Studies on Negation: Syntax, Semantics, and Variation [1 ed.] 9783737005609, 9783847105602

Citation preview

Wiener Arbeiten zur Linguistik

Band 3

Herausgegeben von Alexandra N. Lenz, Melanie Malzahn, Eva-Maria Remberger und Nikolaus Ritt

Advisory Board: Peter Auer, Universität Freiburg, Deutschland Ina Bornkessel-Schlesewsky, Universität South-Australia, Australien Olga Fischer, Universität Amsterdam, Niederlande Junko Ito, UC Santa Cruz, USA Hans Kamp, Universität Stuttgart, Deutschland Johanna Laakso, Universität Wien, Österreich Michele Loporcaro, Universität Zürich, Schweiz Jim McCloskey, UC Santa Cruz, USA John Nerbonne, Universität Groningen, Niederlande Peter Trudgill, Universität Agder, Norwegen

Silvio Cruschina / Katharina Hartmann / Eva-Maria Remberger (eds.)

Studies on Negation Syntax, Semantics, and Variation

With 23 figures

V&R unipress Vienna University Press

Bibliografische Information der Deutschen Nationalbibliothek Die Deutsche Nationalbibliothek verzeichnet diese Publikation in der Deutschen Nationalbibliografie; detaillierte bibliografische Daten sind im Internet über http://dnb.d-nb.de abrufbar. ISSN 2365-7731 ISBN 978-3-7370-0560-9 Weitere Ausgaben und Online-Angebote sind erhältlich unter: www.v-r.de Veröffentlichungen der Vienna University Press erscheinen im Verlag V&R unipress GmbH. Gedruckt mit freundlicher Unterstützung des Rektorats der Universität Wien. © 2017, V&R unipress GmbH, Robert-Bosch-Breite 6, D-37079 Göttingen / www.v-r.de Alle Rechte vorbehalten. Das Werk und seine Teile sind urheberrechtlich geschützt. Jede Verwertung in anderen als den gesetzlich zugelassenen Fällen bedarf der vorherigen schriftlichen Einwilligung des Verlages. Titelbild: Idee und Realisierung: Silvio Cruschina & Eva-Maria Remberger

Contents

Silvio Cruschina, Katharina Hartmann & Eva-Maria Remberger Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

7

Anne Breitbarth Jespersen’s Cycle = Minimize Structure + Feature Economy . . . . . . . .

21

Karen De Clercq The nanosyntax of French negation: A diachronic perspective . . . . . . .

49

Cecilia Poletto Negative Doubling: In favour of a “Big NegP” analysis . . . . . . . . . . .

81

Adam Ledgeway Marking presuppositional negation in the dialects of southern Italy . . . . 105 Maria Barouni Challenging the strict vs. non-strict distinction of Negative Concord: A syntactic proposal . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131 Jakob Steixner Focus Intervention and Double Negation in Bavarian

. . . . . . . . . . . 157

Doris Penka Splitting at most . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 185 Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 213

Silvio Cruschina, Katharina Hartmann & Eva-Maria Remberger

Introduction “Knowledge is two-fold, and consists not only in an affirmation of what is true, but in the negation of that which is false.” Charles Caleb Colton

1.

Presentation of the volume

Negation is a universal feature of human language that has long been the subject of linguistic research from a variety of perspectives thanks to its inherently logical nature, its typologically multiform manifestations, its fundamental role in the mapping from syntactic structure to semantic interpretation, and its recurrent patterns of diachronic change. Work on the functional, lexical, and discursive characteristics of negation, including formal and contrastive approaches and diachronic and synchronic analyses, as well as language acquisition studies and pragmatic and sociolinguistic investigations have all contributed to the broadening of the domain of negation research. Current work on negation owes much to two pioneering monographs: Jespersen (1917) and Klima (1964). The first is an extensive study of negation, addressing a range of aspects, but particularly renowned for the delineation of the cyclical pattern that characterizes the historical development of the expression of negation in a variety of languages and that has now come to be known as Jespersen’s Cycle (cf. Dahl 1979). Klima (1964) is the first attempt at providing a comprehensive grammar of negation within a generative framework, inspiring much subsequent work on sentential negation, relative scope and configurational relations, and the licensing of Negative Polarity Items (NPIs). These two studies provide the model for more recent work on negation that is based on developments in formal syntax and semantics and attempts to account for the distribution and semantics of the various negative expressions (cf., e. g., Horn 1989 and references therein). There is also a long tradition of typological work which, building on the first surveys of Dahl (1979), Payne (1985), and Dryer (1988), and more recently Miestamo

8

Silvio Cruschina, Katharina Hartmann & Eva-Maria Remberger

(2005, 2007) and Dryer (2011), has highlighted the variety of means that natural languages possess to express negation.1 The aim of this volume is not to compete with the vast body of literature that has been published on negation, but to complement it by providing a further contribution to research on this topic and offering a set of cutting-edge studies on negation. All the contributions are related to recent questions or controversies bearing on the syntax and semantics of negative elements and the variation in their form, and follow the central assumption that a proper understanding of the multifaceted and varied expression of negation is central to our understanding of the grammar as a whole. With this in mind, we have deliberately attempted to include different approaches and a variety of empirical and analytic details. The event that led to the conception of this edited volume was an interdisciplinary workshop on the syntax and semantics of negation held at the University of Vienna in April 2014, jointly organized by the Departments of German Studies, Romance Studies, and Linguistics. Four of the participants in the workshop are also authors in the present volume (Adam Ledgeway, Doris Penka, Cecilia Poletto and Jakob Steixner); three additional contributors joined the project following the invitation of the editors (Maria Barouni, Anne Breitbarth and Karen De Clercq).

2.

Themes of the volume

Since Klima (1964) studies on the grammar of negation have largely concentrated on sentential negation (cf. Zanuttini 2001), with the associated issues of scope relations and the licensing of NPIs (cf., a.o., Lasnik 1972; Ladusaw 1980, 1996; Laka 1990; Aoun & Li 1993; Acquaviva 1997; Giannakidou 2011), and the syntax and semantics of sentential structures with more than one negative element (cf. Zeijlstra 2004, a.o.). Within these major research domains, this volume is organized around three specific key themes: 1) The internal structure and featural composition of negative markers, from both a diachronic (Breitbarth, De Clercq) and a synchronic (De Clercq, Poletto) perspective; 2) The variation in the expression of negation and negative meanings, in particular with respect to lexical and morphosyntactic diatopic variation (Poletto, Ledgeway) and to specific phenomena such as Negative Concord (NC) and Double Negation (DN) (Barouni, Steixner).

1 On other aspects of negation, see, among many others, the chapters in Horn (2010) and the references therein. See Givón (1978, 2001: Ch.8) for work within a functionalist perspective.

Introduction

9

3) Scope readings in negative contexts, and more specifically scope readings in the presence of an intervening focus (Steixner), as well as the interaction of negative expressions and other scope bearing elements (Penka). In the following sections we present a brief overview of the work carried out on these topics (cf. sections 2.1–2.3), followed by a synopsis of the papers (cf. section 3) contained in this volume.

2.1

Syntactic representation of negation

In current syntactic theory, the central issue has been to establish exactly how to represent sentential negation in syntactic structure and whether this formal representation should be associated with the presence of a specific projection (Pollock 1989; Laka 1990; Ouhalla 1990) or with a structural relation between specific elements (Haegeman & Zanuttini 1991; Haegeman 1995; Acquaviva 1997). Since Pollock (1989), in particular, the syntactic representation includes a fully-fledged functional projection hosting sentential negation within the functional spine of the clause. Based on comparative evidence, mainly from English and French, Pollock (1989) arrives at the so-called split IP hypothesis: focusing on word-order variations and on the relative position with respect to negation and adverbs, Pollock concludes that functional elements such as agreement, tense and negation should be viewed as independent heads projecting their own phrases. Negation is therefore the head of NegP. In his analysis, the position of negation (and hence of NegP), as well as that of adverbs, functions as a reference point for the parameterization of verb movement: in finite clauses, lexical verbs have to precede the main negation pas in French (cf. 1a), while they can only follow the negation not in English (cf. 1b) (see also Belletti 1990). (1) a. Pierre ne mange pas. (French) Pierre neg eat.prs.3sg neg b. Pierre does not eat. (English)

French exhibits a case of discontinuous negation in which a clitic-like pre-verbal negative element (i. e. ne) co-occurs with the post-verbal negative marker pas (in fact, ne is optional, although this optionality depends on several factors; cf., e. g., De Clercq this volume). In Pollock’s analysis, pas is assumed to sit in the specifier position of NegP (as not in English), while ne is the head of NegP that cliticizes onto the verb. Discontinuous negation is just an example of the different methods that languages may adopt to express negation. Negative inflection, for instance in Turkish (see the suffix -mi- in gelmiyor ‘he doesn’t come’), is an

10

Silvio Cruschina, Katharina Hartmann & Eva-Maria Remberger

example of morphological negation (for a further classification of negative markers into strong, weak, clitic and affixal, cf. Breitbarth this volume). An important development of Pollock’s influential analysis is presented in Zanuttini (1997), who distinguishes at least four distinct syntactic positions for sentential negative markers (i. e. NegPs) inside the clause. Her proposal is motivated by the distribution of preverbal, discontinuous and postverbal negative markers in Romance, which interact in various ways with other syntactic phenomena (cf. Poletto this volume for a brief summary). Zanuttini focuses on the synchronic stages of the negation types that can be seen as the consequences of a diachronic tendency for pre-verbal negation to be phonologically weakened, gradually reinforced and eventually replaced by a postverbal negative marker that was originally a polarity item or a minimizer (i. e. an element stemming from an indefinite or classifier-like noun indicating a small amount such as Fr. pas from Latin passum ‘step’). This diachronic tendency, known as Jespersen’s Cycle, has been shown to follow a cyclical development in several languages such as English, German and French (cf. 2) (see Hansen 2011: 52; cf. also Jäger 2008, van Gelderen 2008, 2011, and the recent articles in Meisner, Stark & Völker 2014, a special issue of Lingua): (2) je ne dis > je ne dis (pas) > je ne dis pas > je (ne) dis pas > je dis pas (French) ‘I do not say …’

Despite the superficial structural distinctions in the syntactic expression of negative force, semantically, negation is a unitary phenomenon. Studies on the syntax of negation, therefore, have also attempted to capture the relationship between syntactic variation in the manifestation of negation and the semantic negative operator (i. e. Ø), deepening our understanding of both parametric variation and the syntax-semantics interface (Haegeman 1995; Zanuttini 1997; Acquaviva 1997; Herburger 2011; a.o.). More recent work links the relevant syntactic configurations with two or more negative markers (e. g. DN or NC) to the underlying interpretation by proposing a distinction between interpretable and uninterpretable (or valued and unvalued) formal syntactic features that entertain specific agreement relations (Zeijlstra 2004; see also Zeijlstra 2014 and Barouni this volume). Along the same lines, another research tradition dealing more with the variety of negative markers, rather than with DN or NC, has concentrated on the internal structure and featural composition of negative markers, from both a diachronic and a synchronic perspective, proposing alternatives to or reformulations of the NegP hypothesis. This is the line of investigation adopted in the first three chapters of this volume, by Breitbarth, De Clercq, and Poletto, respectively.

11

Introduction

2.2

Types of negative markers

Not every negative element or ‘n-word’ (like English nobody, nothing, never, nowhere, no one, etc.) is always interpreted as a logical negative operator; nor are all apparent ‘n-words’ etymologically derived from a negative item (see, in particular, Spanish nadie ‘nobody’ < (homines) nati ‘(people) born.mpl’, and nada ‘nothing’ < (rem) nata(m) ‘(thing.acc) born.fsg(acc)’). Negative elements can just be morphosyntactic indicators of the presence of a logical negation in the structure. Thus, syntactic configurations with more than one negative element might nevertheless be interpreted as containing one single negation: a phenomenon found in many languages of the world called Negative Concord (NC). NC-languages can in turn be divided into strict and non-strict NC-languages, with the latter showing NC only in certain syntactic configurations but not in others. Italian, for example, exhibits obligatory NC with postverbal n-words (cf. 3a), but not when the n-word appears in a preverbal position (3b), in which case the n-word has a negative force per se and need not co-occur with an explicit sentential negation: (3) a. Non ha neg have.prs.3sg b. Nessuno ha nobody have.prs.3sg ‘Nobody called.’

chiamato nessuno. (Italian) call.pst.ptcp nobody chiamato. call.pst.ptcp

In non-NC-languages negation is typically interpreted twice, a phenomenon called Double Negation (DN) (German: Ich habe nicht den Hans nicht gesehen, sondern die Maria ‘It is not John I didn’t see, but Mary’); this, however, depends on several factors, such as the presence of already negated alternatives in the context. In special circumstances, generally related to information structure, a DN reading is also possible in both strict and non-strict NC-languages, despite the fact that the co-occurring exponents of negation typically participate in NC. Phenomena such as NC and DN show that the interpretation of negation can apply at different linguistic levels and that it is highly dependent on context. Negation can apply to the semantic interpretation at the level of the proposition that is simply negated (Øp, e. g. John didn’t buy bread, because he had no money with him), but it could also apply to the pragmatic level of the utterance in the discourse and, thus, be interpreted metalinguistically (Horn 1989). This is the case when a proposition that is corrected is present in the discourse (e. g. John didn’t buy bread, but butter; John didn’t buy bread but he stole it, and John didn’t buy bread, but Mary did in a context containing John bought/had to buy bread). Metalinguistic negation differs from semantic negation in that it fails to incorporate morphologically (This apartment is not clean /*unclean, it is antiseptic) and does not license NPIs (Peter didn’t buy some / *any of the comics, he bought

12

Silvio Cruschina, Katharina Hartmann & Eva-Maria Remberger

them all). Information structure and associated intonational patterns play an important role in determining the interpretation of sentences with more negative elements. Focus-marking (e. g. on buy vs. John vs. bread in the examples mentioned above) has significant effects on the interpretation and on the different scope readings of negation, and must therefore be taken into account in morphosyntactic analyses (cf. e. g. Barouni, Steixner, in this volume). In addition to the main negative adverbs or clitics expressing sentential negation (called ‘negative markers’ or, often, ‘standard negators’, e. g. English not, German nicht, Italian non, etc.) and n-words variously participating in NC according to the specific language and/or syntactic configuration, we also need to distinguish negative items that, semantically, convey additional components of meaning: presuppositional negation, scalar negative markers, negative completive or intensive negators, and intensifying or emphasizing elements. Presuppositional negation characterizes contexts where a positive version of the proposition is implicitly or explicitly present in the context (e. g. He didn’t buy bread after all in the conversational context in which the subject of the proposition was supposed to buy bread). Southern Italian mancu can express presuppositional negation, but its range of uses also includes a scalar function, corresponding to the English scalar negative marker not even (cf. 4), as well as a usage as standard negation (cf. Ledgeway this volume): (4) Isso nun s’ è nformato manco. (Neapolitan) he neg himself= be.prs.3sg inform.pst.ptcp not-even ‘He hasn’t even informed himself.’

Negative completive or intensive negators negate the propositional meaning, adding the information that negation is complete or exhaustive (e. g. He didn’t think about it at all) (cf. Garzonio & Poletto 2010). Additional intensifying elements can combine with the standard negative marker to produce negative reinforcement: these are so-called minimizers (Bolinger 1972, Horn 1989), which are typically derived from lexical items indicating small amounts (e. g. I didn’t eat a thing, She didn’t say a word). When these elements “occur in negative contexts, the negation denotes the absence of a minimal quantity, and hence the presence of no quantity at all” (Horn 1989: 400). Minimizers tend to become neutralized as simple negative markers in diachrony as in the case of French pas, originally meaning ‘step’, which has now become the standard negator, or Italian mica, derived from the word for ‘bread crumb’ and now functioning as a presuppositional negator (Cinque 1976).

Introduction

2.3

13

Scope readings in negative contexts

Relative scope is the semantic relation that holds between two or more logical operators in a sentence. The actual scope reading in the presence of more than one logical operator depends on a wide range of factors, including phonological prominence, syntactic relations, and information structure. This also applies to cases where one of the operators at issue is negation. However, as mentioned in the previous sections, the relative scope of logical negation cannot be directly derived from surface structure. This can be attributed to several factors: first, the negative operator may be morphophonologically empty, so that its scope properties can only be established by semantic tests (cf. Barouni, Steixner this volume); second, the negative element hosting the operator can undergo syntactic movement such that its original scope position must be reconstructed from a lower position in the structure (this is particularly important for DN and NC, but also for discontinuous negation); and, third, negation and its scopal properties can be encoded in one functional morphophonologically realized item, but this element can host more than one operator (e. g. Ø and $), which are then interpreted in an distributed way (split scope effects). The latter case is particularly relevant for the interpretation of negative indefinites (see, again, Barouni, Steixner this volume), but also for elements that contribute a weaker semantic form of negativity like at most, which also lead to ambiguity in their interaction with other logical operators (cf. Penka this volume). The aim of syntactic and semantic research in this domain is therefore to determine to what extent syntactic structures are responsible for the relative scope of an operator and to identify the semantic principles that lie behind scope and the level of representation at which they apply.

3.

Structure and contents of this volume

The volume opens with two chapters on the diachronic evolution of negation which offer new insights into the internal syntactic structure and featural composition of negative markers. The empirical basis of Anne Breitbarth’s paper (‘Jespersen’s Cycle = Minimize Structure + Feature Economy’) is provided by the historical development of negation in Low German, as well as by novel observations on the typology of adverbial negative markers. In light of this, Breitbarth proposes a new account of Jespersen’s Cycle based on two principles: ‘Minimize Structure’ (Cardinaletti & Starke 1999) and ‘Feature Economy’ (van Gelderen 2011). These two principles allow the author to identify parallels and similarities between the distribution and typology of pronouns – for which Cardinaletti & Starke’s (1999) Minimize Structure was

14

Silvio Cruschina, Katharina Hartmann & Eva-Maria Remberger

originally proposed – and the diachronic stages and synchronic categorization of negative markers across languages. Depending on their formal features, which are subject to the principle of Feature Economy, negative markers are thus classified into strong, weak, clitic and affixal. In addition to making correct predictions regarding the diachronic paths of development and types of negative marker that are possible crosslinguistically, Breitbarth’s analysis dispenses entirely with the traditional idea that the negation is syntactically associated with one or more functional projections (i. e. the NegP hypothesis). A different set of data and an alternative approach to the historical evolution of negative markers is adopted in the chapter by Karen De Clercq (‘The nanosyntax of French negation: A diachronic perspective’). Based on French data, De Clercq adopts a nanosyntactic approach (cf. Starke 2009) to the structural composition of negative markers to account for both their diachronic development and synchronic stylistic variation. In particular, under this analysis De Clercq arrives at an explanation for the obligatory presence of the preverbal negator in a specific register/style of French, known as le bon usage French (Grevisse & Goosse [1936] 1993), and its disappearance from spoken or colloquial language. The author acknowledges that negation is featurally complex and identifies specific patterns that lead to the splitting up of negation into five different syntactic features. Following and refining an idea already present in the literature (cf., e. g., Poletto 2008), she claims that all negations start out as a single constituent and that they project a different range of functional layers according to the class they belong to and the features they realize. This proposal is thus able to capture the transition from one stage to the other, both in terms of diachronic development and synchronic stylistic variation, which is attributed to a change in the makeup and featural size of each negative marker. It also sheds new light on the relationship between sentential negation and constituent or lexical negation. The NegP hypothesis – the idea that negation is syntactically associated with one or more functional projections – is firmly rejected in Breitbarth’s analysis and substantially redefined by De Clercq. In her contribution (‘Negative Doubling: In favour of a “Big NegP” analysis’), in contrast, Cecilia Poletto offers a reformulation of this hypothesis which leads both to a revival of the category NegP and to a refined cartographic account of discontinuous negation. Poletto’s paper, like Breitbarth’s and De Clercq’s, pursues the question of the internal structure of negative makers, but the empirical basis of the study is different: northern Italian dialects from a synchronic perspective. Based on an idea originally proposed for French in Pollock (1989) and successively built upon in Poletto (2008), the author further develops the claim that discontinuous negative markers originate from the same projection. This argument constitutes an extension to the negation domain of the big-DP proposal, origi-

Introduction

15

nally formulated to account for the pronominal doubling of nominal constituents (Kayne 1975; Cecchetto 2000; Belletti 2005). Drawing on evidence from northern Italian dialects, Poletto presents new data and observations in defence of the generalizations outlined in Zanuttini (1997) and questioned in recent work, especially in Manzini & Savoia (2011). The apparent exceptions to Zanuttini’s generalizations are explained here in terms of movement from one single constituent, i. e. the “Big NegP”. Problematic cases of negative doubling in northern Italian dialects are therefore derived on the basis of the author’s cartographic analysis. With the chapter by Adam Ledgeway (‘Marking presuppositional negation in the dialects of southern Italy’), we move from northern to southern Italy and to the rich lexical and morphosyntactic variation in negation found in these dialects, with a contribution that will definitely fill an apparent gap in the literature: while negation has been widely investigated with respect to northern Italian dialects, the variation in negative markers in southern Italian dialects has so far been neglected. Ledgeway starts from the observation that, besides standard negation, there are two types of emphatic negation in Italo-Romance, the ‘at all’type and the ‘after all… not’-type. The range of uses of southern Italian manco (the type ‘after all… not’), mainly in northern Calabrian, is illustrated, together with the theoretical approach that he adopts to capture the interpretation of the syntactic variation concerning negative markers. The discussion then concentrates on Salentino filu, and includes an excursus on Florentine punto (both originally minimizers) as a necessary preliminary step for a contrastive analysis of the Salentino data. Returning to Calabrian, Ledgeway shows that in this dialect the interpretation of emphatic negation depends on syntactic conditions, whereas for other dialects a lexical approach is necessary. In conclusion, three strategies for the variation in the interpretation of emphatic negation in southern Italo-Romance are identified: (i) the grammaticalization of clefts (e. g. in Sicilian), (ii) V-movement (e. g. in northern Calabrian), and (iii) the grammaticalization of minimizers (Salentino). The article introduces a substantial amount of new data on negation in southern Italian dialects, a systematic description of which has previously been missing in the literature; it also represents an important contribution to the empirical and theoretical research into the possible strategies for the implementation of emphatic negation. The next two chapters, by Maria Barouni and by Jakob Steixner respectively, are also principally concerned with crosslinguistic variation. Both discuss DN-readings in languages which are clearly strict NC-languages, and for the analysis of their data both rely, adapting it to their needs, on the seminal work by Zeijlstra (2004). In her paper ‘Challenging the strict vs. non-strict distinction of Negative Concord: A syntactic proposal’, Barouni discusses n-words in Greek, a strict

16

Silvio Cruschina, Katharina Hartmann & Eva-Maria Remberger

NC-language, where some n-words, like oudhepote (‘never’) and oudholos (‘not at all’), do not trigger a strict NC pattern. These data would thus contradict Zeijlstra’s (2004) theory on NC-languages, according to which NC-languages can be strict or non-strict depending on whether they license a negative marker with both preverbal and postverbal n-words (strict NCL) or only with postverbal n-words (non-strict NCL). The hitherto undiscussed set of negative ou-elements, which behave like n-words in a non-strict NCL, provide a good basis to argue for a distribution and specification of negative features in NCL that differ from those proposed by Zeijlstra. Barouni’s claim is then that negative markers (sentential negators like Greek dhen) in NCLs carry [iNEG], which is formal but not semantic in nature, and that n-words in non-strict NCLs (like Italian nessuno and Spanish nadie) and ‘transparent’ negative words in strict NCLs (like the Greek ou-elements) can have both a [uNEG] and an [iNEG] version. With regard to the syntactic head Nego, she assumes that in NCLs it is always filled with an operator containing [uNEG], which can be valued either by the negative marker (such as Greek dhen) or by an n-word with a [iNEG] feature. Steixner’s chapter (‘Focus Intervention and Double Negation in Bavarian’) starts with a similar set of data from Bavarian, another strict NC-language. DNreadings are also possible in Bavarian, but, in contrast to Greek, these readings do not depend on the properties of the lexical entries of particular n-words, but on information-structure configurations. The author rejects the analysis proposed in Brugger & Poletto (1993) and, in particular, in Weiß (1998, 1999) for Bavarian. For the blocking of NC, these accounts rely on the purely syntactic mechanism of c-command, necessary for the negative marker hosting the negative operator to check the other n-word(s) lower in the structure. For Steixner, negative operators in NegP are always silent and thus only interpretable with the help of other visible cues, normally negative markers and n-words, in the syntactic structure. Negative markers in NC-languages, like Bavarian ned, are neither hosts of the logical negation nor head of a NegP, but are adjuncts to vP. As for DN, a second silent operator (i. e. a second NegP) is syntactically merged, but this second logical negation can only be interpreted if the information-structural configuration facilitates an interpretation of the negated proposition as given in the discourse. In this case, the blocking of NC can be interpreted as focus intervention. Thus, the author modifies Zeijlstra’s account insofar as he counts on the availability of NegP not only once but twice in negated structures, depending on information-structural properties provided by the context. Finally, Penka’s paper (‘Splitting at most’) mainly tackles the third theme of the volume, namely, scope readings, in a formal semantic perspective. She discusses the interpretations of the expression at least and at most when used in the context of modal verbs. She first illustrates split scope effects with negative quantifiers (as with English no and German kein) where modal operators are

Introduction

17

interpreted between a negative and an existential quantifier, and shows that split scope readings also arise with weaker negative quantifiers like few, fewer than and at most, the latter being the main topic of the chapter. At most seems to be particularly relevant in a volume on negation, since it displays split scope readings for the operators that it hosts, among them the negation operator, but split scope readings only arise with possibility modals and not with necessity modals. The author bases her proposal on former semantic and pragmatic analyses, like Geurts & Nouwen (2007) and in particular Büring (2008), and Schwarz (2013). Whereas these papers struggle with the asymmetrical behaviour of at least and at most in the modal context, Penka offers a step-by-step derivation of the readings for both at least and at most in both modal contexts in a decompositional analysis that yields a solution to the problem. The conclusion of the paper is that at most is in fact semantically more complex than at least, in that at most involves a form of semantic negation that scopally interacts with other logical operators.

Acknowledgements This volume would not have been possible without the help, commitment and advice of several people. First of all, we would like to thank the participants of the original workshop on the syntax and semantics of negation, held in Vienna in April 2014, for their valuable comments and fruitful discussions. We also thank the anonymous reviewers, all international experts in the field of negation, for their precious help and collaboration during the reviewing process. Furthermore, we are grateful to Michael Dobish for proof-reading the English and to Ruth Steindl for copy-editing all chapters. We would also like to express our gratitude to Susanne Franzkeit, Laura Haase, Oliver Kätsch and Ruth Vachek from the scholarly publishing house Vandenhoeck & Ruprecht, who assisted us in the various stages of the publication process, as well as to the editors of the Wiener Linguistische Arbeiten (Vienna Studies in Linguistics) for the opportunity to publish the present volume within this series. Finally, we gratefully acknowledge the financial support of the Rectorate of the University of Vienna.

18

Silvio Cruschina, Katharina Hartmann & Eva-Maria Remberger

References Acquaviva, Paolo (1997): The Logical Form of Negation. New York: Garland. Aoun, Joseph & Li, Yen-hui Audrey (1993): The Syntax of Scope. Cambridge, MA: MIT Press. Belletti, Adriana (1990): Generalized Verb Movement: Aspects of verb syntax. Turin: Rosenberg & Sellier. Belletti, Adriana (2005): Extended doubling and the VP periphery. Probus 17: 1–35 Bolinger, Dwight (1972): Degree Words. The Hague: Mouton. Brugger, Gerhard & Cecilia Poletto (1993): On Negation in German and Bavarian. University of Venice Working Papers in Linguistics 2 (3): 41–79. Büring, Daniel (2008): The least at least can do. In Chang, C. B. & H. J. Haynie (eds.): Proceedings of the 26th West Coast Conference on Formal Linguistics. Somerville, MA: Cascadilla Press, 114–120. Cardinaletti, Anna & Michal Starke (1999): The typology of structural deficiency: A case study of the three classes of pronouns. In Henk van Riemsdijk (ed.): Clitics in the Languages of Europe (EALT/EUROTYP 20–5). Berlin/New York: Mouton de Gruyter, 145–233. Cecchetto, Carlo (2000): Doubling structures and reconstruction. Probus 12: 93–126. Cinque, Guglielmo (1976): Mica. Annali della Facoltà di Lettere e Filosofia dell’Università di Padova 1: 101–112. Dahl, Östen (1979): Typology of sentence negation. Linguistics 17: 79–106. Dryer, Matthew S. (1988): Universals of negative position. In Hammond, Michael, Edith A. Moravcsik & Jessica Wirth (eds.): Studies in Syntactic Typology. Amsterdam: John Benjamins, 93–124. Dryer, Matthew S. (2011): Negative morphemes. In Dryer, Matthew S. & Martin Haspelmath (eds.): The World Atlas of Language Structures Online: . Munich: Max Planck Digital Library. Garzonio, Jacopo & Cecilia Poletto (2010): Quantifiers as Negative Markers in Italian Dialects. In Van Craenenbroeck, Jeroen (ed.): Linguistic Variation Yearbook 2009. Amsterdam: John Benjamins, 127–152. Gelderen, Elly van (2008): Negative cycles. Linguistic Typology 12: 195–243. Gelderen, Elly van (2011): The Linguistic Cycle: Language Change and the Language Faculty. Oxford: Oxford University Press. Geurts, Bart & Rick Nouwen (2007): ‘At least’ et al.: The semantics of scalar modifiers. Language 83 (3): 533–559. Giannakidou, Anastasia (2011): Negative and positive polarity items. In von Heusinger, Klaus, Claudia Maienborn & Paul Portner (eds.): Semantics: An International Handbook of Natural Language Meaning, Vol. 2, Berlin: de Gruyter, 1660– 1712. Givón, Talmy (1978): Negation in language: pragmatics, function, ontology. In Cole, Peter (ed.): Syntax and Semantics 9: Pragmatics, New York: Academic Press, 69–112. Givón, Talmy (2001): Syntax: An introduction. Amsterdam: John Benjamins. Grevisse, Maurice & Andre Goosse ([1936] 1993): Le bon usage. Paris: Duculot.

Introduction

19

Haegeman, Liliane & Raffaella Zanuttini (1991): Negative heads and the Neg-Criterion. The Linguistic Review 8: 233–252. Haegeman, Liliane (1995): The Syntax of Negation. Cambridge: Cambridge University Press. Hansen, Maj-Britt Mosegaard (2011): Negative cycles and grammaticalization. In Narrog, Heiko & Bernd Heine (eds.): The Oxford Handbook of Grammaticalization. Oxford: Oxford University Press, 570–579. Herburger, Elena (2011): Negation. In Heusinger, Klaus von, Claudia Maienborn & Paul Portner (eds.): Semantics. An International Handbook of Natural Language Meaning. Volume 2. Berlin: Mouton de Gruyter, 1641–1660. Horn, Laurence (ed.) (2010): The Expression of Negation. Berlin: Mouton de Gruyter. Horn, Laurence (1989): A Natural History of Negation. Chicago: University of Chicago Press. Reissued: Stanford, CA: CSLI Publications, 2001. Jäger, Agnes (2008): History of German Negation. Amsterdam: John Benjamins. Jespersen, Otto (1917): Negation in English and Other Languages. Copenhagen: Høst. Kayne, Richard (1975): French Syntax. Cambridge, MA: MIT Press. Klima, Edward S. (1964): Negation in English. In Fodor, Jerry A. & Jerrold J. Katz (eds.): The Structure of Language, Englewood Cliffs, NJ: Prentice-Hall, 246–323. Ladusaw, William (1980): Polarity Sensitivity as Inherent Scope Relations. New York: Garland. Ladusaw, William (1996): Negation and Polarity Items. In Lappin, Shalom (ed.): Handbook of Contemporary Semantic Theory. Oxford: Blackwell, 321–342. Laka, Itziar (1990): Negation in Syntax: On the Nature of Functional Categories and Projections. PhD thesis, MIT. Lasnik, Howard (1972): Analyses of Negation in English. PhD thesis, MIT. Manzini, Maria Rita & Leonardo M. Savoia (2011): Grammatical Categories: Variation in Romance Languages. Cambridge: Cambridge University Press. Meisner, Charlotte, Elisabeth Stark & Harald Völker (eds.) (2014): Jespersen Revisited: Negation in Romance and Beyond. Special issue of Lingua 147. Miestamo, Matti (2005): Standard Negation: The negation of declarative verbal main clauses in a typological perspective. Berlin: Mouton de Gruyter. Miestamo, Matti (2007): Negation: An overview of typological research. Language and Linguistics Compass 1: 552–570. Ouhalla, Jamal (1990): Sentential negation, relativized minimality and the aspectual status of auxiliaries. The Linguistic Review 20: 183–231. Payne, John R. (1985): Negation. In Shopen, Timothy (ed.): Language Typology and Syntactic Description, Vol. I: Clause structure. Cambridge: Cambridge University Press, 197–242. Poletto, Cecilia (2008): On negative doubling. Quaderni di lavoro ASIT 8: 57–84. Pollock, Jean-Yves (1989): Verb movement, universal grammar, and the structure of the IP. Linguistic Inquiry 20: 365–424. Schwarz, Bernhard (2013): At least and Quantity Implicature: Choices and Consequences. In Aloni, Maria, Michael Franke & Floris Roelofsen (eds.): Proceedings of the 19th Amsterdam Colloquium, 187–194. Starke, Michal (2009): Nanosyntax: a short primer to a new approach to language. Nordlyd 36 1: 1–6.

20

Silvio Cruschina, Katharina Hartmann & Eva-Maria Remberger

Weiß, Helmut (1998): Syntax des Bairischen. Studien zur Grammatik einer natürlichen Sprache. Tübingen: Niemeyer (Linguistische Arbeiten 391). Weiß, Helmut (1999): Duplex negatio non semper affirmat: A theory of double negation in Bavarian. Linguistics 37: 819–846. Zanuttini, Raffaella (1997): Negation and Clausal Structure: A Comparative Study of Romance Languages. Oxford/New York: Oxford University Press. Zanuttini, Raffaella (2001): Sentential negation. In Baltin, Mark & Chris Collins (eds.): The Handbook of Contemporary Syntactic Theory, Oxford: Blackwell, 511–535. Zeijlstra, Hedde (2004): Sentential negation and Negative Concord. PhD thesis, University of Amsterdam. Zeijlstra, Hedde (2014): On the uninterpretability of interpretable features. In Kosta, Peter et al. (eds.): Minimalism and Beyond: Radicalizing the interfaces. Amsterdam: John Benjamins, 109–128.

Anne Breitbarth

Jespersen’s Cycle = Minimize Structure + Feature Economy

Abstract: In the current paper, I outline a new account of Jespersen’s Cycle, mainly concentrating on data from historical Low German, but embedding the proposal into a more general typology of adverbial negative markers and their diachronic connections. Building on proposals by Cardinaletti & Starke (1999) and Grosz (2005; 2007) and Cardinaletti (2011) regarding the distribution of pronouns and modal particles, I propose to distinguish four classes of negative markers with different amounts of internal structure, representing different diachronic stages in the development of such markers. I argue that the NegP-hypothesis is (a) not necessary to account for the observed distribution, and (b) in fact makes empirically wrong predictions, which are avoided under the proposed NegP-free account. Keywords: NegP-hypothesis, Jespersen’s Cycle, Old Saxon, Middle Low German, third-factor principles

1.

Overview

Since Pollock (1989), the structural locus of sentential negation within generative approaches to syntax has been assumed to be a functional projection, NegP. The NegP-hypothesis has also been very successful in accounts of the historical development of negation in various languages, more precisely, of Jespersen’s Cycle ( JC) (after Jespersen 1917), as it offers head and specifier positions that can be targeted by new or old negative markers during their grammaticalization. However, while there is general agreement on the existence of NegP, there is much division about the exact number of NegPs, their syntactic position, and precise contribution to interpretation. In the present paper, taking JC in historical Low German as the empirical point of departure, I will argue for a NegP-free approach to JC. The proposed account starts from the observation of similarities between the typology, distribution and

22

Anne Breitbarth

grammaticalization clines of pronouns/agreement markers and adverbs/modal particles on the one hand, discussed in the literature, and the typology and diachronic stages of negation markers on the other. I argue that both the typology and the diachronic development of negative markers in languages using adverbial negation particles can be accounted for by the interplay of two third-factor principles, Minimize Structure (Cardinaletti & Starke 1999) and Feature Economy (van Gelderen 2011).1

2.

Jespersen’s Cycle in historical Low German

The empirical background of the current paper is the development of the expression of negation in the oldest attested stage of Low German, Old Saxon (Old Low German; c. 800–1250) (OS), and the subsequent stage, Middle Low German (c. 1250–1650) (MLG). Like all other West Germanic languages, Low German has undergone JC, the development in which a single expression of negation (stage I) is first reinforced by a new negation marker (stage II) and ultimately replaced by it (stage III), and has also undergone changes in the interaction between sentential negation and (negative) indefinites in its scope related to the development of negative markers ( JC). In OS, sentential negation is expressed by a pre-finite negative particle ni/ne positionally covarying with the finite verb. In (1) the combination of ni/ne + finite verb is found clause-initially, in (2) clause-medially/in verb-second position, and in (3), clause-finally. (1) ni bium ic […] that barn godes neg am I the child God.gen ‘I am not the child of God.’ (Heliand 915) (2) thu ni uuest the maht godes the ik gifrummien scal you neg know the power God.gen rel I serve shall ‘You do not know the power of God that I am to serve.’ (Heliand 3102–3) (3) Ic thes uuirdig ne bium […] that thu an min hus cumes. I that.gen worthy neg am that you to my house come ‘I am not worthy that you come to my house.’ (Heliand 2104–5)

OS is firmly in stage I of JC (Breitbarth 2014a), but there are a number of cases showing the emergence of an adverbial reinforcer (io)uuiht ‘anything’ or 1 I am not going to address languages using negative auxiliaries or other strategies in the present paper.

Jespersen’s Cycle = Minimize Structure + Feature Economy

23

niouuiht ‘nothing’. The bridging context seems to be a class of verbs allowing an optional extent argument or measure phrase, such as verbs meaning ‘harm’ or ‘benefit’, see (4). (4) Ne ik thi geth ni deriu (neo)uuiht quad he2 and.not I you also neg damage nothing said he ‘I will also not harm you (at all), he said.’ (Heliand 3892)

In the entire OS corpus (695 negative clauses), I found one occurrence of niet < niouuiht used as constituent (narrow focus) negation, in the 11th c. Gregorius glosses, see (5). (5) (illorum non solum animae. Sed caro quoque) thuo niet ekir iro selon neuen ok then neg only their soul but also ‘then not only their soul, but also …’ (GG 63, 15–16)

After a gap in attestation of at least 200 years, the expression of negation in MLG3 had already reached the transition from stage II to stage III of JC; the former prefinite negation particle, now weakened to ne/en, is increasingly optional and by the time the MLG is replaced by Early New High German as the written language in the area around the middle of the 16th century, it has all but disappeared. The standard, i. e., neutral and productive, negation marker is the adverbial nicht (6–7). (6) We des nicht en-wete de lat=is sik berichten. who this.gen neg neg-know rel let=it refl report ‘Whoever doesn’t know this should get informed about it.’ (Braunschweig 1349) (7) Iß he ohme nicht euenbordig, so mag he idt nicht theynn. is he him neg equal so may he it neg tithe ‘If he is not equal to him, he may not tithe it.’ (Braunschweig 1553)

Present-day Low German has remained in stage III of JC; the adverbial marker, nich, continues to be used in much the same way as nicht was in MLG, (8), with nich occurring at the right edge of the middle field.

2 In the older Monacensis (c. 850), it is neouuiht ‘nothing’, while in the Cottonianus (10th c.) manuscript, the form uuiht ‘(any)thing’ is used. 3 The corpus used is described in Breitbarth (2014a).

24

Anne Breitbarth

(8) … dat sik de Akadeemsche Senaat nich versammeln kunn. … that refl the academic senate neg convene could ‘… that the academic senate was not able to convene.’ (http://www.radiobremen.de/bremeneins/serien/plattdeutsche_nachrichten/ plattnachrichten104_date-20150128.html, accessed 29/01/2015)

Before addressing the question of how the different stages are connected, diachronically, it is necessary to look at the interaction between the expression of negation and (morphologically) negative indefinites in the scope of negation at the various stages. This interaction provides important information regarding the interpretability of formally negative markers (Zeijlstra 2004, Penka 2011). OS for instance with its affix-like negative particle attaching to the finite verb might be expected to have negative concord (NC) (cf. Rowlett’s 1998 Jespersen’s Generalization, after observations in Jespersen 1917). Contrary to what one might expect, however, certainly given the typological overview in Zeijlstra (2004), OS NC is not very frequently attested at all in the two largest surviving texts, the Heliand epic (c. 830 CE) and the Genesis fragments (c. 840 CE). It does not occur in the Genesis at all (9a), and only in c. 20 % of the possible cases in the Heliand, (9b). Only in the minor texts (10th and 11th centuries CE), all attested indefinites in the scope of negation show concord with the negative marker, but there are only five such occurrences, cf. Table 1.4 Table 1: Expression of negation and indefinites in the scope of negation in OS5

Heliand Genesis Minor texts6 Total

ni

ni alone

NMI

NFI

617 37 37

582 37 32

35 0 5

143 12 0

Total negative clauses 620 37 38 695

In addition, NC only seems to take the shape of negative doubling – the cooccurrence of the marker of sentential negation with a negative indefinite – in OS; where more than one indefinite occurs in the scope of negation, at most one of them is morphologically negative, (9c–d). 4 The Heliand is by far the largest body of Old Low German text with c. 6,000 lines of alliterative verse (c. 80 % of the extant material; Sanders 2000: 1277). The Genesis comprises c. 335 lines of alliterative verse. The minor texts (incl. a tax list, a number of glosses, fragmentary psalm translations (from Latin) and commentaries, two short texts on scrolls, a part of a translation (from Latin) of a Homily by Bede, a confession, Baptismal Vows, blessings, and short inscriptions) are of varying length, ranging from a few words to a few pages each. 5 Abbreviations: NMI = n-marked, i. e. morphologically negative, indefinite, NFI = n-free indefinite. 6 From the minor texts, only finite negative clauses were considered. That is, glosses were only taken into account if they formed a full finite clause.

