for the ‘missing reading’ in fragments with a modal antecedent.

Conclusions Although negative quantifiers in languages without negative concord have usually been considered not to show any subject-object asymmetry, it seems that there are indeed some distributional differences. We saw that the polarity of the tag questions in English is sensitive to the position of the negative quantifier. In fact, negative tags are obligatory only if the quantifier is in a VP external position. Moreover, also in German and Norwegian, two languages without negative concord, the interpretation of negative quantifiers varies if they remain in their base position. All these phenomena can be related to the well know asymmetries found in some concord languages as Italian and to the impossibility to have the negative quantifier interpretation for n-words in post-verbal position. In this chapter I propose a way to unify the two kinds of languages based on a general constraint blocking sentential negation inside the VP. The only difference is that in non negative concord languages the structure is saved by the possibility to have the interpretation of constituent negation. In asymmetric concord languages, instead, negative quantifiers need to move outside the VP. This constraint has been formulated by saying that negative quantifiers in Romance need to enter into a syntactic chain with an interpretable position, being it NegP or ForceP. Although this proposal is similar to the Neg-Criterion, the idea that postverbal n-words are NPI allows us to avoid any additional operation as for example the Neg Factorization proposed in Zanuttini (1991). Concord reading are instead accounted for by reducing them to NPI licensing, capturing the important fact that concord readings are typically found in NPI contexts.

Negative Arguments

105

 An open issue which remains to be addressed is the nature of NPIs and, more importantly, why NPI n-words are excluded in preverbal position: if we consider NPIs to be licensed in the scope of a negative operator, Negation Raising should also licence NPIs. I’ll discuss this in the next chapter, where I will consider also strict negative concord languages. I’ll try to push further the idea that concord is nothing else than a polarity phenomenon and I’ll propose that n-words must be considered as a special kind of NPIs.

CHAPTER SIX DOMAIN WIDENING AND STRICT NEGATIVE CONCORD

In this last chapter, I will consider two topics related to polarity. The first concerns some problematic interactions between negative polarity items and the operation of Negation Raising. I will try to tighten here some loosen ends about the distribution of NPIs in subject position. The second issue regards instead a ‘hole’ left unfilled in the typology of negative nominal syntagms. In the previous chapter, we dealt with asymmetric and non concord languages, largely ignoring the properties of strict negative concord languages. I kept these two topics for this final part, since I believe that they can be treated in similar ways by considering the semantic analysis of NPI as in Kadmon & Landman (1993) and Chierchia (2004). Let me illustrate first the potential problem for Negation Raising posed by preverbal subjects. In order to derive the inverse scope readings presented in chapter three, I propose an LF operation able to provide an additional interpretive site to the negative operator, higher than NegP. A tentative proposal is that this extra site can be put in relation with Clause Typing. This means that ForceP can be the highest interpretable site for negation (215)

ForceP

[OP¬j

AgrSP

[

NegP

[ tj VP[ ...]]]]

According to the very general schema in (215), once Negation Raising takes place, the negative operator scopes over the whole sentence including the subject position in AgrSP. The problem is that since Ladusaw (1979) it is less controversial that NPIs are licit only in the scope of an appropriate operator52 and negation is one of them. We then expect, 52

The semantic characterization of the class of NPI licensor it is still object of much debate. See among others Zwart (1995), Giannakidou (1997), von Fintel (1999). I’ll follow here the traditional analysis in Ladusaw (1979) and consider DE contexts as the ones allowing NPIs.

108

Chapter Six

contrary to facts, to find NPIs in subject position if Negation Raising applies: (216)

*Anybody didn't come

However in (216) nothing forces Negation Raising and the problem is only apparent, since Negation Raising should be allowed only to derive scope-shift. A more serious challenge is given by sentence (217): (217)

*Anybody need not come

Here subject NPIs are still illicit, a fact which is unexpected if the wide scope of negation over modality is obtained by Negation Raising in ForceP (218)

ForceP

[OP¬j

AgrSP

[Anybody

ModP

[ need

NegP

[ tj VP[ ...]]]]

Before continuing the discussion, I wish to stress the fact that sentences as (217) do not constitute a counter evidence for the operation of Negation Raising in itself: in fact it is perfectly conceivable that Negation Raising, instead of triggering the topmost structural position, simply triggers a position above modality. However, I believe that (217) doesn’t force us to abandon the link with ForceP and the clause typing hypothesis. In fact, its ungrammaticality may follow from general mechanisms related with the semantic of NPIs. I will consider the problem posed by (217) in section 6.1. The second central issue of this chapter concerns instead the status of n-words in strict negative concord languages. Strict negative concord languages notably differ from asymmetric concord languages in that their n-words exhibit the same behaviour regardless to their clausal position. Russian, Serbo-Croatian, Japanese, Greek and Romanian are well known examples of strict negative concord languages. Three basic features characterize languages belonging to this group and differentiate them from asymmetric negative concord languages: a. n-words are licit only if accompanied by the sentential negative marker53; b. double negation readings are never allowed; c. there is no pre/post verbal asymmetry.

53

Or elements as ‘without’.

Domain Widening and Strict Negative Concord

109

 In the following examples (Brown 1999 for Russian, Giannakidou 1997 for Greek, Progovac 1994 for Serbo-Croatian, Watanabe 2004 for Japanese) I exemplified these three main properties: (219) a. John-wa nani-mo tabe-(nak)*-atta John-top what-mo eat-neg-past ‘John didn’t eat anything’

(Japanese)

b. Dare-mo monku-o wa-(nak)*-atta who-mo complaint-acc say-neg-past ‘nobody complained’ (220) a. (Dhen)* ipe o Pavlos TIPOTA not said.3s the Paul anything ‘Paul said nothing’

(Greek)

b. KANENAS (dhen)* idha anybody not saw.3s ‘I saw nobody’ (221) a. Ja (ne)* videl ni-kogo I neg saw anybody ‘I saw no one’

(Russian)

b. ni-kto (ne)* zvonil anybody neg called ‘No one called’ (222) a. Milan (ne)* vidi nista Milan neg see.3s anything ‘Milan cannot see anything’

(Serbo-Croatian)

b. Ni(t)ko (ne)* vidi Mario anybody neg see.3s Mario ‘Nobody can see Mario’ The examples above show that negation is always required to license nwords, both in preverbal (219b), (220b), (221b), (222b) and in post verbal (220a), (221a), (222a) position. In addition, none of the previous sentences can be possibly interpreted with a double negation reading.

110

Chapter Six

The term strict negative concord ultimately captures the fact that negative concord is obligatory, regardless to the position of the n-word. Furthermore, notice that the effect is maintained for each additional nword we add to the sentence, resulting always in the concord reading, as in (223) and (224) below: (223)

KANENAS *(dhen) ipe POTE TIPOTA KANENAN (Greek) anybody neg said.3s ever anything to anybody ‘nobody ever said anything to anybody’

(224)

Nikto nikogda nigde ni s kemkem *(ne) tancuet (Russian) anybody anytime anywhere with anybody neg dances ‘Nobody dances with anybody at anytime anywhere’

All these examples cast serious doubts on the fact that n-words posses a negative feature on their own, since their interpretations looks similar to multiple NPI readings54 as in (225) (and in the glosses of previous examples) (225)

I did not see anything in any place

Although there are strong similarities between n-words in strict negative concord languages and NPIs, things are not straightforward and n-words cannot be easily reduced to a phenomenon of polarity. In fact, they present some recalcitrant properties. First, any is excluded outside the c-command of negation (226a) while Greek emphatics, considered to be nwords, are licit: (226)

a. *anybody didn’t call b. KANENAS dhen idha anybody neg saw.3s ‘I saw nobody’

(Greek)

Second, NPIs are also forbidden in fragment answers (227a), contrary to Greek emphatics:

54

N-words are licensed only in negative statements, as ‘strong’ NPIs in sentences as the one below: (1) I didn’t earn a red cent in weeks

Domain Widening and Strict Negative Concord

111

 (227)

a. Q: Who did you see? A: *Anybody b. Q: Ti idhes? what saw-2s ‘What did you see?’

(Greek)

A:TIPOTA ‘Nothing’ Third, while NPIs in English can be licensed at long-distance, Greek emphatics must be accompanied with a clause-mate negation (228)

a. John does not claim that Mary knows anybody b *I Ariadhni dhen ipe oti idhe TIPOTA the Ariadne neg said.3s that saw.3s anything ‘Ariadne didn't say that she saw anything’

(Greek)

The previous examples in (226b), (227b) and (228b) show that this series of n-words does not respect the usual distribution of polarity items and the same properties might be found also in other strict negative concord languages. To sum up, n-words in strict concord languages present characteristics of negative quantifiers since they appear in fragments and in subject positions. But it is also evident that they, as polarity items, need a licensor. In section 6.2, I’ll propose an analysis for n-words in strict negative concord languages considering them as a special sort of NPIs, combining insides from Giannakidou (1998, 2000) on the quantificational force of Greek n-words with the semantic of NPI as in Kadmon & Landman (1993) and Chierchia (2004).