Jespersen’s Cycle = Minimize Structure + Feature Economy

25

(9) a. that is ênig seg ni ginas that it.gen any man neg was.saved ‘that no one was saved from it’ (Genesis 322) b. that thar nenig gumono ni ginas that there no man neg was.saved ‘that no one was saved there’ (Heliand, 4369–70) c. N-is thes tueho enig gumono nigienumu. neg-is this.gen.sg.n. doubt any men.gen.pl none.dat.pl ‘There is no doubt about it to any of the men.’ (Heliand 3190–1) d. It ni mag iu te enigoro frumu huuergin uuerdan te enigumu uuilleon. it neg can you to any benefit at.all redound to any happiness ‘It is not able to do you any good at all, nor bring you any happiness.’ (Heliand 1854–5)

MLG, on the other hand, had extensive NC. Like other languages undergoing JC (cf. Haspelmath 1997: 203), negative doubling in MLG is restricted to morphologically negative indefinites co-occurring with the old pre-finite particle ne/ en (in 476 out of 1269 cases7); they do not co-occur with the new adverbial negator nicht. Negative spread – marking negation on more than one or all indefinites in the scope of negation (Den Besten 1986) – is the norm, (10) (in 98 out of 107 cases of more than one indefinite per clause). (10) Und we en-willet noch en-schullet nummermer neyn slot darin buwen and we neg-want nor neg-shall never.more no castle there.in build ‘And we neither intend nor shall build any castle in it.’ (Uelzen 1397)

The standard view concerning how the stages of JC in languages like historical Low German are connected is that there is a functional projection NegP, typically taking vP as its complement, which provides head and specifier positions that the weakening and strengthening negative particles can target during their grammaticalization.

3.

NegP approaches to Jespersen’s cycle

Pollock (1989), comparing English and French, first proposed to split IP into a T(ense)P(hrase), a Neg(ation)P(hrase) and an Agr(eement)P(hrase), see (11). (11) [TP [NegP [AgrP [VP ]]]] 7 Besides, there are 787 cases of n-marked indefinites without negative doubling, marking the loss of the preverbal particle from negative clauses with indefinites in the scope of negation.

26

Anne Breitbarth

AgrPs were soon shown to be empirically inadequate (Iatridou 1990, Julien 2002), and since functional elements without semantic import are no longer considered able to project (Chomsky 1995: 240), they also no longer have a place in the theory. The same is not applicable to NegP, as the logical negation operator (however lexicalized in a language) is clearly interpretable in that it has an effect on the propositional meaning of the clause. Today, three main types of approaches can be distinguished: (i) There is a single NegP in negative clauses in all languages, reflecting the presence of a logical (possibly covert) negation operator (e. g. Roberts & Roussou 2003, van Gelderen 2008; 2011, Roberts 2007, Willis 2012). (ii) There is a single NegP only in those languages in which there are overtly marked syntactic dependencies (i. e., NC), that is, the presence of NegP may vary diachronically (e. g. Zeijlstra 2004, Wallage 2005). (iii) There are several NegPs in each language, accommodating the typological and diachronic variation concerning the position of negative particles with respect to other material in the clause such as verbs, adverbs and other operators (e. g. Zanuttini 1997, Poletto 2008), though only one of these positions carries interpretable negation features. Only proponents of the first kind of approach really seem to have proposed a diachronic model, instead of – as is more common – proposing analyses of separate synchronic stages of JC without explicating the processes that connect these stages. Cartographic approaches (particularly Poletto 2008) suggest a connection between the diachronic source of a negator and its position in the clausal hierarchy, but do not acknowledge a diachronic connection between the different assumed NegPs (not just between one etymological source and one NegP, e. g. MinimizerP). Under such an approach, it is unclear what triggers the ‘activation’ of a given NegP at a given diachronic stage, and at least one false prediction seems to be made: Several new negators should be able to enter a language or be grammaticalized at the same time, targeting different NegPs. According to those NegP approaches which have offered accounts of the diachronic processes creating and affecting negative markers, the head of NegP is typically assumed to have an interpretable/valued negation or polarity feature at stage I of JC.8 The future adverbial negator (for instance, Old French pas) initially has no negation or polarity features at this stage, but is simply a negative polarity item or has an ‘uninterpretable operator feature’ that needs to be licensed by a suitable operator (for the development of French pas cf. Roberts 2007: 72, 79–80). Later, this adverbial element is integrated into the negative system and acquires an uninterpretable/unvalued negation/polarity feature, (12a). At stage II of JC, this licensing relationship is reversed by another reanalysis, which leaves the preverbal marker with

8 Approaches vary as to which model of feature checking or valuation is assumed.

Jespersen’s Cycle = Minimize Structure + Feature Economy

27

the uninterpretable/unvalued negation/polarity feature and the new adverbial element with the interpretable feature licensing it, (12b). (12) a. [NegP [AdvP pas uPOL[NEG] ] [Neg’ [Neg ne POL[NEG] ]]] → b. [NegP [AdvP pas POL[NEG] ] [Neg’ [Neg ne uPOL[NEG] ]]] (after Willis 2012: 96)

Once this step has been completed, the new interpretable negation marker may be considered sufficient to express sentential negation, leading to the loss of the old preverbal marker. According to Roberts & Roussou (2003) and Willis (2012), these reanalyses affect both the head and specifier positions of NegP, that is, they affect the material filling these positions as well as whether head or specifier carry the interpretable features. Under such approaches, the question is what triggers and what restricts such changes. A concrete proposal has been made by van Gelderen (2008; 2011). To account for cyclical change in language including the negative cycle ( JC) and grammaticalization, she makes extensive use of so-called third-factor principles (Chomsky 2005), in particular Feature Economy (FE) (13), supplementing and partially replacing the older Head Preference (HPP) (14), and Late Merge Principles (LMP) (15). (13) Feature Economy: Minimize the semantic and interpretable features in the derivation, for example: Adjunct/argument Specifier (of NegP) Head (of NegP) affix semantic > [iNEG] > [uNEG] >— (van Gelderen 2011:299) (14) Head Preference Principle: Be a head, rather than a phrase. (van Gelderen 2011:298) (15) Late Merge Principle: Merge as late as possible. (van Gelderen 2011:298)

Under van Gelderen’s proposal, the LMP accounts for the reanalysis of former arguments and adjuncts as SpecNegP, while the HPP is made responsible for the reanalysis of a specifier as a head, for instance within NegP. FE, finally, connects the changes governed by the LMP and the HPP and introduces a trigger for the lexical renewal typical of cyclical changes by postulating that uninterpretable features are more economical than interpretable ones. Given that uninterpretable features always need to be licensed by interpretable ones to prevent a derivation

28

Anne Breitbarth

from crashing, this causes the grammaticalization of new interpretable material. (16) is a representation of JC based on van Gelderen (2011: 304; her Fig. 8.4).9 (16) a. [NegP __ [Neg’ Nego[uNEG] [VP XP[NEG] ]]] | ↑ b. [NegP XP[NEG] [Neg’ Nego[uNEG] [VP ]]] → [NegP XP[iNEG] [Neg’ Nego [Ø] [VP ]]] c. [NegP __ [Neg’ Nego [iNEG] [VP ]]] | ↑ d. [NegP __ [Neg’ Nego [iNEG] [VP ]]] → [NegP __ [Neg’ Nego [uNEG] [VP ]]]

(by LMP) (by FE) (by HPP) (by FE)

This approach aligns the stages of JC with observations regarding the feature content and phrase structural status of negative markers from a typological point of view (e. g., Zeijlstra 2004: 175). In brief, the observation is that there are essentially three types of negative markers, (i) head negators with an interpretable negation feature (in non-strict NC languages according to Zeijlstra’s 2004 typology) 10, (ii) head negators with an uninterpretable negation feature (in strict NC languages), and (iii) phrasal negators with an interpretable negation feature.11 Regarding the triggers of such changes, Van Gelderen explicitly takes NC to be indicative of the imminent renewal of the expression of sentential negation.12 This is because the ‘weakening’ of the original negative marker that Jespersen (1917) takes to be the first step in the renewal, is seen as a change in the interpretability of the [NEG] feature of the negative marker, with [uNEG] being weaker than [iNEG], because a [uNEG] negator needs to rely on an [iNEG] (or semantically negative) licenser. The renewal of negative markers is therefore explicitly seen as inextricably intertwined with the availability of NC: if a language has NC, it means that its negator is a [uNEG] head requiring strengthening, 9 ‘[NEG]’ in (16) symbolizes semantic negativity, cf. (13). 10 The terminology is due to Giannakidou (1998). Non-strict NC languages are languages in which a concording indefinite cannot co-occur with the marker of sentential negation if it precedes the negative marker (but must co-occur with it when it follows), while concording indefinites have to co-occur with the negative particle regardless of their position in strict NC languages. 11 Zeijlstra (2004) further distinguishes between two types of languages with phrasal negators, those in which the negator bears an [iNEG] feature (a formal syntactic feature), and enters an Agree relation with bearers of a [uNEG] feature, e. g. a head negation particle or concording indefinites (n-words), and those in which the negator is semantically negative. I reject such a distinction, as an element bearing an [iNEG] feature should be indistinguishable from a “semantic” negator at LF. 12 This reflects the common claim that there is a connection between the phrase-structural status of a negative marker (head or phrase) and the availability of negative concord (cf. also Rowlett 1998, who calls it Jespersen’s Generalization after observations by Jespersen 1917, or Zeijlstra 2004). See Wallage (2005) and Breitbarth (2013) for arguments against this generalization.

Jespersen’s Cycle = Minimize Structure + Feature Economy

29

which the concording element is thought to provide, and is regarded as being about to embark on JC (cf. van Gelderen’s 2011: 337–338 brief allusion to the triggers of JC).13 This is problematic, however. Although there may be [iNEG] or semantically negative indefinites, these are not typically the kind of indefinites involved in NC – more common are [uNEG] n-words.14 A comparison of the historical development of indefinites across languages suggests that the latter often arise from NPIs (and potentially formerly positive elements), which become increasingly restricted to the scope of negation, and typically stop at acquiring a [uNEG] feature (that is, their strong NPI-hood is reanalysed as a checking requirement with an [iNEG] negation operator). Developing an [iNEG] feature is extremely rare, and furthermore unexpected under FE; [iNEG]-indefinites typically arise through univerbation with an [iNEG] negator (cf. Haspelmath 1997). Besides, if indefinites in NC languages uniformly strengthened a [uNEG] negator, one could not account (a) for the distinction between strict and non-strict NC, nor (b) for the fact that negative doubling (of any kind) most commonly correlates with negative spread (the co-occurrence of [iNEG] indefinites would lead to double logical negation). Furthermore, the new negators arising through JC are not necessarily former concordant negative indefinites, a prediction that van Gelderen’s proposal seems to make. Finally, the impression arises that syntactic change occurs in a deterministic fashion: weakening under FE is inevitable, triggering – also inevitably – cyclic renewal, making use of the LMP and HPP. However, as variously observed in the literature on JC, languages can be remarkably stable, and remain in a certain stage for a long time, certainly a stage with an emphatic adverbial emphasizer preceding the actual initiation of JC (see e. g. Posner 1985, Breitbarth et al. 2013). Finally, a general problem for accounts of JC assuming a changing occupation of the specifier and head positions of NegP is the fact that new fillers of SpecNegP either invariably follow the old Nego15, or are compatible with an analysis as 13 Note that Zeijlstra (2004) is less strict in this respect, he acknowledges that there are NC languages with [iNEG] phrasal negators. See Wallage (2005) and Breitbarth (2013) for arguments against a direct dependency between JC and NC. 14 The term n-word is used here to denote semantically non-negative indefinites in the scope of negation that can co-occur with each other or with a sentential negator, but which can also express negation alone in elliptical answers: (i) An expression α is an n-word iff: a. α can be used in structures containing sentential negation or another α-expression yielding a reading equivalent to one logical negation; and b. α can provide a negative fragment answer. (Giannakidou 2006: 328) 15 Cf. Hirschbühler & Labelle’s (1993) criticism of Pollock’s (1989) analysis of French ne and pas as Nego and SpecNegP, respectively, arguing that the surface order ne pas must also be the base order.

30

Anne Breitbarth

adjoined to vP, raising questions for language acquisition: Why should a functional projection be postulated if the PLD (Primary Linguistic Data) does not unambiguously indicate its position – unless one assumes functional projections and their positions are not acquired, but hard-wired into UG? I would like to question whether the assumption of a NegP is indeed necessary to capture the data. In the next section, I will outline a NegP-free account of JC.

4.

Jespersen’s Cycle without NegP

4.1

Penka’s NegP-free approach to NC and negation typology

Several authors have opposed the NegP-hypothesis, whether only for English (e. g., Ernst 1992) or universally (Penka 2011). Penka (2011) argues that sentential negation interacts with other scope-taking material in the clause and proposes to treat negation as an operator OP¬ of type < t, t >, requiring an argument of type < t >, that is, a constituent denoting a truth value, returning the inverse truth value to the input truth value. (17) not is of type < t, t > ! 1 if t = 0 [[not]]= λt ∈ Dt 0 if t = 1 (Penka 2011: 10)

According to Penka, the smallest projection of type < t > is vP.16 This approach attributes the ordering restrictions between projections to independently holding semantic requirements, avoiding their doubling in a fixed hierarchy of functional projections in the syntax. Under Penka’s approach (as under Zeijlstra’s), OP¬ can be covert as long as overt material in the sentence indicates its presence, for instance in (constructions in) languages where negative markers appear in surface positions that do not coincide with the semantic scope position, or in case of concording (morphologically, but not semantically) negative indefinites in NC languages (Penka 2011:51).17 As it stands, however, Penka’s approach can only account for two of the three types of negation markers predicted by van Gelderen (2011) or Zeijlstra 16 Other positions are not excluded, however (cf. Penka 2011: 61). 17 Cf. Ladusaw’s (1992) distinction between expressing and licensing the expression of negation: “[The negation operator] need not be part of a lexical meaning: it may be constructional, in the sense that it is associated with some structural feature not necessarily visible in the clause.” (Ladusaw 1992:251–252)

Jespersen’s Cycle = Minimize Structure + Feature Economy

31

(2004), viz. (i) languages with a weak [uNEG] negative marker or affix on the finite verb and a covert [iNEG] OP¬ in SpecNegP (e. g. Czech), and (ii) languages with an [iNEG] phrasal/adverbial negator, which is in SpecNegP in NC-languages (e. g. Bavarian), or adjoined to vP in non-NC languages (e. g. Standard German), according to Zeijlstra’s analysis. In the first type of language, a NegP is not needed under Penka’s approach, according to which [uNEG] features are always licensed under c-command by an [iNEG] licenser. A simple adjunction of the covert OP¬ to vP (or higher) as proposed by Penka would suffice to account for such languages; the [iNEG] operator would c-command the [uNEG] feature on the verb just as it would under Zeijlstra’s proposal involving NegP. Also languages of the second type can be accounted for without NegP, since in Penka’s view, the overt adverbial [iNEG] element can be adjoined to any category of type < t >, licensing any [uNEG] features in its scope. However, there is a third type of language. If the negative marker is an [iNEG] head, as Penka, following Zeijlstra, assumes for non-strict NC languages like Italian, it must attach to the clausal spine, because it, carrying an [iNEG] feature, overtly realises the negation operator, which under Penka’s analysis (17) needs to take a type < t > complement. Though Penka argues against the NegPhypothesis, an [iNEG] head taking a vP or TP complement is de facto hard to distinguish from the head of a NegP. Assuming instead a covert [iNEG] adjoining to vP or TP would imply that the overt sentential negator be [uNEG], which would obliterate the distinction between strict and non-strict NC languages.

4.2

Jespersen’s Cycle (and negation typology) = Minimize Structure + Feature Economy

As is the case with other instances of grammaticalization, the development of negation particles under JC follows an ordered chain of stages, a so-called cline, each representing a higher degree of grammaticalization. In a first step, a nominal minimizer, generalizer or (negative) indefinite is reanalysed as a negative polarity adverb, and from there as a new (adverbial) negative marker. Such phrasal negative markers can then go on to become clitics and ultimately affixes, (18). (18) argument > adverbial emphasizer > phrasal negator > clitic > affix > ∅

We have seen above that the affix-like negation particle of OS becomes reinforced by an adverbial emphasiser arising from a former argument, and that by MLG, this reinforcing element had become a neutral phrasal negation marker.18 The 18 It is not (as van Gelderen’s approach predicts) inevitable that an affixal negator become

32

Anne Breitbarth

similarity between the cline that negation particles passing through JC follow and the development of, for instance, agreement morphology (Givón 1976, van Gelderen 2011), (19), is striking. (19) noun/demonstrative > pronoun > clitic > affix > ∅

Under generative approaches to grammaticalization (Roberts & Roussou 2003, van Gelderen 2011, Kiparsky 2012), the strict directionality from less grammatical to more grammatical, from less abstract to more abstract, and from independent to increasingly dependent is a consequence of universal properties of language, namely the hierarchical (scopal) order of functional categories, combined with so-called third factors (Chomsky 2005) such as economy principles. The functional hierarchy restricts the possible pathways, while the economy principles effect the ‘upwards’ character of grammaticalization (Roberts & Roussou 2003) and the formal cline from full phrase via clitic to affix and eventually zero (van Gelderen 2011). Under a NegP-free approach to the syntax of negation and to JC, the question arises as to which position in the hierarchy newly grammaticalizing negative markers should aim for, if a designated structural position for sentential negation markers is absent. I propose to look at another parallelism with pronouns here. Cardinaletti & Starke (1999) observe that many languages have synonymous pairs or even triplets of pronouns, which are differentiated by contrasting syntactic distributions. The full or ‘strong’ class is syntactically independent, its members can be coordinated, modified, contrastively stressed, and occur in sentence-initial position, like elle ‘she’ in (20a–e). A second, deficient or ‘weak’, class on the other hand does not have these three properties and is restricted to the middle field. The third class, clitics, are finally also prosodically dependent on a host, like la (l’) in (20a). (20) French a. Je ✓l’ ai aidé (*elle). b. Je *LA ai aidé (✓ ELLE). I HER have helped HER c. Je *☞ la ai aidé. (☞elle) I her have helped her d. Je *la et l’ autre ai I her and the other have e. Je seulement ∗la ai aidé I only her have helped (Cardinaletti & Starke 1999:152)

(☞ = ostension) aidé (✓ helped (seulement ✓ only

elle et l’autre). her and the other elle). her

reinforced in this way, as there are NC languages (like Czech) with such negators which never seem to embark onto Jespersen’s Cycle.

Jespersen’s Cycle = Minimize Structure + Feature Economy

33

Based on data from various Romance and Germanic languages, Cardinaletti & Starke (1999) argue that the three classes of pronouns – strong, weak, and clitic – are distinguished by the amount of internal structure they have. Weak pronouns realise less structure than strong pronouns, and clitics are even more structurally impoverished than weak pronouns. Similar to sentential syntax, Cardinaletti & Starke postulate three layers of functional structure for lexical items, C, Σ, and I (i. e. CLP, ΣLP and ILP, where L = any lexical category). The C-layer encodes referentiality and case on pronouns, the Σ-layer prosodic information, and the I-layer φ-features. Therefore, weak pronouns, lacking the Clayer, can only occur in positions where, for example, their case information can be recovered; they suggest the specifier of agreement projections. Clitics, additionally lacking prosodic information, have to attach to a functional head able to license them prosodically. Furthermore, Cardinaletti & Starke (1999: 198) formulate the economy principle Minimize Structure (MS), (21), requiring the use of the most structurally reduced element possible, that is, unless it leads to ungrammaticality (=‘up to crash’). (21) Economy of Representations: Minimize Structure. Only if the smaller structure is independently ruled out, is the bigger alternative possible. (Cardinaletti & Starke 1999: 47)

Ungrammaticality ensues if the omitted functional structure cannot be recovered. This derives the characteristic distributional and prosodic restrictions on strong, weak, and clitic pronouns exemplified in (20). In (20a), for instance, the stronger form elle is ruled out, as the more deficient form la is available. Cardinaletti (2011), following a proposal for German by Grosz (2007), extends this analysis of the three classes of pronouns to modal particles in Italian and German, arguing that they are in fact deficient adverbs, lacking the highest structural layer, which accounts for their distributional restrictions. Cardinaletti shows that, just like weak pronouns, modal particles cannot be coordinated (22a) or modified (22b); they are positionally fixed, (23), but they are still prosodically independent. (22) a. *Gehen Sie doch und mal zum Arzt. go you mp and mp to.the doctor ‘Just go to the doctor.’ (Cardinaletti 2011: 495) b. Gehen Sie (sehr) ruhig zum Arzt! go you very quietly/*mp to.the doctor ‘Go very quietly to the doctor.’ (manner reading only) ‘*Go ahead, just go to the doctor!’ (*MP reading)

34

Anne Breitbarth

(23) a. Das Auto ist klein: billige Autos sind eben so. the car is small cheap cars are mp so ‘The car is small: cheap cars are indeed like that.’ b. *Das Auto ist klein: eben sind billige Autos so. The car is small mp are cheap cars so (Cardinaletti 2011: 495)

Further support for her proposal comes from the observation that some modal particles have arguably progressed further in their grammaticalization19, such as dn/n < denn ‘than’ (e. g. Grosz 2005; 2007). As seen in (24), dn/n is prosodically dependent, and behaves like a C-clitic, appearing in the Wackernagel position. ‘Weak’ denn, like the other modal particles, occurs in a lower position, within the middle field. (24) was schenkst (dn)/(*denn) du ihr (*dn)/(denn) zum Geburtstag? what give mp you to-her mp for-the birthday ‘What are you going to give her for her birthday?’ (adapted from Cardinaletti 2011: 501)

In parallel to the three classes of pronouns identified by Cardinaletti & Starke (1999), Cardinaletti (2011) argues that adverbs can be structurally deficient, too. Like weak pronouns, weak adverbs, that is, modal particles, need to adjoin to a functional projection that can recover their missing features. Like clitic pronouns, clitic adverbs need to adjoin to a functional head able to license them prosodically. I propose to extend this analysis to the cross-linguistic typology and grammaticalization cline of negative markers. That is, I argue that there are different classes of negative markers distinguished by the amount of internal structure they have.20 19 For arguments that (German) modal particles historically derive from full adverbs by grammaticalization see e. g. Molnár (2002). 20 One of the reviewers points out that one difference between pronouns and modal particles is that pronouns all have the same function regardless of their “size”, while modal particles, though historically derived from adverbs, only share their modificational function, and develop rather different functions. Negative particles would therefore be more similar to pronouns than to adverbs/modal particles. Two remarks on this. First, Minimize Structure can be understood diachronically, as is done in the present article (see main text), i. e., the different sizes represent different historical stages. Second, there are observations in the literature pointing to the fact that clitic negative particles, in cases where they are not entirely replaced by new strong > weak ones, tend to undergo exaptation. That is, they are reanalysed and assume a new function, probably because speakers want to make sense of the continued presence of two negative particles. This has for instance happened in (West) Flemish (Breitbarth & Haegeman 2014), where en has been reanalysed as a discourse particle expressing polarity emphasis and in Middle Low German / Middle Dutch, where ne/en used on its own fed into the development of a new connector introducing exceptive clauses (Breitbarth 2014b), which in the case of Dutch led to the grammaticalisation of a new subordinating complementizer, complete with clausefinal verb placement, viz. tenzij ‘unless’ < ’t en zij ‘it NEG be’).

Jespersen’s Cycle = Minimize Structure + Feature Economy

35

MS can be understood diachronically, as an effect of grammaticalization (contra Cardinaletti 2011, but in line with Grosz 2007). This predicts that there may be a period in which older (less grammaticalized) and newer (more grammaticalized) negative markers are simultaneously present in a language. In research on grammaticalization, this is called divergence (Hopper 1991: 22). This is indeed borne out. As is the case for pronouns, of which strong and weak forms may be homophonous in a language, for instance French elle (ILP) and ELLE (CLP), some languages have homonymous forms of negative markers with different amounts of internal structure under the present approach. This offers an explanation for the apparent ‘dual’ status negation markers seem to have in some languages, for instance English or Norwegian, in terms of partially homophonous forms with different degrees of structural deficiency. English has in fact three ‘sizes’ of negative markers, which show a different distribution. While English n’t (25a) has been argued to be an inflectional affix on finite auxiliaries by Zwicky & Pullum (1983), the long-standing controversy about whether unreduced not is a head or a phrase can be settled by assuming that when used to express neutral sentential negation (25b), it is a deficient form.21 (25) a. Has-n’t John read the book? b. Has John not read the book? (Christensen 2003)

Besides that, there is a stressed contrastive form NOT, which is modifiable (cf. absolutely not), and which can therefore be analysed as a strong adverb, distinguished from the deficient element by its distribution. This strong form is the one used in focus and constituent negation. Evidence comes from the difference in LFscope (and hence, by compositionality, underlying syntactic position) between neutral and contrastive not (cf. Cormack & Smith 2002). Neutral not can only scope above possibility modals, and can be reduced to n’t. Contrastive NOT can scope above as well as below possibility modals, and cannot be reduced to n’t, (26). While the reducible ‘neutral’ not expresses sentential negation, as shown by the positive question tag (cf. Klima 1964) in (26a), the strong narrow focus negator NOT in (26b) does not have sentential scope, as the negative question tag indicates. (26) a. Edwin can’t/cannot climb trees, can he? neg > mod ‘Edwin is not permitted/able to climb trees.’ b. Edwin can NOT climb trees, can’t he? mod > neg ‘Edwin is permitted not to climb trees.’ (after Cormack & Smith 2002: 137)

21 Something will have to be said about do-support, which also occurs with (weak) not and (strong) NOT. I restrict myself here to pointing out the general typology of negative markers.

36

Anne Breitbarth

In sum, English has an affix n’t, a weak form not (Cormack & Smith’s Pol[NEG]), and a homophonous strong form NOT (Cormack & Smith’s Adv[NEG]). Likewise, Norwegian ikke has what appear to be affixal or clitic realizations, besides more phrasal ones. Johannessen (1997), who takes this to be evidence of ikke’s general head status, signals that ikke productively assimilates with mostly monosyllabic verbs in verb-second clauses where the finite verb precedes the negator, leading to sometimes far-reaching phonological restructuring, (27). (27) a. har ikke > hakke ‘has not’ b. går ikke > gåkke ‘goes not’ c. syr ikke > sykke ‘sews not’ (from Johannessen 1997: 3)

The strong form ikke can be contrastively stressed and modified, as in English, (28a), and be put in clause-initial position in V2-contexts if strong contrastive focus is involved, (28b). At the same time, it can also arguably be a weak adverb in neutral contexts. (28) a. Jeg aner slettes ikke hva du I know absolutely neg what you ‘I absolutely don’t know what you mean.’ ( Johannessen 1997:4) b. Ikke har han problemer med alkoholen neg has he problems with alcohol.the ‘Neither has he problems with alcohol.’ (Christensen 2003)

mener mean heller. either

The facts for phrasal or adverbial negators like German nicht or Dutch niet are analogous to the Norwegian ones; these negators are ambiguous between ‘strong’ and ‘weak’ adverbs in the proposed classification. The strong forms can be contrastively stressed and modified, (29), while in neutral negative clauses, the weak form is used, which is restricted to a low middle field position at the edge of vP. (29) German a. Das stimmt ganz und gar nicht. that adds.up whole and wholly not ‘That doesn’t add up at all.’ b. Claudia hat A, B und C getan. NICHT hat sie D getan. Claudia has A B and C done not has she D done ‘Claudia has done A, B, and C. Contrary to that, she has NOT done D.’ (Breitbarth 2014a: 135)

If English n’t, Norwegian -kke, or, as Zeijlstra (2004) argues, Czech ni-, are affixal negators, and if neutral not (English), niet (Dutch) and nicht (German) are weak negators, and their contrastive counterparts strong negators, are there also ‘clitic’ negators to complete the typology? Recall that the only problem with

Jespersen’s Cycle = Minimize Structure + Feature Economy

37

Penka’s NegP-free proposal is the existence of preverbal [iNEG] head negators in a number of Romance languages. In many Italian dialects, they form clusters with pronominal clitics, like -l and me in (30). I therefore propose to analyse such negators as clitics on the finite verb in T.22 (30) Venetian No=l me piaze neg=it me likes ‘I do not like it’ (Poletto 2008: 66)

The assumption that T prosodically licenses clitic negators in some Romance languages lends support, for instance, to the assumption that (true) imperatives lack T (Zanuttini 1996). If such clitic negators are dependent on T for licensing, the absence of true (morphological) imperatives in negative clauses in languages like Spanish or Italian (31), and the requirement to use a surrogate form (e. g. subjunctive or infinitive) (Zanuttini 1996, Zeijlstra 2004), fall out naturally. (31) (*Non) telefona! neg call.imp ‘Don’t call!’ (after Zanuttini 1996: 188)

The question arises what information should be encoded in the structural shells of negative markers stripped by MS, and how it is recovered. I propose that the ‘C’-layer corresponds to focus. A ‘full’ or strong negator equipped with it can be used to express narrow focus of negation, a ‘weak’ negator lacking it needs to adjoin to vP, the domain of neutral wide information-structural focus; it can thus not express a different focus itself. The ‘Σ’-layer, just as for pronominal elements and modal particles, encodes prosodic information. A clitic negator, lacking this layer, needs to attach to a functional head in order to recover this information. I just argued that this head is T. Finally, the ‘I’-layer houses an interpretable formal negation feature. An affixal negator, lacking this layer, may be [uNEG], but may also eventually lose a formal negation feature under FE. Table 2 sums up the properties of the proposed grammaticalization stages of negative markers.

22 This may be a simplification, cf. Zanuttini’s (1997) distinction of two positions for clitic negators depending on their distribution with respect to other clitics. Also, clitic negators may be clitic on other functional projections than T in other languages, for instance C.

38

Anne Breitbarth

Table 2: The four classes of negative markers and their properties neg marker strong weak clitic affix

4.3

size CAdvP ΣAdvP IAdvP AdvP

distribution free, constituent neg adjoined to vP clitic (e. g. on T) affix on verb

formal feature [iNEG] [iNEG] [iNEG] > [uNEG] [uNEG] > Ø

Back to Jespersen’s Cycle in historical Low German

We are now in a position to take a new look at the development of the expression of negation in historical Low German. As negative spread seems to be absent in OS (9c–d), we can postulate that the negative marker ni/ne immediately preceding the finite verb had an [uNEG] feature, while negative indefinites were [iNEG] in OS (and hence could not co-occur with each other). Given that the preverbal marker covaries positionally with the finite verb, I would like to propose that it is a [uNEG] affix on the finite verb. (32) is the structural representation of (the relevant parts of) (2). (32) [vP OP¬[iNEG] [vP [DP thu] [v’ … [v ni[uNEG]-bist]]]] (Heliand 919)

As we have seen above, the new negative reinforcer is recruited from a (negative) indefinite argument, (n)iouuiht ‘any-/nothing’, (33a), which can be used as an optional extent argument with certain predicates (33b), and from there be reanalysed as a strong adverb, which can be used in narrow scope and focus negation, (33c). (33) a. [vP [VP [DP niouuiht([iNEG])] imu biholan] [v ni[uNEG]-is]] (Heliand 1577–8) b. [vP [vP ik [VP thi (ti) tV] [v ni[uNEG]-deriuV ]] (Heliand 4865) [DP/CAdvP neouuiht[iNEG]](i) ] c. [DP [CAdvP niet[iNEG]] [DP ekir io selon]] (GG 63, 15–16) d. [vP [CAdvP nicht[iNEG]] [vP … ]]

Once established as an adverbial negator (CAdvP), niouuiht > nicht becomes a weak adverbial negator adjoined to vP, (34). During the MLG period, the former affixal marker on the finite verb is lost, after FE stripped it of any [uNEG] features. (34) [vP [ƩAdvP nicht [iNEG] [vP … [v en[Ø]-wete]]]] (Braunschweig 1349)

These stages are connected by an interplay between MS and FE. MS favours the insertion of more deficient forms where there is a choice between two available forms, that is, it requires the use the smallest synchronically available adverb – where possible – and the diachronic reduction of the size of an available adverb – where possible. FE requires the reduction of interpretable features ([iNEG]) to

Jespersen’s Cycle = Minimize Structure + Feature Economy

39

uninterpretable features ([uNEG]) to nothing (Ø) – again, where possible. This raises the question what determines the diachronic rise of a choice between two forms. Given the role of inertia in language change (cf. Keenan 2002, Longobardi 2001), other (and robust) triggers are required to set off structural change. That is to say that MS does not create more deficient forms to be used by itself. Neither can FE, under inertia, apply without being triggered. Clearly, MS is also not designed to account for the cyclic renewal of the expression of negation with new phrasal markers. The grammaticalization of new negation markers is a complex interplay of semantic and syntactic properties of input items, lexical bridging contexts, and the loss of original syntactic and semantic distributional restrictions (Breitbarth et al. 2013). With respect to JC in historical Low German, I propose that the rise of [iNEG] CAdvP niouuiht in OS was triggered by the presence of [iNEG] negative indefinites, in fact, one such indefinite was reanalysed as a negation strengthener (‘at all’ > ‘not’) exploiting the pragmatic strategy of reinterpreting a low endpoint on a pragmatic scale (‘anything’ / ‘nothing’) in a bridging context where both the argumental and the adverbial reading are possible – with predicates allowing an optional extent argument (Breitbarth et al. 2013). This new adverbial marker lost its C-layer, turning it into a ‘weak’ ƩAdvP negator adjoined to vP under MS, at the point when new adverbial emphasizers were established (e. g. MLG mit nichte ‘not at all’ lit. ‘with nothing’). At the same time, the former prefinite negation particle, ni > ne/en, lost its [uNEG] feature under FE. Table 3 gives an overview over the developments in historical Low German: Table 3: The development of negation and NC in historical Low German pre-OS: OS-1 (Hel,Gen): OS-2 (minor texts): MLG:

5.

IAdvP [iNEG] ni + NPI indefinites AdvP [uNEG] ni + NPI / [iNEG,iQ] indefinites AdvP [uNEG] ni + CAdvP [iNEG] niouuiht + [iNEG,iQ] > [uNEG,uQ] indefinites AdvP [uNEG] > Ø ne/en + ƩAdvP [iNEG] nicht + [uNEG,uQ] indefinites

No one’s problem

One of the empirical facts that has often been cited in favour of the NegPhypothesis is the observation that negative indefinites in many languages have to evacuate their base position inside VP and move to a position that has in the literature been identified as SpecNegP. There are two sets of facts. First, in some NC languages (like West Flemish (WF) or Bavarian), negative indefinites have to move out of their base position in order to take sentential scope and avoid a

40

Anne Breitbarth

double negation (DN) interpretation (Haegeman & Zanuttini 1991, Weiß 2002, Svenonius 2002), e. g. (35). (35) Bavarian a. da Sepp war [mid nix]i ned ti zfriem. (NC) the Joe was with nothing neg satisfied ‘Joe was not satisfied with anything.’ b. da Sepp war ned [mid nix] zfriem. (DN) the Joe was neg with nothing satisfied ‘Joe was not satisfied with nothing.’ (Weiß 2002: 147)

Second, in the Scandinavian languages, negative indefinites are fine in postverbal position in verb second clauses as long as the verb is not complex (36a). In case it is, the finite auxiliary is in the verb-second position, while the non-finite main verb remains low. Even though the Scandinavian languages are SVO, a negative indefinite may not remain in its base position in such a case (36b) (Christensen 1986).23 That is, it either has to shift out of VP (vacuously satisfied in (36a)), which however is only fully grammatical in case of complex verbs (36c) in Icelandic, Faroese and West Jutlandic, and marginal in (literary) Norwegian, or negation has to be spelled out split from the indefinite (36d).24 (36) Norwegian (Bokmål) a. Jon leser ingen Jon reads no b. *Jon har lest Jon has read c. % Jon har ingen Jon has no d. Jon har ikke Jon has neg e. Ingen romaner har no novels has (Christensen 1986)

romaner. novels ingen no romaneri novels lest read Jon Jon

romaner. novels lest t i. read noen romaner. any novels lest. read

Both the Bavarian/WF data and the Scandinavian data have in common that negative indefinites may not occur inside VP. Haegeman & Zanuttini (1991) have argued for the former group of data that negative indefinites can only take sentential scope if they raise out of VP to “the specifier of a functional head bearing a negative feature” (thus, SpecNegP), under a condition they call the Neg-Criterion, (37), which has to be satisfied by LF at the latest (subject to cross-linguistic variation).

23 The phenomena studied by Ingham (2000) also belong here. 24 The patterns of variation are complex, cf. Engels (2012).

Jespersen’s Cycle = Minimize Structure + Feature Economy

41

(37) The Neg-Criterion (a) Each Nego must be in a Spec-Head relation with a Negative operator; (b) Each Negative operator must be in a Spec-Head relation with a Nego. (Haegeman & Zanuttini 1991: 244)

There are two problems with the original proposal: first, Haegeman & Zanuttini (1991) assume negative indefinites to be semantically negative, needing to undergo a process of Factorization and Absorption in SpecNegP to achieve a single sentential negation interpretation. This has been argued not to be true universally, in particular not in NC languages like Bavarian and WF, for which this mechanism was designed (cf. Zeijlstra 2004, Penka 2011). Furthermore, the movement also occurs in languages without NC (Christensen 1986, Weiß 2002). A second, more theory-internal problem is that in recent versions of the theory, licensing in specifier-head-configurations has been replaced by the operation Agree, which creates a syntactic dependency between a Probe and a Goal c-commanded by the Probe, and can, but does not need to be followed by movement, which needs to be independently triggered (Chomsky 2000, 2001). Note that under Zeilstra’s analysis of the WF and Bavarian facts (38), NegP does not even have anything to do with the shift of the indefinite (which he vaguely attributes to scrambling for scope reasons, an option that would not be available in the Scandinavian languages, as they do not have scrambling). (38) a. S’Maral woid koane (ned) hairadn the=Mary.dim wanted no.one neg marry ‘Mary did not want to marry anyone.’ b. [NegP OP[iNEG] Neg [ s’Maral [koane[uNEG] [vP ned[uNEG] [v woid] [VP hairadn]]]]] (Zeijlstra 2004: 256)

Finally, the approach does not carry over to the Scandinavian phenomena, as Neg-shift exactly to the clause-medial position, which could be identified with SpecNegP, is not possible or degraded in many varieties. The empirical facts of course still stand, so in fact, both a NegP-free account as well as NegP-accounts need a new explanation for these data (or, to turn around the common saying, no one’s problem is everyone’s problem now). I do not have the space here to elaborate a full-fledged alternative account, but I would like to suggest a possible analysis which essentially takes Neg-Shift to be overt QR with partial reconstruction (cf. Zeijlstra 2011). This sketch of an alternative account departs from the assumption that negation is not a Boolean operator (as it is for Penka 2011 for instance), but a quantifier, consisting of operator, restriction, and scope (cf. Weiß 2002 and literature cited there) (39). (39) NEGx [event(x)][P(…x…)]

42

Anne Breitbarth

Adopting this view, I would like to propose that negation, in order to take sentential scope (i. e., over the event variable), needs to reach the syntactic domain that is mapped to the operator of the tripartite quantifier, and leave the domain that is mapped to nuclear scope, i. e. VP (Diesing 1992). Note furthermore that nuclear scope is the domain associated with focus, while elements in the restrictor clause are interpreted as background (Partee 1991). A further essential ingredient for my sketch of an account is Zeijlstra’s (2011) analysis of negative indefinites (NIs) in languages like Dutch and German as being syntactically composed of a negative operator and non-negative indefinites (like nwords), which undergo QR and may then partially reconstruct (only the indefinite part). This can account for split-scope readings such as (40).25 (40) Du musst keine Krawatte anziehen. you must no tie wear ‘It is not required that you wear a tie.’ (¬ > must > $) (Zeijlstra 2011: 113)

Next, note that languages with Neg-Shift in NC-languages discussed in the literature are all languages for which there is a standard variety with NIs which behave like NQs, such as Bavarian (German) and West Flemish (Dutch). That is, their morphologically negative indefinites are possibly NQs (with more internal structure as proposed by Zeijlstra) when inside VP, but n-words (without an extra layer for the negative operator) when shifted.26 The Scandinavian languages have Neg-Shift, but do not have NC. Assuming that NIs in the Scandinavian languages also syntactically decompose in the manner outlined by Zeijlstra (2011) gives us a handle on the enormous variation in Neg-Shift (Christensen 1986, Svenonius 2002, Engels 2012) in the Scandinavian languages: where (for reasons which are irrelevant here) complete shift of the NI to the clause-medial position is not possible or degraded27, the indefinite is spelled out separately from the negative operator. If the negative operator is interpreted inside nuclear scope (=VP), it will be interpreted together with the indefinite as a negative quantifier. The fact that non-shifted NIs tend to have a contrastive, focal interpretation squares with nuclear scope being associated with focus (Partee 1991). This is not, in fact, restricted to the interaction of negation and indefinites in its scope. Dornisch (2000) observes that in Polish, postverbal negative as well as non-negative 25 As also discussed in Penka (2011) and earlier publications of hers, the other two readings with the NI as a negative quantifier (¬ > $ > must and must > ¬ > $) are possible, too, but less natural; see also Penka in this volume. 26 Note that the VP-internal position is also the one where only “strong” pronouns under MS can occur. 27 Complete shift is always possible to sentence-initial position.

Jespersen’s Cycle = Minimize Structure + Feature Economy

43

quantifiers are only felicitous under special information structural conditions: they require heavy contrastive stress (41a). The unmarked position for both types of quantifier is the shifted one (41b).28 If Neg-Shift cannot be generally attributed to the need to satisfy a feature in NegP, but is only an instantiation of a more general pattern of quantifier shift, there is no requirement that the target of the movement of negative indefinites be NegP either, casting further doubt on the necessity of assuming a NegP. (41) Polish a. Anna ✓ nikogoi nie Anna no.one NEG b. Anna ✓ cos´i widziała ti / Anne something saw ‘Anna saw no one / something.’ (Dornisch 2000)

widziała ti / %NIKOGO. saw no one %COS´. something

The proposed alternative to a NegP-account is therefore that it is not a property of a higher Neg-head which triggers the evacuation of VP, but properties of the shifting phrases themselves. That is, Neg-Shift is a form of foot-driven movement. Under a neutral, sentential scope interpretation, the negative operator is interpreted in the shifted position, and the indefinite part is interpreted inside nuclear scope, where it is existentially closed. Under a focussed interpretation, however, the negative operator can stay low, giving rise to the narrow focus/ constituent negation reading (cf. also the observations by De Clercq 2011). Finally, it is worth pointing out the diachronic dimension of this proposal. Recall that new negation markers entering JC are first used as focus or constituent negators in incipient stage II of JC as seen in (5)/(33c), that is, in positions where MS predicts “strong” elements to occur. Recall furthermore that OS seems to develop morphologically negative indefinites only slowly, by univerbation of a non-negative indefinite and a negation marker, and that they originally could not co-occur with each other, indicating that they were interpreted like e. g. German NIs. Under present assumptions, FE later turned them into [uNEG] n-words, while MS removed the extra OP¬-layer, on the way to MLG. That is, the sketched account of Neg-Shift can very well be reconciled with the proposed account of JC, assuming that under MS, negative indefinites can only be inserted inside VP if they are “strong”, that is, contain an extra structural layer with their own OP¬, while they can receive the “split” interpretation, or be n28 In Icelandic, while Neg-shift is obligatory, also (non-negative) quantifiers optionally leave VP, (i) (Svenonius 2002 and literature cited there). (i) Ég hef (mörgum bókum) skila Jóni (mörgum bókum). I have many books returned John many books ‘I have returned many books to John.’ (adapted from Svenonius 2000)

44

Anne Breitbarth

words (without the additional layer) in the medial position, and need to be inserted (under MS) in the absence of focus.

6.

Conclusion

The main claim defended in the current paper was that it is possible to capture the stages of JC, as well as the typology of adverbial negative markers, cross-linguistically, by a combination of two third-factor principles, Cardinaletti & Starke’s (1999) Minimize Structure and van Gelderen’s (2011) Feature Economy, without at the same time adopting the NegP-hypothesis. Based on observations on the development of negation in historical Low German as well as the distribution of negative particles in other languages, it was shown that negative markers across languages and historical stages of languages show a distribution comparable to the distribution of pronouns or modal particles observed in the literature, which is determined by differing amounts of internal structure, and which can be linked to a diachronic cline. The same way there are strong, weak and clitic pronouns (Cardinaletti & Starke 1999), which can eventually grammaticalize further to become agreement affixes (van Gelderen 2011), there are strong, weak, clitic and affixal negative markers. In addition to the amount of internal structure they have, which is subject to MS, and which determines their syntactic distribution, they bear formal negation features which are subject to FE. Among others, this determines how they interact with indefinites in the scope of negation. It was further argued that these two principles do not apply automatically, as syntactic change only happens given solid triggers, otherwise, syntax is inert (Keenan 2002, Longobardi 2001). The proposed account makes clear predictions concerning the kinds of diachronic stages and language types that are possible, and extends Penka’s NegP-free approach to non-strict NC languages by identifying a class of clitic negative adverbs, besides the less controversial affixal and phrasal ones. Finally, contrary to common belief, the NegP-hypothesis does nothing to elucidate the two different kinds of NegShift facts, as the evidence that SpecNegP is the position targeted is insufficient. I sketched an alternative account of Neg-Shift that starts from the assumption that negation is a quantifier in the scope of which only focussed material or existentially closed indefinites can stay. In order to obtain sentential scope, any negative operators need to be interpreted outside this domain, but that does not mean that they need to be in a specific specifier position for that.

Jespersen’s Cycle = Minimize Structure + Feature Economy

45

References Besten, Hans den (1983): Double negation and the genesis of Afrikaans. In Muysken, Peter & Norval Smith (eds.): Substrates versus Universals in Creole Languages. Amsterdam/Philadelphia: John Benjamins, 185–230. Breitbarth, Anne (2013): Indefinites, negation and Jespersen’s Cycle in the history of Low German. Diachronica 30: 171–201. Breitbarth, Anne (2014a): The history of Low German negation. Oxford: Oxford University Press (Oxford Studies in Diachronic and Historical Linguistics 13). Breitbarth, Anne (2014b): Exceptive negation in historical Low German. Paper presented at Diachronic Generative Syntax 16, Budapest 3–5 July 2014. Breitbarth, Anne, Christopher Lucas & David Willis (2013): Incipient Jespersen’s Cycle: The (non-)grammaticalization of new negative markers. In Fleischer, Jürg & Horst Simon (eds.): Comparing Diachronies/ Sprachwandelvergleich. Tübingen: Niemeyer, 143–164. Breitbarth, Anne & Liliane Haegeman (2014): The distribution of preverbal en in (West) Flemish: syntactic and interpretive properties. Lingua 147: 69–86. Cardinaletti, Anna (2011): German and Italian modal particles and clause structure. The Linguistic Review 28: 493–531. Cardinaletti, Anna & Michal Starke (1999): The typology of structural deficiency: A case study of the three classes of pronouns. In van Riemsdijk, Henk (ed.): Clitics in the Languages of Europe (EALT/EUROTYP 20–5). Berlin/New York: de Gruyter, 145–233. Christensen, Kirsti Koch (1986): Norwegian ingen: A case of postsyntactic lexicalization. In Dahl, Östen & Anders Holmberg (eds.): Scandinavian Syntax. Stockholm: Institute of Linguistics, University of Stockholm, 21–35. Christensen, Ken Ramshøj (2003): On the synchronic and diachronic status of the negative adverbial ikke/not. Working Papers in Scandinavian Syntax 72: 1–53. Chomsky, Noam (1995): The Minimalist Program. Cambridge, MA: The MIT Press. Chomsky, Noam (2000): Minimalist Inquiries: The Framework. In Martin, Roger et al. (eds.): Step by Step: Essays on Minimalist Syntax in Honor of Howard Lasnik. Cambridge, MA: The MIT Press, 89–155. Chomsky, Noam (2001): Derivation by phase. In Kenstowicz, Michael (ed.): Ken Hale: A life in language. Cambridge, MA: The MIT Press, 1–52. Chomsky, Noam (2005): Three factors in language design. Linguistic Inquiry 36: 1–22. Cormack, Annabel & Neil Smith (2002): Modals and negation in English. In Barbiers, Sjef et al. (eds.): Modality and its Interaction with the Verbal System. Amsterdam: John Benjamins, 133–163. De Clercq, Karen (2011): SQUAT, zero and nothing: syntactic negation vs. semantic negation. In Elenbaas, Marion & Rick Nouwen (eds.): Linguistics in the Netherlands 2011. Amsterdam/Philadelphia: Benjamins, 14–24. Diesing, Molly (1992): Indefinites. Cambridge, MA: The MIT Press. ´ ski, Piotr & Adam Dornisch, Ewa (2000): Overt Quantifier Raising in Polish. In Ban Przepiórkowski (eds.): Proceedings of GliP-1. Warszawa: Polish Academy of Sciences, Institute of Computer Science, 47–58.

46

Anne Breitbarth

Engels, Eva (2012): Scandinavian Negative Indefinites and Cyclic Linearization. Syntax 15: 109–141. Ernst, Thomas (1992): The phrase structure of English negation. The Linguistic Review 9: 109–144. Gelderen, Elly van (2008): Negative cycles. Linguistic Typology 12: 195–243. Gelderen, Elly van (2011): The Linguistic Cycle: Language Change and the Language Faculty. Oxford: Oxford University Press. Giannakidou, Anastasia (1998): Polarity sensitivity as (non)veridical dependency. Amsterdam: John Benjamins. Giannakidou, Anastasia (2006): N-words and negative concord. In Everaert, Martin et al. (eds.): The Blackwell companion to syntax, vol. iii. Oxford: Blackwell, 327–391. Givón, Talmy (1976): Topic, pronoun and grammatical agreement. In Li, Charles N. (ed.): Subject and Topic. New York: Academic Press, 149–188. Grosz, Patrick (2005): dn in Viennese German: The syntax of a clitic version of the discourse particle denn. Diplomarbeit, Universität Wien. Grosz, Patrick (2007): German discourse particles are weak sentence adverbs. In Picchi, Maria Cecilia & Alan Pona (eds.): Proceedings of the XXXII Incontro di Grammatica Generativa, Firenze, 2–4 March 2006. Allessandria: Edizioni dell’Orso, 79–91. Haegeman, Liliane & Raffaella Zanuttini (1991): Negative heads and the Neg-Criterion. The Linguistic Review 8: 233–252. Haspelmath, Martin (1997): Indefinite Pronouns. Oxford: Oxford University Press. Hirschbühler, Paul & Marie Labelle (1993): Le statut de (ne) pas en français contemporain. Recherches Linguistiques 22: 31–58. Hopper, Paul (1991): On some principles of grammaticization. In Closs Traugott, Elizabeth & Bernd Heine (eds.): Approaches to Grammaticalization, vol. I. Amsterdam: John Benjamins, 17–36. Iatridou, Sabine (1990): About AgrP. Linguistic Inquiry 21: 551–576. Ingham, Richard (2000): Negation and OV order in Late Middle English. Journal of Linguistics 36: 13–38. Jespersen, Otto (1917): Negation in English and Other Languages. Copenhagen: A. F. Høst (Historisk-filologiske Meddelelser I,5). Johannessen, Johanne Bondi (1997): The status of the negation ikke in Norwegian. Ms., University of Oslo. Julien, Marit (2002): Syntactic Heads and Word Formation. Oxford: Oxford University Press. Keenan, Edward (2002): Explaining the creation of reflexive pronouns in English. In Minkova, Donka & Robert Stockwell (eds.): Studies in the History of the English Language: A Millennial Perspective. Berlin/New York: de Gruyter, 325–354. Kiparsky, Paul (2012): Grammaticalization as optimization. In Jonas, Dianne et al. (eds.): Grammatical Change. Origins, Nature, Outcomes. Oxford: Oxford University Press, 15–51. Klima, Edward (1964): Negation in English. In Fodor, Jerry A. & Jerrold Katz (eds.): The Structure of Language: Readings in the Philosophy of Language. Englewoods Cliffs: Prentice-Hall, 246–323.

Jespersen’s Cycle = Minimize Structure + Feature Economy

47

Ladusaw, William A. (1992): Expressing Negation. In Barker, Chris & David Dowty (eds.): Proceedings from Semantics and Linguistic Theory (SALT) II. Working Papers in Linguistics 40. Columbus: Ohio State University, 237–259. Longobardi, Giuseppe (2001): Formal syntax, diachronic minimalism, and etymology: The history of French chez. Linguistic Inquiry 32: 275–302. Molnár, Anna (2002): Die Grammatikalisierung deutscher Modalpartikeln. Frankfurt: Peter Lang. Partee, Barbara (1991): Topic, focus, and quantification. In S. Moore & A. Wyner (eds.): Proceedings from Semantics and Linguistic Theory (SALT) II. Cornell Working Papers in Linguistics 10. Ithaka: Cornell University, 159–187. Penka, Doris (2011): Negative Indefinites. Oxford: Oxford University Press. Poletto, Cecilia (2008): On negative doubling. Quaderni di Lavoro ASIt 8: 57–84. Pollock, Jean-Yves (1989): Verb Movement, Universal Grammar, and the Structure of IP. Linguistic Inquiry 20: 365–424. Posner, Rebecca (1985): Post-verbal negation in non-standard French: A historical and comparative view. Romance Philology 39: 170–197. Roberts, Ian (2007): Diachronic Syntax. Oxford: Oxford University Press. Roberts, Ian & Anna Roussou (2003): Syntactic Change: A Minimalist Approach to Grammaticalization. Cambridge: Cambridge University Press. Rowlett, Paul (1998): Sentential Negation in French. Oxford: Oxford University Press. Sanders, Willy (2000): Die Textsorten des Mittelniederdeutschen. In Besch, Werner et al. (eds.): Sprachgeschichte: Ein Handbuch zur Geschichte der deutschen Sprache und ihrer Erforschung. 2. Teilband. Berlin/New York: Walter de Gruyter, 1276–1282. Sells, Peter (2000): Negation in Swedish: Where it’s not at. In Butt, Miriam & Tracy H. King (eds.): Proceedings of the LFG ’00 Conference. Stanford: CSLI Publications, 244–263. Svenonius, Peter (2002): Strains of Negation. Working Papers in Scandinavian Syntax 69: 121–146. Wallage, Philipp (2005): Negation in Early English: Parametric Variation and Grammatical Competition. PhD thesis, University of York. Weiß, Helmut (2002): A quantifier approach to negation in natural languages. Or why negative concord is necessary. Nordic Journal of Linguistics 25: 125–153. Willis, David (2012): A minimalist approach to Jespersen’s Cycle in Welsh. In Jonas, Dianne et al. (eds.): Grammatical Change: Origins, Natures, Outcomes. Oxford: Oxford University Press, 93–119. Zanuttini, Raffaella (1996): On the relevance of tense for sentential negation. In Belletti, Adriana & Luigi Rizzi (eds.): Parameters and Functional Heads: Essays in Comparative Syntax. New York: Oxford University Press, 181–207. Zanuttini, Raffaella (1997): Negation and Clausal Structure: A Comparative Study of Romance Languages. Oxford/New York: Oxford University Press (Oxford Studies in Comparative Syntax). Zeijlstra, Hedde (2004): Sentential Negation and Negative Concord. PhD thesis, Universiteit van Amsterdam. Zeijlstra, Hedde (2011): On the syntactically complex status of negative indefinites. Journal of Comparative Germanic Linguistics 14: 111–138. Zwicky, Arnold M. & Pullum, Geoffrey K. (1983): Cliticization vs. inflection: English n’t. Language 59: 502–513.

Karen De Clercq

The nanosyntax of French negation: A diachronic perspective*

Abstract: This paper presents a nanosyntactic account for negation in le bon usage French (Grevisse & Goosse [1936] 1993, Rooryck 2010) and in colloquial French. The analysis captures the obligatory presence of ne and pas for the expression of sentence negation in le bon usage French, whilst also accounting for the loss of ne in colloquial French. The difference between ne and pas in le bon usage French boils down to a structural difference: ne consists of only one feature or syntactic layer, a polarity feature, whereas pas consists of four features: focus, degree, quantity and negation proper. Only the combination of these five features can yield sentential negation. The analysis of colloquial French is that the lexical item pas starts spelling out five features, thus making ne redundant for the expression of sentential negation. Given that the use of negative markers in le bon usage French and colloquial French represents two important and well-described stages in Jespersen’s Cycle (Dahl 1979) this account is clearly relevant from a diachronic perspective. Keywords: negation, nanosyntax, French, Jespersen Cycle, semantics, morphology, syntax

* I want to express my gratitude to Michal Starke and Guido Vanden Wyngaerd for comments on my dissertation, which helped me refine and improve on the analysis provided in this paper. I further want to thank the audience of the Jerusalem Workshop on Syntax and Semantics: Negation and Polarity at the Hebrew University (8–10 February 2015), and more in particular Elitzur Bar-Asher Siegal, Hedde Zeijlstra, and Anne Breitbarth for useful comments. Final thanks go to the anonymous reviewers of my paper for their useful comments and suggestions. All errors are of course mine.

50

1.

Karen De Clercq

Introduction

In this paper I propose a nanosyntactic account for French negation in what is referred to as le bon usage1 French (Grevisse & Goosse [1936] 1993, Rooryck 2010) and in colloquial French, two forms still present in French today. Whereas ne and pas are both obligatorily present in le bon usage French to express sentence negation, as in (1a), it is optional or lost in colloquial French, illustrated in (1b). (1) a. Je n’ ai pas faim. I neg have neg hunger ‘I’m not hungry.’ b. J’ ai pas faim. I have neg hunger ‘I’m not hungry.’

The usage of ne . . . pas in le bon usage French is known as bipartite, embracing or discontinuous negation. The term bipartite negation as it is used in this paper does not refer to the co-occurrence of a negative marker and one or more negative indefinites or negative polarity items, as in (2). (2) Je ne veux PAS que PERSONNE vienne. I neg want neg that personne come-SUBJ ‘I don’t want anyone to come.’ (Kayne 1982: 40)

The analysis proposed in this paper captures the obligatory presence of ne and pas for the expression of sentence negation in le bon usage French, but will not deal with negative arguments (such as rien ‘nothing’, aucun ‘no’, personne ‘noone’) or temporal adverbs ( jamais ‘never’). The core issue that I address is the fact that both ne and pas seem to convey negation and yet the fact that they co-occur does not lead to a double negation effect. Whilst the inherent negativity of pas is evident from the fact that ne can be deleted without losing negative meaning, the inherent negativity of ne is less immediately obvious. However, a clear example showing the inherent negativity of ne comes from the obligatory presence of ne with plus to express negative meaning in (3). (3) Julie *(ne) boit plus. Julie neg drinks more ‘Julie no longer drinks.’

1 With the term le bon usage French I refer to the written literary form of French described in the prescriptive grammar by Grevisse ([1936] 1993). Rooryck (2010) states that le bon usage French is a form of French that was spoken by the middle and upper class between 1840 and 1960. In le bon usage French standard clausal negation is expressed by means of the bipartite structure, ne…pas, which was the regular way to negate a clause from the 17th century onwards and is still a common way to negate a clause (Hansen 2013: 63).

The nanosyntax of French negation: A diachronic perspective

51

One way to account for the fact that ne and pas both convey negation without leading to a double negation effect is by means of an asymmetric approach: either only pas is inherently negative and ne becomes negative via Dynamic Agree (Rowlett 1998: 28), or Agree (Roberts & Roussou 2003: 154–155, Roberts 2007: 64–81), or pas is only semantically negative and ne is an NPI (Zeijlstra 2009). An immediate problem with respect to the latter approach is that if ne were an NPI, then its obligatory presence in data like (3) remains unexplained. Whereas nothing in the current asymmetric approaches forces ne to be present for the expression of sentential negation, i. e. its presence is stipulated, the proposal developed in this paper manages to capture the difference between ne and pas and accounts for its obligatory co-occurrence in le bon usage French. Two crucial ingredients of the present proposal are, firstly, that the expression of negation is featurally complex; and secondly, that the presence of ne follows from the syntactic structure of the lexical items involved in the expression of sentential negation. In the present proposal the difference between ne and pas in le bon usage French thus boils down to a structural difference: ne consists of only one feature or syntactic layer, a polarity feature, whereas pas consists of four features, focus, degree, quantity and proper negation. Only the combination of these four features can yield what is conceived of as sentential negation. With respect to colloquial French this analysis implies that at some point the syntactic structure of pas encoded in a post-syntactic lexicon started spelling out five features, thus making ne redundant for the expression of sentential negation. Given that the use of negative markers in le bon usage French and colloquial French represents two important and well-described stages in what is referred to as Jespersen’s Cycle (Dahl 1979) this account is also relevant from a diachronic perspective. I shall therefore discuss and analyse these diachronic changes in terms of my proposal. The outline of this chapter is as follows. First, I explain briefly what Jespersen’s Cycle is and in what way the use of negative markers in present-day French is important from a diachronic perspective as well. Second, I demonstrate how I classified negative markers in four different groups. Third, I discuss the nanosyntactic framework I used. Fourth, I apply the system to negation in le bon usage and colloquial French, and then I conclude.

2.

Jespersen’s Cycle

The bipartite pattern displayed in le bon usage French can be explained as a wellknown development of the expression of sentential negation referred to as Jespersen’s Cycle. The term Jespersen’s Cycle was first used by Dahl (1979) to refer to the evolution described by Jespersen (1917: 4):

52

Karen De Clercq

The history of negative expressions in various languages makes us witness the following curious fluctuation: the original negative adverb is first weakened, then found insufficient and therefore strengthened, generally through some additional word, and this in its turn may be felt as the negative proper and may then in course of time be subject to the same development as the original word.

The evolution of negative markers has since then often been represented by means of three stages: (4) a. Stage I: Preverbal expression of sentential negation. b. Stage II: Discontinuous expression of sentential negation. c. Stage III: Postverbal expression of sentential negation. (De Swart 2010: 114)

In general we can say that the preverbal negator in Stage I gets strengthened at some point by an emphatic element, which leads to the development of bipartite negation or Stage II once the emphasizer gets semantically bleached and becomes a regular negator. The preverbal element then disappears due to the presence of this new negator and gets lost, thus entering Stage III (Willis et al. 2013: 6–7).2 The three-way split, as in (4), is a simplification for French, because there are intermediate stages in the development of negation in French (Rowlett 1998: 96, Zeijlstra 2004: 56, Van der Auwera 2009, Breitbarth & Haegeman 2010), in which one of the two elements turns out to be optional. Therefore, it has been proposed, depending on the language under consideration, that there can be five or more stages (cf. Van der Auwera 2009 for an overview) or intermediate stages (Willis 2011) of the cycle. In (5) the different stages of the Jespersen cycle are represented for French. When a negator is put between brackets it means it is optional at this point. (5) French a. Jeo ne di. (→ 1600) b. Je ne dis (pas). (1600 → 1700) c. Je ne dis pas. (Standard written French) d. Je (ne) dis pas. (Standard spoken French) e. Je dis pas. (Colloquial French) ‘I don’t say.’ ( Jespersen 1924: 335–336; Rowlett 1998: 90)

Stage II in French involves a period in which the postverbal negative element seems optional. Willis (2011: 94) labels this as Stage IIa, and the period in which the postverbal negator becomes compulsory as Stage IIb. The transition from stage II to stage III also involves periods in which the preverbal element seems optional or the 2 I discuss the intermediate stages of the Jespersen cycle, during which the presence of the emphatic element pas or the preverbal marker ne is optional, at the end of section 5.2 below, arguing that they instantiate co-existing grammars in speakers.

The nanosyntax of French negation: A diachronic perspective

53

two stages co-exist. In modern spoken French for instance, the marker ne has not yet fully disappeared, even though it is not obligatory anymore.3 In spite of the huge amount of literature on French negation, the analysis of the respective role of the two negative components in Stage II of the Jespersen cycle is not easily captured within a formal system. There is discussion as to whether stage II should be seen as symmetric or asymmetric (Breitbarth & Haegeman 2010: 68). Under the symmetric approach, both elements are considered to have the same weight in the expression of sentential negation, i. e. both are negative markers. However, it seems hard to capture this in a formal system without ending up with a double negation at the semantic level. Under an asymmetric approach to stage II, ‘the two elements present are not both at the same time related to the expression of negation’ (Breitbarth & Haegeman 2010: 68): one element is conceived of as a semantic negator and the other is seen as, for instance, a negative polarity element (Zeijlstra 2009); alternatively, pas is inherently negative and makes ne negative via Dynamic Agreement (Rowlett 1998: 28, Rizzi 1996: 76); another possibility is that pas has interpretable features and checks the uninterpretable features of ne (Roberts & Roussou 2003: 154–155). An asymmetric approach seems to be more easily rendered within a formal system. Nevertheless, it is a fact that there is a moment in the development of negative markers where both elements are equally necessary to render sentential negation (cf. Willis’s (2011: 94) stage IIb). In present day formal written French, i. e. le bon usage French, both ne and pas are equally necessary ingredients for the expression of sentential negation. Whereas nothing in the current asymmetric approaches forces ne to be present for the expression of sentential negation, i. e. its presence is stipulated, I will show that the present account manages to capture the asymmetry between the negative markers in Stage IIb, whilst still accounting for the necessary presence of ne to express sentential negation. Even though the trigger for change from one stage to another will not be the central topic of this paper, I want to consider for a moment how the evolution from stage I to stage II and from stage II to stage III might have been triggered. For French, this means that the original preverbal negator ne was at some point considered too weak to express negation on its own. This “weakness” is sometimes considered a consequence of the phonological weakening of the preverbal negative marker, the so-called pull-chain approach (Breitbarth 2009: 85), as suggested by Jespersen (1917), or as a consequence of the use of a new emphatic

3 Similarly, the transition from stage III back to stage I can also involve an intermediate stage in which the postverbal negator co-occurs with its weakened form. This stage cannot be described for French, but has been described for English. The clitic n’t which attaches to the finite verb cooccurs with the full negator not in some varieties of English, such as African American English (Zeijlstra 2004: 55).

54

Karen De Clercq

negator, the so called push-chain approach (Breitbarth 2009: 86), as already suggested by Meillet (1912) (cf. Hansen 2013: 51–53). Under the pull-chain approach the preverbal marker remains negative until it disappears. Arguments against this approach have come from Posner (1985: 177), who argues that phonological weakening does not necessarily lead to the development of a new negator, as is the case with the negator in South Central Italian dialects. Under a push-chain approach the preverbal negator is pushed away due to a newly emerged emphatic element. This empathic element is first optionally used with the preverbal negator, until it becomes a compulsory element to express sentential negation in Stage II. When the newly emerged emphatic negator loses its emphatic function and starts functioning as a regular negator, it pushes away the preverbal negator and the language enters stage III (Willis et al. 2013: 1–50). It is due to the fact that the emphatic construction is overused that the negative construction gets neutralised over time and ultimately replaces the original preverbal negation (Detges & Waltereit 2002, Kiparsky& Condoravdi 2006). With respect to French this means that the strengthening of the preverbal ne is optional at first. However, at some point the generic noun pas is used as an emphasizer for negation (next to other emphasizers). When this noun grammaticalises as a new category and consistently starts co-occurring as a negative adverb with the preverbal negator, and until it behaves like a full negative adverb, it pushes aside the preverbal negator ne (Willis et al. 2013: 1–50). The analysis to be proposed below assumes a push-chain approach: it is due to the changing nature of pas that ne eventually becomes redundant in Stage III. Under the present proposal negation is conceived of as being featurally and structurally complex. In Stage IIb the new emphatic negator pas starts spelling out a part of the negative features responsible for the expression of sentential negation. The other part is spelled out by ne. As such, ne and pas complement each other for the expression of sentential negation in Stage IIb. In Stage III, when pas has grown structurally and spells out all negative features, ne is no longer necessary and becomes redundant.4 In the following sections I first elaborate on my classification of negative markers, then I explain the nanosyntactic framework I use, after which I embark upon my analysis of sentential negation in le bon usage French and modern French.

4 For more information with respect to potential triggers which lead from one stage to another I refer the reader to the extensive literature on the Jespersen Cycle and the triggers underlying grammaticalisation and change in French and other languages, such as Schwegler (1988: 45–46), Winters (1987: 27–52), Horn (1989: 452–462), Zeijlstra (2004: 47–57), Van der Auwera (2009), de Swart (2010), Willis et al. (2013), Breitbarth et al. 2013.

The nanosyntax of French negation: A diachronic perspective

3.

55

Classifying negative markers

On the basis of meaning distinctions, functional and morphological distinctions and stacking properties, De Clercq (2013) proposes that there is reason to classify negative markers in four different types. De Clercq’s (2013) classification starts from the observation that the abstract proposition in (6) is true for all sentences in (7) and (8).5 However, English uses several different negative morphemes to express this, as illustrated in (7), whereas other languages, like Czech (8) always use the exact same marker, irrespective of the scope the negative marker has to take. (6) ¬ P (x) (7) a. b. c. d. e. f.

Mary isn’t happy. Mary is not happy. Mary is unhappy. Mary is disloyal. Mary is non-American. Mary is un-American.

(8) a. Ja ne- jsem sˇt’astný. I neg- am happy. ‘I am not happy.’ b. Ja jsem ne- sˇt’astný. I am neg- happy. ‘I am unhappy.’ c. Je ne- loajální. Is neg- loyal ‘He is disloyal.’ d. Je ne- americký. Is neg- American ‘He is un-American.’ ‘He is non-American.’

The Czech negative marker is thus syncretic across different scope positions for negators, whereas this is not the case for English. Depending on where the negative marker takes scope in English, the morphological realisation differs. 5 Of course these negative markers take scope in different positions. This will be discussed at length in this paper. The fact that these negative markers have different scopal properties is not only visible in the surface position of these markers, but is also reflected in the scope these markers can or cannot have over adjuncts. The markers in (7a–b) can take scope over a clausal constituent following the adjective (see ia), whereas this is not the case for the negative markers in (7c–f). (i) a. Mary isn’t happy because John is here, but because his mother has left. b. *Mary is unhappy because John is here, but because his mother left. Differences in scope can be accounted for under the system proposed in this paper and follow from the different internal make-up of the negative markers. Nevertheless, I will not go into these particular data at this point.

56

Karen De Clercq

Crucially, the difference in scope position and/or morphological realisation can also coincide with meaning distinctions. More in particular, the negative markers in (7a,b,e) tend to give rise to contradictory negation, whereas the negative markers in (7c,d,f) often give rise to contrary negation. In what follows I briefly explain the distinction between contradictory and contrary negation. Two sentences are each other’s contradictories when the Law of Contradiction (LC) and the Law of the Excluded Middle (LEM) apply. The LC states that a proposition (p) cannot be true and false in the same circumstances. The Law of the Excluded Middle says that any proposition is either true or false. (9) Law of Contradiction (LC) ∼(p & ∼p) (Russell 1940: 259) (10) Law of the Excluded Middle (LEM) p ∨ ∼p (Russell 1940: 259)

When we apply this to a concrete example, then sentence (11a) and its negation (11b) cannot both be true, nor both be false at the same time. They respect the LEM and the LC and are thus each other’s contradictories. (11) a. Leila is married. b. Leila is not married.

In contrast, for a pair like (12a) and its negative (12b) the LC applies, i. e. they cannot both be true at the same time, but the LEM does not hold: (12a) and (12b) can both be false at the same time, as illustrated by (12c). When two propositions can be false together, they are each other’s contraries. (12) a. She is happy. b. She is unhappy. c. She is neither happy, nor unhappy.

However, if p is a set of possible situations, those in which p is true, and U is the Universe of possible situations of which p is a subset, then whether q is a contradiction or contrary of p is relative to U (Borschev et al. 2006: 6). Consequently, it is hard to pin down when negation gives rise to contradictory or contrary negation, because, as noted by Borschev et al. (2006), it is highly influenced by pragmatic factors. For the sake of the present paper I follow Jespersen (1917: 144), Zimmer (1964: 21–45) and Horn (1989: 273–86; 2005: 331–337) in assuming that low scope negators like un- or dis- give rise to contrariety more easily than the other negators. The deeper reason for why this is the case I leave to future research.6 6 I do not want to claim here that contrariety negation is inherent to the semantics of affixal un-, nor is contradictory negation inherent to the semantics of not. Both contrariety and con-

The nanosyntax of French negation: A diachronic perspective

57

The morphological distinctions also coincide with functional differences. Negative markers like English n’t, which scope over tensed predicates, predominantly have the function of denying a previous utterance. Horn (1989: 203) states that “the prototypic use […] of negation is indeed as a denial of a proposition previously asserted, or subscribed to, or held as plausible by, or at least mentioned by, someone relevant in the discourse context.” Other markers have a modifying function, as in (13a) or a contrastive function, as in (13b). (13) a. a not very happy man, not long ago b. John was not happy, but sad.

Still other markers like non- in English have a classifying function (Warren 1984: 101, Kjellmer 2005), illustrated in (14), whereas the function of markers like English un-, in, dis- is characterizing (Funk 1971, Kjellmer 2005), as shown in (15). (14) Use non-fat milk instead of whole milk. (Corpus npr/07.) (Kjellmer 2005: 162) (15) Some parents say children in Sarajevo have become increasingly disobedient and difficult to control during this wartime. (Corpus: npr/07.) (Kjellmer 2005: 162–163)

A negative marker is classifying when it creates a binary opposition. This is exactly the kind of opposition that sets introduce between members and non-members of the set. The negative marker in (14), for instance partitions the universe into things that contain fat and things that are non-fat. Negative markers like un-, dis, iN- on the other hand are characterizing, i. e. they denote some point on the scale. Semantically, they attach to adjectives that denote degree functions: functions from individuals to degrees (Kennedy 1999; Kennedy & McNally 2005). With respect to the example in (15), the negative marker dis- in disobedient does not trigger a classification of disobedient children in opposition to obedient children. It only wants to situate children with regard to their position on a scale of obedience. Furthermore, within one clause some negative markers can be stacked, as shown in (16a), whereas others cannot, as (16b,c) show. When one combines several negative markers in one clause, it is possible to combine them in a certain order, as in (16a), but not as in (16b,c). This may seem an arbitrary fact, but it is not, because these data support the different scope positions for negative markers and the importance of the associated morphological distinctions.

tradiction are possible with all possible negative markers, depending on the context. Nevertheless, it seems that affixal negation combines more easily with gradable predicates and therefore gives rise more easily to contrary negation. For discussion of this issue, I refer the reader to Horn (1989, chapter 5).

58

Karen De Clercq

(16) a. She isn’t NOT unhappy. b. *She not isn’t unhappy. c. *She un is not n’t happy.

By looking at data from nine different languages (Greek, English, French, Chinese, MS Arabic, Persian, Moroccan Arabic, Hungarian and Czech) in terms of the properties distinguished above, De Clercq (2013: 28–38) distinguishes four different types of negative markers, which she labels as Negative Polarity (PolNeg), Focus (FocNeg), Degree (DegNeg) and Quantity (QNeg) markers. The labels correspond to syntactic positions in the clausal spine where these negative markers take scope. Table 1 gives an overview of the different types of negative markers, their properties and the traditional labels that correspond to them.7 Table 1: Classification PolNeg

FocNeg

DegNeg

QNeg

Scope

Predicate denial Tensed predicate

Predicate negation Predicate

Term negation Predicate term

Stacking Semantics

Foc, Deg, Q Contradiction

Deg, Q Contradiction

Q Contradiction

Contrariety

Function

Denial

Contrast/modifying

Classifying

Characterizing

This classification deviates slightly from the more traditional three-way classification of negative markers into sentential negation (predicate denial), constituent negation (predicate negation) and lexical negation (Penka, to appear: 2) (predicate term negation), because it splits up lexical or predicate term negation in two different groups. Since four different types of negative markers can be distinguished on functional, semantic and morphological grounds in some languages like English, the Czech negative marker ne can be looked upon as a marker that is syncretic for these four different types. From a cross-linguistic perspective one can thus say that the morphology of negation in some languages (like English) supports the scopal, functional and semantic differences between negative markers, whereas other languages (like Czech for instance8) underscore the underlying similarity between these negative markers, i. e. the fact that all negative markers render the abstract proposition in (6) above, repeated here:

7 See Horn (1989) for a discussion of the terms predicate denial, predicate negation and predicate term negation in the logic, philosophical and linguistic history of negation. 8 The presence of syncretic negative markers does not prevent stacking of these markers though. Czech for instance can also stack its negative markers with the expected different interpretations for the negative markers.

59

The nanosyntax of French negation: A diachronic perspective

(6) ¬ P (x)

A cross-linguistic look at negative markers in nine languages reveals the pattern in Table 2 (De Clercq 2013: 67). Even though the sample of languages is not big, it is relatively varied and languages from four different language families are represented. Moreover, the relatively modest scope of the typological sample still allows ample room for generalisations and theoretical work (Baker & McCloskey 2007). Crucially, the negative markers in the languages under investigation can be ordered in such a way that syncretisms across the four different types of markers always target contiguous regions (or cells in the table). More concretely, a systematic pattern arises in which there are no ABA-patterns, i. e. syncretisms never skip a cell, or all syncretisms are contiguous. The sequence that arises when we order the negative markers in such a way is either PolNeg, FocNeg, DegNeg and QNeg or its mirror image. Importantly, not only can the syncretisms be ordered in such a way that no ABA-patterns arise across these nine languages, but the negative sequence which is based on the syncretisms also parallels the natural semantic scope order of negative markers, i. e. from wide to narrow scope or from narrow to wide scope. Table 2: Syncretism patterns

Greek English

PolNeg-marker dhen not

FocNeg-marker oxi not

DegNeg-marker mi non

QNeg-marker aun-

French Chinese

pas bù

pas bù

non fe¯i

iNfe¯i

MS Arabic Persian

laa na

laa na

Moroccan Arabic Hungarian Czech

ma (sˇi) nem ne-

musˇi nem ne

ghayrqueyrmusˇi

ghayrqueyrmusˇi

nem ne-

-tElEn ne-

Therefore, it seems that morphology follows or parallels semantic scope. Put differently, the relationship between these morphological exponents of negation is not arbitrary, but governed by syntactic principles. Since the syncretisms show a contiguity relationship between negative markers, there is reason to assume that these negative markers are structurally related and together constitute what can be thought of as negation. In the next section I explore how this pattern can be captured in nanosyntax, and how the framework can account for negation in English. In section 5 I then proceed to an analysis of French.

60

4.

Karen De Clercq

The Nanosyntax of negation

Nanosyntax (Starke 2009, 2011a,b, Caha 2009, Pantcheva 2009, 2011, Fábregas 2009, Taraldsen 2012) is a Late Insertion theory9 that finds its origins in the cartographic framework (Cinque 1999, 2010, Rizzi 1997, Kayne & Pollock 2001, Cinque & Rizzi 2010, Shlonsky 2010). The cartographic research programme is to provide a detailed structural map of natural language syntax (Cinque & Rizzi 2008: 42). Within nanosyntax, syncretisms are considered surface indications of hidden layers of syntactic structure within what is usually considered as one unit, the morpheme ( Jakobson 1962, Caha 2009), in this case a negative morpheme. From a nanosyntactic point of view the pattern detected within the domain of negative markers leads to splitting up negation into five different syntactic features and hence to the underlying idea that negation is featurally complex, an idea also present in Poletto (2008) and Haegeman & Lohndal (2010: 199), though applied there to negative indefinites.10 These four features can be represented by means of a hierarchical structure, consisting of four different heads which I label Q°, Deg°, Foc° and Pol°, and which come on top of a phonologically empty negative Neg° head. The nanospine that corresponds to the spell-out of the negative morpheme not is illustrated by the tree structure in (17).

9 In this respect Nanosyntax is like Distributed Morphology (Halle & Marantz 1993, Harley & Noyer 1999). I refer the reader to these references for an introduction to DM. I will only focus on Nanosyntax. 10 The fact of splitting up negation in different features or different projections is not new as such and is also present in Poletto (2008, this volume). She also splits up NegP in several projections and base-generates it low on the predicate. However, there are crucial differences between her work and mine, which I elaborate on in footnote 16 of the present paper. Zanuttini (1997) and Cormack & Smith (2002) also postulate different projections for negation in the clausal spine. However, under their approach these projections are structurally unrelated and it remains unexplained why negation surfaces in different positions. Moreover, their approaches focus on sentential negation (Zanuttini 1997) or on sentential negation and what I label as Foc-negation (Cormack & Smith 2002). Nevertheless, similar subdivisions as distinguished in my classification are present in their work.

The nanosyntax of French negation: A diachronic perspective

61

(17)

However, based purely on the syncretisms it is impossible to say whether Pol° or Q° are structurally highest. Theoretically the structure in (18) is also a possible representation of the inner complexity of the negative morpheme. However, I argue this is not the case and the structure in (17) is the correct representation. (18)

Supporting evidence for this claim can be given on the basis of synchronic and diachronic arguments and arguments of morphological containment. Support from synchronic arguments comes from scope and stacking itself. Negative markers like n’t or not can scope over structurally complex constituents potentially containing negative markers like non- or un-, which reflects the fact that the former must be structurally higher, as illustrated in (19). (19) He isn’t happy with her unconventional lifestyle.

A diachronic argument for the containment relationship in (19) is that un-, aand -iN are derived from Proto-Indo-European (PIE) *n-, which is a variant of *ne- (Harper 2013). Non- on the other hand is a univerbation of PIE *ne and the

62

Karen De Clercq

Latin word oinum, meaning ‘one’ (Horn 1989: 453). Not (Harper 2013) is also morphologically bigger than un-, iN- and a-. It is the unstressed variant of naught, which consists of PIE *ne and Old English (OE) wiht, which means ‘person, creature, thing’ (Horn 1989: 455, Harper 2013). Based on etymological data non- and not are clearly morphologically bigger than un-, iN- and a-. Assuming morphological complexity to reflect greater featural complexity, I conclude that the highest feature in the spine yields PolNeg-markers like not, and that the lower features are contained within PolP. With the order of the spine in place, it is still unclear where this nanospine should be generated. I propose it is generated in the specifier of a NegP in the clausal spine, whose Neg° head carries an uninterpretable negative feature [uNeg]. This uninterpretable feature gets checked off when the spine given in (17) above, with at the bottom a phonologically empty head, carrying [iNeg], is inserted.11 However, at this point it is not yet clear how this nanospine can actually lead to the spell-out of these negative markers, and how these syncretisms are derived in syntax. In what follows I explain which nanosyntactic tools help to derive the syncretisms in syntax. I explain the nanosyntactic system by applying it to English before I apply the system to the case under discussion, i. e. French ne… pas in le bon usage French and pas in colloquial French. As mentioned before, nanosyntax is a Late Insertion theory, which means that lexical items are only inserted after Merge creates syntactic structure with morphosyntactic features. In nanosyntax the building blocks that Merge operates on are very small, even submorphemic. It is precisely due to these submorphemic building blocks that spell-out is phrasal.12 It is “only after some steps of derivation that a constituent large enough to correspond to a morpheme is created” (Starke 2011a: 4). That morpheme is then a phrasal constituent. After each Merge step, spell-out is mandatory, i. e. there is Cyclic Full Interpretation.13 In order to spell out these phrasal constituents nanosyntax makes use of a post-syntactic lexicon: each lexical item (LI) consists of information related to the item’s phonology, a lexical tree, and conceptual information, as illustrated in (20). (20) [phonology, lexical tree, conceptual information]

11 For other approaches making use of phonologically empty but semantically interpretable operators, see amongst others Zeijlstra (2004, 2009), Penka (2007, 2011). 12 For non-nanosyntactic implementations of phrasal spell-out, see McCawley (1968), Weerman & Evers-Vermeul (2002) and Neeleman & Szendroi (2007). 13 The spell-out domain in nanosyntax is thus very small, even smaller than in approaches which are considered to take a rigidly cyclic approach to spell out, like Epstein & Seely (2002), Bosˇkovic´ (2002, 2005) and Mueller (2004), who argue that spell-out domains or phases are not bigger than one phrase, and unlike Chomsky (1995, 2001) who considers vP and CP spellout domains or phases.