6.1 Domain widening in DE contexts The issue of polarity items licensing is a complex one, in that different semantic and syntactic factors seems to be involved. My goal here is not an attempt to disentangle the different forces ruling the distribution of NPIs but to consider a specific view on the surface c-command licensing condition based on a scalar analysis of NPIs. Following Ladusaw (1996), the NPI problem can be approached by decomposing it into three sub-problems: [i] the semantic of polarity items,

112

Chapter Six

[ii] the semantic of the licensor and [iii] the relation which holds between them. I’m mainly interested in [iii] since my goal is to understand why subject NPIs, c-commanding the negative operator, are excluded and why NPIs must be in the surface scope of their licensor. This relation is captured by the well-known descriptive generalization in (229): (229)

An NPI must be licensed in the c-command of an appropriate operator at PF

A desirable results will be to derive (229) from the semantic of the licensor and the licensee, in a way that the point in [iii] will follow from [i] and [ii]. This is the spirit of the proposal in Kadmon & Landman (1993), further developed in Chierchia (2004). I will briefly go through the argument, in order to present the characteristics associated with the semantics of NPI. For what concerns the quantificational force of NPIs, they are considered to be indefinites with existential meaning (Chierchia 2004, Giannakidou 1997, Kadmon & Landman 1993, Ladusaw 1979, Linebarger 1980, Progovac 1994). However, it is clear that they have something more than plain indefinites: some special characteristic which rules them out in positive contexts. The idea of Kadmon and Landman is to relate NPI occurrences of any with special free-choice readings of any as in (230) (230)

Any owl hunts mice

Here any is not in the scope of a DE operator and it is best interpreted as having a universal quantificational force: by uttering (230) we are saying that hunting mice is a property in common to every owl. According to Kadmon and Landman, this result is obtained since any permits to widen the interpretive domain of the quantifier in a way that it will include every possible and relevant owl. In order to illustrate this point, consider the following exchange: (231)

a. Speaker A. Will there be French fries tonight? Speaker B. No, I don’t have potatoes. b. Speaker A. Maybe you have just a couple of potatoes that I could take and fry in my room? Speaker B. Sorry, I don’t have ANY potatoes.

Domain Widening and Strict Negative Concord

113

 The first speaker in (231)a asks for some potatoes, where the context makes relevant a sufficiently large quantity or kind of potatoes: clearly speaker A is asking for a sufficient amount of edible potatoes. The answer of speaker B has to be than intended in relation to this implicit domain. The successive conversational move of A in (231)b is to clarify what is the relevant quantity of potatoes (a couple). The answer of B is such as to include either great or small quantity of potatoes. How any reaches this result? Kadmon and Landman reason in the following way: first, the contextual relevant quantity of potatoes is defined by the domain (D) of the quantifier. It is basically assumed that D = large quantity of potatoes However in the successive exchange in (231)b the dimension of D is explicitly varied and D is enlarged D' = small ∧ large quantity of potatoes the speaker B must now deal with this extended domain and must use an element able to enlarge the domain from D to D’ and any does exactly this trick. Kadmon and Landman call this domain-enlargement operation widening, which is the distinctive feature of NPIs: (232) WIDENING In a NP of the form any CN, any widens the interpretation of the common noun phrase (CN) along the contextual dimension. If we consider domain widening a peculiar feature of NPIs, a natural way to push the analysis will be to relate the semantic properties of the NPIs to the semantic of the licensor, in a way that domain widening will be possible only given a determined semantic context. In doing this, licensor/licensee will be the two faces of the same coin. This amounts to ask why any and its related domain widening function are restricted to certain semantic environments. A way to answer this question requires us to consider the "informativeness" or the "informational strength" of the sentence. According to Kadmon and Landman, widening is not an arbitrary move but must comply with conversational principles. In general, widening should be allowed only when it actively contributes to the felicitousness of the sentence. To

114

Chapter Six

capture this fact, they propose that widening obeys to the following condition: (233) STRENGTHENING Any is licensed only if the widening it introduces creates a stronger statement According to (233), any should be licensed only if it appears under the scope of an operator which creates a statement with stronger truth conditions. Stronger with respect to the equivalent sentence where some is instead used. The distribution of any is then regulated by the interaction of the two principles in (232) and (233). Since the core of Kadmon and Landman proposal has been maintained and further refined in Chierchia (2004), I’ll use his reformulation of strengthening and widening in order to capture the interaction between them. Domain widening is thought as a function g which takes as argument the current domain of quantification D and increases its size to D’ (234) DOMAIN WIDENING (Chierchia 2004) Let g be an increasing function from sets into sets (i.e. for any set D, g(D) ⊇ D). In this framework, NPIs introduce a variable ranging over increasing functions and the function must be universally closed at some point in the derivation by an operator ∀g∈Δ, where Δ is the possible excursion for g(D). This universal closure is tightly connected with the notion of strengthening, and it is possible only when the strengthening condition as in (235) is satisfied: (235) Strengthening/Blocking (Chierchia 2004) Domain expansions must be universally closed. Such closure must lead to strengthening with respect to the meaning of the plain indefinite where strengthening is defined in terms of Horn’s scales (Horn 1989) according to (236)

Domain Widening and Strict Negative Concord

115

 (236) Strengthening formulated via asymmetric entailment Given a set of alternatives A and a member of such set β, Sβ(A) will be the weakest member of A that asymmetrically entails β Asymmetric entailment is based on the idea of a scale whose member are ranked. According to the definition in (236), many asymmetrically entails some and in turn all asymmetrically entails many. These quantifiers are then ranked in strength in this way: all > many > some. Given this premise, we are ready to consider why only operators which create a stronger meaning might license the widening function associated with NPI. Let us consider what happens in positive declarative sentence and compare how the two variants any/some are ordered with respect to (236). In (237a) the standard existential meaning of some is reported, in relation to the domain D. In (237b), instead, the widened existential any is used, associated with the widening function g which is universally closed (237) a. someD(man)(walked in) b. ∀g∈Δ[anyg(D)(man)(walked in)] Consider first (237a). It corresponds to the sentence some man walked in. In uttering this sentence, probably the speaker is referring to the existence of a certain man in the domain D. For example the owner of a restaurant can utter (237a) referring only to potential customers, excluding its employees. The variant in (237b) corresponds instead to the sentence any man walked in, where the domain is widened in order to include mankind in general. This means that (237a) implies the variant (237b) (238)

someD(man)(walked in) ĺ∀g∈Δ[someg(D)(man)(walked in)]

In this case the domain widening function introduced by any does not induce any strengthening. This is what happens in general in contexts upward entailing, where the entailment goes from subset to superset. Consider also (239), where the narrower set of ‘red haired man’ entails the truth of the same preposition where we replace ‘red haired man’ with the wider set of man in general: (239) some red haired man entered the room ||∃x(man x)|| ⊆ ||∃x(red haired man x)|| Ÿ some man entered the room

116

Chapter Six

It is easy to see that in upward entailing contexts no strengthening takes place when we replace D with the widen D’. This correctly predicts that NPIs should be normally excluded in UE contexts. Now, consider what happens in Downward Entailing contexts. Here the picture reverses and entailment goes from supersets to subsets. If we replace under the scope of negation the wider domain of ‘men’ with the narrower domain of ‘red haired men’, we have a stronger statement where the introduced variant is entailed by the original sentence (240) It is not true that some man entered the room ||∃x (man x)|| ⊆ ||∃x (red haired man x)|| Ÿ it is not true that some red haired man entered the room We then expect that in negative statement the strengthening condition in (236) will be satisfied and that the any variant in (241a) should result stronger than the plain existential some (241b) (241) a. ¬ someD(man)(walked in) b. ∀g∈Δ [¬ someg(D)(man)(walked in) ] This is confirmed by the fact that (241b) entails (241a): (242)

∀g∈Δ[¬someg(D)(man)(walked in)]ĺ¬someD(man)(walked in)

At this point, through a semantic characterization of the properties of the licensor and the licensee, we have a way to capture their relationship and the original descriptive generalization proposed in Ladusaw (1979): (243)

Licensing condition for PIs: Į is a trigger for negative polarity items in its scope if Į is downward entailing

The merit of Kadmon & Landman (1993) and Chierchia (2004) proposal is to have provided an explanation, based on the meaning of NPIs, to the descriptive generalization in (243). Let us consider now how this mechanism is built in the derivation. This lead us to the point [iii], which is directly relevant for dealing with subject NPIs. We saw that an NPI introduces the function g(D) which is roofed by a universal quantifier ∀g (where ∈Δ indicates the range of possible variation of D). But this universal closure is subject to the strengthening constraint

Domain Widening and Strict Negative Concord

117

 which is satisfied only in DE contexts. This means that universal closure, to be successful, must wait until the right operator has been introduced. To illustrate how the system works, take sentence (244) with an object NPI (244)

John does not see anyone

Assuming that the sentential negative marker sits in NegP, universal closure has to be inserted after the negative operator, only when the DE context has been created: (244)’ AgrP Johni

XP NegP OP∀g not OP¬

VP

ti see

anyone ∃g(D)

This means that universal closure must wait until the proper DE operator has been encountered. But how long is it necessary to wait? Until the end of the sentence or as soon as possible? Although it is not easy to address the matter from an empirical angle, there are certain cases which suggest an “as soon as possible” view. According to Chierchia, one piece of evidence comes from double negative sentences. Consider the sentences in (245): (245) a. It is not true that there aren’t any potatoes b. There are potatoes The presence of two negations in (245a) makes it truth conditionally equivalent to (245b), where the two negative operators cancel each other out (245)a: ¬ ¬∃x [potatoes(x)]