The nanosyntax of French negation: A diachronic perspective

63

When the lexical tree matches the tree in syntax, spell-out is possible. However, when there is no lexical tree that corresponds to the structure in syntax, movement is allowed as a Last Resort. At each node, there is Cyclic Override, which means that the new spell-out overrides the previous spell-out. Two other principles that manage the insertion of lexical items are the Superset Principle, in (21), and the Elsewhere Condition, in (22): (21) Superset Principle A lexically stored tree matches a syntactic node iff the lexically stored tree contains the syntactic node. (Starke 2009: 3) (22) Elsewhere Condition In case two rules, R1 and R2, can apply in an environment E, R1 takes precedence over R2 if it applies in a proper subset of environments compared to R2. (Caha 2009: 18, based on Kiparsky 1973)

For more theoretical discussion and background with respect to nanosyntax I refer to Starke (2009, 2011a, 2011b), Caha (2009), Pantcheva (2009, 2011), Fábregas (2009), Taraldsen (2012), De Clercq (2013), and Rocquet (2013). In what follows I will illustrate how this works for English cases of negation. The lexicon of English contains the following lexical items for negation: (23) LI 1

(24) LI 2

(25) LI 3

64

Karen De Clercq

In order to derive the English PolNeg–marker not, first Neg° is merged. The lexicon is checked at NegP. The LI in (23) matches the structure in syntax by the Superset principle. There are many more items containing NegP, but due to the Elsewhere Principle LI 1 is inserted and spelled out as un-. Q° is merged and the lexicon is checked at QP. (23) matches the structure in syntax. QP can be spelled out as un- and overrides the previous spell-out, which happens to be the same. Deg° is merged, the lexicon is checked again at DegP. (24) is a matching item: DegP can be spelled out as non, overriding un-. Foc° is merged and the lexicon is checked at FocP. Due to the Superset Principle (25) is a candidate. Not spells out FocP and overrides non. Finally, Pol° is merged. (25) can be inserted and PolP is spelled out as not. This negative nanospine is generated in the specifier of a NegP in the clausal spine (Pollock 1989, Haegeman 1995). However, unlike most NegPs in the literature, this NegP is merged low on the predicate. The head of this clausal NegP functions as a probe. I assume that it carries an uninterpretable [uNeg] that gets checked off when the negative spine is generated in its specifier,14 and that the features in the nanospine all carry interpretable features that correspond to the heads of the nanospine. The structure as derived up until now is represented in (26).

14 An issue which I leave unresolved for now is how [uNeg] on Neg° in the clausal spine Agrees with [iNeg] in the nanospine. There are two options to resolve this. One possibility is that the negative spine is first merged lower in the structure and is attracted to SpecNegP. This would be a standard case of Agree. The derivation obtained in this way would however involve many more movements. In order to be able to spell out, AP would always have to move to the specifier of the newly merged negative layer. Since movement is less economic than Stay, it doesn’t seem the best option. Moreover, negative morphemes cross-linguistically tend to precede the predicate. Consequently, it seems counterintuitive to end up with a structure that has AP at the top and the nanospine below it, even though new movements could be postulated to end up with the right word order. Another option is that the base-generation of the nanospine in SpecNegP is comparable to basegenerating a subject in SpecvP. The Agree mechanism in this case is comparable to what Rizzi (1996: 76) calls Dynamic Agreement: the uninterpretable feature on the head can be checked off by an interpretable feature in the specifier.

The nanosyntax of French negation: A diachronic perspective

65

(26)

The underlying assumption here is that all negation starts out as a predicate negator (see also Horn 1989: ch. 7, Collins & Postal 2014: 23–27, Poletto 2008). The main argument for this claim comes from the syncretisms. If the syncretisms point to the fact that these negative features are structurally connected, then the features that give rise to QNeg-markers and the features that give rise to PolNeg-markers should be merged together in one position in order to capture this relation. Given that QNeg-markers, the markers which take lowest scope in all languages under scrutiny, tend to appear in a position very close to the predicate, often even attached to the predicate, as illustrated by the pair in (27), I argue that all features are merged in a position immediately above the predicate, i. e. as predicate negators.15 (27) a. She is very unhappy. b. *She is un very happy.

In spite of the fact that the features are merged together in a low position, they take scope in different positions in the clausal spine.16 The clausal positions have 15 The merge position of these negative features is therefore mostly not the position in which they eventually take scope or where they appear. Movement will ensure that the negative markers end up in their expected surface and/or scope position. 16 There are parallels between my account and Zanuttini’s (1997), who also has four clausal positions for negation; also with Poletto’s (2008, this volume), who advocates a low complex NegP which is attracted to Zanuttini’s projections in the clausal spine. However, there are also many differences. One crucial distinction is that Poletto’s (2008) system does not provide a mechanism to distinguish between, for instance, French ne and Italian non, two negative markers which can be considered as PolNeg-markers. Under her account these negative markers would be attracted to the same projection, her ScalarP. Further differences, like the fact that non can occur on its own, whereas ne needs to co-occur with pas, would have to be lexical differences. Under the present account the combination of multiple projections

66

Karen De Clercq

the same labels as the phrases in the nanospine: QuantifierP (Corver 1997), DegreeP (Corver 1997), low FocusP (Belletti 2001a, 2004, Jayaseelan 2001, 2008) and PolarityP (Laka 1994, Cormack & Smith 2002, Poletto & Zanuttini 2013). Under the present account this is explained: the markers are syntactically different. The clausal positions function as probes and carry uninterpretable features: [uPol], [uFoc], [uDeg], [uQ]. In order to derive Polnegation for a sentence as in (28), the nanospine, derived in a way already described above, is inserted into SpecNegP with its five features. (28) She is not happy.

Due to the structure of the lexicon, the presence of all features leads to insertion of not, as detailed above. Since happy is a gradable predicate17 a QP is projected in the clausal spine. [uQ] on the clausal Q° probes and gets deleted by the interpretable feature [iQ] feature in the nanospine. When Deg° is merged in the main spine, [uDeg] on Deg° probes and agrees with [iDeg] in the nanospine. Low FocP is projected18 and [uFoc] probes and agrees with [iFoc] in the nanospine. There is an EPP on Foc° in English, which attracts the entire negative spine to SpecFocP. When Pol° is merged, [uPol] probes and finds [iPol] in the nanospine, so its uninterpretable features are checked and the widest scope of negation is established. The structure for this derivation is in (29).

and a postsyntactic lexicon allows negative markers to be classified according to their scope position, whereas the postsyntactic lexicon allows for flexibility with respect to the strength of an individual negative marker. The difference between French ne and Italian non, both PolNegmarkers and thus high in the nanospine, would be reflected in the size of the lexical trees in the lexicon. 17 In the absence of a gradable predicate, I assume that QP will not be projected. Given that the features in the nanospine are interpretable this is not a problem. 18 The use of SpecFocP for negative scope markers makes predictions with respect to the information structural properties of negative clauses and the fact that low FocP has been argued to host postverbal new information subjects in Italian and the subject of certain clefts in French (Belletti 2001a, 2004, 2008, 2009: 242–265).

The nanosyntax of French negation: A diachronic perspective

67

(29)

In this way the scope of the semantically interpretable predicate negation in Neg° is extended via the scope markers in its specifier. The internal constellation of the negative features in the specifier correlates with the scope they eventually take. Negation is only interpretable once for the semantics as negation, but is featurally complex. From the present perspective the fact that negation can surface in different positions in the clausal spine, as already described in cartographic proposals by Zanuttini (1997) and Cormack & Smith (2002), can be explained from the present perspective: it follows from the fact that these negative markers are syntactically different.

68

5.

Karen De Clercq

French

In this section I first provide a nanosyntactic analysis for le bon usage French or Stage II of the Jespersen Cycle and thus an answer to the question why in formal written French negation needs two components to express sentential negation. Second, I show how the same nanosyntactic system can also account for why in present-day colloquial French, i. e. Stage III of the Jespersen cycle, this ne is no longer necessary. Third, I explain how the direction of the change such as it manifests itself the Jespersen cycle can be viewed from the perspective of the present proposal. Finally, I briefly consider negative arguments, a topic hitherto untouched.

5.1

Le bon usage French

I propose that in le bon usage French ne and pas are each responsible for the spell-out of different parts of the negative spine. The lexical tree of ne is very small and only spells out the Pol feature. Since ne spells out a feature of the nanospine that is not spelled out by any other negative marker, it cannot be made redundant.19 Put differently, ne has become structurally deficient (Cardinaletti & Starke 1999).20 The lexicon of le bon usage French contains the following lexical items for negative markers. (30) a. b. c. d.

LI1 LI2 LI3 LI4

[Deg° [Q° [ Neg°]]], neg> [Foc° [Deg° [Q° [Neg°]]]], neg> [Pol°], neg>

(31) Il n’ est pas heureux. he neg is neg happy ‘He is not happy.’

To derive the negation in a sentence like (31), Neg° is merged in syntax and the lexicon is checked. There is no LI that matches the syntactic structure. However, by the Superset Principle and the Elsewhere Condition, (30a) can spell out Neg° as iN-. Q° is merged and at QP the lexicon is checked again. (30a) corresponds to the structure in syntax and as such the lexical item iN- is inserted. Deg° is merged in syntax and again the lexicon is checked at DegP. There is an LI, (30b), which 19 I assume that in Old French or Stage I of the Jespersen Cycle ne spelled out all layers of the negative spine. 20 This term can be traced back to Cardinaletti & Starke (1999) and their related concept of Minimize Structure (1999: 47). However, under their approach to French pronouns structurally deficient items are preferred until they are independently ruled out. This is not the case in the analysis proposed here.

The nanosyntax of French negation: A diachronic perspective

69

matches the structure in syntax. Non is inserted in DegP and overrides the spellout of QP. FocP is merged and again the lexicon is checked. (30c) matches the syntactic structure and so it gets spelled out as pas. When the final negative layer PolP is merged, the lexicon is checked in the usual way. However, at this point there is no lexical item in the lexicon that corresponds to the structure in syntax, represented by the structure in (32). (32)

So at this point there is no spell-out for the structure in (32). There is a lexical item though that spells out PolP alone, namely (30d).21 When Merge does not lead to spell-out, phrasal movement can be called upon in order for the structure to be spelled out. The complement of Pol° thus moves to SpecPolP, allowing the spellout of the newly merged feature. Upon consultation of the lexicon, it is clear that there is a LI (30d), which contains a syntactic tree that can spell out PolP. The feature is spelled out as ne. The resulting structure we get then is in (33).

21 I assume that it is due to the deficiency of ne that ne can also be used as an expletive negative marker in le bon usage French. Rowlett (1998: 28) mentions the following examples of expletive ne: (i) a. Marie est plus grande que n’ est son frère. Marie is more tall than neg is her brother ‘Marie is taller than her brother is.’ b. Elle a peur que tu ne sois là. She has fear that you neg be-subj there ‘She’s worried you might be there.’ c. Je doute qu’ il ne soit là. I doubt that he neg be-subj there ‘I doubt he’s there.’ Since the LI for ne only consists of the feature [iPol], I propose it can be immediately merged in its scope position in the clausal spine as well. When it is merged independently from NegP it loses its semantic negativity. Italian non can also be used as an expletive negative maker (Belletti 2001b), but cannot be said to be deficient in the same way as French ne.

70

Karen De Clercq

(33)

Finally, the layers QP and DegP of the negative spine check off [uQ] and [uDeg] on the Q- and Deg-probe in the clausal spine, as shown in the structure in (34) below. When FocP is merged in a position dominating vP, [uFoc] on the Foc°probe gets checked and deleted via Agree with [iFoc] in the nanospine.22 The subspine containing FocP is extracted from its base-position and moves to SpecFocP (due to the presence of an [EPP] feature on Foc°).23 Pol° is merged and probes down and Agrees with [iPol] on Pol°. Due to an [EPP] feature on Pol° the constituent spelling out ne gets attracted and moves to SpecPolP. I assume that the subject moves to SpecFinP (Cardinaletti & Roberts 2002, Haegeman 2012), a position higher than PolP. The structure is illustrated in (34).

22 Rowlett (1998) also proposes that pas starts out low. It starts out as an adjunct to VP in modern spoken French. It then needs to move to SpecNegP in order to give rise to sentential negation and to make ne negative via Dynamic Agreement. If it doesn’t move to SpecNegP it expresses constituent negation. 23 The use of SpecFocP for negative scope markers makes predictions with respect to the information structural properties of negative clauses: it suggests that in negative clauses low FocP is not available for other constituents. However, Belletti (2008, 2009: 242–265) proposes that the subject of clefts in French moves to SpecFocP, in order to give rise to new information focus. Therefore, I assume that in negative subject clefts the subject of the cleft remains within the embedded CP.

The nanosyntax of French negation: A diachronic perspective

71

(34)

Summarizing, under the present approach two elements which express sentential negation together can both be considered negative and still need each other to express sentential negation due to the fact that they form two parts of a negative spine, which is in its entirety responsible for the expression of sentential negation. That the PolNeg- marker ne cannot express sentential negation on its own is due to the fact that it has become structurally deficient: it spells out only the top layer of the negative spine, PolP. The lower negative features need to be spelled out by another lexical item.

72 5.2

Karen De Clercq

Colloquial French

In present day spoken French pas has become the real sentence negator. Within the present system I account for this change by proposing that the lexical tree for pas has grown in size and has thus become a PolNeg -marker. The evolution of pas is visible in the lexical items: pas spells out five layers now. As a consequence, ne, still part of the lexicon, becomes redundant and leaves the negative spine. The account proposed here is thus exactly compatible with a push-chain approach towards Jespersen’s Cycle. (35) a. b. c. d.

LI1 LI2 LI3 LI4

[Deg° [Q° [ Neg°]]], neg> [Pol° [Foc° [Deg° [Q° [Neg°]]]]], neg> [Pol°], [neg]>

The derivation for the nanospine in a sentence like (36) goes in the same way as described for the structure in (34) until the final layer Pol° is merged. (36) Je suis pas heureux.

When the final negative layer Pol° is merged, the lexicon is checked in the usual way. The LI responsible for spelling out Foc°, i. e. (35c), matches with the syntactic structure and leads to spell-out. As such, the lexical item for ne, (35d), which also consists of Pol°, becomes redundant. In order to spell out ne the computation would have to take recourse to phrasal movement. Given that a more economic option, namely no movement, is possible, the LI for ne is banned from the negative spine. Once the split NegP is fully merged and spelled out, the derivation proceeds in the same way as described above. The Q- and Deg-probe in the clausal spine Agree with the corresponding features in the nanospine in the manner described in the previous subsection. When FocP is merged in a position dominating vP, Foc° probes and Agrees with [iFoc] in the nanospine to have its uninterpretable features deleted. The entire negative nanospine, corresponding to pas moves to SpecFocP (due to the presence of an [EPP] feature on Foc°). Clausal Pol° Agrees with [iPol] and checks off its uninterpretable feature. There is no [EPP] feature on Pol° and the negative marker stays in SpecFocP. The derivation for sentential negation in colloquial French is in (37).

The nanosyntax of French negation: A diachronic perspective

73

(37)

Due to the fact that the LI for pas has grown in size, presumably triggered by the fact that the emphatic negator becomes a neutral negator due to overuse (Kiparsky & Condoravdi 2006), the LI for ne becomes redundant and is no longer used for the spell-out of the negative spine. As such, it loses its function as a negative marker in colloquial French, and we have arrived in Stage III of Jespersen’s Cycle.

74

Karen De Clercq

As a final note I will say something about the intermediate stages of Jespersen’s Cycle, those in which one of the other negative marker is optional. I want to argue that the optionality can be understood as a consequence of co-existing grammars or diglossia (Lightfoot 1999: 92). In Stage IIa (Willis 2011: 94), when pas is optionally present, the grammar of Stage I, in which the LI of ne was still able to spell out the entire negative spine, co-exists next to the new grammar of Stage IIb. Similarly, when the language undergoes a change from Stage IIb to Stage III, there is a period in which the grammars as described above co-occur. At this stage, both grammars are accessible to all speakers and are mutually intelligible.

5.3

A note on diachronic change

Grammatical reanalysis is generally thought of as involving upward change along the functional hierarchy, i. e. from a lexical item to a functional item, and is generally thought of as leading to structural or featural simplification (cf. Roberts & Roussou 2003, van Gelderen 2004). The Jespersen cycle works similarly: new negative elements start out lower in the functional hierarchy, as predicate negators, and they slowly climb up along the spine, until they reach the propositional level (Chatzopolou 2013). However, the present system offers a novel way of looking at diachronic change, and more in particular the Jespersen cycle, in that upward change does not only involve structural simplification (pace Roberts & Roussou 2003). Rather it shows that diachronic change involves both growing complexity and simplification of lexical items at the same time. For example, the introduction of pas as a negator in French in stage II of the Jespersen Cycle leads to a competition with ne, which in turn leads to a loss of features in the lexical item for ne, i. e. the tree it spells out becomes smaller. This goes hand in hand with the tree of pas becoming bigger, however. As the tree of pas becomes progressively bigger, ne may even completely disappear from the system. The process of grammaticalisation in general could be viewed as one involving the loss of grammatical features in certain lexical items, with a concomitant gain of these same features in different lexical items. This is because the features in themselves are not lost, but there is a redistribution in the way they are expressed. One might say that there is a law of the Conservation of Features, which ensures that in the grammatical system as a whole, features are not lost. While feature loss may occur at the level of individual lexical items, it needs to be compensated for in other lexical items, which will come to express the features lost in the former. Needless to say, this is an issue with ramifications that go well beyond the scope of the present paper. I must therefore leave it as a matter for further research.

The nanosyntax of French negation: A diachronic perspective

5.4

75

A note on negative arguments

An important issue that was not addressed in this paper, and that I will not elaborate on here, is how negative arguments like rien ‘nothing’, personne ‘noone’, aucun ‘no’, jamais ‘never’, etc. are treated within the present account. Essentially, I think a similar approach as taken for regular negative markers is possible, provided it is taken into account that these negative arguments also spell out features like [thing] (rien), [person] (personne) and [time] ( jamais). I assume that these negative arguments incorporate in FocP of the negative spine, as new emphatic negators, as soon as they become compulsory elements to express sentential negation. As long as they still need ne they consist of only four negative layers, i. e. NegP, QP, DegP, FocP. As soon as they can spell out sentential negation on their own, they consist of all five layers. Evidence for this approach comes from the partial syncretisms between the negative marker not and negative indefinites like no, nothing, never and nobody in English, and between Czech ne and the indefinite nikdo ‘no one’. More cross-linguistic comparative research is needed to see whether this approach can indeed be maintained.

5.5

Summary

The analysis showed how the diachronic evolution of negative markers, known as Jespersen’s Cycle, can be captured well by a nanosyntactic system. Whereas other theoretical approaches struggle to account for the obligatory presence of ne in Stage IIb (Willis 2011), a nanosyntactic approach manages to explain the obligatory nature of the preverbal negator. The fact that two markers together express sentential negation in le bon usage French is reflected in the lexical entries of the negators. The LI for the old preverbal negative marker has become ‘structurally deficient’: it only spells out one negative feature, Pol°, and cannot spell out the features below Pol°.24 However, since it is the only negative marker in the lexicon that can spell out Pol°, it still is inserted when the syntax merges Pol in the nanospine. As a consequence, both the PolNeg- and FocNeg-marker combine to spell out sentential negation together. Spoken or colloquial French on the other hand has entered stage III of Jespersen’s Cycle. The postverbal negative marker has become the new negative marker, because the size of the lexical item pas has grown in structure. It now spells out all negative features, including Pol°. As a consequence, when PolP is merged in the nanospine, pas is inserted, because inserting pas does not require movement and is thus the most economical option. As such, ne becomes redundant as a negative scope marker in the 24 See Breitbarth (2014) for a different proposal.

76

Karen De Clercq

spine and leaves the spine. Ne remains present in the language as an expletive marker of negation, losing its real negative meaning.

6.

Conclusion

This case study of French negation showed that the nanosyntactic system is well equipped to account for diachronic change and more in particular for Jespersen’s Cycle. The change from bipartite negation in Stage IIb of Jespersen’s Cycle to the use of a sole postverbal negative marker in Stage III boils down to a change in the structural size of the lexical items. The account supports a push-chain approach: changes in the emphatic postverbal negator make the old preverbal negator redundant. The preverbal negative marker in Stage IIb is structurally deficient and cannot spell out the lower negative features in the spine anymore due to the emergence of the new emphatic negator pas that now spells out these lower features. Consequently, in this stage sentential negation can only be expressed by means of two items, ne and pas, that complement each other. Stage III is characterized by the fact that the new negator pas has grown in size and spells out all negative features, pushing away the preverbal negator. Finally, the proposal developed here sheds a new light on the relationship between markers for sentential negation and constituent or lexical negation. The pattern of syncretism uncovered within the realm of negative markers shows a structural relationship between negative markers giving rise to sentential negation and those giving rise to constituent and lexical negation.

References Baker, Mark & Jim McCloskey (2007): On the relationship of typology to theoretical syntax. Linguistic Typology 11: 285–296. Belletti, Adriana (2001a): Inversion as focalization. In Hulk, Aafke & Jean-Yves Pollock (eds.): Subject inversion in Romance and the theory of universal grammar. New York: Oxford University Press, 60–90. Belletti, Adriana (2001b): Expletive negation in Italian comparative clauses. In Cinque, Guglielmo & Giampaolo Salvi (eds.): Current studies in Italian syntax. Essays offered to Lorenzo Renzi. Amsterdam: Elsevier, 19–37. Belletti, Adriana (2004): Aspects of the low IP area. In Rizzi, Luigi (ed.): The Structure of IP and CP: The cartography of syntactic structures, Vol. 2. Oxford: Oxford University Press, 16–51. Belletti, Adriana (2009): Structures and strategies. London: Routledge. Belletti, Adriana (2008): The CP of clefts. Rivista di Grammatica Generativa 33: 191–204.

The nanosyntax of French negation: A diachronic perspective

77

Borschev, Vladimir, Elena Paducheva, Barbara Partee, Yakov Testelets & Igor Yanovich (2006): Sentential and constituent Negation in Russian BE-sentences revisited. In Filip, Hana, Steven Franks, James Lavine & Mila TassevaKurktchieva (eds.): Formal approaches to Slavic Linguistics: The Princeton Meeting 2005 (FASL 14). Ann Arbor: Michigan Slavic Publications, 50–65. Bosˇkovic´, Zˇeljko (2002): A-movement and the EPP. Syntax 5: 167–218. Bosˇkovic´, Zˇeljko (2005): On the locality of move and agree: eliminating the activation condition, generalised EPP, inverse case filter, and phase impenetrability condition. In Herdan, Simona & Miguel Rodríguez-Mondoñedo (eds.): University of Connecticut occasional papers 3, Connecticut: University of Connecticut Linguistics Club. Breitbarth, Anne (2009): A hybrid approach to Jespersen’s Cycle in West Germanic. Journal of Comparative Germanic Linguistics 12: 81–114. Breitbarth, Anne & Liliane Haegeman (2010): Continuity is change: The long tail of Jespersen’s Cycle in Flemish. In Breitbarth, Anne, Christopher Lucas, Sheila Watts & David Willis (eds.): Continuity and change in grammar. Amsterdam: Benjamins, 61–75. Breitbarth, Anne, Christopher Lucas & David Willis (2013): Incipient Jespersen’s Cycle: The (non-)grammaticalization of new negative markers. In Fleischer, Jürg & Horst Simon (eds.): Comparing Diachronies/ Sprachwandelvergleich. Tübingen: Niemeyer, 143–164. Breitbarth, Anne (2014): The history of Low German negation. Oxford: Oxford University Press. Caha, Pavel (2009): The nanosyntax of case. PhD thesis, University of Tromsø. Cardinaletti, Anna & Ian Roberts (2002): Clause structure and X-second. In Cinque, Guglielmo (ed.): Functional structure in DP and IP. Oxford: Oxford University Press, 123–166. Cardinaletti, Anna & Michal Starke (1999): The typology of structural deficiency: a case study of the three classes of pronouns. In van Riemsdijk, Henk (ed.): Clitics in the languages of Europe. New York: de Gruyter, 145–233. Chatzopolou, Katerina (2013): Re(de)fining Jespersen’s Cycle. University of Pennsylvania Working Papers in Linguistics 19 (1), 29–40. Chomsky, Noam (1995): The minimalist program. Cambridge, Mass.: MIT Press. Chomsky, Noam (2001): Derivation by phase. In Kenstowicz, Michael (ed.): Ken Hale: a life in language. Cambridge, Mass.: MIT Press, 1–52. Cinque, Guglielmo (1999): Adverbs and functional heads: a cross-linguistic perspective. Oxford: Oxford University Press. Cinque, Guglielmo (2010): The syntax of adjectives: a comparative study. Cambridge, Mass.: MIT Press. Cinque, Guglielmo & Luigi Rizzi (2008): The cartography of syntactic structures. Studies in Linguistics (CISCL Working Papers) 2: 42–58. Cinque, Guglielmo & Luigi Rizzi (2010): The cartography of syntactic structures. In Bernd, Heine and Heiko Narrog (eds.): The Oxford handbook of linguistic analysis. Oxford: Oxford University Press, 51–65. Collins, Chris & Paul Postal (2014): Classical NEG raising. Cambridge, Mass.: MIT Press.

78

Karen De Clercq

Cormack, Annabel & Neil Smith (2002): Modals and negation in English. In Barbiers, Sjef, Frits Beukema & William van der Wurf (eds.): Modality and its interaction with the verbal system. Amsterdam: Benjamins, 133–163. Corver, Norbert (1997): Much-support as a last resort. Linguistic Inquiry 28: 119–164. Dahl, Östen (1979): Typology of sentence negation. Linguistics 17: 79–106. De Clercq, Karen (2013): A unified syntax of negation. PhD thesis, Ghent University. Detges, Ulrich & Richard Waltereit (2002): Grammaticalization vs. Reanalysis: a semantic-pragmatic account of functional change in grammar. Zeitschrift für Sprachwissenschaft 21 (2): 151–195. De Swart, Henriette (2010): Expression and interpretation of negation: an OT typology. Dordrecht: Springer. Fábregas, Antonio (2009): An argument for phrasal spell-out: indefinites and interrogatives in Spanish. Nordlyd 36: 129–168. Funk, Wolf-Peter (1971): Adjectives with Negative Prefixes in Modern English and the Problem of Synonymy. Zeitschrift für Anglistik und Amerikanistik 19: 364–386. Gelderen, Elly van (2004): Economy, innovation and prescriptivism. Journal of Comparative Germanic Linguistics 7: 59–98. Grevisse, Maurice &Andre Goosse ([1936] 1993): Le bon usage. Paris: Duculot. Haegeman, Liliane (1995): The Syntax of Negation. Cambridge: Cambridge University Press. Haegeman, Liliane (2012): Adverbial clauses, main clause phenomena, and composition of the left periphery. Oxford: Oxford University Press. Haegeman, Liliane & Terje Lohndal (2010): Negative concord and multiple agree: A case study of West Flemish. Linguistic Inquiry 41: 181–211. Halle, Morris & Alec Marantz (1993): Distributed morphology and the pieces of inflection. In Hale, Kenneth & Samuel Jay Keyser (eds.): The view from building 20. Cambridge, Mass.: MIT Press, 111–176. Hansen, Maj-Britt Mosegaard (2013): Negation in the history of French. In Willis, David, Christopher Lucas & Anne Breitbarth (eds.): The History of Negation in the Languages of Europe and the Mediterranean. Oxford: Oxford University Press, 51–76. Harley, Heidi & Rolf Noyer (1999): State-of-the-article: distributed morphology. GLOT 4: 3–9. Harper, Douglas (2013): Online etymology dictionary. . Horn, Laurence (1989): A natural history of negation. Chicago: University of Chicago Press. Horn, Laurence (2005): An un-paper for the un-syntactician. In Mufwene, Salikoko et al. (eds): Polymorphous Linguistics: Jim McCawley’s Legacy. Cambridge, Mass.: MIT Press, 329–365. Jakobson, Roman (1962): Beitrag zur allgemeinen Kasuslehre: Gesamtbedeutungen der Russischen Kasus. In Jakobson, Roman: Selected writings. Vol. 2. The Hague: Mouton de Gruyter, 23–71. Jayaseelan, Karattuparambil (2001): IP-internal topic and focus phrases. Studia Linguistica 55: 39–75. Jayaseelan, Karattuparambil (2008): Topic, focus and adverb positions in clause structure. Nanzan Linguistics 4: 43–68. Jespersen, Otto (1917): Negation in English and other languages. København: A. F. Høst & Søn.

The nanosyntax of French negation: A diachronic perspective

79

Jespersen, Otto (1924): The philosophy of grammar. London: Allen and Unwin. Kayne, Richard (1982): Predicates and arguments, nouns and verbs. In GLOW newsletter 8, 24. Kayne, Richard & Jean-Yves Pollock (2001): New thoughts on stylistic inversion. In Hulk, Aafke & Jean-Yves Pollock (eds.): Subject inversion in Romance and the theory of universal grammar. Oxford–New York: Oxford University Press, 107–162. Kennedy, Christopher (1999): Projecting the adjective: The syntax and semantics of gradability and comparison. New York: Garland. Kennedy, Christopher & Louise McNally (2005): Scale structure, degree modification, and the semantics of gradable predicates. Language 81: 345–381. Kiparsky, Paul (1973): ‘Elsewhere’ in phonology. In Kiparsky, Paul & Steven Anderson (eds.): A Festschrift for Morris Halle. New York: Holt, Rinehart and Winston, 93–106. Kiparsky, Paul & Condoravdi, Cleo (2006): Tracking Jespersen’s Cycle. In Janse, Mark, Brian D. Joseph & Angela Ralli (eds.): Proceedings of the 2nd international conference of Modern Greek dialects and linguistic theory. Doukas: Mytilene, 172–197. Kjellmer, Göran (2005): Negated adjectives in modern English. Studia Neophilologica 77: 156–170. Laka, Itziar (1994): On the syntax of negation. New York: Garland Publishing. Lightfoot, David (1999): The development of language: Acquisition, change and evolution. Oxford: Blackwell. McCawley, James D. (1968): Lexical insertion in a grammar without deep structure. In Darden, Bill J. et al. (eds.): Papers from the fourth Regional Meeting of the Chicago Linguistic Society. Chicago: University of Chicago Linguistic Department, 71–80. Meillet, Antoine (1912): L’évolution des formes grammaticales. Scientia 12: 384–400. Müller, Gereon (2004): Phase impenetrability and wh-intervention. In Stepanov, Arthur, Gisbert Fanselow & Ralf Vogels (eds.): Minimality effects in syntax. Berlin: Mouton de Gruyter, 289–325. Neeleman, Ad & Kriszta Szendroi (2007): Radical pro drop and the morphology of pronouns. Linguistic Inquiry 38: 618–714. Pantcheva, Marina (2009): Directional expressions cross-linguistically. Nanosyntax and lexicalization. Nordlyd 36: 7–39. Pantcheva, Marina (2011): Decomposing path. The nanosyntax of directional expressions. PhD thesis, University of Tromsø. Penka, Doris (2007): Negative indefinites. PhD thesis, University of Tübingen. Penka, Doris (2011): Negative indefinites. Oxford: Oxford University Press. Penka, Doris (to appear): Negation and Polarity. In Riemer, Nick (ed.): Routledge Handbook of Semantics. London: Routledge. Poletto, Cecilia (2008): On negative doubling. Ms. University of Venice. Poletto, Cecilia & Raffaella Zanuttini (2013): Emphasis as reduplication: evidence from sì che/no che sentences. Lingua 128: 124–141. Pollock, Jean-Yves (1989): Verb Movement, Universal Grammar and the Structure of IP. Linguistic Inquiry 20: 365–424. Posner, Rebecca (1985): Post-verbal negation in non-standard French: A historical and comparative view. Romance Philology 39: 170–197.

80

Karen De Clercq

Rizzi, Luigi (1996): Residual verb second and the wh-criterion. In Belletti, Adriana & Luigi Rizzi (eds.): Parameters and functional heads. Oxford–New York: Oxford University Press, 63–90. Rizzi, Luigi (1997): The fine structure of the left periphery. In Haegeman, Liliane (ed.): Elements of Grammar. Dordrecht: Kluwer, 281–337. Roberts, Ian & Anna Roussou (2003): Syntactic change. A minimalist approach to grammaticalization. Cambridge: Cambridge University Press. Roberts, Ian (2007): Diachronic syntax. Oxford: Oxford University Press. Rocquet, Amélie (2013): The nanosyntax of agreement. PhD thesis, Ghent University. Rooryck, Johan (2010): A compositional analysis of French negation. Ms., Leiden University. Rowlett, Paul (1998): Sentential negation in French. Oxford: Oxford University Press. Russell, Bertrand (1940): An inquiry into meaning and truth. New York: Routledge. Shlonsky, Ur (2010): The cartographic enterprise in syntax. Language and Linguistics Compass 4: 417–429. Schwegler, Arnim (1988): Word-order changes in predicate negation strategies in romance languages. Diachronica 5: 21–58. Starke, Michal (2009): Nanosyntax: a short primer to a new approach to language. Nordlyd 36 (1): 1–6. Starke, Michal (2011a): Nanosyntax. Lecture series presented at GIST, Ghent University, November 2011. Starke, Michal (2011b): Towards elegant parameters: language variation reduces to the size of lexically stored trees. Ms., Tromsø University. Taraldsen, Tarald (2012): *ABA and the representation of features in syntax. Talk presented at the 7th Brussels Conference of Generative Linguistics, 16–18 December 2012. Van der Auwera, Johan (2009): The Jespersen cycles. In Gelderen, Elly van (ed.): Cyclical change. Amsterdam: Benjamins, 35–71. Warren, Beatrice (1984): Classifying adjectives. Gothenburg: Acta Universitatis Gothoburgensis. Weerman, Fred & Jacqueline Evers-Vermeul (2002): Pronouns and case. Lingua 112: 301–338. Willis, David, Christopher Lucas & Anne Breitbarth (2013): The history of negation in the languages of Europe and the Mediterranean. Oxford: Oxford University Press. Willis, David (2011): A minimalist approach to Jespersen’s Cycle in Welsh. In Jonas, Dianne, John Whitman & Andrew Garrett. Grammatical change: Origins, natures, outcomes, Oxford: Oxford University Press, 93–119. Winters, Margaret E. (1987): Innovations in French negation: a cognitive grammar account. Diachronica 4: 27–52. Zanuttini, Raffaella (1997): Negation and clausal structure: a comparative study of Romance languages. Oxford: Oxford University Press. Zeijlstra, Hedde (2004): Sentential negation and negative concord. PhD thesis, Utrecht University. Zeijlstra, Hedde (2009): On French negation. Ms., . Zimmer, Karl (1964): Affixal negation in English and other languages: an investigation of restricted productivity. London: William Clowes and Sons.

Cecilia Poletto

Negative Doubling: In favour of a “Big NegP” analysis

Abstract: In this work, I reformulate Pollock’s (1989) analysis of French discontinuous negation and argue in favour of a cartographic approach to the category called NegP. I propose that it is a complex category which contains at least four different projections and I will show that the various types of discontinuous negation found in several Northern Italian dialects can be analysed in the same way as DP doubling has been analysed by Kayne (1975) and Belletti (2005), i. e. as independent movement of structural portions of an originally unitary constituent containing all negative markers first merged inside the VP. This proposal explains the numerous exceptions to Zanuttini’s (1997) analysis noted by Manzini & Savoia (2011) without rejecting the core of her proposal. Furthermore, this view is potentially interesting in a wider perspective if we assume that each of the elements identified here expresses structural projections relevant to semantics and could ultimately lead to a different way of conceiving negation as something different from the operation of formal logic which negates propositions. Keywords: discontinuous negation, Northern Italian dialects, DP doubling, negative concord, functional projections

1.

Introduction

In this work, I will examine the distribution of negative markers in Italian dialects and will provide arguments in favour of an old idea that is already present in Pollock’s (1989) analysis of negation in French: he proposes that the two sentential negative markers ne and pas in French originate as the head and the specifier of a single projection located in between AgrP and TP. On the other hand, subsequent work like Rowlett (1998, 2007) has entertained the hypothesis that the negative marker pas is actually located much lower in the structure of the clause than Pollock had originally proposed, in a position which can either be the one of the direct object or adjoined to the vP. A similar proposal, i. e. that

82

Cecilia Poletto

the negative marker is first merged in the object position, is made by Bayer (2009) and Manzini & Savoia (2011). The view that ne and pas form originally a constituent and the view that pas is actually merged lower than the position where we see it are not incompatible. If they are combined, they actually achieve more predictive power. They can explain some of the exceptions to Zanuttini’s (1997) generalizations noted by Manzini & Savoia (2005, 2011) with respect to the distribution of sentential negative markers in Northern Italian dialects without throwing away the whole bulk of empirical generalizations originally provided by Zanuttini’s work. More specifically, I will propose that all negative markers occurring in languages with discontinuous negation start out as a unit, and that this unit is first merged inside the vP, not necessarily in the object position, but definitely in an argumental one.1 Pollock’s (1989) original intuition about negation and much old and recent work on doubling starting from Kayne (1975), Uriagereka (1995), Cecchetto (2000), Belletti (2005), Poletto (2008a,b), a.o., can be reinterpreted in the following way: all elements that appear to have a unitary function in the clause are first merged together. This is valid for clitic doubling of arguments, where, despite appearances, the clitic and the DP are merged as a unitary nominal expression with a single thematic role and are then independently moved to different positions, but it is also valid for negative markers when they do not give rise to double negation. This means that although we apparently see two or even three negations in the clause, there is actually only one constituent performing the function of negation. I will therefore use the term “negative doubling” with the meaning of two or more sentential negative markers which do not give rise to double negation; the term is not identical to discontinuous negation, since it also encompasses cases in which doubling is not obligatory, but adds sentence implicatures to the original negative marker related to the speaker/hearer expectations. The paper is split according to the two conceptual steps of the argumentation: I will first present Zanuttini’s analysis of Northern Italian negative markers and will address the exceptions to her generalizations noted by Manzini & Savoia (2011) and by Garzonio (2008), showing that her analysis is correct if complemented by the idea that all negative markers start out as a syntactic unit. In section 3, I will show that negative doubling and clitic doubling are two instances of the same phenomenon, as they share crucial syntactic properties. In section 4, I will take into account the various cases of doubling found in the Northern Italian dialects (NIDs) and show that they can all be derived by means of the cartographic approach to NegP proposed 1 An alternative would be to assume that negation takes the first argument merged as its complement, in verbs which do not have any thematic role associated to them; this could be the event itself. I will leave this possibility open, since I do not have any direct evidence that the negative marker is always located in the object position.

Negative Doubling: In favour of a “Big NegP” analysis

83

here. Section 5 is a tentative account of a cartographic approach to negation and provides some insights into future perspectives concerning non-standard negation. Section 6 concludes the article.

2.

On the internal layering of NegP

In order to answer the question about the number and properties of the markers involved I will first summarize Zanuttini’s (1997) results on NIDs. She shows that inside the Northern Italian domain there are at least four types of negative markers, whose syntactic distribution varies. On the basis of the test with adverb positions in Cinque (1999), she notices that we have to assume that there are at least four distinct positions for sentential negative markers inside the sentence structure, since they are interspersed inside the FPs in whose specifiers we see different classes of adverbs:2 a) NegP1 is the position of the preverbal negative marker of the type non, also found in standard Italian and typically found in North Eastern Italian dialects which displays typical clitic properties in all the Romance languages in which it occurs: (1) No so. (Venice) neg1 know ‘I do not know.’

b) NegP2 is the position of negative markers of the type pas, also found in French. This type of elements is located above the adverb already marking AnteriorT in Cinque’s theory of sentence structure. In this position we generally find elements that were originally minimizers, i. e. classifier-like nouns indicating a small quantity, like French pas (literally ‘step’), Rhaetoromance buca ‘morsel’, Florentine punto ‘dot’, and Emilian brisa and Lombard mia (originally ‘crumble’). (2) A su mia. (Bagnolo S. Vito, MN) 3 I know neg2 ‘I do not know.’

c) NegP3 is the position of negative markers whose etymological origin is related to the element meaning ‘no-thing’, like Piedmontese nen or Rhaetoromance nia, etymologically similar to the German or English negative markers. 2 I use the exact same terminology here as Zanuttini. 3 Here I use the same reference system I used in Poletto (2000) to refer back to dialects, i. e. I provide the name of the town and the abbreviation of the ‘province’, the territorial administrative unit used in Italy. Only the cities that are the head of a province (as for instance in (1)) are without this additional indication, because the name is the same. This will provide an idea of the geographical area where the dialect is spoken.

84

Cecilia Poletto

(3) A sogh nèn. (Borgo S. Martino, AL) I know neg3 ‘I do not know.’

d) NegP4 is the last type of negative marker and the lowest one, located in the VP space, below all adverbs and in some dialects also below arguments in sentence final position. This element etymologically corresponds to the pro-sentence negative marker. (4) Su no. (Milan) know neg4 ‘I do not know.’