=

(245)b: ∃x [potatoes(x)]

But notice that, once the global meaning of (245a) is computed, the context is not DE anymore: NPIs should be unlicensed, contrary to facts. This suggests that universal closure is inserted as soon as the first DE

118

Chapter Six

operator is encountered and once the relation is satisfied, it will be impermeable to successive changes. The representation in (244’) seems to be substantially adequate and supports the idea that ∀g is inserted as soon as possible in the first available scope site. We are finally in the position to consider the problem posed by subject NPIs and look back at our first examples in (216) and (217), repeated here as (246) and (247) (246) *Anybody didn't come (247) *Anybody need not come Sentence (246) is derived in a way similar to (244) for what concerns both the position of negation and the successive insertion of ∀g. The only difference is the final position of the NPI. Assuming that subjects are generated VP internally, the structure of (246) is the following: (246)’ AgrP anyonei ∃xg (D)

TP XP

did OP∀g

NegP VP

not OP¬ ti

come

Here the domain variable g c-commands ∀g, its binder, violating usual constraints on vacuous quantification. And the same thing happens in (247)

Domain Widening and Strict Negative Concord

119

 (247)’ AgrP TP

anyonei ∃xg (D)

XP

need OPƑ OP∀g

NegP VP

not OP¬

come

ti

Even if we apply Negation Raising, OP∀g will be again c-commanded by its variable and the representation is ruled out: (247)’’ ForceP AgrP

OP¬

TP

anyonei ∃xg (D)

XP

need OPƑ

NegP

OP∀g

VP

not OP¬ ti

come

From this general mechanism associated with domain widening, the general requirement of overt c-command follows. The impossibility of a configuration as (248) will ultimately exclude NPIs in subject positions and in the left periphery in general. (248)

Anyg(D)

∀g OP ¬

t

Now, we are left with the second issue, namely with the treatment of nwords in strict negative concord languages. As we saw in the beginning of this chapter, a constraint as the one blocking NPIs c-commanding their DE operator is not active for this kind of n-words. This means that we need to say something more to capture the dependency between n-words and the

120

Chapter Six

sentential negative marker. In the following section, I’ll try follow the same logic used for NPIs, trying to derive the distributional constraints from the semantic of the n-words.

6.2. Strict negative concord N-words in strict negative concord languages are remarkably different from NPIs in that thay may appear outside the surface c-command domain of the negative operator. This means that the mechanism ruling the ccommand requirement for NPI does not work for n-words. A first possibility is to abandon the NPI analysis. But we saw that this is not trivial, at least if we embrace the alternative analysis considering them as negative quantifiers. In this case, their dependency from the sentential negative marker will be problematic. A second option is instead to pursue the polarity account, trying to derive the different distribution of n-words by considering their special semantics. I’ll try to follow this direction, inspired by the work of Giannakidou (1997, 2000) on Greek emphatics. In particular, the observations on the quantificational force of this specific set of n-words could help us to shed light on strict negative concord in general55. Let me briefly recapitulate the gist of the proposal presented in the previous section. We saw that NPIs can be considered as associated with a domain widening function and that their distribution follows from the properties of DE contexts. These are the two core elements of the proposal: a DE licensing operator and a strengthening requirement on domain widening functions. Notice that very few relies on the quantificational strength of NPIs and although NPIs might be considered to be built on the shoulders of indefinites, nothing prevents that other elements, with a different quantificational force, develop polar dependencies. Putting it in a nutshell, there is no reason to associate g only to indefinites. In fact, it is perfectly plausible to imagine that also universal quantifiers might be associated with a domain widening function and carry the variable g. I want to recast n-words in strict negative concord languages exactly in these terms. We know that existential and universal quantifiers might be mutually exchangeable in combination with negation only by varying the relative 55

Although the analysis could be extended to other strict negative concord languages, this will require a careful cross-linguistic analysis of the properties of the different series of n-words. I leave this for further research, limiting the discussion to Greek emphatics only (see Moscati 2006a for Italian).

Domain Widening and Strict Negative Concord

121

 scope of the operators. This means that the two sentences below (249) and (250) are truth conditionally equivalent: (249)

It does not exist an immortal man

(250)

All the man are not immortal

(251)

¬∃x ( man(x) ∧ immortal(x)) = ∀x ( man(x) ∧ ¬immortal(x))

In order to obtain negative polarity items with universal quantificational force, we can associate also to universal quantifiers a domain widening function. A universal NPI will be an item as follows: (252)

∀xg(D)

The system will work exactly as in the case of existentials: universal NPIs associated with g are licensed only if they respect strengthening. This means that the domain widening function obeys to the very same conditions illustrated in the previous section. It follows that ‘every’ might be substituted by ‘every g(D)’ only if this last option asymmetrically entails the former. Let us see what happens in UE contexts: (253) a. everyD(man)(walked in) b. ∀g∈Δ[everyg(D)(man)(walked in)] in (253a) the domain taken into consideration is the one made salient by the discourse, while (253b) constitute the ‘widened’ variant. It is easy to see that in UE contexts, where the entailment goes from sets to supersets, (253a) asymmetrically entails (253b) (254) everyD(man)(walked in)ĺ∀g∈Δ[everyg(D)(man)(walked in)] For this reason, the widened version of ‘everyg(D)’ does not respect the condition imposed by strengthening and it is excluded. In the DE contexts, instead, where the entailment goes from supersets to subsets, the picture reverses. Consider again the two variants, both taking wide scope over negation56: 56

Importantly, this is required to keep the parallel with the truth conditions of existentials by virtue of the equivalence in (251).

122

Chapter Six

(255) a. everyD(man) ¬(walked in) b. ∀g∈Δ [every g(D)(man) ¬(walked in)] here the entailment between (255a) and (255b) reverses: (256)

∀g∈Δ[everyg(D) (man) ¬(walked in)]ĺ everyD (man) ¬(walked in)

At this point it seems that the same mechanism proposed by Kadmon & Landman (1993) and Chierchia (2004) can be applied with minor modifies also to universal quantifiers. This is true in abstract. Can we find empirical support to the existence of polar universal quantifiers? In the case of existentials, we easily find semantic differences encoded in the lexicon. A parallel case for universal will support the existence of universal NPIs. Furthermore, an appropriate analysis should capture at least other two basic properties of n-words. The first is that n-words differentiate from NPIs not only for the absence of c-command between the licensor and the licensee, but also for the fact that this relation is clause bounded. The second property is that, while NPIs are disallowed in fragments, n-words can be felicitously uttered even without pronouncing the negative marker. I’ll consider next the case of Greek and the set of properties related to the emphatic series.

6.3. Greek emphatics as negative polarity universal quantifiers According to Giannakidou (2000), Greek can be considered to belong to the strict negative concord language group. In particular, she considered a special class of Greek n-words as universal quantifiers with a polar dependency. Leaving aside her characterization of the appropriate licensing operator, Greek n-words look very similar to polarity items, even if with some major differences in their distribution. Again, a welcome result would be to derive these differences from their semantics. Before turning our attention to the interpretive properties of this class of items, a preliminary distinction has to be made in relation to the n-word paradigm. This set of elements may or may not carry ‘emphatic’ accent, varying their distribution in accordance to this special kind of stress. Giannakidou treats emphatic accent as a special kind of morphological marking, distinguishing in this way two sets of n-words:

Domain Widening and Strict Negative Concord

123

 (257) kanenas kanenas N tipota pote puthena

KANENAS KANENAS N TIPOTA POTE PUTHENA

‘anybody’ ‘any N’ ‘anything’ ‘ever’ ‘anywhere’

where emphasis is reported in (257) with capital letters. An important difference between emphatics and non-emphatics is that only this last set of items respects the distribution of weak NPIs. Emphatics present instead a different behavior and they have at least three properties which differentiate them from non-emphatics and NPIs: [i] emphatics do not require overt c-command by their licensor; [ii] they might appear in fragments; [iii] they require a clause-mate negation. These properties are exemplified by the following examples: (258)

KANENAS dhen idha n-person neg saw.3s ‘I saw nobody’

(259)

Q: Ti idhes? ‘What did you see?’ A:TIPOTA ‘nothing’

(260) *I Ariadhni dhen ipe oti idhe TIPOTA the Ariadne neg said.3s that saw.3s n-thing ‘Ariadne didn't say that she saw anything’ In (258) the emphatic appears in preverbal position, c-commanding the sentential negative marker. Example (259) shows instead that emphatics are admitted in fragment answers. Finally, the last example (260) is ungrammatical since the n-word is in the subordinate clause while negation appears in the matrix. Given this distribution, although emphatics are dependent from negation, they cannot be considered weak NPIs. However, NPIs are not a homogeneous class and elements with different licensing properties might be included in the set of polarity items (van der Wouden 1997). I believe, following Giannakidou, that also Greek emphatics could be considered as a special case of NPI and that their distribution might follow from the semantics. I’ll try to follow this idea, keeping the main ingredients of Chierchia’s analysis and introducing only

124

Chapter Six

a minimal variation. The widening function, the strengthening condition and the DE contexts are the core features of the proposal and must be left unvaried. However, very few relies on the quantificational force and in Chierchia’s system nothing excludes the possibility to apply domain widening functions to universal quantifiers. Giannakidou (2000) proposed that Greek emphatics should be treated as universal quantifiers and I report here some of her arguments, remanding to the original work for a complete discussion. A first analogy between negative emphatics and universal quantifiers concerns the modification by adverbials as almost57. As the English (261) shows, almost is grammatical only with universals and it is excluded with existentials: (261) a. *Electra was willing to accept almost something b. Electra was willing to accept almost everything If the semantic of adverbials as almost allows them to combine only with universal quantifiers, we expect that Greek emphatics could be modified by such adverbs if they have universal force. This is exactly what happens: (262)

Dhen idha sxedhon KANENAN/*kanenan neg saw.1s almost n-person ‘I saw almost nobody’

Notice that, while emphatics behave as universal quantifiers, nonemphatics are not allowed to be modified by almost, suggesting that they are instead indefinites. Apart from almost, other adverbials reveal to be sensitive to quantificational force. Some of them, for example, only select indefinites. For example, ‘ook maar’ in Dutch or ‘anche solo’ in Italian, both used to express small quantities, combine only with indefinites and existential. Consider ‘ook maar’ in (263): the example shows that it can only modify the existential iemand and that it is excluded with universal quantifiers as iedereen. (263)

57

Wil jij ook maar iemand / *iedereen zien? want.2s too only somebody / everybody see ‘Do you want to see anybody?’

On the reliability of these tests, see also Giannakidou (2000) but also Horn and Lee (1995) and Depréz (1997).

Domain Widening and Strict Negative Concord

125

 If Greek emphatics are to be treated as universals, we expect that they will be excluded with the Greek equivalent of ‘ook maar’. The adverbial ‘ke’ has exactly this meaning and it is excluded with emphatics, as illustrated in (264) (264)

Dhen ipe ke tipota/*TIPOTA spudheo neg said.3s adv n-thing important ‘He didn't see anything important’

The examples (262) and (264), based on adverbial modification, are not the only kind of evidence showing a parallel between emphatics and universal quantifiers. Among other symmetries, anaphora resolution also indicates that emphatics in Greek behave as universals, not being able to be a suitable antecedent for a pronoun. In (265a) the pronoun to ‘it’ cannot be referred to kathe vivlio ‘every book’ and the same behavior can be found if we substitute the universal with an emphatic, as in (265b) (265) a. *I fitites pu aghorasan kathe vivlio1, na to1 ferun mazi tus the students that bought.3p every book su it bring with them b. *I fitites pu dhen exun TIPOTA1 na pun, as min to1 pun tora the students that have nothing1 to say, let them not say it1 now If we consider the similar behaviour between emphatics and universal quantifiers exemplified in the examples (264) and (265), we might account for the first property of Greek n-words, namely the need for a local dependency between n-words and a clause-mate negation. Emphatics, as universal quantifiers, must respect the constraints in force on quantifier raising, which is clause bounded. Consider the following Greek example: (266) Kapjos fititis ipe oti kathe kathijitis tis sxolis apolithike some student said that every professor the department got-fired.3s ‘Some student said that every professor in the department got fired’ this example can be only interpreted according to the meaning where a single student says that all the professors have been fired. Thus kathe kathijitis ‘every professor’ cannot take scope over the existential quantifier. This is a familiar phenomenon suggesting that QR and the scope of universals is clause-bounded.

126

Chapter Six

If emphatics are licensed through a mechanism related to QR, we expect that they should be excluded where QR is impossible. And in fact, emphatics are excluded if a clause boundary separates them from negation: (267)

*I Ariadhni dhen ipe oti idhe *TIPOTA/tipota the Ariadne neg said.3s that saw.3s n-thing ‘Ariadne didn't say that she saw anything.’

Sentence (267) shows that emphatics are forbidden in clausal complements introduced by the indicative complementizer oti58. Notice that this block is not absolute but selective on the quantificational force, since in (267) non emphatics are perfectly licit. The question is why emphatics should be related to QR. I believe that the connection is not arbitrary. Consider the desired interpretation of nwords. If we assume that they are universal quantifiers, the only reading allowed is ‘every not’ in which the universal quantifier takes wide scope over negation. This is the exact opposite of the scope relation between polar indefinites and negation, where negation takes wide scope giving the only possible interpretation ‘not any’. Notice that these two equivalent readings ‘every not’ and ‘not any’ are special in that they constitute the more informative extreme of a ‘Horn scale’ only in DE contexts. Under this view, universal n-words must escape from the c-command domain of negation and undergo to QR. However, this is only a first useful hint to account for the lack of the surface c-command requirement between n-words and negation and things are more complicate than that. In fact, the domain widening function is possible only once a DE context is introduced. This means that universal closure must wait until an appropriate operator has been encountered, in a position c-commanding

58

The constraint on clause boundedness disappears in subjunctive clausal complements introduced by the complementizer na: (1) I Ariadhni dhen theli na dhi KANENAN the Ariadne neg want.3s su see.3s n-person ‘Ariadne doesn’t want to see anybody.’ In this same environment, also QR is possible: (2) Kapjos kathijitis frondise kathe fititis s’ afti ti lista na vri dhulja some professor made-sure.3s every student in this the list su find.3s job Some professor made sure that every student in this list will find a job Notice that also here emphatics can be licensed, behaving again as universal quantifiers.

Domain Widening and Strict Negative Concord

127

 the variable g. Basically, we must account for the fact that while g should be in the scope of the negative operator, the quantifier must be outside. Let me take stock here and look again to the various ingredients of the system I’m proposing. So far I followed Giannakidou’s proposal in considering Greek emphatics as universal quantifiers with a polar dependency. Now, this polarity can be casted in the general ‘widening’ mechanism by saying that emphatics are paired with a domain widening function g(D) which must be universally closed by ∀g. Remember that ∀g must be inserted as soon as possible, but not too soon: it must wait until negative features are inserted in the derivation. The key question is where the earliest negative feature is encountered. In the case of n-words, it is plausible that this happens inside the DP, as suggested by the fact that n-words generally bear negative morphology. What I’m proposing here is to consider n-words as DPs with a domain widening function plus a negative feature as in (268): (268) DP neg

KANENAS ∀x g(D)

This negative feature does not immediately correspond to an interpretable negative operator, since we know that it not sufficient to convey sentential negation by itself. This means that this feature is a formal one and that it is dependent from the sentential negative marker (which has an interpretable negative feature). We need a way to capture this dependency and the system introduced in chapter four offers us an immediate way to do this. The negative feature of the n-word can be considered as to be only a formal feature u[val] which has to be interpreted by establishing a link with the interpretable one carried by the sentential negative marker. We can refine the structure in (268) labeling the negative feature as u[val] (269) DP neg u[val]

KANENAS ∀x g(D)

The presence of this uninterpretable negative feature insures that another interpretable one must be inserted in order for the derivation to converge. From this, it follows the necessity to have the negative marker.

128

Chapter Six

At this point, the system knows that a DE context is going to be created: we can be sure, already within the DP, that a proper DE operator will be inserted. This will satisfy the necessary requirement of strengthening within the clause boundary. If the ‘promise’ of an adequate DE operator is sufficient, universal closure will be possible very soon, already in the DP just above [+neg u[val]]. The final structure of the n-word DP which I’m proposing is the following (270) DP ∀g neg u[val]

KANENAS ∀x g(D)

At this point, given the DP structure above, we may see that the general constraint in (248), avoiding indefinites NPIs to move across the universal closure is not violated. In fact, the requirement of surface c-command was derived from the impossibility from the variable g to move outside the ccommand of its binder. However, if the binder and the bindee move together, nothing will block movement of the DP in (270). To make this point clearer, consider the sentence below (271)

KANENAS dhen idha n-person neg saw.3s ‘I saw nobody.’

here the internal arguments moves from its VP internal position to some projection (Topic) outside the VP. In (271) it is the whole DP which is moved, with the widening function already universally closed. The structure of (271) is represented in (271)’ below:

Domain Widening and Strict Negative Concord

129

 (271)’ TopP NegP

DPi

TP

dhen neg i[val]

VP

idhak

∀g neg u[val]

KANENAS ∀x g(D)

pro tk

ti

The representation in (271’) shows that universal closure has been made already possible by the negative feature internal to the DP. This feature signals that the appropriate DE context will be soon created, since the negative feature in the DP is in need for interpretability. From the properties of this system, it follows that movement above negation is possible and necessary for Greek emphatics, in order to have the strong DE reading ‘every not’. Emphatics must then escape from the scope of negation, overtly or covertly. Given this state of affairs, emphatics are free to appear in structural positions higher that NegP and the fact that that they are licit in subject position or in the left periphery of the clause will be unsurprising. This also explains why they are possible in fragment answers, where the IP has been phonologically deleted by sluicing.