Using the test on adverbs first proposed by Cinque (1999) to determine the position of other elements in the clause, the structure Zanuttini (1997) proposes on the basis of the distribution of negative markers with respect to adverbs is the following one: (5) [NegP1 non [TP1 V+Agr [NegP2 mica [ TP2 [AdvP already] [NegP3 niente [ Asp perf Vpast part [Asp gen/progr [AdvP always] [NegP4 NO]]]]]]]]

This structure is not the one for standard Italian, but the abstract representation of all possible negative markers found in NIDs. Furthermore, Zanuttini also discusses several properties of the various negative markers, which clearly show that they are located in different positions since they interact with elements located at different heights in the sentence structure. I will only briefly summarize them here and refer to her work for a more detailed discussion and more data. NegP1 is the only type of negative marker that can block subject clitic inversion, i. e. T to C in North Eastern Italian dialects, as the following examples show:4 (6) a. *No vienlo? neg1 comes-he ‘Isn’t he coming?’ b. Magnelo mina? (S. Anna, VE) eats-he neg2 ‘Isn’t he eating?’ c. Vàste nia? (Corvara, BZ) go-you neg3 ‘Aren’t you going?’

All negative markers are too low to block T to C except for NegP1, hence this property singles out the preverbal negative marker.

4 This test cannot be carried out for NegP4, because in all the dialects where this negative marker occurs subject clitic inversion has been lost.

Negative Doubling: In favour of a “Big NegP” analysis

85

NegP1 and in some dialects NegP2 are the two negative markers that can trigger a change in the verb morphology of the imperative in cases of a negative imperative. The suppletive form of the verb can either be an infinitival, a subjunctive or a gerund depending on the dialect, but the phenomenon is the same: NegP1 and NegP2 are the only two negative markers that can display incompatibility with a true (i. e. morphologically univocally marked) imperative form, while NegP3 and NegP4 are always compatible with true imperative in all dialects observed to this point: (7) a. *No va. neg1 go-imp ‘Don’t go!’ a′. No laòra! neg1 work ‘Don’t work!’ b. Movat mia! move neg2 ‘Don’t move!’ b′. Movrat mia! move-infinit.yourself neg2! ‘Don’t move!’ c. Parla nen! talk neg3! ‘Do not talk!’ d. Vusa no! shout-imp neg4 ‘Don’t shout!’

(Venice) (Cortina, BL) (S. Antonino, CH) (Albinea, RE) (Zanuttini 1997: 4, (20b)) (Milan)

The third phenomenon that singles out different types of negative markers is the position with respect to clitics: while NegP1 is located inside the structural domain of clitics (and the ordering between them depends on the dialect), NegP2 can be the host of clitics only in those (rare) dialects which have generalized enclisis like the dialect of Borgomanero systematically investigated in Tortora (2015). NegP3 and NegP4 never interact with clitics. (8) a. Nol vien. (Venice) neg1-he comes ‘He is not coming.’ b. At crumpulu opura at crumpi millu? (Borgomanero, NO) you buy.it or you buy neg2.it ‘Will you buy it or not?’

The fourth phenomenon that shows distinctions is negative concord:5 while NegP1 always requires negative concord, NegP2 frequently does not (though not

5 I make use of the standard definition of negative concord, namely those cases in which two or

86

Cecilia Poletto

in all dialects), NegP3 does only in very limited contexts, and NegP4 has never been reported to allow for negative concord: (9) a. No ‘l è lugà nogugn. not he is come nobody ‘No one came.’ b. E’n m’a briza / mia vest endsun. scl neg1 me has neg2/neg2 seen nobody b′. E’n m’a vest entsun. scl neg1 me has seen nobody ‘Nobody saw me.’ c. A parla nen cun gnun. he speaks neg3 with nobody ‘He does not speak with anybody.’ d. L’è rivà nisun. it is come nobody ‘No one came.’

(Cencenighe Agordino, BL) (Zocca, MO) 6 (Zocca, MO) (Piedmontese, Zanuttini 1997: 3, (55)) (Milan)

We can summarize all these properties in the following table: Table 1: Types of negative markers and their properties

Position

NegP1 NegP2 NegP3 NegP4 preT preAnteriorT pregenericAsp prevP

V to C interference Negative concord

+ +

Compatibility with true imperatives – Reorders with clitics +

– +/–

– –/(+)

– –

+/– –/(+)

+ –

+ –

Therefore, there are rather strong empirical arguments to believe that the four negative markers occupy different positions. However, according to Zanuttini’s view, every element is merged in the position illustrated above in (5) without movement. In what follows, I will show that there are clear indications that negative markers do move. Furthermore, all the exceptions to (5) found by Manzini & Savoia (2005, 2011) can be dealt with if we complement Zanuttini’s proposal with the idea that all negative markers start out inside a unit located at the vP border and that (5) is the final position of the various negative markers and not the first-merge one.

more negative elements do not cancel each other out but amount to a single negation in the interpretation of the sentence. 6 The two possibilities depend on the speaker investigated.

Negative Doubling: In favour of a “Big NegP” analysis

2.1

87

Movement of negative markers

It is well known that negative markers can move around in the clause. This was first noticed in Cinque (1976) in his work on the postverbal negative marker mica, which can also occur (especially in Central Italian varieties) in preverbal position. In colloquial Northern Italian, we also find alternations as the following one: (10) a. Mica ti ho detto di telefonargli. neg2 you I.have told to phone.him b. Non ti ho mica detto di telefonargli. neg1 you I.have neg2 told to phone.him ‘I did NOT tell you to phone him.’

Notice that when the postverbal negative marker is moved to the preverbal position, the preverbal one, i. e. non, disappears. This is by no means an indication that mica occupies the position of non, as all n-words in preverbal position, and the adverb mai (‘never’), cannot co-occur with non when they are in preverbal position.7 Furthermore, there are clear indications that both the preverbal (cf. (11a)) and the postverbal negative marker of colloquial Italian (cf. (11b)) can occur in the CP layer in structures like the following, since the negative marker is located at the left of the complementizer: (11) a. Non che sia stupido, è solo che non studia. (colloquial Italian) neg1 that be stupid, is only that not studies ‘He is not stupid, he just does not study enough.’ b. Mica che sia stupido, è solo che non studia. (colloquial Italian) neg2 that be stupid, is only that not studies ‘He is not stupid, he just does not study enough.’

Zanuttini also notices that the Piedmontese postverbal negative marker of the NegP3 type can be found in the CP layer: (12) par nen ch’a s stofeissa (Piedmontese) so neg3 that he himself get-tired ‘in order for him not to get tired’

Therefore, these data can be perfectly handled in Zanuttini’s approach simply by assuming that all the negative markers found in Italian varieties are not subject to any specific requirements other than those generally applied on second merge/ movement. Evidently, one has to explain why negative markers move to the CP domain, i. e. identify the feature that triggers the movement.

7 Since non is always analysed as a head while mica is clearly a specifier, the two elements cannot occupy exactly the same position. Evidently, it is plausible that mica goes to the specifier of the position of which non is the head.

88

Cecilia Poletto

Further cases of movement of negative markers are provided by Manzini & Savoia (2011: 25–26) and reported below. They show that both negations of the NegP3 and of the NegP4 types can occur higher than the position where they are supposed to occur in Zanuttini’s structure: (13) a. i an durmi:d no ben they have slept neg4 well ‘They haven’t slept (well)’ b. i an no semper durmi:d they have neg4 always slept ‘They haven’t (always) slept.’

The problem for Zanuttini’s analysis is the following: here, we have a negative marker that according to the schema in (5) should occur in the domain of the vP, i. e. lower than all adverbs, but which actually occurs higher than manner adverbs like ben ‘well’ and also aspectual adverbs like semper ‘always’. Notice that, although Zanuttini does not mention any cases like these, they can still be treated in her system simply by assuming that negative markers can raise to a higher position than the one where they are merged, as must in any case be assumed for colloquial Italian preverbal mica. The only point to be cleared is once again the exact feature that triggers the movement of the negative marker to the position of a higher one, but cases like those above are in principle compatible with (5). However, Manzini & Savoia (2005, 2011) report cases of exceptions which are more difficult to explain when starting from a structure like (5), because in their examples the negative marker occurs in a position which is lower than the first merge-position proposed by Zanuttini and therefore cannot be treated as an instance of movement out of the merge position illustrated in (5) as in the cases mentioned above. I only consider examples of movement of the negative marker across adverbs, and not with respect to past participle or infinitival forms, as it is well known that verbal forms can raise to different positions inside the clause in each variety (as already shown by Cinque 1999), and this might blur the data or require a preliminary investigation of the movement span of past participle/infinitive in each dialect considered. Therefore, the respective order of verbal forms and negative markers is not safe enough as a counterargument to (5). As adverbs do not move, they are a far more reliable test. The following Rhaetoromance case quoted by Manzini & Savoia (2011) is a case of a NegP2, i. e. minimizer negation (bo = bucca), which occurs lower than the adverb ‘yet’ while according to Zanuttini’s structure in (5) it is merged higher than this type of adverb: (14) jau dormel aun bo (Münstertal) I sleep yet neg2 ‘I don’t sleep yet.’

Negative Doubling: In favour of a “Big NegP” analysis

89

This cannot be a case of movement of the adverb, since adverbs do not generally move to other positions.8 Furthermore, it cannot be a case of constituent negation where the negative marker modifies the adverb, since constituent negation occurs on the left of the modified constituent and not on the right in NIDs.9 Hence, we are forced to conclude that this is a real exception to Zanuttini’s analysis, which should not exist, if a system like (5) is correct. Manzini & Savoia notice that the same type of data are found with NegP3, which should occur higher than the adverb ‘yet’, but can actually occur lower: (15) i dormu ŋku naint they sleep yet neg3 ‘They don’t yet sleep.’

Further problematic cases are those analysed in Garzonio (2008) concerning the negative marker punto in Florentine, which is once again a minimizer and should occur in NegP2, but does not, as the following data show:10 (16) Un ha ancora dormito punto (Florence) not has yet slept neg2 ‘He has not slept yet.’

The fact that the negative marker occurs after the past participle is not a problem per se, since Cinque (1999) has shown that the past participle can raise to different positions in Italian varieties. The problem here is again represented by the adverb, which does not move and is on the left of the negative marker, although it should be on the right according to (12). Hence, if we want to keep Zanuttini’s original generalization, which indeed has the merit of accounting for the vast majority of the data found across the 8 Cinque’s (1999) account also analyses cases of movement of adverbs to a Focus position, and cases in which adverbs are ambiguous between two readings and could therefore occur in two different positions. None of these “exceptions” apply to this example. Manzini & Savoia mention several other cases, some of which could be seen as cases of constituent negation applying onto the adverb as the following one: (i) El dorme no semper. (Manzini & Savoia 2011: 100, 27) he sleeps not always ‘He doesn’t always sleep.’ They also discuss the element più in depth, though I do not deal with that here. 9 One might wonder whether this is a case of remnant movement of the FP including the vP and the adverb to the left of the negative marker, but this would imply that the order of all adverbs has to be reversed, because remnant movement of the FP should also be possible in other contexts. Since FP remnant movement is not otherwise attested, I will consider this as a real counterexample to (5). 10 Once again, this can be treated as a case of remnant movement of the FP containing ancora and dormito, but since there is no further evidence of the possibility of remnant movement in the low portion of the sentence structure and the order of the adverbs is fixed according to Cinque’s hierarchy, this is also to be considered as a real counterexample to (5).

90

Cecilia Poletto

NIDs and at the same time explains Manzini & Savoia’s exceptions, we need to complement (5) with an additional assumption,11 namely that all negative markers are merged inside a single constituent located in the vP area,12 as Manzini & Savoia (2011) also assume, and then raised to the specifiers of functional positions in the clausal spine according to the standard mechanism of probe and goal (see section 5 for a first sketch of a proposal). This explains the first set of movements discussed above, where postverbal negative markers like colloquial Italian mica occur preverbally like in (10), it explains cases where any type of negative marker is found in the CP layer above the complementizer (cf. 11, 12), but also cases of negative markers which occur in a position apparently lower than the one where they are merged according to (5). Therefore, my proposal is that (5) is correct, but is to be seen as the result of second merge/movement of the negative markers in the majority of the dialects, not as the merge positions of all negative markers. The strongest counterexamples to Zanuttini’s schema discussed by Manzini & Savoia (2005, 2011) (see 13–16), i. e. cases where a given negative marker occurs lower than the position where Zanuttini places it in the structure, can only be explained by admitting that the negative marker remains in its original merge position. (5) has thus to be complemented in the following way: (17) [NegP1 non [TP1 V+Agr [NegP2 mica [ TP2 [AdvP already] [NegP3 niente [ Asp perf Vpast part [Asp gen/progr [AdvP always] [NegP4 NO] [vP [VP… [NegP [mica [non [niente]]]]]]]]]]]]]

The assumption that negation is first merged in the VP gives us the possibility to explain doubling in a straightforward way if we assume that the “NegP” found inside the VP is the unit that contains all the negative markers in the clause before splitting in order to reach the various positions where Zanuttini has identified them. In what follows, I will show that the syntax of doubling also supports the view that all negative markers are merged inside a “Big NegP” located inside the VP.

11 Manzini & Savoia actually propose that the negative marker is a noun and is always generated as a nominal constituent inside the position of the direct object. Whether this is the only possible position is still an open question, as the negative marker might be located in different argumental positions according to the thematic structure of the verb. This is rather clear if one observes the distribution of negative indefinite articles in German, which do not only occur in the direct object position, but also on dative objects, if the verb selects a dative complement. Therefore, I will not identify a specific position inside the VP, as Manzini & Savoia do, what I am interested in here is the internal layering of the category called “NegP”. 12 As mentioned in section 1, the idea that the sentential negative marker starts out in object position is present in Rowlett’s work (1998, 2007) on French and is proposed by Bayer (2009) for German ‘nichts’. German actually shows rather clearly that the negative marker can be on the object, since it has negative indefinite articles.

Negative Doubling: In favour of a “Big NegP” analysis

3.

91

Negative doubling and clitic doubling

It is a well known fact that the arguments of the verb can occur twice in Romance, once as a DP and once as a clitic. The phenomenon was first analysed by Jaeggli (1982, 1986) for Spanish, but is found with a different distribution in French, where tonic subject pronouns must be doubled by a clitic (see Kayne 1975); in Southern Italian dialects, which display a distribution related to differential object marking similar to the one of Spanish; and in Northern Italian dialects, where different types of subjects can or must be doubled depending on the dialect (see Poletto 2000). Furthermore, North Eastern Italian dialects display an obligatory doubling of dative arguments of all sorts, including wh-items and quantifiers (Cordin 1993). Belletti (2005) has shown that also standard Italian has doubling, not of clitics, but of tonic pronouns in emphatic structures13 and she also analyses quantifier floating as a case of doubling. There are essentially two views on the phenomenon, the first one, originally proposed by Jaeggli, is that the clitic is a type of agreement marker and therefore does not really count as an argument. This explains doubling in languages where the clitic is always obligatory, as agreement markers are, but not those languages where the clitic appears only under some circumstances (for instance only with a subclass of nominal expressions, like with pronouns or definite DPs but no quantifiers). Furthermore, this analysis does not extend to cases where neither of the two elements involved in the doubling relation is a clitic, as in the cases mentioned by Belletti (2005). The other hypothesis on the phenomenon, which was originally proposed by Kayne (1975), Uriagereka (1995) and Belletti (2005) among others, is that the elements are merged as a unit, but can raise independently to check different features in the clausal spine. This is the analysis I have also tried to defend in Poletto (2008b) and will assume here. The same can be assumed for the negative marker, where the various elements check different features (see below section 5) of a complex adverbial, much in the same spirit of the original proposal made by Pollock. The major theoretical reason why cases of doubling pose a potential problem is that we have two elements which express the same thematic role and case. A similar problem arises when we have several negative markers in the clause, since they should cancel each other out, but do not. Notice that the problem already exists in cases of negative concord between the negative marker and an n-word, but in this case it is possible to assume (and has actually been proposed) that the 13 These are cases like: (i) Gianni va lui a parlare col capo Gianni goes him to talk to the boss ‘Gianni will himself go to talk to the boss.’

92

Cecilia Poletto

n-word is not a negative quantifier but a negative polarity element.14 This is less plausible in the case of two negative markers; however, it has been proposed that the French preverbal negative marker ne is not a real negative element anymore but a sort of polarity element (see for instance Breitbarth 2014; cf. also Breitbarth, this volume). This is rather difficult to assume for a language like Italian, where both non and mica can occur alone and negate a clause, but they can also occur together with no double negation effect. Therefore, the problem we face with cases of several negative markers is similar to the one of DP-doubling; we have several elements which perform the same function in the clause. In what follows I will show that the syntactic restrictions found with negative doubling are also identical to those of DP-doubling, and I will conclude that the analysis must also be the same.

3.1

NegP doubling has the same properties as DP doubling

The first argument in favour of an idea that there is a clear parallel between the doubling of negative markers and doubling phenomena in nominal expressions is precisely the one mentioned above. Notice that in the Northern Italian dialects (NIDs) several negative markers can occur, we find examples with two (cf. 18) or three negations (cf. 19) and combinations of all types identified by Zanuttini (1997) (see below section 3 for a discussion of their properties):

14 Negative doubling between two negative markers has to be kept apart from negative concord, where the negative marker occurs with quantifiers, and from negative spread where several nwords occur together. Evidence in favour of this distinction is that the three phenomena can occur in different languages and do not overlap as we would expect if they were instances of the same process. There are languages that allow for negative concord and/or negative spreading but do not allow for negative doubling, for example Bavarian (see Biberauer & Zeijlstra 2012 for an analysis that keeps negative concord and negative spread distinct). (i) a. daß da Hons koa Buach (nit) glesn hot (Bavarian) that the Hans no book(acc) not read has ‘that Hans did not read any book’ b. daß da Hons koan Freind (nit) ghoifn hot that the Hans no friend(dat) not helped has ‘that Hans did not help any friend’ c. daß eam koa Mensch (nit) gseng hot that him no man(nom) not seen has ‘that nobody saw him’ Here I propose that only negative doubling is to be dealt with like DP doubling, while negative concord and negative spread are instances of a different phenomenon, namely Agree (see Zeijlstra 2004).

Negative Doubling: In favour of a “Big NegP” analysis

(18) a. Nol me piaze. neg1-it me likes b. Nol me piaze miga. neg1-it me likes neg2 c. Nol me piaze gninte. neg1-it me likes neg3 d. Nol me piaze NO. neg1-it me likes neg4 ‘I do not like it.’

93

(Venice) NegP1 + NegP2 NegP1 + NegP3 NegP1 + NegP4

(19) No la go miga magnada NO! Neg1 it have not eaten neg4 ‘I did not eat it!’

These data show that it is not very plausible to assume that only one of the negative markers is the “real” negation and all the others are polarity items, since each negative marker can occur alone and negate the clause, as shown by (20): (20) a. No! (Venice) Neg4 ‘No!’ b. Nol lo ga fato. Neg1-he it has done ‘He did not do it.’ c. Miga che el lo gabia fato. Neg2 that he it has done ‘He did not do it.’

In (19) we see three negative markers, the preverbal no (pronounced with a closed /o/), the postverbal negative marker miga (akin to standard Italian mica, although it occurs in more contexts than it does in standard Italian) and the pro-sentence negator no at the end of the clause (pronounced with an open o /ɔ/). Both the preverbal and the postverbal negative markers can negate the clause alone, as the pro-sentence negative marker must also be assumed to be a negation. This means that negative doubling behaves like the cases of clitic doubling noted by Belletti (2005), where neither element can be analysed as “something else” and therefore, the only plausible analysis is to say that they start out as a unit which is then split in the course of the derivation to check different features both belonging to the original unit but specified onto the two (or three) elements. In the next sections, I will provide further arguments to show that DP doubling and negative doubling are amenable to the same analysis and, in section 5, I will make a first proposal that can account for the features that are checked by different elements.

94 3.2

Cecilia Poletto

Clause bound phenomenon

It is a well known fact that DP doubling is a clause bound phenomenon; with welldefined exceptions, we cannot double a constituent belonging to a different clause.15 The same is true of negation: it is not possible to double a negative marker with another sitting in a different clause. If we do so, the result is either ungrammaticality or a real double negation reading. (21) a. Non mi ha detto che *(non) viene mica. (Italian) neg1 to.me has said that not comes neg2 b. Un m’ha detto che *(un) viene punto. (Florence) neg1 to.me has said that neg1 comes neg2 ‘He has not said that he is not coming.’ c. No credo che *(no) el venia NO. (Venice) neg1 (I) think that neg1 he comes neg4 ‘I do not think that he is not coming’ d. Nol me ga dito che *(no) el dorme gninte. (Venice) neg1-he has said that neg1 he sleeps neg3 ‘He hasn’t told me that he doesn’t sleep.’ (22) Non è che *(non) viene mica. (Italian) neg1 is that neg1 comes neg2 ‘It is not the case that he is not coming.’

If we assume that negative doubling is the result of the split of a unit containing the two (or three) negative markers, this is the expected result.

3.3

Constituent negation

One further rather strong argument to analyse negation doubling as originating from a single constituent (much as DP doubling in Belletti’s 2005 analysis) is the fact that two negative elements can actually be observed to occur as a single constituent, for instance in dialects like Paduan, where constituent negation can be marked by two negative markers together (i. e. no and miga).

15 What I mean here by clitic doubling are cases in which a clitic doubles an XP in its argumental position as, for example, in Spanish. I abstract here from cases of topics, which are not analysed as real cases of movement by all authors (see for instance Cinque 1990), from cases of wh-movement in varieties which also double wh-items through a clitic, since wh-movement is recursive, and from cases of restructuring, where the clitic sits on the modal and not on the infinitival verb, which have convincingly been analysed by Cinque (2006) as monoclausal constructions. Notice that in Italian (as well as in English) there is a phenomenon known as negation raising, which closely parallels clitic climbing.

Negative Doubling: In favour of a “Big NegP” analysis

95

(23) a. no miga tutti (Padua) neg1 neg2 all ‘not everyone’ b. no miga pochi neg1 neg2 few ‘not few’ c. no miga tanti neg1 neg2 many’ ‘not many’

This clearly shows that the two negative markers can at least occur as a constituent, and crucially do so when the structure does not provide any clausal projection the two elements can move to. Manzini & Savoia (2011) also report cases of two adjacent negative markers inside clauses. Data of this type are interesting for two main reasons: they show that two negative markers can occur together, and the order of the two negative markers is the opposite of the one we would expect, as the element non, which is usually found in a preverbal position comes after the negative marker mia and not in front of it: (Sant’Angelo Lodigiano, MI; Manzini & Savoia 2011: 27) (24) El ciami mia non anmo. him I-call neg2 neg1 yet ‘I am not calling him yet.’

It is not self-evident that in this case the two negative markers have to be analysed as still being a unit; it could be the case that they occur in adjacent positions simply because their target positions are. However, cases like this are to be analysed, the fact that something like (24) is possible clearly shows that at least two negative markers can occur as a unit in NIDs. On the basis of the three arguments mentioned above, I conclude that negative doubling displays properties similar to those of DP doubling and as such can potentially be analysed in the same way.

4.

Possible combinations between negative markers

If all negative markers start out inside the vP and are then raised to a different position, as I claim here, this means that all combinations between all negative markers should be allowed. In the majority of the dialects doubling or tripling has semantic/pragmatic import in the sense that in addition to negation, it expresses the speaker’s attitude towards the event not taking place or is related to an additional conversational implicature (see Cinque 1976 for an analysis of postverbal mica in colloquial Italian). However, there are also dialects where the

96

Cecilia Poletto

combination of two negative markers does not give rise to any special reading (as it is the case of standard French).16 As seen above in section 3, in Veneto NegP1 can be combined with all other types of negative markers. The combination of NegP1 and NegP2, similar to French, is attested in the Emilian area, rather in the central part of Northern Italy. (25) A n magn menga la cherna. (Carpi, MO) scl neg1 eat neg2 the meat ‘I do not eat meat.’

The combination between NegP1 and NegP3 is attested in the Rhaetoromance area (in the Badia valley). (26) Dytaurela n el nia gny. (S. Leonardo di Badia, BZ) yet neg1 is-he neg3 come ‘He has not come yet.’

The combination between preverbal negation and NegP4 is attested in the Trentino area (although this type of negation goes back to a system where only preverbal negation is found, or focus negation is only used in special contexts), and is known to have existed in Milanese in the XVI century (see Vai 1996), which has nowadays only NegP4. (27) No se dis cosi no. (Val di Non, TN) neg1 it says so neg4 ‘We do not say so.’

As far as I know, there are no dialects where it is possible to combine NegP2 and NegP3, NegP2 and NegP4 or NegP3 and NegP4 without involving any special semantics (i. e. without giving rise to non-standard negation).17 This might be due to a historical accident, as all dialects started out with a “high” type of negation, or it might be a meaningful hole in the paradigm. With the data base used here, it 16 The geographical distribution of standard negation (i. e. non presuppositional negation) is as follows: In the Eastern Area standard negation is provided by a preverbal negative marker of the no(n) type. This area includes Veneto, Friuli, Trentino and partly Romagnolo, and the Rhaetoromance dialects of the Fassa Valley and Cortina. In the Western area standard negation is provided by a postverbal negative marker of different types (in Lombard Focus negation, in Piedmont quantifier negation) except for the Ligurian area which patterns with the East. Emilian dialects display doubling negation with a combination of scalar and minimizer negation of the standard French type, which is also a stage attested in all dialects which nowadays have postverbal negation, and this incidentally confirms Jespersen’s cycle. Some Rhaetoromance dialects (Badia and Gardena) also display discontinuous negation but of the type scalar+ quantifier negation type. Hence, there is no real homogeneous trend from East to West, although this is the rough situation at first sight. 17 Each type of negative marker can give rise to different types of sentence implicatures related to the expectations of the speaker or of the hearer. The matter is too complex to be handled here.

Negative Doubling: In favour of a “Big NegP” analysis

97

is not possible to make a choice between these two options, because other language types should be taken into account.18 However, the combinations which are not found for standard negation are all attested in different dialects when a special attitude of the speaker is signalled, which I will refer to as “non-standard negation”.19 Although it is not attested as standard negation, the combination between NegP2 and NegP3 is by no means impossible in other dialects. In Piedmontese, where NegP3 is the standard negative marker, the combination with NegP2 is indeed possible, so there is no a priori ban against this combination, it just depends on the implicature they are associated with in the relevant dialect. (28) Fa pa nen sulì. (Zanuttini 1997: 46) do neg2 neg3 that ‘Don’t do that.’

As mentioned in section 3, tripling exists, and is obtained either by the combination of NegP1, NegP2 and NegP4 or by NegP1, NegP3 and NegP4: (29) a. No la go miga magnada NO! neg1 it have neg2 eaten neg4 ‘I did not eat it!’ b. No-l me piaze gninte NO! neg1-it me likes neg3 neg4 ‘I do not like it at all!’

The conclusion we can draw is that any type of negative marker is compatible with any other depending on the dialect. In some cases the combination gives rise to standard negation, in others to non-standard negation (see below for an analysis of the difference between standard and non-standard negation). This means that the Big NegP must contain at least four distinct projections where each of the four negative markers is sitting.20

18 For instance some Dutch dialects display the occurrence of what seems to be quantified and focus negation as standard negation. Therefore, the lack in the paradigm would be accidental and due to the fact that we are looking at languages which are all undergoing a similar diachronic process. 19 I will make use of the distinction between standard negation and non-standard negation, referring by the latter to those negative markers that can only occur under certain pragmatic conditions related to the speaker’s or addressee’s expectations. 20 However, it is possible that there are more than four, the typological literature on negation reports several languages which use up to three negations (see Alsenoy & van der Auwera 2014), and some rare cases of four negative markers. Johan van der Auwera (p.c.) tells me that there even exist cases of five negative markers, a language that has this phenomenon is Kiranti. I will not entertain a systematic comparison with typological work on negation here, but it is clear that macro- and microvariation should complement each other when it comes to a detailed analysis of the type of projections located inside the Big NegP.

98

Cecilia Poletto

If negative doubling is an instance of a process of splitting a unit due to independent feature checking of several features, one might ask how many and what type of positions the unit which we refer to as NegP contains. In this work, I determined how many positions there are and the elements that belong to these four categories while leaving the exact semantic value of these positions to future work, since it requires detailed semantic tests. Nonetheless, it is possible to make some preliminary remarks concerning the approximate semantic value of the elements contained in the Big NegP, since the etymology of the negative markers is rather transparent. The two clearest cases are those of Neg2 and Neg3, the first always derives from a classifier-like element representing a minimizer, which provides the smallest quantity on a scale. NegP3 is clearly the n-word meaning ‘nothing’, which contains, in addition to an nmorpheme another type of light noun, the word meaning ‘thing’, as English also transparently shows. This element can be interpreted as an existential light noun which can be found in the nominal spine of several polarity items like ‘everything’, ‘something’ or ‘anything’ (see Garzonio & Poletto 2015 on existential elements inside quantifiers in the Southern Italian dialects). As for NegP4, this seems to be related to Focus structures, since the element no occurs in the first contexts when negation is focused (see Poletto 2008a). The hardest element to identify in terms of its semantic contribution is precisely the preverbal negative marker non, which derives from ‘non-unum’, but was already one single word in Latin. The presence of these elements in the Big NegP might indicate that the way negation is achieved in natural language is not the way that has been supposed up to now, namely the same as the one of formal logic. It might be the case that natural languages do not simply use something like ‘¬P’ to negate a clause, but that the process of negation involves various steps, each of which corresponds to a distinct semantic operation and is represented by one of the four negative markers Zanuttini (1997) originally identified in the NIDs. The fact that an existential, as well as a scalar minimizer and a Focus element can all be used to represent part (in cases of doubling) or the whole (in cases of singleton negation) of the “Big NegP” could simply be seen as the result of a historical process of grammaticalization, but I think that it might be interesting to pursue the alternative line of thought that what we see is a synchronic process which reflects the way in which negation is achieved in the semantics of natural languages. In the following section, I will try to sketch a formalization of the observations made until now.

Negative Doubling: In favour of a “Big NegP” analysis

5.

99

The internal structure of NegP

If it is true that natural languages do not have a single operator in their inventory which corresponds to the ‘¬P’ negation of a proposition of formal logic, one wonders how negation is achieved. My intention here is not to provide a definitive and reliable system of what the semantic operations that are performed to obtain negation really are, but to show how this could be done and concentrate on the syntactic aspects of such a hypothesis. Therefore, in this section I will provide a very tentative analysis of the internal projections located in the NegP; although this proposal has to be further refined or maybe even completely restated from the semantic point of view, a first approximation is nonetheless interesting to give an idea of the way the syntactic mechanism works which also allows us to make more precise syntactic predictions. I propose that the lowest element inside NegP is an existential element. This is rather clear in those varieties that use the n-word meaning ‘nothing’, where one of the two formatives is clearly an existential one, as a standard negative marker. Similar existential elements are also present inside the etymological source of the Germanic negative marker originating German ‘nicht’, English ‘not’ and Scandinavian ‘ikke’, which all contain an existential light noun originally meaning ‘body’. This is also the case of Piedmontese varieties which use the element nen derived from nent (i. e. ‘nothing’) which most probably contains the element ente, ‘thing’ or gente ‘people’, or of Provençal varieties using ren (the counterpart of French rien, from Latin res once again ‘thing’) or of Central Ladin varieties which use nia both for the negative marker as well as for the n-word ‘nothing’. On top of this existential element there must be a projection which morpho-syntactically encodes the fact that this existential is set on a scale. Although I do not have any real evidence for this, I will assume that this is the position of the preverbal negative marker non of standard Italian, Central and Southern Italian dialects as well as North-Eastern Italian dialects. The third element found in the internal structure of the NegP is a minimizer, i. e. what Zanuttini calls NegP2, and is the type of French pas, Emilian brisa, Lombard mia/minga, Veneto mia/ mina/miga, Northern Lombard bo, Florentine punto etc. This element lexicalizes the semantic operation that connects the existential to the scale, i. e. it shows that the element to be taken into consideration is the smallest one of the scale. The last element is an element that lexicalizes an operation similar to the one of Focus, i. e. it univocally identifies an element inside a set of elements with the same property. Suppose that the set in question is the set of the true propositions, then identifying one proposition and taking it out of the set amounts to negating it. Notice that there are again etymological clues of this fact since several negative markers are etymologically related to prepositions meaning ‘out’ (see Sanfelici 2012). What negation does in other words is to assert that something exists, and that it is

100

Cecilia Poletto

the minimal entity on a scale and then takes it out of the set of true propositions. The set of FPs will thus be something like: (30) [FocusP NO [MinimizerP mica [ScalarP non [ExistentialP (ni)ente]]]]

Even though in many languages the whole NegP is represented by only one of these elements, there are languages that have two or three even in cases of standard negation. Devos et al. (2010) investigate the Jespersen cycle in the Bantu languages and mention languages which have two or three negative markers like the following one: (31) (mvûl) kà-nák-ááŋ-áp (kwénd) (Kanincin, Devos et al. 2010: 7) (1.rain) neg1.1sc-rain-tam-neg2 (neg3) ‘it does not rain’

Furthermore, they notice that there exist cases of three negative markers which are only interpreted as standard negation (i. e. negation without any special sentence implicature) like the following: (32) ka-zeby-áandi khúumbu ya ŋgúdy-áani kó (Suundi, Baka 1998: fieldnotes) neg1.1sc-know-neg3 9.name 9.conn 1.mother-poss.1sg neg2 ‘he does not know the name of my mother’

Interestingly, Devos et al. (2010) also discuss the probable etymological source of the possible negative markers in Bantu and reach the conclusion that one (the element p(a) in (31), which for phonological reasons appears as -áp in the example) was originally a minimizer. Other elements like kwend in (31) and ko in (32) in this language group are originally elements expressing a (non-negative) Focus feature, which lends support to the idea that one of the semantic operations performed to obtain negation is the same that is also performed with Focus. This seems to indicate that the etymological sources of the elements marking negation are limited and all have to do with one of the projections sketched above. This is actually a straightforward way to explain the Jespersen cycle: given that negation is not a single projection but a set of projections expressing the basic semantic components of what results as a sentential negation, it is clear that languages can choose between elements that lexicalize one of the four projections, and that in some cases, they can even lexicalize two, three or even all four of them. Even though the semantics of (30) is most probably not correct, the syntactic idea of a set of projections can still be correct. This idea also explains why the negative markers of different etymological origins end up in different positions: they move to the position where they check the feature corresponding to their semantics: elements like Milanese no move to a Focus position (probably the one at the vP border, which was first identified by Belletti 1999) and are therefore located at the vP border. Elements like minimizers move

Negative Doubling: In favour of a “Big NegP” analysis

101

to an aspectual position related to their meaning, i. e. probably inchoative or prospective Aspect. Elements like Piedmontese nen occupy the same position bare quantifiers occupy in French (actually the position of nen is exactly the same position as French argumental rien), elements like standard Italian non and French ne are clitics, and they move to a position inside the clitic field. Let us now briefly see how it is possible to model non-standard negation, i. e. negative markers that are apparently optional but actually negate a conversational implicature as is the case of colloquial Italian mica already analysed as a “presuppositional negative marker” in Cinque (1976). In other words, what is the structural difference between French pas and Italian mica, if both are sitting in the same minimizer projection inside the complex NegP? I propose that the distinction is purely syntactic. In Italian the element mica is extracted from its original position inside the NegP at the vP edge and moved to a position in the Aspectual field which can be interpreted as marking some type of Aspect which is in a sense similar to a minimizer (probably something like inchoative). The element non, which is a clitic, moves to the clitic field as a remnant,21 with all the rest of the projections included inside NegP. This means that mica still retains its original meaning as a minimizer and the sentence implicature is computed on this basis. French does the opposite: ne is extracted out of the NegP and moved alone to the clitic field, while the whole NegP moves with pas to the Aspectual position. This means that, although pas moves to the same position as mica, it does not maintain the same value, since it is embedded under all the projections of the NegP. This explicitly formalizes the idea that pas is indeed the standard negative marker of French. In other words, the proposal I put forth is that the difference between standard and non-standard negation has to do with the amount of pied piping the various elements in the NegP drag along: if the whole NegP is dragged along, the negative marker is interpreted as standard negation, if not, only its original meaning is still accessible and the type of non standard negation is related to it.

6.

Concluding remarks

In this work I have discussed empirical evidence in favour of the idea that discontinuous negation is to be analysed in the same way in which DP-doubling has been analysed, i. e. as the result of the split of a single unit which is first merged in a very low position inside the VP. This view allows us to account for the cases of doubling and tripling of the negative marker and the fact that all combinations are possible and actually attested. Furthermore, this view allows us to explain the 21 See Cattaneo (2009) for the proposal that all clitics are instances of remnant movement of a whole DP.

102

Cecilia Poletto

exceptions to Zanuttini’s hierarchy of negative markers discussed in the literature by Manzini & Savoia and by Garzonio. Hence, we can keep Zanuttini’s original proposal about the existence of four different types of negative markers, if we assume that the positions observed are the final positions and not those where the negative markers are first merged. All negative markers can be moved to different positions inside the clausal spine for feature checking starting from the unique basic position inside the vP.

References van Alsenoy Lauren & Johan van der Auwera (2014): On the relation between double clausal negation and negative concord. In Hansen, Maj-Britt Mosegaard & Jacqueline Visconti (eds.): The Diachrony of Negation. Amsterdam: John Benjamins, 13–46. Baka, Jean (1998): Questionnaire kisuundi (complété). Unpublished. Bayer, Josef (2009): Nominal negative quantifiers as adjuncts. The Journal of Comparative Germanic Linguistics 12: 5–30. Belletti, Adriana (1999): “Inversion” as focalization and related questions. Catalan Working Papers in Linguistics 7: 9–45. Belletti, Adriana (2005): Extended doubling and the VP periphery. Probus 17: 1–35. Biberauer, Theresa & Hedde Zeijlstra (2012): Afrikaans Negation: filling the typological gap. Journal of Semantics 29: 345–371. Breitbarth, Anne (2014): The History of Low German Negation. Oxford: Oxford University Press. Cattaneo, Andrea (2009): It Is All About Clitics: The Case of a Northern Italian Dialect Like Bellinzonese. PhD thesis, New York University. Cecchetto, Carlo (2000): Doubling Structures and Reconstruction. Probus 12: 93–126. Cinque, Guglielmo (1976): Mica. Annali della Facoltà di Lettere e Filosofia dell’Università di Padova 1: 101–112. Cinque, Guglielmo (1990): Types of A′-Dependencies. Cambridge (MA): MIT Press. Cinque, Guglielmo (1999): Adverbs and Functional Heads. Oxford/New York: Oxford University Press. Cinque, Guglielmo (2006): Complement and Adverbial PPs: Implications for Clause Structure. In Cinque, Guglielmo: Restructuring and Functional Heads. The Cartography of Syntactic Structures. 4th vol. New York/Oxford: Oxford University Press, 11–64. Cordin, Patrizia (1993): Dative Clitics and Doubling in Trentino. In Belletti, Adriana (ed.): Syntactic Theory and the Dialects of Italy. Turin: Rosenberg and Sellier, 130–154. Devos, Maud, Michael Kasombo Tshibanda & Johan van der Auwera (2010): Jespersen cycles in Kanincin: Double, triple and maybe even quadruple negation. Africana Linguistica 16: 5–31. Garzonio, Jacopo (2008): A case of incomplete Jespersen’s cycle in Romance. Rivista di Grammatica Generativa 33: 117–135.

Negative Doubling: In favour of a “Big NegP” analysis

103

Garzonio, Jacopo & Cecilia Poletto (2015): How bare are bare quantifiers? Some notes from diachronic and synchronic variation in Italian. To appear in Linguistic Variation. Jaeggli, Osvaldo A. (1982): Topics in Romance syntax. Dordrecht: Foris. Jaeggli, Osvaldo A (1986): Three issues in the theory of clitics: case, doubled NPs, and extraction. In Borer, Hagit (ed.): Syntax and semantics 19: The syntax of pronominal clitics. New York: Academic Press, 15–42. Kayne, Richards. (1975): French Syntax. Cambridge, Mass.: MIT Press. Manzini, Maria R. & Leonardo M. Savoia (2005): I dialetti italiani e romanci. Morfosintassi generativa. 3rd vol. Alessandria: Edizioni dell’Orso. Manzini, Maria R. & Leonardo M. Savoia (2011): Grammatical Categories: Variation in Romance Languages. Cambridge: Cambridge University Press. Poletto, Cecilia (2000): The Higher Functional Field: Evidence from Northern Italian dialects. Oxford/New York: Oxford University Press. Poletto, Cecilia (2008a): On negative doubling. Quaderni di lavoro ASIT 8: 57–84. Poletto, Cecilia (2008b): Doubling as a spare movement strategy. In Barbiers, Sjef, Olaf Koeneman, Marika Lekakou & Margreet van der Ham (eds.): Microvariation in Syntactic Doubling. Bingley: Emerald, 38–68. Pollock, Jean-Yves (1989): Verb Movement, Universal Grammar and the Structure of IP. Linguistic Inquiry 20: 365–424. Rowlett, Paul (1998): Sentential Negation in French. Oxford: Oxford University Press. Rowlett, Paul (2007): The Syntax of French. Cambridge: Cambridge University Press. Tortora, Christina (2015): A comparative Grammar of Borgomanerese. Oxford: Oxford University Press. Sanfelici, Emanuela (2012): The grammaticalization of cen in Irish: A diachronic study. Paper presented at the 26th Irish Conference of Medievalists. University College Dublin, July 5–7 2012. Uriagereka, Juan (1995): Aspects in the Syntax of Clitic Placement in Western Romance. Linguistic Inquiry 26: 79–123. Vai, Massimo (1996): Per una storia della negazione in Milanese in comparazione con altre varietà altoitaliane. ACME 40: 57–98. Zanuttini, Raffaella (1997): Negation and Clausal Structure: A Comparative Study of Romance Languages. Oxford/New York: Oxford University Press. Zeijlstra, Hedde (2004): Sentential Negation and Negative Concord. PhD thesis, University of Amsterdam. Utrecht: LOT Publications.

Adam Ledgeway

Marking presuppositional negation in the dialects of southern Italy

Abstract: In contrast to the dialects of northern Italy which, given their rich variety of grammaticalized minimizers, have been the subject of a considerable number of semantic and syntactic studies, negation in the dialects of southern Italy has received very little attention since they appear to present very little lexical and morphosyntactic variation in this area. In this article, by contrast, I show that the linguistic expression of emphatic negation displays considerable variation in the dialects of southern Italy, including the syntactic marking of presuppositional negation through verb movement in some dialects and through the grammaticalizaton of an original lexical minimizer filu ‘thread’ in others. Among other things, the southern Italian evidence will be shown to support the idea that within the area of emphatic negation the binary opposition between presuppositional negators (‘after all…not’) and negative completive/intensive negators (‘not at all/even’) is more fluid than traditionally assumed, and, following proposals in Garzonio (2008a,b), that these two values can in some southern dialects correspond to two distinct positions within the lower portion of the functional clause structure. Significantly, it will be demonstrated that in certain cases both of these positions can be simultaneously lexicalized through distinct functional items or be derivationally related through movement from one position to the other. Keywords: presuppositional negation, Calabrian, Salentino, verb movement, minimizers

106

1.

Adam Ledgeway

Introduction

Studies on negation in the dialects of Italy have been dominated by investigations of northern varieties which show a number of spectacular developments in terms of discontinuous and postverbal negation strategies,1 with the result that today still very little is known about negation in southern dialects which is assumed to be largely similar to that of standard Italian and hence of little comparative interest (but cf. Damonte 2008; Poletto 2009; Egerland 2010; Garzonio & Poletto 2010b). Nonetheless, it has been claimed that “any language has the resources to express both plain and emphatic negation” such that we may “assume that these functions are so basic that any language must have the means to express them” (Kiparsky & Condoravdi 2006: 7). In relation to emphatic negation, Garzonio & Poletto (2010a: §3.1) claim that in Italo-Romance “[t]here are two main groups of reinforcers of negation: on the one hand, adverbs which can be considered as equivalent to ‘at all’ in English” and “on the other hand, negative morphemes that are used to express the fact that an explicit or implicit assumption made by the interlocutor is wrong.” For instance, in Italian this contrast is signalled respectively through the lexical opposition between affatto/proprio ‘at all’ and mica ‘not…after all’ (Cinque 1976; Zanuttini 1997: 61f.; Pescarini 2005; Penello & Pescarini 2008; on the connection with verum focus, see Romero & Han 2004; Gyuris 2009): (1) a. Non mi piace affatto/proprio. (It.) not me= pleases at.all ‘I don’t like it at all.’ b. Vuole una sigaretta? – Non fumo mica (NIt.) / Mica fumo. (C-SIt.) you.want a cigarette not I.smoke mica mica I.smoke ‘Do you want a cigarette? – But I don’t smoke.’

Similarly, (central) Catalan varieties lexically contrast gens ‘at all’ with pas ‘not… after all’ in the same contexts (Yates 1984: 108–109; Hualde 1992: 155; Wheeler, Yates & Dols 1999: 481): (2) a. No està gens contenta. (Cat.) not she.is at.all happy ‘She’s not happy at all.’ b. Quan vagi a Palma, veuré a en Joan – No viu pas a Palma. (Cat.) when I.go.sbjv to Palma I.will.see acc the Joan not lives pas in Palma ‘When I go to Palma, I’ll see Joan – But he doesn’t live in Palma.’

1 Cf., among others, Schwegler (1990: 151–174), Vai (1996), Parry (1997a,b; 1998; 2013), Zanuttini (1997), Manzini & Savoia (1998; 2002; 2005, III; 2011), Patruno & Sgarioto (2004), Penello (2005), Penello & Pescarini (2008), Pescarini (2005), Poletto (2008; 2010).

Marking presuppositional negation in the dialects of southern Italy

107

It is reasonable therefore to assume that southern Italian dialects too must draw such a distinction (pace Parry 2013: 96f.) and, if so, then it is natural to ask how they mark (emphatic) presuppositional negation. Already following Cruschina (2010: 31; cf. also Garzonio & Poletto 2010b: 78–79.; Poletto 2016: §51.2.3), it is known that Sicilian has grammaticalized an erstwhile cleft structure n(un) è ca… ‘not it.is that…’ > neca as a presuppositional negator:2 (3) a. Neca ci vonsi jiri. (Mussomeli, S. Cruschina p.c.) neca there= they.wanted go.inf ‘In any case, they didn’t want to go there.’ b. Statti quieto, neca scappu. (Corleone, Garzonio & Poletto 2010b) stay.imp calm neca I.escape ‘Don’t worry, I’m not about to run off.’

In what follows we shall show how, on a par with northern Italian dialects and other Romance varieties, the formal opposition between so-called plain and emphatic negation is not unknown in southern Italy, but represents a robust distinction across many southern Italian dialects. In particular, we shall sketch a broader typology of presuppositional marking in southern Italy through the close examination of two other types, one syntactic (from the dialects of northern Calabria) and the other lexical (from the dialecs of Salento), which to date have barely received any attention in the literature. At the same time, we shall use these data to throw further light on the syntax of presuppositional negation and its structural representation within the clause.

2.

Southern Italian negator MANCO

We begin our discussion with the southern Italian negator mancu/manco/manchə (henceforth manco). Traditionally, it is claimed (cf. Rohlfs 1969: 294) that manco is an emphatic negator which functions as a scalar negative adverb with the meaning ‘not even’. As such, it can occur in postverbal (4/5a) or preverbal (4/ 5b) position or, alternatively, in a left-peripheral position (4/5c), as illustrated by the following Neapolitan (cf. Ledgeway 2009: 692) and Sicilian (S. Cruschina p.c.) examples: (4) a. Isso nun s’ è nformato manco. (Nap.) he not self= is informed not.even ‘He hasn’t even informed himself.’

2 In spoken Italian, the cleft structure non è che… ‘not it.is that…’ is also frequently used to mark presuppositional negation (cf. Bernini 1992).

108

Adam Ledgeway

b. Cresce senza ca isso manco sape comme. (Nap.) it.grows without that he not.even knows how ‘It’s growing and he doesn’t even know how.’ c. Manco ll’ acqua ce putimmo vévere! (Nap.) not.even the water us= we.can drink.inf ‘We can’t even drink the water!’ (5) a. Un sapi mancu cantari. (Sic.) not he.knows not.even sing.inf ‘He can’t even sing.’ b. Ci vuliva purtari na nguantera di cannola e mancu c’ era to.him= I.wanted take.inf a tray of cannoli and not.even there=he.was dintra! (Sic.) inside ‘I wanted to take him a tray of cannoli and he wasn’t even at home!’ c. Mancu iddu u sapiva! (Sic.) not.even he it= knew ‘Even he didn’t know!’

However, it has sometimes been noted that manco may also function as a plain negator (Ledgeway 2009: ch.18; S. Cruschina p.c.), witness the Neapolitan and Sicilian examples in (6a-b), as well cases of constituent negation such as (6c) and its use in isolation in examples like (6d). This situation has been taken to the extreme in the Basilicatese dialect of Rionero in Vulture (6e), where manco has now entirely replaced all previous reflexes of non (Poletto 2016: §51.2.3): (6) a. Chella manco capesce chello che amm’ ’a fa’. (Nap.) she manco understands that which we.have to do.inf ‘She doesn’t understand what we’ve have to do.’ b. Javi du uri ca pruavu a chiamarla pi telefunu ma mancu it.has two hours that I.try to call.inf=her for phone but mancu arrispunni. (Sic.) replies ‘I have been trying to call her on the phone for two hours, but she doesn’t (even) answer.’ c. Manco pe ssuonno! (Nap.) manco for dream ‘Not on your life!’ d. M’ ’o vuo’ da? – Manco. – Dammillo! (Nap.) me= it= you.want give.inf manco give.imp=me.it ‘Will you give it to me? – No. – Give it to me!’ e. Ii pens ca Gianni manc vene. (Rionero in Vulture) I think that Gianni manco comes ‘I think that Gianni is not coming.’

The tendency for manco to develop into a marker of so-called plain negation (cf. also north-eastern Italian mica discussed in Pescarini 2005; Penello & Pescarini 2008: 44) can perhaps be explained by the fact that

Marking presuppositional negation in the dialects of southern Italy

109

reinforced structures, introduced for pragmatic and semantic reasons arising from the controversial nature of denying an active expectation to the hearer’s face […] are bound to lose their impact with increased use (Dahl 2001) and eventually lead to their being used for less contentious acts of denial (i. e. less discourse-active), especially if other (e. g. structural) factors make the traditional negator a less efficient negating strategy. (Parry 2013: 96)

As for the origin of southern manco (Tekavcˇic´ 1980: 482; Ledgeway 2009: 226; Parry 2013: 80, n.6), it represents a continuation of Latin mancus, originally an adjective meaning ‘maimed (in the hand), lame(-handed)’ but which soon developed the related figurative meanings ‘defective, imperfect(ive)’. This sense of ‘incomplete’ no doubt lies behind the early southern Italo-Romance meaning ‘less’ (7a) and the subsequent semantic development ‘not at all/even’ (7b) according to a semantic path which we can characterize as ‘maimed’ > ‘less’ > ‘not at all/even (anything)’ > ‘not’. (7) a. quatro oy cinque caporali manco potienti de nuy (ONap.) four or five corporals less powerful of us ‘four or five corporals less powerful than us’ b. Ma però ca foru fimini eu nu ndi maravilyu mancu (OSic.) but because that were women I not of.it= I.marvel mancu ‘But because they were women I’m not at all surprised’

Indeed, the development from ‘less’ > (plain/emphatic) negator is not limited to southern Italy, but is found both within Latin and other Romance varieties, witness the role of Latin/Italian minus/meno ‘less’ as a negator in (8a-b) (cf. also It. nemmeno ‘not even’ < ne- ‘neither’+ meno ‘less’): (8) a. epistulam Caesaris misi, si minus legisses (Lat., Cic., Att. 13.22) letter.acc Caesar.gen I.sent if less you.had.read.sbjv ‘I have sent you Caesar’s letter in case you hadn’t read it’ b. Non so se accettare o meno. (It.) not I.know whether accept.inf or less ‘I don’t know whether to accept or not.’

Following this brief overview of southern Italian manco, we now turn to consider its distribution and meanings in northern Calabria as part of our first case study.

2.1

Northern Calabrian mancu

Following preliminary analyses in Ledgeway and Lombardi (2005), northern Calabrian (Cosentino) mancu licenses the typically scalar intensive negative reading ‘not even’ when in postverbal position (9a-c) or when in a left-peripheral

110

Adam Ledgeway

position under focus-fronting (10a-c) giving rise to a case of ‘non-strict’ negative concord (Zeijlstra 2004: 258). (9) a. Un cci dorma mancu cchiù. (NCal.) not there= he.sleeps not.even anymore ‘He doesn’t even sleep there anymore.’ b. Un m a mancu salutatu. (NCal.) not me= he.has not.even greeted ‘He didn’t even greet me.’ c. Un pulizzava mancu a casa. (NCal.) not he.cleaned not.even the house ‘He didn’t even clean the house.’ (10) a. Mancu not.even b. Mancu not.even c. Mancu not.even

cci there= m’ me= (a the

dorma he.sleeps a he.has casa) houses

cchiù. (NCal.) anymore salutatu. (NCal.) greeted pulizzava a casa (NCal.) he.cleaned the house

Besides this scalar intensive negative reading, mancu also licenses the presuppositional reading ‘not…after all, in any case…not’ if it occurs immediately before the finite verb (11a-c), from where it follows the plain negator (n)un ‘not’ and any accompanying complement clitics (cf. 11b): (11) a. Te’, mintati ssu magliune! – Ma un mi mancu fa friddu. (NCal.) here put.imp=you this jumper but not me= mancu makes cold ‘Here you are, put on this jumper! – But I’m not cold (in any case).’ b. Stativi tranquilli, un vi mancu fazzu pagà. (NCal.) be.imp=yourselves quiet not you= mancu I.make pay.inf ‘Don’t worry, I wont charge you for it after all.’ c. Comunque un cci parrà, un mancu capisciadi. (NCal.) however not to.him.it= speak.inf not mancu he.understands ‘In any case don’t speak to him, after all he doesn’t understand.’

We thus see that, whereas languages like Italian mark the distinction between scalar and presuppositional negative readings lexically (e. g. nemmeno/neppure/ neanche ‘not even’ vs mica ‘after all/in any case…not’), in northern Calabrian the two readings are marked by the same lexical item but are superficially distinguished syntactically by the position of the verb with respect to mancu, as demonstrated by the minimal contrasts in (12–13). (12) a. Un mi canuscia mancu (/ Mancu mi canuscia). (NCal.) not me= he.knows mancu mancu me= he.knows ‘He doesn’t even know me.’ b. Un mi mancu canuscia. (NCal.) not me= mancu he.knows ‘After all, he doesn’t know me.’

Marking presuppositional negation in the dialects of southern Italy

111

(13) a. Un ni ’nteressava mancu (/ Mancu ni ’nteressava). (NCal.) not us= it.interested mancu mancu us= it.interested ‘We weren’t even interested in it.’ b. Un ni mancu ’nteressava. (NCal.) not us= mancu it.interested ‘We weren’t interested in it any way.’

As to the question of how a single lexical item, viz. mancu, comes to mark two different values, it will suffice to recall for the moment how the negative scalar intensive reading ‘not even’ may often simultaneously imply the presuppositional reading (cf. Manzini & Savoia 2005, III: 154–155), witness the examples in (14): (14) a. a’ a’’ b. b’ b’’

Give him a beer! – But he doesn’t even drink (alcohol)! Offrigli una birra! – Ma non beve nemmeno (l’alcol)! (It.) Dunacci na birra! – Ma un viva mancu (ll’alcol) / Ma mancu viva (ll’alcol)! (NCal.) Take your umbrella! – Why, it’s not even raining! Porta l’ombrello! – Perché, non piove nemmeno! (It.) Pòrtati l’ombrellu! – Picchì, mancu chiova (/ un chiova mancu)! (NCal.)

On the basis of such examples, we may conclude that the close relationship between both readings and their development from a single lexical item in northern Calabrian is quite natural, insofar as scalar intensive readings often imply presuppositional readings. 2.1.1 Northern Calabrian Presuppositional Negation: Syntactic Marking We now turn to look in greater detail at the syntax of mancu in northern Calabria where we have already seen that the scalar intensive vs presuppositional readings are marked syntactically, rather than lexically. To understand the relevant facts, we start from the premise that the basic functional structure of the clause can be conceived of in terms of the three broad areas outlined in (15), where the v-VP complex is progressively embedded in a lower adverb space (LAS), a higher adverb space (HAS), and finally the left periphery (for an overview, see Cinque 1999; Ledgeway 2012: ch. 4; Cruschina & Ledgeway 2016). Of these, we are directly interested here in the LAS which, following Cinque (1999; cf. also Zanuttini 1997), we present in some considerable detail in (16). (15) [Left Periphery [HAS S Adv (probably, perhaps…) [LAS Adv (already, still…) [v-VP Vlexical O]]]] (16) [HAS… [mica Neg1presuppositional [already T(Anterior) [no longer Aspterminative [still Aspcontinuative [always Aspperfect [guère ‘hardly’ Neg2 [ just Aspretrospective [soon Aspproximative [briefly Aspdurative [characteristically Aspgeneric/progressive [almost Aspprospective [completely AspSgCompletive [everything AspPlCompletive [well Voice [fast/early Aspcelerative(process) [again Asprepetitive(process) [often Aspfrequentative(process) [completely AspSgCompletive(process) [v-VP…

112

Adam Ledgeway

The structure in (16) proves essential to understanding both Calabrian verb movement and clitic placement (cf. Ledgeway & Lombardi 2005; 2014). In particular, we note that the finite verb typically targets a low position lexicalizing one of the functional heads contained within the space delimited by Neg1 (mica) to the left and AspSgCompletive (completely) to the right. This straightforwardly explains why the verb chiamu in (17a-b) follows all the lower pre-VP adverbs. (17) a. Unn’i propiu / cchiù / mai chiamu. (Neg1/Asptermin./Aspperf.) not them= at.all anymore never I.call ‘I won’t call them at all / anymore / ever.’ ammalappena / subbitu / quasi / tantu piacìa (Neg2/Aspprox./Aspprosp./AspSgCompl.) b. Ti you= hardly soon almost much it.pleased ‘You barely / soon / almost / thoroughly liked it.’

However, the finite verb must raise a little within the LAS since it has to precede the lowest pre-VP adverbs, as exemplified in (18): cuntu tuttu / buonu / torna (*cuntu). (AspPlCompl/Voice/Asprepet./process) (18) Ciu to.him-it= I.tell everything / well / again I.tell ‘I’ll tell him everything / well / again.’

As for object clitics, following Ledgeway & Lombardi (2005) these may move as phrases to the LAS (19a) or, more frequently, they raise even higher as heads to a clause-medial head position (YP) from where they are consequently separated from the finite verb by any intervening lower pre-VP adverbs (19b). (19) a. [HAS Giuvanni…] [YP [LAS sempe ni rumpa] [v-VP rumpa ni]] b. [HAS Giuvanni…] [YP ni [LAS sempe ni rumpa] [v-VP rumpa ni ]] Giovanni us= always us= annoys ‘Giovanni always annoys us.’

Consequently, the basic clausal order in northern Calabrian is Clitic + Adverb + Verb (20). Accordingly, we predict that if two (21a) or more (21b) lower adverbs co-occur then they will follow any object clitics and that they will strictly follow the linear order predicted in (16). (20) A mia [YP un mi [LAS mancu (piacìa) cchiù mi piacìa] [v-VP piacìa mi.]] to me not me= mancu it.pleased anymore it.pleased ‘In any case I didn’t like it any more.’ (21) a. Un si mancu ancora / *ancora mancu canuscianu. not selves= mancu yet yet mancu they.know ‘In any case they don’t know one another yet.’ b. Un vi mancu cchiù sempe fissìa cum’ a na vota. not you= mancu anymore always he.mocks like to a time ‘In any case he no longer always makes fun of you as in the past.’

Marking presuppositional negation in the dialects of southern Italy

113

In light of these facts, we now have a natural way of understanding the syntax of northern Calabrian mancu and how its respective presuppositional (‘not…after all’) and scalar intensive (‘not even’) readings are licensed. In particular, we assume that mancu invariably lexicalizes the specifier position of Neg1P, but that the two readings are a by-product of variable V-movement: under the presuppositional interpretation the finite verb raises to its usual low position within the LAS (22a), whereas the scalar intensive reading obtains whenever the verb raises to the clause-medial position to the left of mancu (22b). (22) a. [YP Un mi [LAS mancu mi duna] [v-VP duna mi not me= mancu he.gives ‘In any case he won’t give me the money.’ b. [YP Un mi duna [LAS mancu mi duna] [v-VP duna mi not me= he.gives not.even ‘He won’t even give me the money.’

ri sordi.]] the money ri sordi.]] the money

Following Ledgeway & Lombardi (2005: 80, 98–100), we conclude that in the dialects of northern Calabria the two types of emphatic negation are marked by the same lexical item but that they are differentiated syntactically through Vmovement, the extent of which unambiguously discriminates between the two readings.

3.

Salentino filu

Having examined a case of syntactic marking of presuppositional marking in southern Italian dialects, we now turn to a different strategy involving the grammaticalized lexical marking of presuppositional negation in the dialects of Salento. On the question of the grammaticalization of minimizers, Parry (2013: 84) notes that “[s]outhern Italian dialects in particular show a far lower degree of extension of minimizers beyond the original semantic context (Ledgeway 2009: 687–689), while those varieties that have progressed beyond the optional reinforcement of stage 2 are the Gallo-Italian varieties.”3 Indeed, this seems to be borne out by the available data, inasmuch as examples of minimizers in the 3 ‘Minimizers’ refer to those scalar negative markers which derive from original lexical items indicating a small quantity. These frequently emerge in the cyclical development of negation as first noted by Jespersen (1917: 4) who highlighted how in “[t]he history of negative expressions in various languages […] the original negative adverb [Stage 1, e. g. OFr. ne+V; A.L.] is first weakened, then found insufficient and therefore strengthened, generally through some additional word [Stage 2, e. g. St.Fr. ne+V+pas; A.L.], and this in its turn may be felt as the negative proper [= Stage 3, e. g. coll.Fr. V+pas; A.L.].” For discussion, see Parry (1997a; 2013: §3.2), Zanuttini (1997), Poletto (2016: §51.2.2.1), and Willis, Lucas & Breitbarth (2013: §1.2).

114

Adam Ledgeway

dialects of southern Italy such as those illustrated in (23a-d) from the history of Neapolitan (taken from Ledgeway 2009: 687–689) are invariably lexically restricted and prove relatively unproductive and somewhat rare, forcing us to conclude that they have not been grammaticalized as full-fledged negators. (23) a. non se mose uno not self= moved a ‘he didn’t move a step’ b. non te stimo più not you= I.esteem more ‘I no longer respect you at all.’ c. Non aggio potuto trovà not I.have been.able find.inf ‘I couldn’t find a soul.’ d. Non se senteva manco not self= heard not.even ‘You couldn’t hear a thing.’

pillo (ONap.) hair una fava (ONap.) a bean no cane. (Nap.) a dog na mosca. (Nap.) a fly

However, one major exception to this generalization are the dialects of Salento which have grammaticalized the minimizer filu (< filum ‘thread’), a fact which has gone unnoticed in the specialist literature (Mancarella 1998; Parry 1997a; 2013; Zanuttini 1997; Manzini & Savoia 2005, III). In those rare cases in which it is briefly mentioned, its significance is entirely overlooked and its meaning and distribution are incorrectly represented (cf. Garzonio & Poletto 2010a: §3.1). For example, Rohlfs (1956: 234–235) simplistically observes that filu “expresses negation or reinforces the negation; not, nothing, not at all, not one” (cf. also Rohlfs 1969: 305), citing Leccese examples such as (24a-c). (24) a. No ssienti filu? (Lec.) not you.hear filu ‘Don’t you hear?’ b. L’ òmu a Ddiu mò picca e ffilu crede. (Lec.) the man to God now sins and filu believes ‘Today man sins before God and has no faith.’ c. senza filu scherzi (Lec.) without filu jokes ‘without any jokes’

Slightly different, although still vague, are the descriptions given in two other dictionary entries. For instance, in relation to the dialect of Parabita Romano (2009:68) translates filu with Italian affatto ‘not at all’ (e. g., nunn è bbanutu filu ‘he has not come at all’), and similarly Presicce (n.d.; v. filu 2; www.dialettosalentino.it/filu_2.html) describes filu in the dialect of Scorrano as a “negative adverb with reinforcing function” translating it with Italian per niente and niente affatto ‘not at all’ (e. g. Nu llu sapìa filu ‘I didn’t know at all’). As we shall see below (§3.1), such descriptions are ultimately quite far off the mark.

Marking presuppositional negation in the dialects of southern Italy

115

Regarding the development of the negator filu, as briefly noted above, it represents the grammaticalization of the still homophonous lexical item filu ‘thread, yarn’. One does not need to look far to find parallels for similar grammaticalizations (cf. Eng. I haven’t got a thread to wear). Already in Latin, for example, filum ‘thread, yarn, string, cord’ – albeit in its variant form hilum – had been grammaticalized in various negative strategies (cf. Tekavcˇic´ 1980: 486; Bernini & Ramat 1996: 30), notably the negative polarity item ni(hi)l ‘nothing’ < ne ‘not’ + hilum ‘thread’. Moreover, hilum could also be employed as a reinforcer in conjunction with non/ne(c) with the scalar intensive meaning ‘not in the least, not a whit, nothing at all’: (25) a. neque proficit hilum (Lat., Cic. Tusc. 1.10) and.not he.progresses thread ‘and he does not advance an inch’ b. nec defit ponderis hilum (Lat. Lucr. 3.220) and.not is.lacking weight.gen thread ‘and none of the load is missing’

3.1

Salentino presuppositional negation: lexical marking

Having briefly reviewed the existing scant information about Salentino filu, let us now turn to examine the properties of Salentino filu in detail. We begin by noting that filu shows all the hallmarks of a grammaticalized m(inimizer)-negator (Garzonio & Poletto 2010a). First, it obligatorily occurs as a bare fossilized singular indefinite quantifier, no longer able to occur with accompanying determiners or modifiers (asterisked forms confirmed by native speakers): (26) Nu sta scherzu (*nu / *u) filu (*lungu). (Lec., Protopapa 1990: 24) not prog I.joke a the filu long ‘I’m not kidding.’

As part of its grammaticalized status, filu qua negator is also unable to select a PP lexical restriction, unlike its lexical counterpart (cf. nu filu t’oju ‘a drizzle of oil’, nu filu te ientu ‘a breath of wind’; asterisked forms confirmed by native speakers): (27) Filu (*ti) chesciuni / Senza filu (*ti) scherzi (Lec., Rohlfs 1956: 234) filu of sheet(s) without filu of jokes ‘No sheet(s) / without any jokes’

Furthermore, as an autonomous negative marker filu may occur in isolation (28a) and function as a constituent negation (28b):4 4 On the iteration of filu in examples like (28a), see the discussion of examples (36a-e) below.

116

Adam Ledgeway

(28) a. Ca cce simu surdi? – None, filu filu… (Scorrano, Musio 1995: 14) that what we.are deaf no filu filu ‘Are we deaf or something? – No, no, no…’ b. Tie lu pueti influenzare, filu ieu (Lec., Protopapa 1992: 95) you him= can influence.inf filu I ‘You can influence him, not me’

In line with its grammaticalized functional status, filu has also come to occupy a fixed position in the clause which, following Cinque (1999), we can identify with the specifier of Neg1P (equivalent to Zanuttini’s (1997) Neg2P). This can be clearly seen in (29a) where filu immediately follows the past participle raised to a clause-medial position (cf. Ledgeway 2012: 143–144), but obligatorily precedes the continuative aspectual adverb ncora ‘still’ in SpecAspContinuative. By the same token, preverbal filu orders such as (29b) are derived by the verb raising to the head of Neg1P,5 from where it is immediately preceded by filu in the associated specifier position. (29) a. Tantu sordi jeu nun aggiu visti filu ncora. (Matino, Fiorentino so.much money I not have seen filu yet 1998: 55) ‘In any case, I haven’t seen any money yet.’ b. Filu sapimu la libirtà. (Lec., Rohlfs 1956: 234) filu we.know the freedom ‘We don’t know freedom.’

The full-fledged status of filu as a grammaticalized negator is further evidenced by the fact that it may also co-occur with negative polarity items/N-words such as nienti ‘nothing’ and né ‘neither, nor’: (30) a. A ttie la capu nu te dice filu nienti? (Lec.) to you the head not you= says filu nothing ‘Does your head not tell you anything?’ b. Jeu nu’ supportu filu né l’ una né l’ adhu. (Matino, Fiorentino I not stand filu neither the one nor the other 1998: 17) ‘I can’t stand either one or the other.’

Finally, the grammaticalized nature of filu is highlighted by the fact that it has lost all of its original referential content, viz. ‘(not a) thread, yarn, wire’ > ‘not’ (cf. the separate entries for lexical and grammatical uses of filu registered in Presicce (n.d.): www.dialettosalentino.it). As a consequence of this semantic bleaching, its distribution is now lexically unrestricted, no longer limited to occurring in conjunction with semantically related predicates such as sew and spin, but is equally found with such predicates as speak, please and go (31a-c).

5 Cf. the analysis of preverbal Italian mica in Cinque (1999: 214–215 n. 7) and Ledgeway & Lombardi (2005: 87–88).

Marking presuppositional negation in the dialects of southern Italy

117

(31) a. Nu sta pparli filu chiaru. (Lec., Protopapa 1991: 42) not prog you.speak filu clear ‘You’re not speaking clearly.’ b. Dha tosse nu’ ne piace filu. (Matino, Fiorentino 1998: 136) that cough not to.him= pleases filu ‘He doesn’t like the sound of that cough.’ c. Li mònichi no’ vànnu fìlu a lli tiscutèchi cu lli minicònni. (Maruggio) the nuns not go filu to the discos with the mini-skirts ‘Nuns don’t go to discos wearing mini-skirts.’

In light of these observations, as well as the vague definitions in most of the available descriptions of Salentino filu reviewed in §3, it would not, however, be correct to conclude that filu in Salentino is a simple negator which has reached Stage 2 of Jespersen’s Cycle as part of a grammaticalized discontinuous negation strategy. Rather, Salentino filu is an emphatic negator and, in particular, is employed to mark presuppositional negation, an observation which all previous characterizations have failed to capture. This can be clearly seen in examples such as (32a-c) where the verb negated by filu negates an explicit presupposition or expectation in the immediately preceding discourse. With this function, filu may also be used as a constituent negator (32d-e). (32) a. Sta schersi? – Nu’ sta scherzu filu. (Lec., Protopapa 1990: 24) prog you.joke not prog I.joke filu ‘Are you kidding? – I’m not kidding.’ b. Penzene sempre alli fatti de l’ autri e mai alli pinzieri they.think always at.the facts of the others and never at.the thoughts soi. – Nu ssu filu fatti de l’ autri. (Scorrano, Musio 1995: 15) their not are filu facts of the others ‘They’re always interested in other people’s business and never in their own. – But this isn’t other people’s business.’ c. Sˇta scatinàmu la fantasia. – Viti ca no’ è ffilu fantasìa. (Maruggio) prog we.unchain the imagination you.see that not it.is filu imagination ‘We’re letting our imagination run away with itself. – But it’s not our imagination.’ d. Lu malatu iddu ete, filu ieu. (Lec., Protopapa 1992: 55) the patient he is filu I ‘He’s the patient, not me.’ e. Dopu dopu. – Moi, filu dopo. (Scorrano, Musio 1995: 49) after after now filu after ‘Later, later. – Now, not later.’

The presuppositional value of filu can also be seen in examples such as (33a-c) where it marks the denial of a previous implicit presupposition or expectation: (33) a. Me pare ca quistu nu’ sse sta preoccupa me= it.seems that this not self=prog worries filu. (Lec., Protopapa 1990: 36) filu ‘He doesn’t look worried to me after all/despite expectations.’

118

Adam Ledgeway

b. llu farmacìsˇta tissi ca no’ era filu còrpa sua ci èrumu the pharmacist said that not it.was filu fault his if we.were sbaijàtu. (Maruggio) wrong ‘The pharmacist said that in any case it wasn’t his fault if we had made a mistake.’ c. È dittu ca nun c’ olenu filu muti sordi cu ziccamu. (Cellino San Marco) is said that not there= want filu much money that we.start ‘He said that we don’t need much money after all to get things going.’

The presuppositional use of filu also proves very frequent in polar questions where the speaker anticipates a particular answer from which particular conclusions or consequences emerge (roughly corresponding to Eng. By any chance…? Because if you do/did…): (34) a. Mugghièrema sai filu a ddu stae? (Lec., Protopapa 1990: 32) wife=my you.know filu where she.stands ‘By the way, do you happen to know where my wife is?’ b. Nu nc’ è ffilu nna particedda pe mmie? (Lec., Protopapa 1990: 193) not there= is filu a part.dim for me ‘There isn’t by any chance a little part for me, is there? c. E nu’ te toli filu? (Matino, Fiorentino 1998: 11) and not you= it.aches filu ‘But does it not hurt you?’

On a par with the presuppositional use of Italian mica studied by Cinque (1976) where the negated proposition must be ‘discourse-old’ and ‘contextually activated’, Salentino filu is also excluded from those sentences where the conversational implicature is not amenable to negation, including wh-questions, certain types of embedded clause (e. g. infinitival complements) and restrictive relative clauses. Consequently, our corpus does not offer any examples of filu in such contexts. From a comparative perspective, it is interesting to note that, unlike northern Calabrian where the same lexical item, viz. mancu, marked both the presuppositional and the scalar intensive readings, Salentino filu marks the former reading whereas the scalar intensive negative reading is typically marked by mancu, yielding the lexical distinction filu vs mancu. (35) a. E nu’ parli mancu? (Lec., Protopapa 1990: 91) and not you.talk not.even ‘And you are not even going to talk?’ b. Jeu nu’ lu visciu mancu. (Matino, Fiorentino 1998: 55) I not him= I.see not.even ‘I can’t even see him.’ c. E mmancu tuni mi cunvìnci. (Maruggio) and not.even you me= convince ‘And even you can’t convince me.’

Marking presuppositional negation in the dialects of southern Italy

119

d. Nu face mancu niente cu lu scunde. (Cellino San Marco) not he.does not.even nothing that it= he.hides ‘He doesn’t even do anything to hide it.’

However, just as we saw with northern Calabrian mancu (cf. 14a-b), in certain cases the presuppositional value of Salentino filu may be extended to include the scalar intensive interpretation (‘after all/in any case…not’ > ‘not at all/not even’), as illustrated by examples such as (36a-c). Such scalar readings are even clearer in examples such as (36d-e) where filu is doubled. (36) a. Nu’ lli sta ccanùsci […]? – Nu’ lli sta ccanuscìa not them= prog you.know not them= prog I.knew filu. (Lec., Protopapa 1990: 14) filu ‘Don’t you recogise them? – I didn’t recognise them at all.’ b. Su’ ppicciccu, me stancu – Nu’ ssi ffilu picciccu. Tieni I.am little me= I.tire not you.are filu little you.have ticiott’ anni. (Lec., Protopapa 1990: 18) 18 years ‘I’m a child, I get tired […] – You’re not a child at all, you’re 18 years old.’ c. me pinzava ca stivi distratta! – No, no… ste cose nu stau filu me= I.thought that you.were distracted no no these things not I.am filu distratta! (Scorrano, Musio 1995: 106) distracted ‘I thought you were distracted! – No, no, I’m not at all distracted by things like these!’ d. Timu cu nnu’ ssia la mentimu intru li uài… A mmie quistu affare nu’ I.fear perhaps her= we.place in the troubles to me this matter not mme piace filu filu. (Lec., Protopapa 1990: 10) me= pleases filu filu ‘I fear we might be putting her in a difficult situation…I don’t like this story one bit.’ e. Ci va te ssetti susu idhu nu’ te vene filu filu ncapu ca lu poti if fut you= seat up it not you= comes filu filu in.head that it= you.can fare? (Matino, Fiorentino 1998: 116) do.inf ‘If you go and sit up there, surely does it not enter your head at all that you can do it?’

In light of this overlap between the scalar intensive and presuppositional negative readings seen here for Salentino filu, not to mention the dual value of northern Calabrian mancu, it is legitimate to ask whether we are dealing with two emphatic negations or just one. As already noted in §1, Garzonio & Poletto (2010a: §3.1) claim that there are two classes of emphatic negative reinforcer, a scalar intensive negative adverb with the meaning ‘not at all/not even’ and a presuppositional negative adverb with the meaning ‘after all/in any case…not’. By contrast, Kiparsky & Condoravdi (2006: 7) identify three functions of emphatic negation: “The first function of emphatic negation is to mark contradiction of a (possibly implicit) assertion. […]; A second function of emphatic negation is to deny a presumption

120

Adam Ledgeway

or an expectation. […] Third, emphatic negation strengthens a negative assertion by lifting contextual restrictions on an indefinite in the scope of negation or by forcing a ‘totality’ reading on a definite argument of a gradable predicate.” [= ‘at all’; AL.].

However, the first and second functions they identify fall together into the single presuppositional value that we have seen lexicalized by preverbal mancu in northern Calabrian and Salentino filu, whereas the third function is clearly that which we have been calling the scalar intensive value (viz. ‘not at all/not even’). Consequently, on the basis of Garzonio & Poletto (2010a) and Kiparsky & Condoravdi (2006) we can conclude that emphatic negation broadly divides into two types associated with the presuppositional and the scalar intensive readings. According to Cinque (1999: 8), all emphatic negators occur in the specifier of Neg1P (cf. Neg2P in Zanuttini 1997 and Garzonio & Poletto 2010a), hence a position variously lexicalized in Italian, for example, by mica ‘after all…not’, nemmeno/neppure/neanche ‘not even’ and affatto/proprio ‘not… at all’. However, as we have seen, not all languages necessarily mark the distinction lexically between different types of emphatic negation. For instance, we have observed how in northern Calabrian the same lexical item (viz. mancu) can license all these meanings, and similarly we have just noted similar pragmatic extensions of presuppositional filu in Salentino to include the scalar intensive interpretation (cf. 36a-e). Nonetheless, there arises a potential problem with this assumption of a single position for all emphatic negators which concerns the behaviour of Salentino propriu ‘not at all, simply not’: (37) a. Lu liettu nun c’ è ppropriu. (Lec., Protopapa 1990: 22) the bed not there is at.all ‘There’s no bed at all.’ b. nu sacciu propriu ce aggiu fare! (Scorrano, Musio 1995: 45) not I.know at.all what I.have do.inf ‘I simply don’t know at all what I’m supposed to do!’

If, however, following Cinque (1999: 8), we take Salentino propriu to lexicalize SpecNeg1P on a par with its Italian equivalent proprio, then it is not clear how the sequence propriu + filu (38a-c) is derived since both negators should be competing for the same position. Indeed, direct translations of this combination in Italian (39a) and northern Calabrian (39b) with proprio/propriu and mica/mancu prove entirely ungrammatical. Yet, as the examples in (38) illustrate, the sequence proprio + filu in Salentino receives a transparently compositional reading ‘after all/in any case (filu)…really not/not at all (propriu)’.

Marking presuppositional negation in the dialects of southern Italy

121

(38) a. Dovresti fingere di essere il padre di Nanà e intrattenere i you.should pretend.inf of be.inf the father of Nanà and entertain.inf the concorrenti.– No, no. Me tespiace, ma a mmie stu schersu nu mme contestants no no me= it.displeases but to me this joke not me= piace propriu filu. (Lec., Protopapa 1990: 150) pleases at.all filu ‘You should pretend to be Nanà’s father and entertain the contestants. – No, no. I’m sorry, but(, despite what you might think,) I don’t like this messing around at all.’ b. Tie spietti cu tturnamu. – Ieu nu nci spettu propriu filu. you wait that we.return I not there wait at.all filu ‘You’re expecting us to come back. – On the contrary, I really don’t expect anything at all.’ (Lec., Protopapa 1991: 53) c. Attentu cu nnu ccati te lu tronu! […] – Ieu nu nci tegnu propiu careful that not you.fall from the throne I not there= hold at.all filu cu ffazzu la recina. (Lec., Protopapa 1991: 137) filu that I.do the queen ‘Watch that you don’t fall from the throne! […] In any case I’m not at all keen on playing the part of the queen.’ (39) a. *Ma but b. *Ma but

questo scherzo non mi piace proprio/affatto mica. (It.) this joke not me= pleases at.all mica ssu scherzu um mi mancu (piacia) propiu piacia. (Cal.) this joke not me= mancu pleases at.all pleases

Somewhat puzzling is the observation that in these same dialects the opposite order filu + propriu is also attested (40a-c), a sequence which once again licenses the same compositional reading as in (38a-c) and raises the same theoretical problems for a single SpecNeg1P. (40) a. Lui mi ha insultato. – None, nu’ n’ imu capiti filu he me= has insulted no not we= have understood filu propriu. (Matino, Fiorentino 1998: 93) at.all ‘He insulted me. – No, we really haven’t understood one another at all.’ b. Lustig vuol dire divertente e tie nu si divertente filu propriu. (Lec.) lustig wants say.inf fun and you not are fun filu at.all ‘“Lustig” means “fun” and you’re certainly no fun at all.’ c. In pigiama culla freve poi non godi filu propriu. (Lec.) in pyjamas with.the fever then not you.enjoy filu at.all ‘But in your pyjamas with a raging fever you certainly won’t be having a good time at all.’

3.2

Florentine punto: A short excursus

At this point, it is instructive to consider Salentino propriu in relation to Florentine punto ‘not at all’ recently studied by Garzonio (2008a: 14, 131–132; 2008b: 27–29, 36–37). He argues that punto does not lexicalize SpecNeg1P, but must

122

Adam Ledgeway

occur in a lower position since it follows both più ‘anymore’ (41a) and mai ‘never’ (41b) which lexicalize SpecAspPterminative and SpecAspPperfect, respectively. Furthermore, he notes that punto proves incompatible with già ‘already’ and tutto ‘everything’ (41c) – SpecTPAnterior and SpecAspPPlCompletive – and, like già but unlike tutto (41d), precedes the active past participle (cf. 41a-b). Considered together, these observations allow Garzonio to conclude that Florentine punto must occupy a low position in the LAS but higher than that lexicalized by tutto, giving rise to the cartographic representation in (41e) where we have informally labelled the functional projection targeted by punto as NegPCompletive. (41) a. Un m’ è (*punto) più punto piaciuto. (Flo.) not me= is at.all anymore at.all pleased ‘I no longer liked it all.’ b. Un m’ è (*punto) mai punto piaciuto. (Flo.) not me= is at.all never at.all pleased ‘I never liked it at all.’ (*tutto). (Flo.) c. Un ho punto (*già) capito not I.have at.all already understood everything ‘I haven’t (already) understood (everything) at all.’ d. Un ho già / *tutto capito tutto. (Flo.) not I.hav already everything understood everything ‘I haven’t understood everything.’ e. [HAS… [Neg1Presup. [*già TAnterior [più AspTerminative [mai AspPerfect [NegCompletive punto [*tutto AspPlCompletive … [v-VP…]]]]]]]]

Garzonio (2008b: 29–30) also highlights a revealing comparison between the Florentine facts and the corresponding Italian evidence. In particular, he notes that, under one of its uses and in accordance with Cinque’s (1999) predictions, Italian affatto ‘not at all’ precedes già ‘already’, più ‘anymore’ and tutto ‘everything’ (42a-c). At the same time, he also identifies a lower instance of affatto ‘not at all’ which, exactly like Florentine punto, follows più (43b) but proves incompatible with già and tutto (43a,c). visto. (It.) (42) a. Non l’ ho affatto già not it= I.have at.all already seen ‘I haven’t already seen it at all.’ b. Non l’ ho affatto più visto. (It.) not it= I.have at.all anymore seen ‘I haven’t at all seen it since.’ c. Non l’ ho affatto preparato tutto. (It.) not it= I.have at.all prepared everything ‘I haven’t by any means prepared it all.’