Conclusion This chapter was intended to conclude the book by considering the distribution of NPIs in relation to Negation Raising and to complete the discussion on negative nominal constituents extending the analysis also to strict negative concord languages. The two issues have been grouped together on the basis of an important similarity between NPIs and nwords: the fact that they both require a licensor. However, this similarity is counterweighted by some relevant differences, mainly related to the absence of overt c-command in the case of n-words. The common need of a licensor can be derived from the system proposed in Kadmon & Landman (1993) and Chierchia (2004). The mechanism required to satisfy the ‘strengthening condition’ could be the

130

Chapter Six

shared characteristic between NPIs and n-words in strict negative concord languages. As we saw, they both require a DE context to get strengthened sentences. For what concerns the differences, a first observation is that usually nwords do have negative morphology, suggesting that they carry some sort of negative feature. I suggested that it is exactly this feature that influences the distribution of n-words, in that it gives to the derivational system an indication that we are in a DE context and that a lexical item associated with a domain widening function can be inserted. All the processes and the dependencies described in Chierchia (2004) are satisfied within the n-DP. This means that, as far it concerns polar dependencies, the n-DP it is free to move everywhere in the clause, even in a position c-commanding the sentential negative marker. The necessity to have negation, in the case of n-words, follows instead from the fact that they have only a formal negative feature. This feature needs to enter in a chain-relation with an interpretable negative feature, which is typically carried by the negative marker. If this proposal is on the right track, it is possible that n-words have been diachronically derived from universal quantifiers. This point deserves further attention, but if diachronic development will be attested, we could assume a general mechanism able to associate domain widening function to quantifiers: it takes both indefinites and universal quantifiers as an input and it gives back NPIs on one side and n-words on the other.

REFERENCES

Aboh, Enoch Olade. 1999. From the Syntax of Gungbe to the Grammar of Gbe. Editions à la Carte SA. Sierre, Suisse. —. 2002. Morphosyntaxe de la péripherie gauche nominale. In Anne Zribi-Hertz & Anne Daladier eds., Recherches linguistiques de Vincennes 31: 9–26, La syntaxe de la definitude. PUV, Paris. Acquaviva, Paolo. 1997. The Logical Form of Negation. New York: Garland. —. 1999. Negation and operator dependencies: Evidence from Italian. Lingua 108:137–174. Aissen, Judith and David Perlmutter. 1983. Clause Reduction in Spanish, in D. Perlmutter ed., Studies in Relational Grammar 1. Chicago, University of Chicago Press. Aoun, Joseph and Li, Yen-hui Audrey. 1993. The Syntax of Scope, Cambridge, Massachusetts, MIT Press. Baker, Mark. 1985. The Mirror Principle and morphosyntactic explanation. Linguistic Inquiry 16: 373-415. Bayer, Josef. 1990. What Bavarian Negative Concord Reveals about the Syntactic Structure of German. In Juan Mascaró and Marina Nespor eds., Grammar in Progress, pp. 13-24, Foris Publications, Dordrecht. Bech, Gunnar. 1955/57. Studien über das deutsche verbum infinitum. København: Det Kongelige Danske Akademie av Videnskaberne. Belletti, Adriana. 1990. Generalized Verb Movement. Turin, Rosenberg & Sellier. —. 2002. Aspects of the low IP area. In L. Rizzi ed., The Structure of IP and CP. The Cartography of Syntactic Structures. vol.2., OUP. —. 2003. Extended doubling and the VP periphery. Ms. University of Siena Beghelli, Filippo and Tim Stowell. 1997. Distributivity and negation: The syntax of each and every. In Anna Szabolcsi ed., Ways of scope taking, 71–107. Dordrecht: Kluwer. Beretta, Claudio. 1980. Contributo per una grammatica del Milanese contemporaneo. Circolo Filologico Milanese, Milano Bianchi, Valentina. 1999. Consequences of Antisymmetry: Headed Relative Clauses. Berlino, Mouton de Gruyter.

132

References

Blaszczak, Joanna. 1999. The Riddle of n-words in Polish. Ms. Humbolt Universität Berlin. Borer, Hagit. 1984. Parametric Syntax: Case Studies in Semitic and Romance Languages. Dordrecht: Foris. Boškovic, Željko. 1999. On multiple feature-checking: Multiple Whfronting and multiple headmovement. In S. Epstein and N. Hornstein (eds), Working Minimalism, 159-187. Cambridge, Mass.: MIT Press. Brody, Michael. 1997. Perfect chains. In Elements of grammar. Liliane Haegeman ed., 139-167. Dordrecht: Kluwer Academic Publishers. Brown, Keith. 1990. Double Modals in Hawick Scots. In Peter Trudgill and J. K. Chambers (eds.) Dialects of English: Studies in Grammatical Variation. London: Longman, 74-103. Brown, Sue. 1999. The syntax of Negation in Russian. A Minimalist Approach. CSLI Stanford Publications. Büring, Daniel. 1997. The Great Scope Inversion Conspiracy. Linguistics & Philosophy 20. 175- 194. Burzio, Luigi. 1986. Italian Syntax. Reidel, Dordrecht. Carnie, Andrew, Heidi Harley and Elizabeth Pyatt. 2000. VSO order as Raising to out of IP. In Carnie and Guilfoyle eds., The Syntax of Verb Initial Languages. Oxford University Press. 39-60. Cecchetto Carlo. 2001. Syntactic or Semantic Reconstruction? Evidence from Pseudoclefts and Clitic Left Dislocation. In C. Cecchetto, G. Chierchia e M. T. Guasti eds., Semantic Interfaces. CSLI publications, Stanford University. Cheng, Lisa. 1991. On the Typology of Wh-questions. Cambridge, Mass.: PhD Dissertation MIT. Cheng, Lisa, James Huang and Jane Tang. 1996. Negative particle questions: A dialectal comparison. In James R. Black and Virginia Motapanyane eds., Microparameter Syntax and Dialect Variation , 4178. Amsterdam: John Benjamins. Chierchia, Gennaro. 2004. Scalar implicatures, polarity phenomena, and the syntax/pragmatics interface. In A. Belletti ed., Structures and beyond. Oxford: OUP. Chomsky, Noam. 1995. The Minimalist Program. Cambridge, MA: The MIT Press. —. 1998. Minimalist Inquiries: the Framework. MIT Occasional Papers in Linguistics 15, MIT. —. 2001. Derivation by phase. In Ken Hale: A life in language, In Michael Kenstowicz ed., 1–52. Cambridge, Mass.: MIT Press.

Negation Raising: Logic Form and Linguistic Variation

133

Christensen, Ken Ramshøj. 2004. On the Synchronic and Diachronic Status of the Negative Adverbial ikke/not. Working Papers in Scandinavian Syntax, WPSS 72, 1-53, December 2003. —. 2005. Interfaces: Negation - Syntax - Brain. PhD dissertation, Department of English, University of Aarhus, Denmark. Chumbow, S. Beban and Pius N. Tamanji. 1994. Bafut. In Peter Kahrel and René van den Berg eds., Typological Studies in Negation. Typological Studies in Language 29. Amsterdam: Benjamins, 211– 236. Cinque, Guglielmo. 1994. On the Evidence for Partial N-movement in the Romance DP. In G.Cinque, J.Koster, J.-Y.Pollock, L.Rizzi and R.Zanuttini eds., Paths Towards Universal Grammar. Studies in Honor of Richard S. Kayne. 85-110. Washington (D.C.): Georgetown University Press —. 1999. Adverbs and functional heads: A cross-linguistic perspective. Oxford: OUP. —. 2001. "Restructuring" and functional structure. In Laura Brugè ed., University of Venice Working Papers in Linguistics 11, 45-127. Venice, University of Venice. —. 2004. Restructuring and functional structure, in A. Belletti ed., Structures and beyond: the cartography of syntactic structures. New York, Oxford University Press. —. 2006. Restructuring and Functional Heads: the cartography of syntactic structure, Vol.4. Oxford University Press. Cinque, Guglielmo and Luigi Rizzi. 2008. The Cartography of Syntactic Structures, in Vincenzo Moscati ed., STiL Studies in Linguistics – CISCL working papers Vol.2, 43-59. Collins, Chris and Thráinsson, Höskuldur. 1996. VP internal structure and object shift in Icelandic. Linguistic Inquiry, 27 (3), 391-444. Corbin, Francis and Lucia Tovena. 2001. On multiple expression of negation in Romance. In D’Hulst et al. eds., Going Romance 1999, Selected Papers. Amsterdam, John Benjamins. Culicover, Peter. 1991. Topicalization, Inversion and Complementizers in English. In D. Delfitto, M. Everaert, A. Evers and F. Stuurman eds., OTS Working Papers. Going Romance and Beyond, 1-45. Department of Linguistics, University of Utrech —. 1996. On distinguishing A'-movements. Linguistic Inquiry 27: 445-63. Culicover, Peter and Robert Levine. 2001. Stylistic inversion and the ‘thatt’ effect in English: a reconsideration. Natural Language and Linguistic Theory. 19.2.283-310. Dahl, Östen. 1979. Typology of sentence negation. Linguistics, 17.79-106.