Marking presuppositional negation in the dialects of southern Italy

(43) a. *Non not b. Non not c. *Non not

l’ it= l’ it= l’ it=

ho I.have ho I.have ho I.have

già already più anymore visto seen

affatto at.all affatto at.all tutto everything

123

visto (It.) seen visto (It.) seen affatto (It.) at.all

These distributional facts lead Garzonio to conclude that in Italian there are two types of affatto, a higher presuppositional negator (henceforth affatto1) and a lower negative scalar intensitivity adverb (henceforth affatto2). I take the former to lexicalize SpecNeg1P and the latter the specifier of the negative completive projection introduced above, as sketched in (44). (44) [HAS… [Neg1Presup. affatto1 [già TAnterior [più AspTerminative [mai AspPerfect [NegCompletive affatto2/punto [tutto AspPlCompletive … [v-VP…]]]]]]]]

Thus, although both instances of affatto can be broadly translated as ‘not at all’, we note that affatto1 and the associated higher position typically license the presuppositional interpretation (‘(in any case) not at all’), whereas affatto2 and the lower position emphasize the scalar intensive reading (‘(really) not at all’). This claim finds support in our previous observations about northern Calabrian mancu (14a-b) and Salentino filu (36a-e) where we also saw how in many cases the presuppositional and scalar intensive negative readings may fall together.

3.3

Salentino filu revisited

In light of these observations regarding Italian and Florentine, let us now return to the Salentino sequences filu propriu and propriu filu. In the former case, we take filu to lexicalize the typical presuppositional negation position SpecNeg1P (cf. affatto1) and propriu to lexicalize the lower position SpecNegPCompletive (cf. affatto2, punto), as exemplified in (45a-b). (45) a. None, nu’ n’ imu capiti filu propriu. (Matino) no not we= have understood filu at.all b. [HAS…nu’ n’imu capiti [Neg1Presup. filu [TAnterior… [NegCompletive propriu [AspPlCompletive… capiti]]]]]

By contrast, in the case of the opposite order propriu filu we take this to be a derived order obtained through raising of propriu to the specifier of filu (46a-b), producing a complex constituent formed through adverb incorporation (viz. [Neg1Presup. [Spec [Spec propriu] filu]]).

124

Adam Ledgeway

(46) a. Nu’ mme piace propriu filu. (Lec.) not me= it.pleases at.all filu b. [HAS…Nu’ mme piace [Neg1PPresup. [Spec propriu] filu [NegPComp. propriu…piace]]]

Though this might seem like an ad hoc solution to the Salentino facts, adverb incorporation of this type has already been independently proposed for other cases of unexpected adverb ordering (Cinque 1999: 9–10; Ledgeway 2009: 696– 697; in press: §2.1.1.2). For example, whereas in French incorporation of jamais ‘(n)ever’ into plus ‘anymore’ is optional (47a), in Italian the equivalent adverb mai obligatorily raises to the specifier of più ‘anymore’ (47b). A similar superficial distortion of underlying patterns is found in French and Italian with encore/ ancora ‘yet’ and pas/mica ‘not’ (48a-b), as well as in many southern regional varieties of Italian where, under the influence of the corresponding dialects (cf. Ledgeway 2009: 696–697), the standard orders più ‘anymore’ + niente ‘nothing’ and più + negative quantifiers nessuno ‘not one’ and manco ‘not even’ are frequently reversed (cf. 48c), suggesting that incorporation into the specifier of più in these varieties proves more liberal than in the standard (namely, più + niente/ nessuno/manco ⇒ [[niente/nessuno/manco] più]). Ultimately, though, all such incorporations are licensed by the properties of individual lexical items, rather than the structural positions they occupy, since different adverbs of the same class do not generally permit the same operation within the same language or across closely related languages, witness the ill-formedness of incorporation of Italian ancora ‘still’ (in contrast to più ‘anymore’) into affatto ‘not at all’ (viz. *[[ancora] affatto]). (47) a. Il n’ a [AspPTerm [Spec ([ jamais]) he not has never b. Non ha [AspPTerm [Spec [mai] not he.has never ‘He’s never won again since.’

plus] [AspPPerfect ([Spec anymore più] [AspPPerfect [Spec anymore

jamais]) never mai] never

(48) a. Je n’ ai [Neg1P [Spec ([encore]) pas] [TPAnterior ([Spec encore]) I not have yet not yet b. Non ho [Neg1P ([Spec [ancora]) mica] [TPAnterior ([Spec ancora]) not I.have yet not yet ‘(In any case) I haven’t yet understood.’ c. Non me ne importa [AspPTerm [Spec ([niente]) più] not me= thereof= it.matters nothing anymore ([AspPPlCompl. niente.])] (sth. It.) nothing ‘I couldn’t care less.’

[gagné.]]] (Fr.) won [vinto.]]] (It.) won

[compris.]]] (Fr.) understood [capito.]]] (It.) understood

If we now compare Salentino and Italian, we see that whereas Salentino readily allows incorporation of propriu into filu, viz. ⇒ [[propriu] filu] (cf. 38a-c), the equivalent incorporation is ruled out in Italian, viz. ⇒ *[[proprio/affatto2] mica]

Marking presuppositional negation in the dialects of southern Italy

125

(cf. 39a). By contrast, when propriu/proprio are left in their base position, both the Salentino sequence [filu…propriu] (cf. 40a-c) and its Italian equivalent [mica…affatto2] (49a-c) prove grammatical. (49) a. Non mi dispiace [Neg1P [Spec mica] [NegPCompletive [Spec affatto.]] (It.) not me= it.displeases mica at.all ‘In any case, I’m not at all sorry.’ b. Ma al di là di questo non è [Neg1P [Spec mica] [NegPCompletive [Spec affatto]] male! (It.) but beyond this not it.is mica at.all bad ‘But other than that it’s really not bad at all!’ c. A però… non scherzate [Neg1P [Spec mica] [NegPCompletive [Spec affatto]]! (It.) ah but not you.joke mica at.all ‘But on the other hand…don’t joke about it at all!’

4.

Calabrian mancu revisited

In light of our analysis of Salentino filu and propriu and, in turn, the two types of Italian affatto, it is tempting to revisit northern Calabrian to see whether the same analysis can be applied to mancu. On this view, under the scalar intensive reading (50a) mancu would lexicalize the lower SpecNegPCompletive position on a par with Salentino propriu (and Florentine punto, Italian affatto2), but would lexicalize SpecNeg1P when used as a presuppositional negator (50b) on a par with Salentino filu (and Italian affatto1). Accordingly, verb movement is predicted to be identical in both cases, but with the two values distinguished by mancu lexicalizing two distinct positions within the LAS on either side of the finite verb, as sketched in (51). (50) a. Un mi canuscia mancu. (NCal.) not me= he.knows mancu ‘He doesn’t even know me.’ b. Un mi mancu canuscia. (NCal.) not me= mancu he.knows ‘After all, he doesn’t know me.’ (51) [HAS… [Neg1presup. mancu1 [ggià TAnterior [cchiù AspTerminative [mai AspPerfect… VFinite [NegCompletive mancu2 [tuttu AspPlCompletive … [v-VP…]]]]]]]]

If this analysis of the northern Calabrian facts is correct, then it makes a number of empirical predictions that we can test directly. First, analogously to the Salentino discontinuous sequence filu…propriu (cf. 45a-b), we should a priori expect both mancu adverbs to be able to co-occur, contrary to fact (52a; cf. also Penello & Pescarini 2008: 49–50). Second, on a par with Salentino [[propriu] filu] (cf. 46a-b), it is legitimate to expect mancu under the scalar intensive reading to be able to raise and incorporate into mancu qua presuppositional negator

126

Adam Ledgeway

yielding the complex constituent [[mancu] mancu], again contrary to fact (52b).6 Third, under the scalar intensive reading we would expect mancu from its position in SpecNegPCompletive to follow higher pre-VP adverbs such as ggià ‘already’, mai ‘never’ and cchiù ‘anymore’. However, once again this prediction is not borne out, insofar as only the reverse order is attested (52c).7 Finally, given what we saw above for Florentine punto and Italian affatto2, we should a priori expect mancu under its scalar intensive reading to prove incompatible with tuttu ‘everything’, but once again this prediction is not borne out (52d). (52) a. *Un si mancu parranu cchiù mancu. (NCal.) not selves= mancu they.speak anymore not.even ‘After all they don’t even speak to each other anymore.’ b. *Un si [[mancu] mancu] parranu cchiù mancu. (NCal.) not selves= not.even mancu they.speak anymore (*ggià/*cchiù/*mai) mancu ggià/cchiù/mai pagatu. (NCal.) c. Unn’ a not he.has already/anymore/ever not.even already/anymore/ever paid ‘He hasn’t even already paid / even paid anymore / even ever paid.’ d. Unn’ aju capitu mancu tuttu. (Cal.) not I.have understood not.even everything ‘I didn’t even understand everything.’

As the examples in (52a-d) unambiguously show, this lexical approach to northern Calabrian mancu is simply not empirically supported, leading us to reaffirm that the distinction between the presuppositional and scalar intensive readings is syntactically marked by verb movement as outlined in §2.1.1.

5.

Conclusion

We have seen that, despite a lack of previous interest, there are significant things to be said about negation in the dialects of southern Italy. In particular, the varieties under investigation here have allowed us to establish a broader typology of marking of emphatic presuppositional negation in southern Italy, for which we have identified three distinct strategies: 6 This is all the more surprising since, as illustrated in Ledgeway (2009: 696–697; in press: §2.1.1.2), mancu/manco is one of those adverbs in southern Italy which does otherwise allow adverb incorporation, witness the contrast between (i.a) and (i.b): (i) a. Un ni mancu (chiama) cchiù chiama. not us= mancu he.calls no.longer he.calls b. Un ni [[cchiù] mancu] chiama. not us= anymore mancu he.calls ‘In any case he no longer call us anymore.’ 7 Note that the incorporated order [[cchiù] mancu] illustrated in footnote 6 above is not a counterexample, since the latter order is only possible under the presuppositional reading of mancu, not under its scalar intensive reading.

Marking presuppositional negation in the dialects of southern Italy

127

(53) a. Grammaticalization of cleft: Sicilian neca (Garzonio & Poletto 2010b) b. V-movement wrt Calabrian mancu: (V)…[Neg1Presup. mancu…(V) [v-VP V]] c. Grammaticalization of minimizer: Salentino (nu’…) filu ‘thread’ > ‘after all not (at all)’

In terms of emphatic negation more generally, we have seen that the distinction between negative presuppositional readings (‘not…after all’) and negative scalar intensive readings (‘not at all/not even’) is not discrete, but more fluid than traditionally thought. More specifically, we have witnessed how across different varieties scalar intensive negators can frequently license presuppositional readings and vice versa. A case in point are the dialects of northern Calabria where it was shown that the same lexical item (viz. mancu) which marks scalar intensive readings also functions as the marker of presuppositional negation, with the two readings mediated in most cases by the differential extent of verb movement (to a clause-medial position immediately above the LAS in the former case and to a lower position within the LAS in the latter case). This is not accidental, but highlights a more general tendency for both types of emphatic negation to be formally related. Whereas in northern Calabrian they are lexically related, in other varieties they may, following Garzonio (2008a,b), be derivationally related by the activation of a lower negative completive aspectual position (SpecNegPCompletive) alongside the higher (canonical) emphatic negation position (SpecNeg1P). This is the case in Salentino where both types of negation are spelt out by two distinct positions that can be simultaneously lexicalized filu…propriu (cf. It. mica…affatto), giving rise to a compositional emphatic negative reading ‘after all (filu)…not at all (propriu)’. However, we have also seen that the two positions may be derivationally related by overt raising of the scalar intensive negator to incorporate into the presuppositional negator, namely [[propriu] filu]. Given such movements, it is tempting to view the presuppositional readings often licensed by scalar intensive negators in isolation (cf. 14a-b) and, by the same token, the scalar intensive readings frequently licensed by presuppositional negators in isolation (36a-e) as the result of raising of the relevant negator from SpecNegPCompletive to SpecNeg1P, thereby licensing the observed compositional reading.

128

Adam Ledgeway

References Bernini, Giuliano (1992): Forme concorrenti di negazione in italiano. In Moretti, Bruno, Dario Petrini & Sandro Bianconi (eds.): Linee di tendenza dell’italiano contemporaneo. Atti del XXIV Congresso della Società di Linguistica Italiana. Rome: Bulzoni, 191–215. Bernini, Giuliano & Paolo Ramat (1996): Negative Sentences in the Languages of Europe. A Typological Approach. Berlin: de Gruyter. Cinque, Guglielmo (1976): Mica. Annali della Facoltà di Lettere e Filosofia dell’Università di Padova 1: 101–112. Cinque, Guglielmo (1999): Adverbs and Functional Heads. A Cross-linguistic Perspective. Oxford: Oxford University Press. Cruschina, Silvio (2010): Aspetti morfologici e sintattici degli avverbi in siciliano. Quaderni di lavoro dell’ASIt 11: 19–39. Cruschina, Silvio & Adam Ledgeway (2016): Structure of the Clause. In Ledgeway, Adam & Martin Maiden (eds.): The Oxford Guide to the Romance Languages. Oxford: Oxford University Press, 556–574. Dahl, Östen (2001): Inflationary Effects in Language and Elsewhere. In Bybee, Joan & Paul Hopper (eds.): Frequency and the Emergence of Linguistic Structure. Amsterdam: Benjmanins, 471–480. Damonte, Federico (2008): Modo e negazione nei dialetti calabresi meridionali. Quaderni di lavoro dell’ASIt 8: 85–100. Egerland, Verner (2010): On the Interaction between aspect and negation in dialectal Italian. Quaderni di lavoro dell’ASIt 10: 1–17. Fiorentino, Fernando (1998): Vernacolo salentino. Galatina: Congedo Editore. Garzonio, Jacopo (2008a): A case of incomplete Jespersen’s Cycle in Romance. Rivista di grammatica generativa 33: 117–135. Garzonio, Jacopo (2008b): Diacronia e sincronia della sintassi di punto in fiorentino. Quaderni di lavoro dell’ASIt 8: 21–41. Garzonio, Jacopo & Cecilia Poletto (2010a): Quantifiers as Negative Markers in Italian Dialects. In Van Croenenbroeck, Jeroen (ed.): Linguistics Variation Yearbook 2009. Amsterdam: John Benjamins, 127–152. Garzonio, Jacopo & Cecilia Poletto (2010b): Alcuni fenomeni di negazione nei dialetti siciliani. Quaderni di lavoro dell’ASIt 11: 69–82. Gyuris, Beáta (2009): Quantificational contrastive topics with verum/falsum focus. Lingua 199 (4): 625–649. Hualde, José Ignacio (1992): Catalan. London: Routledge. Jespersen, Otto (1917): Negation in English and Other Languages. Copenhagen: Høst. Kiparsky, Paul & Cleo Condoravdi (2006): Tracking Jespersen’s Cycle. In Janse, Mark, Brian Joseph & Angela and Ralli (eds.): Proceedings of the Second International Conference of Modern Greek Dialects and Linguistic Theory. Patras: University of Patras, 172–197. Ledgeway, Adam (2009): Grammatica diacronica del napoletano. Tübingen: Max Niemeyer Verlag (Beihefte zur Zeitschrift für romanische Philologie Band 350).

Marking presuppositional negation in the dialects of southern Italy

129

Ledgeway, Adam (2012): From Latin to Romance. Morphosyntactic Typology and Change. Oxford: Oxford University Press. Ledgeway, Adam (in press): The Verb Phrase. In Longobardi, Giuseppe (ed.): The Syntax of Italian. Cambridge: Cambridge University Press. Ledgeway, Adam & Alessandra Lombardi (2005): Verb movement, adverbs and clitic positions in Romance. Probus 17: 79–113. Ledgeway, Adam & Alessandra Lombardi (2014): The Development of the Southern Subjunctive: Morphological Loss and Syntactic Gain. In Benincà, Paola, Adam Ledgeway & Nigel Vincent (eds.): Diachrony and Dialects. Grammatical Change in the Dialects of Italy. Oxford: Oxford University Press, 25–47. Mancarella, Giovan Battista (ed.) (1998): Salento. Monografia regionale della ”Carta dei Dialetti Italiani”. Lecce: Edizioni del Grifo. Manzini, Maria Rita & Leonardo Savoia (1998): The Position of Clitic and Adverbial Negation in Italian Varieties. In Benincà, Paola & Cecilia Poletto (eds.): Atti della terza giornata italo-americana di dialettologia: La negazione nelle lingue romanze. Quaderni di lavoro dell’ASIt 2: 39–61. Manzini, Maria Rita & Leonardo Savoia (2002): Negative Adverbs are neither Adv nor Neg. In Hirotani, Masako (ed.): Proceedings of NELS 32. Amherst: University of Massachusetts, 327–346. Manzini, Maria Rita & Leonardo Savoia (2005): I dialetti italiani e romanci. Morfosintassi generativa. Vol. 3. Alessandria: Edizioni dell’Orso. Manzini, Maria Rita & Leonardo Savoia (2011): Sentential negation in Piedmontese dialects. Quaderni di lavoro dell’ASIt 13: 135–159. Musio, Corrado (1995): Il Salento in vernacolo. Galatina: Congedo Editore. Parry, Mair (1997a): Negation. In Maiden, Martin & Mair Parry (eds): The Dialects of Italy. London: Routeldge, 179–185. Parry, Mair (1997b): Preverbal negation and clitic ordering, with particular reference to a group of north-west Italian dialects. Zeitschrift für romanische Philologie 113: 244–270. Parry, Mair (1998): On Negation in the Ligurian Hinterland. In Benincà, Paola & Cecilia Poletto (eds.): Atti della terza giornata italo-americana di dialettologia: La negazione nelle lingue romanze, Quaderni di lavoro dell’ASIt 2: 85–114. Parry, Mair (2013): Negation in the History of Italo-Romance. In Willis, David, Christopher Lucas & Anne Breitbarth (eds.): The History of Negation in the Languages of Europe and the Mediterranean. Oxford: Oxford Uiversity Press, 77–118. Patruno, Barbara & Laura Sgarioto (2004): Osservazioni sull’imperative negativo in alcune varietà del ladino centrale. Quaderni di lavoro dell’ASIt 4: 105–137. Penello, Nicoletta (2005): Aspetto e negazione: l’elemento ‘altro’ in Veneto. Quaderni di lavoro dell’ASIt 5: 1–19. Penello, Nicoletta & Diego Pescarini (2008): Osservazioni su mica in italiano e alcuni dialetti veneti. Quaderni di lavoro dell’ASIt 8: 43–56. Pescarini Diego (2005): Mica nell’area metropolitana di Verona. In Marcato, Gianna (ed.): Dialetti in città. Padua: Unipress, 283–288. Poletto, Cecilia (2008): On negative doubling. Quaderni di lavoro dell’ASIt 8: 57–84. Poletto, Cecilia (2009): I quantificatori e la negazione nei dialetti calabresi del progetto ASIt. Quaderni di lavoro dell’ASIt 9: 25–36.

130

Adam Ledgeway

Poletto, Cecilia (2010): The syntax of focus negation. Quaderni di lavoro dell’ASIt 10: 39–62. Poletto, Cecilia (2016): Negation. In Ledgeway, Adam & Martin Maiden (eds.): The Oxford Guide to the Romance Languages. Oxford: Oxford University Press, 833–846. Presicce, Giuseppe (ed.) (n.d.): Vocabolario salentino. Repertorio del dialetto scorranese. www.dialettosalentino.it/home.html. Protopapa, Raffaele (1990): Teatro dialettale leccese. Volume secondo. L’acchaitura. La riforma. Eulalia. Le mbrogghie te lu requenzinu. Le corne d’argentu. Cavallino di Lecce: Capone. Protopapa, Raffaele (1991): Teatro dialettale leccese. Volume terzo. L’àncura. La furtuna. Lu rre te Lecce. Filippu e ppanaru. Nna causa alla pretura. Cavallino di Lecce: Capone. Protopapa, Raffaele (1992): Teatro dialettale leccese. Volume quarto. Li do’ cumpari. Le do’ cummari. Salvatore lu rapinatore. Epigrammi. Cavallino di Lecce: Capone. Rohlfs, Gerhard (1956): Vocabolario dei dialetti salentini (Terra d’Otranto). Vol. 1, A–M. Munich: Verlag der bayerischen Akademie der Wissenschaften. Rohlfs, Gerhard (1969): Grammatica storica della lingua italiana e dei suoi dialetti. Sintassi e formazione delle parole. Turin: Einaudi. Romano, Antonio (2009): Vocabolario del dialetto di Parabita. Lecce: Grifo. Romero, Maribel & Chung-hye Han (2004): On negative yes/no questions. Linguistics and Philosophy 27: 609–658. Schwegler, Armin (1990): Analyticity and Syntheticity. A Diachronic Perspective with Special Reference to Romance Languages. Berlin/New York: de Gruyter. Tekavcˇic´, Pavao (1980): Grammatica storica della lingua italiana. Vol. 2. Morfosintassi. Bologna: Il Mulino. Vai, Massimo (1996): Per una storia della negazione in Milanese in comparazione con altre varietà altoitaliane. ACME 40 (1): 57–98. Wheeler, Max W., Alan Yates & Nicolau Dols (1999): Catalan: A Comprehensive Grammar. London: Routledge. Willis, David, Christopher Lucas & Anne Breitbarth (2013): Comparing Diachronies of Negation. In Willis, David, Christopher Lucas & Anne Breitbarth (eds.): The History of Negation in the Languages of Europe and the Mediterranean. Oxford: Oxford Uiversity Press, 1–50. Yates, Alan (1984): Catalan. London: Hodder and Stoughton. Zanuttini, Raffaella (1997): Negation and Clausal Structure. A Comparative Study of Romance Languages. New York: Oxford University Press. Zeijlstra, Hedde (2004): Sentential Negation and Negative Concord. Utrecht: LOT Publications.

Maria Barouni

Challenging the strict vs. non-strict distinction of Negative Concord: A syntactic proposal*

Abstract: This paper focuses on a class of n-words in Greek which so far has been neglected in the discussion of Negative Concord (NC henceforth) and which displays properties similar to the elements of non-strict Negative Concord Languages (NCLs henceforth). I show that, in its current form, one prevalent syntactic theory, that of Zeijlstra (2004, 2008a,b), cannot account for these facts. For this reason, I propose a new analysis, based on Zeijlstra (2004, 2008a,b), but deviating from it in crucial assumptions. More precisely, I claim that the [NEG] feature that distinguishes strict from non-strict NC languages is not semantic, but formal in nature. I assume that a covert semantic operator introducing negation is always present at Nego bearing [uNEG]. An element with [iNEG] has to check its [uNEG] feature and possible candidates to check this [uNEG] of the silent operator are negative markers (NMs) of both strict and nonstrict NCLs which are argued to carry interpretable features [iNEG], as well as preverbal n-words with [iNEG]. A theory with these assumptions, which are empirically motivated, can adequately account for ‘hybrid’ languages, such as Greek and Hungarian. Overall, the current analysis supports the view that there is no clear distinction between strict and non-strict NCLs, in that any strict NCL may have n-words with non-strict NC properties.

* I would like to particularly thank Winfried Lechner, Elena Anagnostopoulou, and Katharina Hartmann for their invaluable comments, as well as two anonymous reviewers. All errors are mine.

132

Maria Barouni

1.

Introduction

1.1

Negative Concord in Greek1

NCLs fall into two groups, strict and non-strict NCLs (after Giannakidou 2000). These two types of languages differ in the distribution of NMs (Negative Markers) with preverbal negative words (n-words) (terminology adopted from Laka 1990). While strict NCLs require the presence of the NM both in preverbal and postverbal position (1–2), non-strict NCLs exclude the NM with preverbal nwords, but require it when n-words appear postverbally (3–4): Strict NCL, Czech (1) Dnes *(ne)volá nikdo. today NM.call n-body ‘Today nobody is calling.’ (2) Dnes nikdo *(ne)volá. today n-body NM.call ‘Today nobody is calling.’ (Biberauer & Zeijlstra (B&Z henceforth) 2012: 347) Non-strict NCL, Italian (3) Gianni *(non) ha telefonato a nessuno. Gianni NM has called to n-body ‘Gianni didn’t call anybody.’ (4) Ieri nessuno (*non) ha telefonato (a nessuno). yesterday n-body NM has called to n-body ‘Yesterday nobody called (anybody).’ (B&Z 2012: 347)

Greek has been considered a typical strict NCL (Giannakidou 1998, 2000, 2006) as the presence of the NM, dhe(n) 2 in indicative clauses (5), and mi(n) in subjunctive (6), is obligatory both with preverbal and postverbal n-words.3 (5) *(dhen) irthe . n-one NM came n-one ‘No one came.’

Indicative

1 ‘Greek’ always refers to ‘Modern Greek’. The two terms are used interchangeably. 2 Final -n is conditioned by the initial vowel or consonant of the element that appears after the NM. See Barouni (in preparation). 3 Greek n-words have been argued to come in two versions: they can bear emphatic stress or not (Veloudis 1982; Giannakidou 1998 and subsequent work). Emphatics receive a negative meaning, while their non-emphatic counterparts have been analyzed as English style NPI elements (Giannakidou 1993, 1998). In what follows, ‘regular n-words’ will be taken to refer to ‘emphatic’ n-words which always require the presence of the NM.

Challenging the strict vs. non-strict distinction of Negative Concord

(6) na *(min) erthi . n-one PRT NM come n-one ‘No one should come.’

133 Subjunctive

However, I will argue that there is another class of elements that has been ignored in the literature of NC, which suggests that the empirical basis is more complex than previously thought.

1.2

New evidence. Class of ou- elements

This new group to be referred to as ou-class consists of n-words which start with the prefix ou. Ou was the NM for indicative clauses in Ancient Greek, and it was also the prefix of a class of n-words in Ancient Greek with non-strict NC properties (7–8): Ancient Greek (7) oudhis (*ouk) idhe ton Sokrati. n-one NM saw the Socrates ‘No one saw Socrates.’

(modified from Horrocks 2014: 44)

(8) O Sokratis *(ouk) idhen oudhena. the Socrates NM saw n-one ‘Socrates saw no one/didn’t see anyone.’

The ou-class was extensive but only some of its elements have survived to the present. What is important is that ou-elements which are used in Modern Greek still keep their non-strict NC properties. When ou-elements appear preverbally the NM is excluded (9–10), while postverbally the NM is obligatory (11–12). Thus, in Modern Greek, the non-strict ou-class coexists with ‘regular’ n-words, n-words that require the presence of the NM in preverbal position, as it is expected for the n-words of a strict NCL.4 Modern Greek (9) Oudhepote (*dhe) zitise ti gnomi mu. n-ever NM asked the opinion mine ‘Never did he ask my opinion.’ (10) Oudholos (*dhe) me endhiaferi. n-at.all NM clitic interests ‘I am not interested at all.’

4 Giannakidou (1998: 200, footnote 7) refers to oudhis (‘n-one’) as being inherently negative and incompatible with negation, but she considers it formal and limited in use, so she leaves it aside in her analysis.

134

Maria Barouni

(11) *(Dhe) me endhiaferi oudholos. NM clitic interests at.all ‘I am not interested at all.’ (12) *(Dhe) zitise ti gnomi mu oudhepote. NM asked the opinion mine n-ever ‘He never asked my opinion.’

The difference between these two classes extends to the phenomenon of ‘negative spread’ (den Besten 1986). In negative spread environments, the negative feature is distributed among any number of indefinite expressions without there being an NM. N-words of non-strict NCLs typically license negative spread, as seen in the case of Italian (13) and Ancient Greek (14) – in (14) there are even three n-words participating in negative spread. Regular n-words in Modern Greek do not license negative spread (15). By contrast, ou-elements admit negative spread (16), similarly to n-words of non-strict NCLs (13–14). Italian (13) Ieri nessuno (*non) ha telefonato (a nessuno). yesterday n-body NM has called to n-body ‘Yesterday nobody called (anybody).’ Ancient Greek (example modified from Horrocks 2014: 61) (14) phronisin ke noun…oudhois popote out’ ipar out’ onar aishron idhen… wisdom and mind n-one ever neither awake nor asleep disgraceful saw ‘No one ever, either asleep or awake, saw of wisdom and mind…’ (Plato, Philebus 6) Modern Greek (15) Pote *(dhe) zitise tipota. n-ever NM asked n-thing ‘He never asked for anything.’ (16) Oudhepote (*dhe) zitise tipota. n-ever NM asked n-thing ‘He never asked for anything.’

Ou-elements are in competition with regular n-words, with the latter group being preferred in postverbal constructions. Ou-elements are preferred preverbally when the speaker wants to express emphasis. As can be seen in Table 1, there are elements like tipota and puthena (‘n-thing’ and ‘n-where’ respectively) which have no ou-counterparts with non-strict NC properties. Interestingly, and most importantly for present purposes, there are elements like ‘neither/nor’ and ‘even’ which exclusively have an ou manifestation and lack a regular n-word counterpart. The fact that this split is systematic will be seen to constitute crucial evidence that a theory of Greek NC also needs to cover ou-elements (contra Giannakidou (cf. fn.4)). In the next section, I will test whether a prominent

Challenging the strict vs. non-strict distinction of Negative Concord

135

theory of NC – Zeijlstra’s (2004) – can account for ‘hybrid’ languages such as Greek which have both elements with strict and non-strict properties. Table 1. Regular n-words and ou-elements in Modern Greek Regular n-words (strict NC properties)

Ou-elements (non-strict NC properties)

n-body/ n-one masculine feminine neutral n-ever

kanis5 kamia kanena pote

oudhis oudhemia oudhen oudhepote

not at all n-where

katholou pouthena

oudholos ––

n-thing neither/ nor

tipota –– –– –– ––

–– oute oudhe mite midhe

even

––

oute

2.

ZEIJLSTRA’s theory and the class of ou-elements

2.1

ZEIJLSTRA (2004, 2008a,b)6

Zeijlstra (2004, 2008a,b) presents a syntactic theory which is based on the core idea that NC is an instance of multiple agreement (Ura 1996, Hiraiwa 2001 a.o.). In this approach, multiple [uNEG] features are checked by a single element which carries [iNEG]. Feature valuation between interpretable and uninterpretable features takes place in a ‘downward’ direction such that all n-words with [uNEG] features must be in the c-command domain of an element with [iNEG] features. In a more recent paper, Zeijlstra (2012: 493) argues that Agree7 more generally works as specified in (17): 5 Giannakidou (1998) argues that Greek n-words come with special intonation, emphatic stress, which is not related to focus (1998: 227–231). To distinguish them from NPIs she uses uppercase. Since my discussion is limited to n-words, I omit the uppercase notation, but the reader should keep in mind that regular n-words always bear emphatic stress (see also fn. 2). 6 I leave aside the question whether negative elements of Double Negation (DN) languages should be treated as negative quantifiers. For criticism of the quantificational approach of negative elements see Penka (2011). 7 Agree is the operation that establishes a relation between two features of the same kind. An element with an uninterpretable feature, [uF], needs to have its [uF] valued by an element

136

Maria Barouni

(17) α can Agree with β iff a. α carries at least one unvalued and uninterpretable feature and β carries at least one matching interpretable and valued feature b. β c-commands α c. β is the closest goal to α

An obvious question that arises is which elements carry [iNEG] features and which ones are associated with the corresponding [uNEG]. Since multiple nwords can appear in a sentence without introducing semantic negation, Zeijlstra assumes that n-words (in both strict and non-strict NCLs) are semantically non-negative indefinites which bear [uNEG] features. The difference between strict NC and non-strict NCLs is located in the different status of the NMs. NMs in non-strict NCLs carry an [iNEG] feature, while NMs in strict NCLs are marked by [uNEG] (the criteria for the different status of the operator between strict and non-strict NC languages are discussed in section 3). When an NM carries [iNEG], Agree takes place between the [iNEG] of the NM and the [uNEG] of n-words (18– 19). On the other hand, in strict NC languages in which both the NM and the nwords carry [uNEG], a covert semantic operator with [iNEG] has to be assumed (20–21). What is important to keep in mind is that the way the theory is formulated, [NEG] is a semantic feature.8 Elements with [iNEG] introduce semantic negation, while elements with [uNEG] do not. Non-strict NCL, Italian (18) Gianni *(non) ha telefonato a nessuno. Gianni NM has called to n-body ‘Gianni didn’t call anybody.’

(B&Z 2012: 347)

(19) [TP Gianni [NegP non[iNEG] ha telefonato [vP a nessuno[uNEG] ]]] Strict NCL, Czech (20) Dnes nikdo *(ne) 9 volá. today n-body NM.call ‘Today nobody is calling.’

(B&Z 2012: 347)

(21) [Dnes Op¬[iNEG] [TP nikdo [uNEG] nevolα[uNEG]]]

bearing a feature of the same kind which is interpretable, [iF]. The element with [uF] is said to look/probe for a goal, an element with the respective [iF]. 8 Zeijlstra (2014), in a recent paper, distinguishes formal from semantic features and argues that an element with [iF] does not mean that it introduces a semantic feature. Nonetheless, in the analysis of NC [iNEG] always introduces semantic negation. 9 The NM appears as a clitic on the verb.

Challenging the strict vs. non-strict distinction of Negative Concord

2.2

137

The negative operator

Regarding the covert negative operator which is assumed to have elements with an [uNEG] feature, Zeijlstra (2004) argues that it is an adverbial operator which introduces Boolean negation and binds all free variables under existential closure (22). In this way, all free variables introduced in vP or below are existentially closed off within the scope of negation that is introduced by Op¬. Among these variables bound by the negative operator is the event variable e. The event variable has to be bound by the negative operator in order for sentential negation to arise (Acquaviva 1997, Giannakidou 1999). (22) [[Op¬]] = ¬ (∃) (23) Dhen irthe kanenas.10 NM came n-body ‘Nobody came.’ ¬ ∃e, x came’(e, x)

So what is the explanation for NC? NC is the result of, on the one hand, a multiple Agree operation between the elements with [iNEG] features (the covert Op¬ in strict NCLs and the NM in non-strict NCLs) 11, and the n-words with [uNEG] features, on the other hand. Consider first how the theory applies to regular nwords in Greek. In Greek, which is a strict NCL, neither the NM nor the n-words carry [iNEG]. The operator, which carries [iNEG], is located in SpecNegP (25). Agree is subject to locality conditions, so the NM and the n-word have to be in the same phase as the negative operator, or at the phase edge of the lower phase. Therefore, the object has to move to a vP adjoined position, resulting in the syntactic representation in (25). Multiple Agree allows the NM and the n-word to have their negative features checked against the same Op¬ and the [uNEG] features on Nego and tipota (n-thing) are deleted. (24) Dhen idha tipota. NM saw n-thing ‘I didn’t see anything.’

10 See footnote 5. 11 Non-strict NCLs also trigger a covert operator when n-words appear preverbally; since nwords bear [uNEG] and the NM with [iNEG] is excluded in this construction (see (4) above), a covert operator is assumed with [iNEG] (Zeijlstra 2004).

138 (25)

Maria Barouni NegP 3 Op¬ [iNEG]

NegP 3 Nego dhen NM [uNEG]

TP 3 TP 3 idha ‘saw’

vP 3 tipota ‘n-thing’ [uNEG]

2.3

vP 3 VP 3 idha tipota

Applying ZEIJLSTRA’s (2004, 2008a,b) and B&Z’s (2012) theory

First, let us examine whether an analysis along the lines of Zeijlstra (2004, 2008a,b) can also account for the new Greek data introduced in section 2.1. Remember that all speakers require the presence of the NM with regular n-words (26), but exclude the NM with preverbal ou-elements in verb modification12 (27) and that there is no speaker variation regarding these facts. Let us therefore assume that all n-words in Greek have [uNEG], i. e. both regular n-words and ouelements, and that the NM carries [uNEG] just like the NMs in strict NCLs. While successful with regular elements, the analysis does not extend to ou-elements. If both oudhepote and the NM bear [uNEG], why is it the case that the NM is excluded in (27), but required when a regular n-word appears preverbally (26)? (26) Pote *(dhen) tilefonise. n-ever NM called ‘(S)he never called.’ (27) Oudhepote (*dhen) tilefonise. n-ever NM called ‘(S)he never called.’

Greek is not unique in this respect. Hungarian shows the same ‘hybrid’ properties as Greek in that it has two categories of n-words with distinct characteristics.

12 I refer to (27) as verb modification because the associate element of the ou-expression is the verb. As we will see things are more complex when the associate of the ou-element is a DP (cf. (41)).

Challenging the strict vs. non-strict distinction of Negative Concord

139

Therefore, Surányi13 (2006) argues that Hungarian is a ‘hybrid’ NC language with two distinct classes of n-words that display strict NC and non-strict NC properties, respectively. Elements of one class, the s-class (Surányi’s mnemonic for the initial morphological s(e) particle of all n-words), behave like regular nwords of a strict NCL, as seen by the fact that the NM is obligatory: (28) *(Nem) jött el senki. NM come PREF14 nobody ‘Nobody came along.’

(adapted from Surányi 2006: 275)

(29) Senki *(nem) jött el. nobody NM come PREF ‘Nobody came along.’

On the other hand, n-words modified by the element sem, pattern with negative elements of a non-strict NCL. Preverbally they exclude the NM, postverbally they require it: (30) Senki sem (*nem) jött el. nobody SEM NM come PREF ‘Nobody came along.’ (31) *(Nem) jött el senki sem. NM come PREF nobody SEM ‘Nobody came along.’

In sum, it has been shown that the distribution is more complex than generally assumed; even strict NCLs may include n-words with non-strict NC properties. Therefore, the system must be modified to predict and account for such cases. In the next section, I argue that changing some basic assumptions in Zeijlstra’s system will enable us to account for ‘hybrid’ languages such as Greek and Hungarian. More precisely, the next chapter presents arguments against the different status between NMs of strict and non-strict NCLs. I show that the criteria on which this assumption is based are not solid. Then, I lay out which consequences this entails for the analysis of NC.

13 Surányi adopts a quantificational analysis for Hungarian n-words which I am not pursuing here. Importantly, the analysis faces problems in accounting for postverbal sem-elements. For criticism see Barouni (in preparation). 14 El is a verbal prefix (PREF). The verb in negative clauses appears to its left (Surányi 2006).

140

Maria Barouni

3.

Status of the NMs in NCLs

3.1

Criteria for the status of NMs

As pointed out above, Zeijlstra (2004, 2013, a.o.) argues that the difference in behaviour between strict and non-strict n-words is due to the different status of the NMs. The NM in strict NC languages bears [uNEG] and a covert operator taking scope over the clause has to be assumed. In support of the presence of the covert operator in strict NC languages, but not in non-strict ones in which the NM bears [iNEG], Zeijlstra provides evidence from the following:15 (a) differences in readings with quantificational DPs (b) optionality of the NM with preverbal n-words in strict NCLs

In the next sections I argue that the above criteria do not constitute solid evidence for the different status of the NMs between strict and non-strict NC languages. 3.1.1 Different readings with quantificational DPs A first argument for the different status of the NMs in strict NC and non-strict NCLs is related to scope facts. Zeijlstra (2004, 2008a) claims that strict NC and non-strict NCLs induce different scope orders in contexts where an NM is ccommanded by a quantificational DP. More precisely, Zeijlstra observes that in Czech, a strict NCL, the quantificational DP can only be interpreted under the scope of negation (32). The NM is in the surface c-command domain of moc, but semantically it takes scope over moc. According to Zeijlstra, this reading is due to the presence of a covert negative operator with [iNEG] which c-commands moc. If negation were interpreted in its surface position, the reverse reading, in which negation scopes under the quantificational DP (the (b) reading), would be expected to be available:

15 Zeijlstra (2013) shows that the [uNEG] status of the NM in strict NC languages can account for the fact that n-words are licit in elliptical constructions (i) with a negative interpretation. Ellipsis takes place under semantic identity (Merchant 2001), which is problematic if the NM introduces semantic negation. If the NM is taken to bear [uNEG] this problem is circumvented, because semantic identity is preserved. (i) Q: Ti ides? A: TIPOTA *(dhen ida) (Giannakidou 2006) what saw nothing ‘What did you see? Nothing!’ I omitted this criterion from the main discussion because it is not directly related to a comparison between NMs of strict and non-strict NC languages and changes nothing to the argumentation above.

Challenging the strict vs. non-strict distinction of Negative Concord

141

(32) Milan moc nejedl. Czech Milan much NM.eat a. ¬ > much: ‘Milan didn’t eat much.’ b. *much > ¬: ‘There is much that Milan didn’t eat.’

On the other hand, Zeijlstra (2004, 2008a) argues that in non-strict NCLs, negation cannot take semantic scope over the quantificational DP. This is evidence that these languages lack a covert operator, as the only available reading is the one in which the quantificational element molto takes scope over the NM non (the (b) reading) (Zeijlstra 2004:245). The example is from Italian, a non-strict NCL (33): (33) Molto non ha mangiato Gianni. Italian much NM has eaten Gianni a. *¬ > much: ‘Gianni didn’t eat much.’ b. much > ¬: ‘There is much that Gianni didn’t eat.’

What is crucial in this account is that NMs do not reconstruct; the syntactic position of the NM is accordingly identical to its semantic interpretation. Zeijlstra argues, based on the interpretation in (33) that the Italian NM must have [iNEG] features as negation is interpreted below the quantificational element. On the other hand, the wide scope requirement for negation in strict NCLs (32) is evidence for the assumption that the NM is assigned a [uNEG] feature (as the NM is syntactically ccommanded by the quantificational element). In the next section I argue, first, that the description above only covers parts of the data; and second, that some of the predictions of the theory are not confirmed empirically. 3.1.2 Data in conflict: evidence from Greek, Romanian, Italian and Spanish Let us turn to the interaction of negation and quantifiers in strict NCLs. First, contrary to expectations, two strict NCLs that I have tested, Romanian and Greek, admit not only the expected wide scope reading in (34a, 35a), but also the low scope reading in (34b, 35b), where the negative element is interpreted below the quantificational one. Romanian (34) Ion mult n-a mâncat.16 John much NM has eaten a. ¬ > much: ‘John didn’t eat much.’ b. much > ¬: ‘There is much that John didn’t eat.’