134

References

—. In press. Typology of negation, in Horn, L. R. ed., The Expression of Negation. Mouton, Berlin. De Haan, Ferdinand. 1997. The interaction of modality and negation. A typological study, Garland, New York. Delahunty, Gerald. 1983. But sentential subjects do exist. Linguistic Analysis, 12, 379-398. Den Besten, Hans. 1986. Double Negation and the Genesis of Afrikaans. in Pieter Muysken & Norval Smith eds., Substrata versus Universals in Creole Genesis. Papers from the Amsterdam Creole Workshop, April 1985. Amsterdam/Phildelphia: Benjamins. Creole Language Library 1: 185-230. —. 1989. Studies in West-Germanic syntax. Dissertation, University of Tilburg. Den Dikken, Marcel. 1996. The Minimal Links of Verb (Projection) Raising. In Abraham,W., Epstein, S., Thrf nsson, H. & Zwart, J.-W. eds., Minimal Ideas. Amsterdam: John Benjamins. Déprez, Viviane. 1997. A Non-unified Analysis of Negative Concord, in Danielle Forget et al. eds., Negation and Polarity: Syntax and Semantics, John Benjamins, Amsterdam, pp. 53–75. —. 2000. Parallel (A)symmetries and the internal structure of negative expressions. Natural Language and Linguistic Theory 18: 253-342. Diesing, Molly. 1992. Indefinites. Cambridge, MA: MIT Press. Drubig, Hans Bernhard. 2001. On the syntactic form of epistemic modality.www.sfb441.uni-tuebingen.de/b2/papers/DrubigModality. Duffield, Nigel. 1995. Particles and projections in Irish syntax. Dordrecht: Kluwer. Ellegård, Alvar. 1953. The Auxiliary Do: the Establishment and Regulation of its Use in English. Gothenburg Studies in English, vol.II. Stockholm, Almqvist, Wiksell: Göteborg. Enç, Mürvet. 1996. Tense and Modality. In Shalom Lappin ed., The Handbook of Contemporary Semantic Theory. Oxford: Blackwell, 345358. Espinal Maria Teresa. 2000. On the semantic status of n-words in Catalan and Spanish. Lingua, 110, 557-580. Fanselow, Gisbert. 2001. Features, theta-roles, and free constituent order. Linguistic Inquiry 32:405-437. Fintel, Kai von. 1999. NPI-Licensing, Strawson-Entailment, and ContextDependency. Journal of Semantics, 16(2): 97–148 Fintel, Kai von and Sabine Iatridou. 2003. Epistemic Containment. Linguistic Inquiry, Vol. 34, Number 2, Spring 2003 173–198.

Negation Raising: Logic Form and Linguistic Variation

135

Fox, Danny. 1995. Economy and Scope. Natural Language Semantics, 3, 283-341. —. 2000. Economy and semantic interpretation. Cambridge, Mass.: MIT Press. Frampton, John and Sam Gutmann. 2000. Agreement is feature sharing. Ms.North.Univ.Boston. www.math.neu.edu/ling/pdffiles/agrisfs.pdf. Geurts, Bart. 1996. On No. Journal of Semantics 13, 67-86. Giannakidou, Anastasia. 1997. The Landscape of Polarity Items. Ph.D. Dissertation, University of Groningen. —. 1998. Polarity Sensitivity as (Non)Veridical Dependency. John Benjamins, Amsterdam. —. 2000. Negative ...Concord? Natural Language and Linguistic Theory 18: 457-523. Ginzburg, Jonathan and Ivan Sag. 2000. Interrogative investigations: The form, meaning, and use of English interrogatives. Center for the Study of Language and Information: Stanford, Ca. Giorgi, Alessandra and Pianesi, Fabio. 2001. Imperfect Dreams, Probus, 12, 3 Giorgi, Alessandra and Pianesi, Fabio. 1997. Tense and Aspect: From Semantics to Morphosyntax. Oxford: OUP. Gordon, Lynn. 1986. Maricopa Morphology and Syntax. University of California Publications in Linguistics 108. Berkeley: University of California Press. Grewendorf, Gunther and Joachim Sabel. 1999. Scrambling in German and Japanese: Adjunction versus Multiple Specifiers. Natural Language and Linguistic Theory 17: 1-65. Grimshaw, Jane. 1979. Complement selection and the lexicon. Linguistic Inquiry 10, 279-326 Haegeman, Liliane. 1995. The Syntax of Negation. Cambridge University Press. —. 1998a. V-Positions and the Middle Field in West Flemish. Syntax 1: 259-299. —. 1998b. Verb Movement in Embedded Clauses in West Flemish. Linguistic Inquiry 29, 631-656. —. 2000. Inversion, non-adjacent-inversion, and adjuncts in CP. Transactions of the Philological Society 98. 121–160. —. 2000a. Remnant Movement and OV-Order. In Svenonius, P. ed., The Derivation of VO and OV. Amsterdam: John Benjamins, pp. 69-96. —. 2000b. Negative Preposing, Negative Inversion, and the Split CP. In Horn, L. & Kato, Y. eds., Negation and Polarity. Oxford: Oxford University Press, pp. 21-61.

136

References

—. 2001. Antisymmetsy and Verb-Final Order in West Flemish. Journal of Comparative Germanic Linguistics 3, 207-232. —. 2002. West Flemish Negation and the Derivation of SOV-Order in West Germanic, Nordic Journal of Linguistics 25 (2), 154-89 Haegeman, Liliane and Raffaella Zanuttini. 1991. Negative heads and the negative criterion. The Linguistic Review 8, 233-251. Haegeman, Liliane and Raffaella Zanuttini. 1996. Negative concord in West Flemish. In Adriana Belletti and Luigi Rizzi eds., Parameters and functional heads, 117–179. Oxford: Oxford University Press. Haider, Hubert. 2003. V-clustering and clause union: Causes and effects. In Verb constructions in German and Dutch. In Pieter Seuren and Gerard Kempen eds., 91-126. Amsterdam/Philadelphia: John Benjamins. Hamann, Cornelia. 1993. Notes on Negation in German. MS University of Geneva Han Chung-hye. 2000. The evolution of do-support in English imperatives. In Pintzuk, Tsoulas and Warner, eds., Diachronic Syntax: Models and Mechanisms. Oxford University Press. Hankamer, Jorge. 1979. Deletion in coordinate structures. Garland: New York. Haspelmath, Martin. 1997. Indefinite Pronouns. Clarendon Press, Oxford Heim, Irene. 1982. The semantics of definite and indefinite noun phrases. Doctoral dissertation, University of Massachusetts, Amherst. Heim, Irene and Angelika Kratzer. 1998. Semantics in Generative Grammar. Basil Blackwell, Oxford. Hinterholzl, Roland. 1999. Restructuring Infinitives and the Theory of Complementation. Doctoral dissertation, University of Southern California. Hiraiwa, Ken. 2005. Dimensions of symmetry in syntax: agreement and clausal architecture.Unpublished Ph.D. dissertation, MIT. Hoeksema, Jack. 2000. Negative Polarity Items: Triggering, Scope and CCommand. In L.R. Horn and Y. Kato eds., Negation and Polarity. Syntactic and Semantic Perspectives. Oxford University Press, Oxford, 115-146. Hoijer, Harry. 1933. Tonkawa, an Indian language of Texas. Handbook of American Indian Languages, vol. III. New York: Columbia University Press. Holmberg, Anders. 1999. Remarks on Holmberg’s generalization. Studia Linguistica 53, 1-39. —. 2001. The Syntax of Yes and No in Finnish. Studia Linguistica 55, 141-175.

Negation Raising: Logic Form and Linguistic Variation

137

Horn, R. Laurence 1972. On the semantic properties of logical operators in English. Mimeo, Indiana University Linguistics Club, Bloomington, IN. Horn, R. Laurence. 1989. A Natural History of Negation. Chicago: University of Chicago Press. Horn, R. Laurence and Yung-Suk Lee. 1995. Diagnosing the Diagnostics. Unpublished manuscript, Yale University. Hornstein, Norbert. 1995. The Grammar of Logical Form from GB to Minimalism. London: Blackwell. Huang, C.-T. James. 1993. Reconstruction and the structure of VP: Some theoretical consequences. Linguistic Inquiry 24: 103-138. Huddleston, Rodney. 1984. Introduction to the Grammar of English. Cambridge: Cambridge University Press. Jackendoff, S. Ray. 1972. Semantic Interpretation in Generative Grammar. MIT Press. Jacobs, Joachim. 1980. Lexical Decomposition in Montague Grammar. Theoretical Linguistics 7, 121-136. Jäger, Agnes. 2008. History of German negation. Amsterdam/ Benjamins. Jacobs, Joachim. 1980. Lexical Decomposition in Montague Grammar. Theoretical Linguistics 7, 121-136. —. 1991. Negation. In A. von Stechow & D. Wunderlich eds., Semantics: An International Handbook. Berlin: de Gruyter, 560-596. Jespersen, Otto. 1917. Negation in English and other languages. Reprinted in Selected Writings of Otto Jespersen (1962), pp. 3-151, George Allen and Unwin Ldt. Jones, Michael Allan. 1993. Sardinian Syntax, London: Routledge. Kadmon, Nirit and Fred Landman. 1993. Any. Linguistics and Philosophy 16.353-422. Kayne, S. Richard. 1983. Connectedness. Linguistic Inquiry 14: 223-249. —. 1994. The Antisymmetry of Syntux, MIT Press, Cambridge. Mass. —. 2000. Parameters and Universals. Oxford University Press, Oxford. —. 2005. Some notes on comparative syntax, with special reference to English and French. In Guglielmo Cinque and Richard Kayne eds., The Oxford Handbook of Comparative Syntax, 3-69. New York: Oxford University Press. Kas, Mark. 1993. Essays on Boolean Functions and Negative Polarity. PhD thesis, University of Groningen, Dissertations in Linguistics 11. Kemenade Ans van. 1999. Sentential negation and clause structure in Old English. In: I. Tieken-Boon van Ostade, G. Tottie and W. van der Wurff, eds., Negation in the History of English, Mouton de Gruyter, Berlin, pp. 147–165.

138

—.

References

2000. Jespersen’s cycle revisited: formal properties of grammaticalisation. In S. Pintzuk, G. Tsoulas and A. Warner eds., Diachronic Syntax: Models and Mechanisms, Oxford University Press, Oxford (2000), pp. 51–74. Kiparsky, Paul. 1973. Elsewhere in phonology. In Stephen Anderson and Paul Kiparsky eds., Feestschrift for Moris Halle. Holt, Rinehart and Winston. New York, pp. 93-106. Klima, S. Edward. 1964. Negation in English. In Jerry Fodor and Jerold Katz eds., The Structure of Language, pp. 246-323, Englewood Cliffs, Prentice Hall. Koopman, Hilda and Anna Szabolcsi. 2000. Verbal Complexes. Cambridge, MA: MIT Press. Koster, Jan. 2003d. All Languages are Tense Second. In Jan Koster and Henk van Riemsdijk eds., Germania et Alia: A Linguistic Webschrift for Hans den Besten. —. 1994. Predicate Incorporation and Word Order in Dutch. In Cinque, G., Koster, J., Pollock, J.-Y., Rizzi, L. & Zanuttini, R. eds., Paths Towards Universal Grammar. Georgetown, DC: Georgetown University Press, pp. 189-213. —. 2000. The Word Orders of English and Dutch. Ms. University of Groningen. Kroch, S. Anthony. 1989. Reflexes of grammar in patterns of language change. Language Variation and Change 1, 199–244. Kuno, Susumo and Ken-ichi Takami. 1997. Remarks on Negative Islands. Linguistic Inquiry 28: 553-676. Ladusaw, A. William. 1979. Polarity Sensitivity as Inherent Scope Relations. PhD Dissertation, University of Texas at Austin. —. 1992. Expressing negation. In C. Barker and D. Dowty, eds., SALT II: Pro-ceedings of the Second Conference on Semantics and Linguistic Theory, vol. 40 of Ohio State University Working Papers in Linguistics (OSU WPL), Columbus, Ohio State University. —. 1994. Thetic and Categorical, Stage and Individual, Weak and Strong. In Semantics and Linguistic Theory (SALT) IV, pp. 220-229. Cornell University Press, Ithaca. —. 1996. Negation and Polarity Items. Lappin S. ed., The Handbook of Contemporary Semantic Theory, Basil Blackwell, Oxford. Laenzlinger, Christopher. 2004. A feature-based theory of adverb syntax. In Austin, Jennifer R., Stefan Engelberg and Gisa Rauh eds., Adverbials, , 205–252. Laka, Itzar. 1990. Negation in Syntax: on the Nature of Functional Categories and Projections. PhD. Dissertation, MIT.

Negation Raising: Logic Form and Linguistic Variation

139

Landau, Idan. 2002. (Un)interpretable Neg in Comp. Linguistic Inquiry 33, 465-492. Lasnik, Howard. 1972. Analyses of negation in English. Doctoral dissertation. MIT, Cambridge, MA. Linebarger, Marcia. 1980. The Grammar of Negative Polarity. PhD thesis. MIT. Lyons, John. 1977. Semantics. 2 vols. Cambridge: Cambridge University Press. Longobardi, Giuseppe. 1991. In defense of the correspondence hypothesis: Island effects and parasitic constructions at LF. In C.-T. James Huang and Robert May eds., Logical structure and linguistic structure, 149– 196. Dordrecht: Kluwer. Louwerse, John. 1988. The Morphosyntax of Una in Relation to Discourse Structure. Pacific Linguistics, Series B, No. 100. May, Robert. 1977. The grammar of quantification. Doctoral dissertation, MIT, Cambridge, Mass. May, Robert. 1985. Logical Form: Its structure and derivation. Cambridge, Mass.: MIT Press. Martins, Ana Maria. 2000. Polarity Items in Romance: Underspecification and Lexical Change. In Susan Pintzuk, George Tsoulas and Anthony Warner eds., Diachronic Syntax: Models and Mechanisms, 191- 219. Oxford & New York: Oxford University Press. Matos, Gabriela. 1999. Negative Concord and the Scope of Negation. CatWPL, vol. 7, p. 175-190.. McCloskey, James. 1979. Transformational syntax and model theoretic semantics: A case-study in Modern Irish. Dordrecht: Reidel. —. 2001. The morphosyntax of Wh-extraction in Irish. Journal of Linguistics, 37:67–100. —. 2003. Working on Irish, Glot 7:3, 63-72. Meisel, Jurgen. 1993. Der Erwerb der Satznegation in baskischen und im spanischen. MS. Universität Hamburg. Merchant Jason. 2001. The syntax of silence: Sluicing, islands, and the theory of ellipsis. Oxford University Press: Oxford. —. 2003. Sluicing. In Martin Everaert and Henk van Riemsdijk eds., The syntax companion. London: Blackwell. —. 2004. Fragments and ellipsis. Linguistics and Philosophy 27.6:661738. —. 2006. Why no(t)?. Style. In William Salmon and Charalabos Kalpakidis eds., Special issue edited by as a Festschrift for Haj Ross. vol.10, 20-23.

140

References

Miestamo, Matti. 2003. Clausal negation: A typological study. PhD. dissertation, University of Helsinki. —. 2005. Standard negation: The negation of declarative verbal main clauses in a typological perspective. Empirical Approaches to Language Typology 31. Berlin/New York: Mouton de Gruyter. Montague, Richard. 1973. The proper treatment of quantification in ordinary English. In K. J. J. Hintikka, J. M. E. Moravcsik, and P. Suppes eds., Approaches to natural language, , 221–242. Dordrecht: Reidel. Morgan, Jerry. 1973. Sentence fragments and the notion ‘sentence’. In Braj Kachru, Robert Lees, Yakov Malkiel, Angelina Pietrangeli, and Sol Saporta eds., Issues in linguistics, 719-751. University of Illinois Press: Urbana. Moscati, Vincenzo. 2006a. The scope of Negation. Unpubblished PhD. dissertation, Università degli Studi di Siena. —. 2006b. Parameterizing Negation: Interactions with copular constructions in Italian and English children. In Belletti A., Bennati E., Chesi C., Di Domenico E., Ferrari I. eds., Language Acquisition and development. Cambridge Scholar Press. —. 2006c. Left Edge Licensors in italiano: un argomento in favore di un entrata lessicale disgiunta per ‘niente’ e ‘nessuno’. Atti della XXXI Conferenza della Società di Glottologia Italiana. Il Calamo, Roma. —. 2007. Covert Movement of Negation: raising over modality. STiL, CISCL Working Papers on Language and Cognition, Vol.1. —. 2008. Negative Chains and LF Negation Raising. In Bisetto, A. e Barbieri, F.E. eds., Proceedings of XXXIII Incontro di Grammatica Generativa, 106-120. Bologna, Italy. Nicoli, Franco. 1983. Grammatica milanese. Busto Arsizio, Bramanti Editori Nilsen, Øystein. 1997. Adverbs and A-shift. Working Papers in Scandinavian Syntax 59:1-31. Ouhalla, Jamal. 1990. Sentential Negation, Relativised Minimality, and the Aspectual Status of Auxiliaries. The Linguistic Review, 7: 83-231. —. 1991. Functional categories and parametric variation. Routledge. Payne, R. John. 1985. Negation. In Timothy Shopen, ed., Language Typology and Syntactic Description, vol. I. Clause Structure. Cambridge: Cambridge University Press, 197–242. Palmer, Frank. 1986. Mood and modality. Cambridge: Cambridge University Press. Papafragou, Anna. 2006. Epistemic Modality and Truth Conditions. Lingua 116: 1688-1702.

Negation Raising: Logic Form and Linguistic Variation

141

Parry, M. Mair. 1997. Preverbal negation and clitic ordering, with particular reference to a group of North-West Italian dialects, Zeitschrift für romanische Philologie, 113, 244-270. Pearson, Matt. 2001. Two Types of VO Languages. In Svenonius, P. ed., The Derivation of VO and OV. Amsterdam: John Benjamins, pp. 327364. Penka, Doris & Hedde Zeijlstra. 2005. Negative indefinites in Germanic. Paper presented at the 20th Comparative Germanic Syntax Workshop, Tilburg, June 2005. Pereltsvaig, Asya. 2004. Negative Polarity Items in Russian and the 'Bagel Problem'. In A. Przepiorkowski and S. Brown eds., Negation in Slavic. Bloomington: Slavica Publishers. Pesetsky, David. 1991. Zero Syntax: Vol.2 - Infinitives, unpublished manuscript. web.mit.edu/linguistics/people/faculty/pesetsky/infins.pdf Pesetsky, David and Esther Torrego. 2001. T-to-C Movement: Causes and Consequences. In M. Kenstowicz ed., Ken Hale: a Life in Language, Cambridge (Mass.): The MIT Press, 355-426. Pesetsky, David and Esther Torrego. 2004b. The Syntax of Valuation and the Interpretability of Features, to appear in Clever and Right: a Festschrift for Joe Emonds. Plann, Susan. 1982. Indirect questions in Spanish. Linguistic Inquiry 13.297-313. Pollock, Jean-Yves. 1989. Verb movement, universal grammar, and the structure of IP. Linguistic Inquiry 20, 365-424. Progovac, Ljiljana. 1988. A binding Approach to Polarity Sensitivity. PhD. Dissertation, University of Southern California. —. 1994. Positive and Negative polarity: a binding approach. Cambridge University Press. Puskás, Genoveva. 1998. On the Neg-criterion in Hungarian. Acta Linguistica Hungarica 45, 167-213. —. (ed). 2006. Negation in Finno-Ugric Languages. Lingua Thematic Issue. Volume: 116, Issue: 3 —. 2006. Double negation and information structure: somewhere between old and new. In V. Molnár & S. Winkler eds., The Architecture of Focus. Mouton De Gruyter. Reinhart, Tanya. 1997a. Quantifier scope: How labour is divided between QR and choice functions. Linguistics and Philosophy 20:335–397. —. 2006. Interface strategies. Optimal and Costly Computations. Linguistic Inquiry Monographs , 45. MIT Press. Richter, Frank and Manfred Sailer. 1998. LF Constraints on Expressions of Ty2: an HPSG Analysis of Negative Concord in Polish. In Robert

142

References

D. Borsley and Adam Przepiórkowski eds., Slavic in HPSG: 211-246. CSLI Stanford University. pp. 247-282. Rizzi, Luigi. 1982. Issues in Italian Syntax. Foris, Dordrecht —. 1990. Relativized Minimality. Cambridge, MA: MIT Press. —. 1997. The Fine Structure of the Left Periphery. In L. Haegeman ed., Elements of Grammar, Kluwer, Dordrecht. —. 2001. On the position "Int(errogative)" in the left periphery of the clause. In G. Cinque & G. Salvi eds., Current Studies in Italian Syntax offered to Lorenzo Renzi (pp. 287-296). Amsterdam; New York: Elsevier. —. 2004. On the Cartography of Syntactic Structures. In L. Rizzi ed., The structure of CP and IP. The cartography of syntactic structures, Volume 2. Oxford Studies in Comparative Syntax Roberts, Ian. 2007. Diachronic syntax. Oxford: Oxford University Press. Rooryck, Johan. 1992. Negative and Factive Islands Revisited. Journal of Linguistics 28(2). 343–373 Ross, John R. 1967. Constraints on variables in syntax. PhD thesis, MIT. —. 1969. Guess who? In Robert Binnick, Alice Davison, Georgia Green, and Jerry Morgan (eds.), Papers from the 5th regional meeting of the Chicago Linguistic Society, 252-286. Chicago Linguistic Society: Chicago, Ill. —. 1973. Slifting. In Gross, Halle & Schützenberger ed., The Formal Analysis of Natural Languages. 133-169. The Hague: Mouton. Rowlett, Paul. 1998. Sentential Negation in French. New York/Oxford: Oxford. Rullmann, Hotze. 1995. Geen eenheid. In Tabu 25. 194-197. Santelmann, Lynn. 1994. Evidence for NegP and Object Shift in German. Cornell Working Papers in Linguistics,12: 154-182. Seiler, Walter. 1985. Imonda, A Papuan Language. Pacific Linguistics Series B 93. Canberra: Australian National University. Sobin, Nicholas. 2003. Negative Inversion as Nonmovement. Syntax, 6:2, 183-212. Stainton, Robert. 1997. Utterance meaning and syntactic ellipsis. Pragmatics and Cognition 5: 51-78. Stechow, Arnim von. 1984. Gunnar Bech’s government and binding theory: GB’s GB theory. Linguistics 22, 225–241. Stockwell, Robert, Paul Schachter and Barbara H. Partee. 1973. The Major Syntactic Structures Of English. Holt, Rinehart and Winston, Inc., New York, 1973, pp. 294-418. Stowell, Tim. 2004. Tense and Modals. In J. Guéron and J. Lecarme eds., The Syntax of Time. Cambridge, Mass.: MIT Press, 621-635.

Negation Raising: Logic Form and Linguistic Variation

143

Suñer, Margarita. 1994. Negation and the features of Comp. In H. Grabois, D. Parkinson and D Yaeger eds., Cornell Working Papers in Linguistics 12:Papers on Negation, 183-192, 1994. Surányi, Balázs. 2006. Quantification and focus in Negative Concord. In Lingua Thematic Issue. Vol. 116, Issue 3. Svenonius, Peter. 1994. C-selection as Feature-checking. In Studia Linguistica. 48.2:133-155. —. 2000a. On Object Shift, Scrambling, and the PIC. MIT Working Papers in Linguistics 39. MIT. Cambridge, MA. —. 2000b. Quantifier movement in Icelandic. In Peter Svenonius ed., The Derivation of VO and OV, 255-292. John Benjamins, Amsterdam. —. 2002. Strains of negation in Norwegian. Working Papers in Scandinavian Syntax 69: 121–146. Swart, Henriëtte de. 2000. Scope Ambiguities with Negative Quantifiers’. In von Heusinger, Klaus & Urs Egli eds., Reference and Anaphoric Relations, Dordrecht/Boston/London: Kluwer Academic Publishers: 109-132. Swart, Henriëtte de and Ivan Sag. 2002. Negation and Negative Concord in Romance. In Linguistics & Philosophy 25, 373-415. Szabolcsi, Anna ed. 1997. Ways of scope taking. Dordrecht: Kluwer. Szabolcsi, Anna and Frank Zwarts. 1997. Weak islands and a algebraic semantics for scope taking. In A. Szabolcsi ed., Ways of Scope Taking, pages 217-262. Kluwer. Szabolcsi, Anna and Marcel den Dikken. 2003. Islands. In L. Cheng & R. Sybesma eds., The Second State-of- the-Article Book. Berlin: Mouton de Gruyter. Tovena, Lucia. 1996. Issues in Polarity Sensitivity. PhD. Dissertation, Universtity of Edinburgh. Tryon, Darrel T. 1970. An Introduction to Maranungku (Northern Australia). Pacific Linguistics B 15. Canberra: Australian National University. Vallduvi, Enric. 1994. Polarity items, N-words and minimizers in Catalan and Spanish. Probus 6:263-274. van der Auwera, Johan. 2001. On the typology of negative modals. In Jack Hoeksema, Hotze Rullmann, Víctor Sánchez-Valencia and Ton van der Wouden eds., Perspectives on negation and polarity items, 23– 48. Amsterdam, The Netherlands: Benjamins. Vikner, Sten. 1988. Modals in Danish and Event Expressions. In Working Papers in Scandinavian Syntax 39. Lund, Sweden: Department of Scandivian Linguistics.

144

References

Vries, Lourens de. 1996. Notes on the Morphology of the Inanwatan Language. In Ger P. Reesink, ed., Studies in Irian Languages, part 1. NUSA 40. Jakarta: Universitas Katolik Indonesia Atma Jaya, 97–127. Wallage, Phillip. 2008 Jespersen's cycle in Middle English: parametric variation and grammatical competition. Lingua 118: 643-674 Wason, Peter. 1965. The contexts of plausible denial. Journal of Verbal Learning and Verbal Behavior, 4:7-11. Watanabe, Akira. 2000. Feature copying and binding: Evidence from complementizer agreement and switch reference. Syntax 3:159–181. —. 2001. Decomposing the neg-criterion. In D’hulst, Yves, Johan Rooryck and Jan Schroten eds., Romance Languages and Linguistic Theory 1999, , 383 ff. Selected papers from ‘Going Romance’ 1999, Leiden, 9–11 December 1999 —. 2004. The Genesis of Negative Concord. In Linguistic Inquiry 35, 559612. Wouden, Ton van der. 1997. Negative Contexts. Routledge, London. Wurmbrand, Susi. 2004. Two types of restructuring: Lexical vs. functional. Lingua 114.8:991-1014. —. 1999. Modal verbs must be raising verbs. In: Proceedings of the 18th West Coast Conference on Formal Linguistics (WCCFL 18). Somerville, Mass: Cascadilla Press. —. 2007. How complex are complex predicates? Syntax, vol 10, 243-288. Zagona, Karen. 1990. Times as Temporal Argument Structure. Manuscript, University of Washington, Seattle. Zanuttini, Raffaella. 1991. Syntactic Properties of Sentential Negation: a comparative study of romance languages, PhD. Dissertation, University of Pennsylvania. —. 1997. Negation and Clausal Structure. A Comparative Study of Romance Languages. New York and Oxford: Oxford University Press. Zeijlstra, Hedzer Hugo. 2004. Sentential Negation and Negative Concord. PhD. dissertation, Universiteit van Amsterdam. Zwarts, Frans. 1995. Nonveridical contexts. Linguistic Analysis 25, 286312. Zwart, Jan-Wouter. 1993. Dutch Syntax: A Minimalist Approach. Doctoral dissertation, University of Groningen. —. 1997. Morphosyntax of Verb Movement. A Minimalist Approach to the Syntax of Dutch. Dordrecht: Kluwer. —. 2001. Syntactic and Phonological Verb Movement. Syntax 4, 34-62.