16 Thanks to Gianina Iordachioaia for the examples in Romanian.

142

Maria Barouni

Greek (35) Poli dhen efage.17 much NM ate a. ¬ > much: ‘He didn’t eat much.’ b. much > ¬: ‘There is much that he didn’t eat.’

The existence of an ambiguity has also been confirmed experimentally in Greek. Baltazani (2002) has tested and confirmed the availability of both readings in examples like (36), where the quantificational element at surface structure ccommands the NM.18 What is important in (36) is that the syntactic c-command domain and the locus of semantic interpretation do not match. This is of relevance as it is a crucial assumption in Zeijlstra’s theory that the syntactic position of the element that carries negation also marks its scope.19 This follows because feature checking takes place in a ‘downward’ direction and the element with [iNEG] features has to c-command the elements with [uNEG]. But in (37) we see that negation can also be interpreted below the quantificational element. If the covert operator is the one that checks the [uNEG] features of the NM dhen, that is unexpected. (36) Polla provlimata dhen elisan. [NOT > MANY] many problems NM solved ‘The problems they solved are not many.’ (37) Polla provlimata dhen elisan. [MANY > NOT] many problems NM solved ‘The problems they didn’t solve are many.’ (Baltazani 2002: 112)

17 The two readings are distinguished by different intonational patterns. In the (a) reading, two phonological phrases are formed, as indicated in (i), while in the (b) reading there is only one phonological phrase (ii): (i) (POLI), (DHEN efage) a. ¬ > much: ‘He didn’t eat much.’ (ii) (POLI dhen efage) b. much > ¬: ‘There is much that he didn’t eat.’ 18 Thanks to Stella Gryllia for pointing this out to me. 19 A reviewer points out that Zeijlstra’s theory can predict the ambiguity by allowing semantic negation to take scope from a different locus, as long as the operator c-commands the negative marker as in (i) and (ii): (i) [Op¬ [polla dhen[uNEG]]] (ii) [ polla [Op¬ dhen[uNEG] ]] While correct, this conflicts with Zeijlstra’s claim that no such ambiguity is expected to arise in strict NCLs (Zeijlstra 2013: “Although Czech moc (‘much’) c-commands the negative marker, it is outscoped by negation, both under neutral and focus intonation”). But even assuming that the ambiguity can be accounted for by Zeijlstra’s modified proposal for strict NCLs, there is still an empirical problem with non-strict NCLs for which the validity of the data has been disputed by native speakers (cf. (39–40)). Overall, the quantificational data is inconclusive, as the attested scope orders do not match the predictions. Thus, the data does not serve as a reliable criterion for deciding the status of the NMs.

Challenging the strict vs. non-strict distinction of Negative Concord

143

The above is solid evidence that the interpretation of negation is not functionally determined by its syntactic position; negation is not interpreted necessarily where it surfaces. The fact that a covert operator can take scope over the quantificational element is argued to be an option and not the only available reading. Let me also point out that some of the predictions of Zeijlstra’s theory are not fulfilled either. More precisely, there are languages which have more than one NM. For instance, Greek has two NMs as we have seen, dhe(n) and mi(n). Dhen bears [uNEG] and mi(n), which is used in subjunctives, [iNEG] features (following Zeijlstra 2006). NMs with different semantic status should in principle give rise to different readings within the same languages when these co-occur with quantificational elements, similarly to the difference claimed to exist between strict and nonstrict NC languages. More precisely, it is expected that only the narrow scope reading (38a) should be available with the NM min, as no covert operator is assumed to be present in the structure. Contrary to expectations, once again both readings are attested. In fact the wide scope interpretation (38a) is even preferred: (38) Poli (na) min piis! 20 much PRT NM drink a. ¬ > much: ‘Don’t drink much!’ b. much > ¬: ‘There is much that I order you not to drink.’

To sum up, I have argued that in strict NCLs such as Greek and Romanian, there is no evidence for the presence of a covert negative operator obligatorily taking scope over the quantificational element. Experimental evidence has shown that both readings are permissible, arguing against a theory that equates the syntactic position of negative elements with the point where they are semantically interpreted. Finally, the expectation that different readings should arise within a language due to the different status of the NMs also remained unconfirmed. 3.1.3 Non-strict NC languages The data become more complex when we turn to non-strict NCLs. An Italian informant I have consulted admits the wide scope reading but excludes the narrow scope interpretation which is the only possible reading reported in Zeijlstra (cf. (33)): (39) Moltο non ha mangiato Gianni.21 a. ¬ > much: ‘Gianni hasn’t eaten much.’ b. *much > ¬: ‘There is much that Gianni hasn’t eaten.’

20 Speakers found it easier to perceive both readings with the verb drink. 21 Thanks to Nino Grillo for the data in Italian.

144

Maria Barouni

The same picture is replicated in Spanish, another non-strict NCL. The speaker I have consulted22 also prefers the wide scope reading (a) and considers the narrow reading (b) marked, if not impossible: (40) Mucho no ha comido Juan. Much not has eaten Juan a. ¬ > much: ‘Juan has not eaten much.’ b. ?? much > ¬: ‘There is much that Juan has not eaten.’

So, one important question that arises at this point regards the factual basis. Is there actually a difference between strict and non-strict NCLs and if so, what is this difference? Based on my informants, all NCLs admit the Neg. > Quant. reading, while the reading in which negation takes scope below the quantificational elements seems to be more easily obtained in strict NCLs. In table 2, I contrast the findings of Zeijlstra with the judgments of my informants: Table 2. Scope facts in NCLs Syntactic order: Quant. > ¬ Reported judgments Current findings

strict NCLs

non-strict NCLs

Sem. scope: ¬ > Quant. (Czech)

Sem. scope: Quant. > ¬ (Italian)

Sem. scope: ¬ > Quant. Quant. > ¬ (Romanian, Greek)

Sem. scope: ¬ > Quant. ??/ * Quant. > ¬ (Spanish, Italian)

What we can conclude so far from the data from both strict and non-strict NCLs is the following: – the semantic scope of negation is not computed on the basis of its overt syntactic position. – quantificational data cannot serve as a reliable criterion for deciding the status of the NMs in NCLs (see table 2). – NMs in Greek give rise to the same semantic interpretations which argues against the assumption that they differ semantically (Zeijlstra 2006).

22 Thanks to Ismael Ivan Teomiro García for the Spanish data.

Challenging the strict vs. non-strict distinction of Negative Concord

3.2

145

(Non) optionality of the NM

Zeijlstra (2013) relates semantic (un)interpretability of the NMs with their obligatory (or optional) presence in the structure. Semantic interpretable elements are expected not to be optional, while semantically vacuous elements could be, since their absence would not affect the interpretation of the sentence (i. e. NMs in strict NCLs with preverbal n-words). Note that there is no obvious reason for the presence of the NM when a preverbal n-word is in the structure, as the [uNEG] feature on the n-word is sufficient reason for having to assume the covert operator in the representation. More precisely, Zeijlstra argues that there are cases in which the NM in strict NCLs is optional, which is evidence for its [uNEG] status. This is considered to be the case in (41), with preverbal oute. Although Zeijlstra himself recognizes that optionality is not generally attested either in Greek or in strict NCLs, he takes it to be in support of the [uNEG] status of the NM. (41) Oute kan tin Maria (dhen) proskalese o Pritanis. even PRT the Maria NM invited the dean ‘Not even Maria did the dean invite.’ (Giannakidou 2007: 51)

However, contrary to previous assumptions in the literature, I would like to argue that there is no real optionality; rather, there are systematic differences regarding the presence or not of the NM which are related specifically to the properties of the n-word (and not to the status of the NM). To begin with, it is no coincidence that the example in (41) contains an ou-element, which in verb modification excludes the presence of the NM (9–10). Sentence (41) involves DP modification, a phenomenon which I have left aside at present and which is subject to speaker variation. Crucially, as I argue in Barouni (in preparation), optionality is only apparent, even in contexts of DP modification, in that (41) is licit without the NM only for a certain group of people. Moreover, this group of speakers can drop the NM in DP modification only under certain conditions: (i) with oute kan, (ii) with a certain class of verbs and (iii) with certain prosodic restrictions. Therefore, since judgements are subject to systematic speaker variation, indicating that there are two groups of speakers, [uNEG] cannot always be linked to optionality. Another argument against using optionality as a diagnostic for [uNEG] comes from data we have already seen with regular n-words and ou-elements. The NM dhen is obligatory with preverbal regular n-words in (26), while excluded with ouelements in (27): (26) Pote *(dhen) tilefonise. n-ever NM called ‘(S)he never called.’

146

Maria Barouni

(27) Oudhepote (*dhen) tilefonise. n-ever NM called ‘(S)he never called.’

Therefore, [uNEG] should not be linked to optionality. NMs in strict NCLs do not behave differently from NMs in non-strict NCLs, in that in neither group they are optional.

4.

The proposal

In this section, I will develop my own analysis, which shares basic features with Zeijlstra (2004, 2008a,b), viz. the hypothesis that NC consists in multiple syntactic agreement between an element with [iNEG] and elements with [uNEG]. I also adopt the assumptions that n-words are indefinites and that the negative operator existentially binds indefinites, including the event variable. Further, the negative operator has to appear above the VP (or the TP) level. Most importantly, I will deviate from Zeijlstra (ibid.) by arguing that semantic negation is introduced by a covert operator with [uNEG] at Neg° both in strict and non-strict NC languages. Evidence for the independent position of semantic negation will be provided by quantificational data and negative split facts. The second important deviation is that since neither the NMs nor the n-words introduce negation, what distinguishes strict from non-strict NC languages is a formal feature. This feature, the [NEG] feature, is formal in nature, licensing elements with [iNEG] to participate in syntactic operations. N-words with [iNEG] have to appear at a preverbal position while elements with [uNEG] stay in situ. The above results in a system with the following properties: – NMs in NCLs bear [iNEG]. – N-words in non-strict NCLs, the ou-class in Greek and the sem-class in Hungarian, bear [iNEG] in preverbal position. – The semantic interpretation of negation is not limited to its surface syntactic position. – Agree takes place in a ‘downward’ way as proposed in Zeijlstra (2004). – Feature valuation can be also derivational.

4.1

Status of the NMs

I have already argued that the argument that NMs differ between strict and nonstrict NCLs is disputable. For this reason, I propose that the status of the NMs in both strict and non-strict NCLs is [iNEG] and none introduces semantic neg-

Challenging the strict vs. non-strict distinction of Negative Concord

147

ation. NMs in all NCLs can check the [uNEG] feature of a covert semantic operator which is located in Nego. This is a welcome result as it can account for the ambiguity between negation and quantificational elements. In (42) the Neg > Quant reading arises because poli/‘much’ is interpreted in its initial base position. By contrast, in (43), the quantificational element is interpreted in its surface syntactic position: (42) Poli dhen efage poli. much NM ate much ‘He didn’t eat much.’

Sem. interpret.: Neg > Quant.

Sem. interpret.: Quant > Neg (43) Poli dhen efage poli. ‘There is much that he didn’t eat.’

Assume that the NM carries [iNEG] also in strict NCLs. Then, the NM with [iNEG] and regular n-words with [uNEG] enter into a feature checking relation no matter whether the n-word appears pre- or postverbally. Assuming that feature valuation under c-command can also be derivational (contra Zeijlstra), the [uNEG] features are checked before the n-word moves at a preverbal position in (44). Note that in this analysis, the obligatory presence of the NM also in preverbal position receives a straightforward explanation: it checks the [uNEG] feature of the n-word (44). This fact remained open in Zeijlstra’s analysis (see Penka (2011) for criticism). (44) Kanis *(dhen) irthe n-one NM came ‘Nobody came.’

Let us examine in detail how the derivation develops. If we maintain ‘downward’ valuation, feature checking takes place before the n-word has moved to its surface position. The negative indefinite is bound by the operator at LF, but at surface structure it appears above the NM:

148 (45)

Maria Barouni NegP 3 kanis ‘n-one’

NegP 3 dhen NM

[iNEG]

NegP 3 Nego ¬

TP 3

[uNEG]

TP 3 To irthe ‘came’

vP

3 kanis [uNEG]

vP 3 VP 4 irthe

When the n-words appear postverbally, feature checking proceeds as in (45), with the only difference that the subject remains in the SpecvP position and does not move higher. The question now that arises is whether my proposal can account for ou-elements in Greek and sem-elements in Hungarian.

4.2

Status of n-words: n-words with interpretable versions, [iNEG] or [uNEG]

So far, I have deviated from Zeijlstra’s theory in two ways: (i) the [NEG] feature is not semantic and (ii) a covert operator is always at Nego bearing [uNEG]. In this section I will argue that the difference between strict and non-strict NC languages resides in the different status of n-words. More precisely: (i) all n-words of non-strict NCLs admit two versions, one with [iNEG] and one with [uNEG] features; (ii) in strict NCLs, elements that carry ‘transparent’ morphological negation (e. g. ouelements and sem-elements) have also an [iNEG] version (cf. Herburger 2001 on the ambiguity of n-words).

Therefore, I propose that, in all NC languages, the [uNEG] feature of an n-word can be checked by an overt NM with [iNEG]; when n-words come with an [iNEG] version, then these check the [uNEG] feature of the covert operator and no NM appears in the structure. Supportive evidence for the presence of the covert operator will be provided in the next section. As we have seen, negative expressions that show non-strict NC properties in strict NCLs are morphologically negatively marked (cf. table 1). I assume that this

149

Challenging the strict vs. non-strict distinction of Negative Concord

is required for learnability reasons: for speakers of strict NCLs to receive a cue that an n-word has an [iNEG] version, the n-word has to be morphologically marked with negation. Crucially, negation has to be morphologically ‘transparent’, i. e. speakers should be able to recognize the n-word as negative; otherwise, the element is not an eligible candidate for an [iNEG] feature. This requirement is not necessary in non-strict NCLs, where speakers know from their grammar that all n-words in their language come in two versions. Therefore, transparent negation is not required for learnability reasons. We can now examine how the derivation proceeds with negative expressions that come in two versions, starting with the [uNEG] variant. When such items appear postverbally, nessuno-type elements of non-strict NCLs, ou-elements, and sem-elements behave like regular n-words of strict NCLs, in that they require the presence of the NM with [iNEG]. Their [uNEG] features are checked by the NM which in both strict NC and non-strict NCLs bears [iNEG]. The pertinent structure in (47) is similar to the structure in (45) above: (46) Dhen tilefonise oudhepote. NM called n-ever ‘(S)he never called.’ (47)

CP 3 NegP 3 dhen NM [iNEG]

NegP 3 Nego ¬

TP 3

[uNEG]

TP 3 To

tilefonise ‘called’

vP 3 oudhepote ‘never’

vP 3 v VP [uNEG] tilefonise 3 VP 3 v tilefonise

VP 3 AdvP 5 oudhepote [uNEG]

150

Maria Barouni

Consider the syntax of preverbal negative expressions next (example (27) repeated from above). In this structure, n-words with [iNEG] are fronted; I assume that there is a requirement in both strict and non-strict NCLs that a negative element (with [iNEG]) is present above the VP/TP level which expresses sentential negation (in Double Negation (DN) languages there is no such requirement). This is either fulfilled by the NM, in case n-words appear postverbally, or by preverbal n-words with [iNEG], in which case speakers do not spell out the NM. The n-word endowed with the [iNEG] feature checks the [uNEG] features of other n-words with [uNEG] which may be part of the structure, as well as the [uNEG] features of the silent operator in Nego. If the n-word stays in its initial position, then the derivation crashes as the features of the silent NM stay unvalued. (27) Oudhepote (*dhen) tilefonise. n-ever NM called ‘(S)he never called.’ (48)

CP 3 CP 3 NegP 3 oudhepote NegP ‘never’ 3 [iNEG]

Nego ¬

TP 3

[uNEG]

T

o

TP 3

tilefonise ‘called’

vP

3 vP 3 v tilefonise

VP 3 VP 3 tilefonise

VP 3 AdvP 5 oudhepote [iNEG]

151

Challenging the strict vs. non-strict distinction of Negative Concord

4.3

Is there independent evidence for the existence of the covert operator?

A natural question is whether there is independent motivation for the presence of a silent NM. I propose that such evidence comes from Negative Split phenomena with preverbal ou-elements. It is a well-known fact that the semantic scope of negative expressions can be dissociated from their surface position. First, let’s consider the interaction of a modal with negation. Iatridou & Zeijlstra (2013: 550) point out that prepi is a PPI that takes scope over negation: (49) O Yanis dhen prepi na figi. the John NM must PRT leave ‘John must not leave.’

synt. order: NEG > MODAL sem. scope: MODAL > NEG

The same picture can be replicated with ou-elements. Syntactically, the negative element resides above the modal, but in semantics, the scope relations are reversed. The modal takes scope over negation, providing evidence for the need to interpret negation in a lower position (50c). I assume that this reading is licensed by a silent operator in NegP. Note that the reading in which negation takes scope over the modal (50a) is metalinguistic or contrastive (Iatridou & Zeijlstra 2013). Similar observations have been made about negative split phenomena in German (Bech 1955/57, Geurts 1996, Jacobs 1980, Rullmann 1995, de Swart 2000, a.o.) (51).23 Greek (50) Oudhemia mama prepi na ine parousa. n-one mother must PRT be present a. ‘It is not required that there be a mother present’ ¬ > need > ∃ b. *‘There is no mother who is required to be present’ ¬ > ∃ > need c. ‘It is required that there be no mother present’ need > ¬ > ∃ German (51) Es muss kein Arzt anwesend sein. there must no physician present be a. ‘It is not required that there be a physician present’ ¬ > must > ∃ b. *‘There is no physician who is required to be present’ ¬ >∃> must c. ‘It is required that there be no physician present must > ¬ > ∃

(Zeijlstra 2011: 114)

Based on the findings above, we can conclude that – similarly to DN languages – negation of preverbal elements is interpreted below the position in which it 23 Note that the indefinite in (51) is also interpreted at a lower position, arguing in support of an indefinite analysis of the ou-elements.

152

Maria Barouni

surfaces in strict NCLs. NegSplit facts support the view that negation resides in NegP, providing evidence for the presence of an abstract, covert operator in Nego. The ou-element, which bears [iNEG], checks the [uNEG] features of the covert operator, while semantic negation is located in NegP, resulting in Negative Split. If correct, this suggests that feature interpretability and logical negation do not go hand in hand. A negative expression, such as the ou-element, can have [iNEG], but this does not mean that it needs to introduce logical negation. Further, I claim that an element may have [uNEG], as the silent operator, but still be able to carry semantic features (cf. Zeijlstra’s 2014 assumptions about interpretability and semantic negation).

5.

Predictions of the theory

Does the above theory make any predictions that can be tested? Given current assumptions, elements that incorporate morphologically transparent negation are possible candidates to host an [iNEG], and should then be incompatible with the NM in the preverbal position. The prediction is borne out by Greek mielements, i. e. negative particles which consist of the NM mi(n), which in turn is the standard NM used in subjunctives (and gerunds). All elements in this class exclude the NM when they appear preverbally: (52) Mite (*dhen) irthe. n-either NM came ‘Neither did (s)he come.’ (53) Mite (*dhen) irthe, mite (*dhen) tilefonise. n-either NM came n-either NM called ‘Neither did (s)he come, nor did (s)he call.’ (54) *(Dhen) irthe mite o Petros, mite i Maria. NM came n-either the Petros n-either the Maria ‘Neither did Peter come, nor Mary.’

In future research, it would be interesting to test other strict NCLs and see whether there are elements in these languages which behave similarly to Greek ou- and mi-elements, as well as Hungarian sem-elements.

6.

Summary

Based on the study of a class of negative expressions in Greek (the ou-class) which shows non-strict NC properties, I have argued that the conditions placed on NC are more complex than previously thought. I have claimed that the evidence for the

Challenging the strict vs. non-strict distinction of Negative Concord

153

different status between strict and non-strict NCLs is disputable, and that the difference between the two types of NCLs resides in the different status of the [NEG] feature of n-words. The resulting analysis of NC adopted the following assumptions: a. NMs in NC languages carry [iNEG], which is not semantic in nature. b. N-words of non-strict NCLs and ‘transparent’ negative words of strict NCLs also include an [iNEG] version. c. The Nego is always filled with an operator in NCLs with [uNEG]. This [uNEG] feature is valued either by a NM or by an n-word with [iNEG].

The analysis outlined here has the advantage that it can naturally account for the presence of the NM with preverbal elements in strict NCLs, and offers a possible explanation for languages which are in the process of change, among them Greek and Hungarian. Further, a welcome result is that the difference between strict and non-strict NCLs lies in that all elements of the latter have an [iNEG] version of the formal [NEG] feature. There is nothing prohibiting a strict NCL from having negative expressions that are marked [iNEG], as long as these carry morphological negation. Another benefit is also that the theory makes clear predictions regarding which elements can bear [iNEG] features in strict NCLs (see mi-elements in Greek). Future research will have to extend the analysis to a wider range of languages and also explore the principles underlying typological variation in the negative system (e. g. the differences between NCLs from DN languages) (see Barouni in preparation).

References Acquaviva, Paolo (1997): The Logical Form of Negation: A Study of Operator-Variable Structures in Syntax. New York: Garland. Baltazani, Mary (2002): Quantifier scope and the role of intonation in Greek. PhD thesis, University of California, Los Angeles. Barouni, Maria (in preparation): Syntax and semantics of focus particles in Greek. PhD thesis, University of Crete. Bech, Gunnar (1955/57): Studien über das deutsche verbum infinitum. (Historisk-filologiske. Meddelelser udgivet af Det Kongelige Danske Videnskabernes Selskab; Bd. 35, no. 2 1955, Bd. 37, no. 6 1957). Kopenhagen: Munksgaard. Biberauer, Theresa & Hedde Zeijlstra (2012): Negative Concord in Afrikaans: filling a typological gap. Journal of semantics 29 (3): 345–371. Den Besten, Hans (1986): Double Negation and the Genesis of Afrikaans. Ιn Muysken, P. & N. Smith (eds.): Substrata versus Universals in Creole Genesis. Amsterdam: Benjamins, 185–230. Geurts, Bart (1996): ‘On No’. Journal of Semantics 13: 67–86. Giannakidou, Anastasia (1993): KANIS/ kanis: A case of polarity sensitivity in Modern Greek. In Anastasiadi-Simeonidi, Anna, Mirto Koutitia-Kaimaki & Melita

154

Maria Barouni

Stavrou-Sifaki (eds.): Studies in Greek Linguistics 14: Proceedings of the 14th Annual Meeting of the Department of Linguistics. University of Thessaloniki, 130–144. Giannakidou, Anastasia (1998): Polarity Sensitivity as (Non) Veridical Dependency. Amsterdam: John Benjamins. Giannakidou, Anastasia (1999): Affective dependencies. Linguistics & Philosophy 22: 367–421. Giannakidou, Anastasia (2000): Negative … Concord? Natural Language and Linguistic Theory 18 (3): 457–523. Giannakidou, Anastasia (2006): N-words and Negative Concord. In Everaert, M. et al. (eds): The Blackwell companion to Syntax. 3rd vol. London: Blackwell, 327–391. Giannakidou, Anastasia (2007): The landscape of even. Natural Language and Linguistic Theory 25: 39–81. Herburger, Elena (2001): The negative concord puzzle revisited. Natural Language Semantics 9: 289–333. Hiraiwa, Ken (2001): Multiple Agreement and the Defective Intervention Effect. In Matsushansky, O. et. al. (eds.): The Proceedings of the MIT-Harvard Joint Conference (HUMIT 2000). Cambridge, Mass.: MITWPL, 67–80. Horrocks, Geoffrey (2014). Ouk Ismen Ouden: Negative Concord and Negative Polarity in the History of Greek. Journal of Greek Linguistics 14 (1): 43–83. Iatridou, Sabine & Hedde Zeijlstra (2013): Negation, polarity, and deontic modals. Linguistic Inquiry 44 (4): 529–568. Jacobs, Joachim (1980): Lexical Decomposition in Montague Grammar. Theoretical Linguistics 7: 121–136. Laka, Itziar (1990): Negation in syntax: on the nature of functional categories and projections. PhD thesis, MIT. Merchant, Jason (2001): The Syntax of Silence. Oxford: Oxford University Press. Penka, Doris (2011): Negative Indefinites. Oxford: Oxford University Press. Rullmann, Hotze (1995): ‘Geen einheid’, Tabu 25: 194–197. Surányi, Balázs (2006): Quantification and focus in negative concord. Lingua 116 (3): 272–313. de Swart, Henriette (2000): Scope Ambiguities with Negative Quantifiers. In von Heusinger, K. and U. Egli (eds.): Reference and Anaphoric Relations. Dordrecht: Kluwer, 109–132. Ura, Hiroyuki (1996): Multiple Feature-Checking: A Theory of Grammatical Function Splitting. PhD thesis, MIT. Veloudis, Ioannis (1982): Negation in modern Greek. PhD thesis, University of Reading. Zeijlstra, Hedde (2004): Sentential negation and Negative Concord. PhD thesis, University of Amsterdam. Zeijlstra, Hedde (2006): The ban on true negative imperatives. Empirical issues in syntax and semantics 6: 405–424. Zeijlstra, Hedde (2008a): Negative concord is syntactic agreement. Ms., University of Amsterdam. Zeijlstra, Hedde (2008b): On the syntactic flexibility of formal features. In Biberauer, Theresa (ed.): The Limits of Syntactic Variation. Amsterdam: Benjamins, 143–174. Zeijlstra, Hedde (2011): On the syntactically complex status of negative indefinites. Journal of Comparative Germanic Linguistics 14: 111–138.

Challenging the strict vs. non-strict distinction of Negative Concord

155

Zeijlstra, Hedde (2012): There is only one way to agree. The Linguistic Review 29: 491– 539. Zeijlstra, Hedde (2013): Not in the first place. Natural Language and Linguistic Theory 31: 865–900. Zeijlstra, Hedde (2014): On the uninterpretability of interpretable features. In Kosta, P. et al. (eds.): Minimalism and Beyond: Radicalizing the interfaces. Amsterdam: John Benjamins, 109–128.

Jakob Steixner

Focus Intervention and Double Negation in Bavarian*

Abstract: A number of Negative Concord (NC) languages allow, and in some configurations require, Double Negation (DN) readings, where two or more negative elements are interpreted at LF even though all occurring exponents of negation are otherwise capable of participating in NC. While DN readings in a number of NC languages have frequently been related to information structural constraints, the existing literature on Bavarian (Bayer 1990; Brugger & Poletto 1993; Weiß 1998, 1999), an NC variety of German, focusses on purely syntactic locality restrictions. Negative Indefinites (NIs) are treated as semantically negative, and a strictly local Spec-Head relation is required for their negative force to be factored out or absorbed. The present paper discusses an alternative theory of the observed restrictions on NC. Assuming the Agreement theory of NC (Zeijlstra 2004), I analyse blocking of NC as focus intervention in the sense of Beck (2006). When NC becomes unavailable, an additional silent negative operator is inserted in the structure, leading to a DN reading. To avoid ubiquitous DN readings, the insertion of that second operator is constrained pragmatically. Negative utterances are pragmatically marked: While a semantically single negative utterance (i. e., an NC reading) makes the affirmative utterance it negates accessible in the context, a DN reading requires the availability of an utterance that’s already negative. DN readings are therefore difficult to accommodate unless information structural cues facilitate an interpretation with one negation as discourse old. It is argued that, for NC languages that allow monoclausal DN readings, those readings are structurally always available but suppressed by the more natural NC reading unless the licensing of NC breaks down due to an intervening focus operator. * I am very thankful to the reviewers and editors of this volume whose patient and insightful comments have helped me improve this paper beyond recognition, and to the audiences at the Interdisciplinary Workshop on the syntax and semantics of negation (Vienna 2014) and the Workshop Negation and polarity: interfaces and cognition (International Congress of Linguists, Geneva 2013) for discussions about aspects of this work. I owe special thanks to Katharina Hartmann for detailed comments and discussion. All remaining errors are my own.

158

Jakob Steixner

Keywords: Negative Concord, Double Negation, intervention effects, Bavarian, focus, quantifying adverbs

1.

Introduction

The goal of this article is to present a novel analysis of several classes of double negation (DN) readings in Bavarian, a negative concord variety of German. Section 2 offers a general description of NC in Bavarian. I use Negative Concord in the general sense, as a cover term for all phenomena where several negative elements contribute to a single negation in terms of the interpretation of the sentence as a whole, as illustrated in examples (1) and (2).1 Here, a negative element is any constituent that is capable of expressing negative meaning alone in at least some contexts, such as fragment answers. (3) shows that neamd and nia are negative elements in this sense, and yet the prevalent reading of (1) and (2) is one with only one semantic negation. (1) dass i nia koa Zeidung ned kaft hob that I never no newspaper neg bought aux NC reading: ‘that I never bought a/any newspaper’ ?DN reading: ‘that I never bought no newspapers’ (2) dass am Freidog nia neamd ned kema is that on Friday never nobody neg come aux NC reading: ‘that nobody ever came on Friday’ ?DN reading: ‘that nobody never came on Friday’ (3) a. Wea isn kema? Neamd. Who aux- prt come nobody ‘Who came?’ – ‘Nobody.’ b. Wonn kimmt- a- n? Nia. When comes he prt never ‘When does he come?’ – ‘Never.’

I refer to elements such as nia ‘never’ and neamd ‘nobody’ in (1)–(3) by the traditional term of Negative Indefinites or NIs. This is not to prejudge their theoretical interpretation as semantically negative, or indeed indefinite (cf. Penka 2011, who, despite the title of her work, concludes that they are strictly speaking neither indefinite nor negative). Another common term for the elements in question is n-word (Laka 1990).

1 In Bavarian, as in Standard German, verb-second main clauses have a much freer word order than verb-final embedded clauses. It is for this reason that I frequently use embedded structures in the examples, in order to show the restrictions in base order.

Focus Intervention and Double Negation in Bavarian

159

The Bavarian examples in (1) and (2) show both negative doubling, i. e. the cooccurrence of a sentential negative marker ned with NIs, and negative spread, the co-occurrence of several NIs. Furthermore they show that Bavarian is a strict NC language in the terms of Giannakidou (2000) because it allows NC even when the sentential negative marker does not c-command the concording NIs: ned occurs linearly after and structurally c-commanded by the NIs. This is impossible in non-strict NC languages such as (Standard) Italian and Spanish illustrated in (4), where the sentential negative marker no co-occurs with an NI to its right, but not with one to its left. (4) a. *(No) ha venido neg aux come ‘Nobody came.’ b. Nadie (*no) ha Nobody neg aux

nadie. (Spanish) nobody venido. come

NC is interesting from a theoretical perspective as it poses a challenge to the compositionality of meaning. If all morphosyntactically negative elements are carriers of semantic negation, we expect double negation (DN) interpretations whenever they co-occur. On the other hand, if they are semantically non-negative, the negative interpretation in isolation remains mysterious. Both of these interpretations have been used as vantage points for theories of negative concord, briefly sketched in section 3. Section 4 focusses on constructions in Bavarian that do not receive a full NC interpretation. After introducing the relevant background in section 4.1, narrow focus constructions such as (5) and (6),2 which have not previously received much attention in the literature, are discussed in section 4.2. In these examples, full concord involving all three negative elements (the (b) readings) is excluded. The salient DN readings in (5a) and (6a) instantiate partial concord between two negative elements while the third one is interpreted independently. (5) dass i nia KOA Zeidung ned kaft hob that I never no newspaper neg bought aux a. DN reading: ‘that I never bought no newspapers’ = ‘that I always bought some newspaper’ b. * NC reading: ‘that I never bought any newspaper’ (6) weil am Freitag nia NEAMD ned kema is because on Friday never nobody neg come aux a. DN reading: ‘because it was never the case that nobody came on Friday’ b. * NC reading: ‘because never does anybody come on Friday’

2 Narrow focus will be represented typographically throughout the paper by putting the focus constituent in upper-case letters.

160

Jakob Steixner

Section 4.3 analyses the double negation readings arising in such examples as an instance of intervention effects in the sense of Beck (2006). Section 4.4 returns to the failure of NC in sentences with quantifying adverbs such as ‘often’ (see (7)), a problem previously discussed by Brugger & Poletto (1993) and Weiß (1999). Section 4.4 identifies empirical problems with the analyses presented there and argues that the DN readings observed with adverbs can be explained as an extension of the intervention effect theory developed in 4.3. (7) dass koa Mensch oft ned kema is. that no person often neg come aux ? DN reading: ‘that nobody often didn’t come’ * NC reading: ‘that nobody came often’

The basis for the data presented here is a variant of Bavarian spoken in the Innviertel region of Upper Austria, close to the border with the German state of Bavaria. While the author is a speaker of this variety of Bavarian, all examples have been checked with at least one other speaker of this variety and in some cases with up to 7 speakers. Examples were presented to the informants in context either orally or in an informal orthography closely tracking the local dialectal pronunciation, and with the intended interpretation made explicit.

2.

Basic Properties of Bavarian Negative Concord

2.1

Optionality of NC

Bavarian allows single negative or Negative Concord readings when one (or more) negative indefinites co-occur with an element of clausal negation. The simple case is illustrated in (8), with the NI neamd followed by the clausal negation ned. (8) dass neamd ned kema is that nobody neg come aux ‘that nobody came’

Negative Concord is also possible with two NIs in combination with clausal negation, resulting in three negative elements contributing to one semantic negation, as in (9). (9) dass nia neamd ned kema is that never nobody neg come aux ‘that nobody ever came’

Focus Intervention and Double Negation in Bavarian

161

Clausal negation can be left out without a change of meaning when negation is already marked through an NI, as in (10a). Non-initial NIs can be replaced by their neutral, non-negative counterparts, see (10b). (10) a. dass neamd kema that nobody come ‘that nobody came’ b. dass nia wer that never somebody ‘that nobody ever came’

2.2

is aux kema is come aux

Intervening elements

It is not possible to interrupt a sequence of negative elements with a non-negative indefinite and still have an NC reading, as shown in (11a). This sentence only has a DN interpretation available with an adequate context. However, for at least some speakers,3 NC sequences can be interrupted by a definite argument, as in (11b) from Brugger & Poletto (1993: ex. (104)), with both readings available. I will come back to the intervention effect in section 4.3. (11) a. dass nia wer ned kema is that never somebody neg come is *NC reading: ‘that nobody ever came’ DN reading: ‘that everybody always came’ b. dass neamd1 sei1 Frau ned mitgnumma hod that nobody his wife neg taken_along has NC reading: ‘that nobody1 took along his1 wife’ DN reading ‘that everybody1 took along his1 wife’

Negative indefinites may also occupy SpecCP, giving rise to verb second structures. NIs in the pre-field may form part of a NC chain even if non-negative pronouns or indefinites intervene. This is shown in (12) where the second person dative pronoun da intervenes between the topicalized NI nix and the negative marker ned. Despite the intervention of the pronoun the NC reading is available. Weiß (1999) takes this fact as evidence that negative elements may reconstruct to their assumed base-position in SpecNegP, see the discussion in section 3.

3 One anonymous reviewer notes that he finds example (11b) quite unacceptable. This is not confirmed by my informants, all native speakers of Upper Austrian varieties of Bavarian.

162

Jakob Steixner

(12) Nix ko i da ned dazöhn. nothing can I you.dat neg tell a. NC reading: ‘I can’t tell you anything (at all).’ b. DN reading 1 (¬ > CAN > ¬) : ‘I cannot tell you nothing.’ = ‘I have to tell you something.’ – corresponds to base structure (13a) c. DN reading 2 ( ¬ > ¬ > CAN): ‘There is nothing I can’t tell you.’ = ‘I can tell you anything.’ – corresponds to base structure (13b)

Note that – apart from its NC interpretation – the string in (12) also has two DN readings illustrated in (13). These two interpretations correspond to different base structures of the NI nix in the unambiguous sentences (13a) and (13b). Interpretation (12b) is derived from the base structure in (13a), while interpretation (12c) corresponds to the base structure in (13b). Prosody determines which of the readings of (12) becomes prominent. We will see more on the effects of prosody and information structure on the interpretation of negation in section 4. (13) a. dass i da ned NIX dazöhn ko that i you neg nothing tell can ‘that I cannot tell you nothing’ = ‘that I have to tell you something’ b. dass i da nix NED dazöhn ko that I you nothing neg tell can ‘that there is nothing I can’t tell you’ = ‘that I can tell you anything’

To sum up this section, negative concord in Bavarian is available if an NI in its middle field position precedes the negative marker ned. While the occurrence of other material between NIs and ned is quite limited, strict (surface) adjacency is not always required. Topicalized NIs can reconstruct, yielding NC readings even if ample material intervenes between the two negative elements at the surface.

3.

Syntax of Bavarian Negative Concord

The aim of the present section is to discuss the syntax of negative concord in Bavarian. Section 3.1 discusses the position of the negative marker in the clausal structure, concluding that it has the status of a vP adjunct. Adapting Zeijlstra’s (2004) theory of NC for the analysis of Bavarian, in section 3.2 I propose an analysis of NC as the result of a syntactic agreement relation between a negative operator OPØ and one or more morphosyntactic exponents of negation. Section 3.3 describes how double negation readings can be derived within the theory proposed.

Focus Intervention and Double Negation in Bavarian

3.1

163

A low position of the negation marker ned

In Zeijlstra’s (2004, 2008a) theory of negative concord, the availability of NegP as a dedicated projection in the clausal spine is not a universal feature of human language. Instead, NegP is only posited if there is syntactic evidence for it within a language. Whether Bavarian ned is the head of a negative projection is a matter of debate. I will argue that this is not the case and will develop an analysis of ned as an adjunct to vP. First, let us consider the option that ned is the head of NegP, i. e. Neg°. This is the analysis pursued e. g. by Weiß (1998, 1999). For the languages of the world, negative markers that are commonly analysed as negative heads share a number of features: They are prosodically weak, they are involved in head movement chains by triggering or blocking verb movement, and they often have an affixal status, as in the Slavic languages. Bavarian ned does not fit this pattern. While it is prosodically weak – unless it is contrastively stressed – it neither blocks V-to-C movement nor is it affixed to the verb on its way to C, see (14). (14) Er kimmdi heid woi ned ti he comes today prt neg ‘He probably won’t come today.’

The analysis of ned as the specifier of NegP runs into problems when considering the scope behavior of negation. The position where ned is realized does not correspond to the semantic scope of negation in many examples such as (15) from Bayer (1990: ex. (5a)), also discussed by Penka & Stechow (2001). In this example, the negative phrase keine Rede (‘no speech’) receives a split-scope interpretation: Negation scopes above the modal (also evidenced by the availability of the modal negative polarity item brauchen), while the indefinite part of the phrase ‘a speech’ is interpreted below the modal and nicht remains semantically vacuous. Thus, we observe an NC interpretation. If the position of the negative marker ned does not correspond to the position where negation is interpreted, it seems misleading to assume its position in SpecNegP. In sum, a NegP analysis does not appear to be appropriate for negation in Bavarian. (15) Ich bin froh, dass ich keine Rede nicht halden brauch.4 I am glad that I no speech neg hold need ‘I am glad, that I do not need to hold a speech.’

4 This example, like several others from Bayer (1990), is taken from early 20th century satirical works by Ludwig Thoma. It represents the artist’s interpretation of a semi-literate dialect speaker attempting to write Standard German, thus the more or less Standard German orthography.

164

Jakob Steixner

Instead, I propose that the negative marker ned is adjoined to vP. In this position, it follows the verbal arguments in the middle field as shown in (16a) for definite argument and in (16b) for generic indefinites. (17) illustrates the syntactic structure of (16a). (16) a. I ko da ’s Göd ned gebm. I can you.dat the money neg give ‘I can’t give you the money.’ b. wei ma eam a Auto ned onvertrauen ko because one him a car neg entrust can ‘because one can’t entrust him with a car’ (17)

CP 3 i3 C' 3 ko TP 3 da4 TP 3 t3 T' 3 vP T 3 gebm1 DP vP ‘s Göd2 3 ned vP 6 t 2 t 4 t1

Definite and generic DPs generally scramble in Standard German and Bavarian. However, even existential indefinites (including NIs) do not regularly appear after ned, except for the special case of ‘light negation’ (Schwarz & Bhatt, 2006), and when they are contrastively focused (see section 4). Since non-generic indefinite DPs do not generally scramble in Standard German and Bavarian this instance of NI-scrambling is rather unexpected. A possible reason for this could be that NI-scrambling is a case of a last resort movement. Note that in Standard German and Bavarian the sequence nicht ein X (‘not an X’) is realized as kein X (‘no X’), see Penka (2011) among others. In order to prevent this rule from application, and to create an environment for negative concord, the NI moves to a position preceding the negation. This is shown in (18) with NC and split-scope of the NI as its salient interpretation under neutral intonation. (18) I ko da koa Göd ned gebm. I can you.dat no money neg give NC reading: ‘I can’t give you any money.’ (NC: ¬ > $ > CAN)

Focus Intervention and Double Negation in Bavarian

165

Existential indefinites do not generally scramble and are bound under existential closure in their VP base positions (cf. Heim, 1982). Weiß (1999) uses this to argue for a special movement related to the Neg-criterion of Haegeman & Zanuttini (1996). He takes ned to be the head of a NegP situated at the edge of the VP domain. Weak indefinites cannot remain inside the VP base position under negation because they are opaque for the scope of negation, but being subject to existential closure, they must remain in the (broader) VP domain. As a solution to these conflicting requirements Weiß proposes that the negative existential indefinites move to the specifier of NegP. This solution elegantly explains the position of (negative) indefinites preceding ned. As a downside, it requires multiple specifiers for NegP since NC may involve several NIs. A further problem of Weiß’ proposal is that it builds on universal semantic constraints to the extent that he predicts that all languages show NC. As a consequence he needs to stipulate PF deletion of noninitial negative elements in non-NC languages. These problems, along with the possibility of interpreting negation in a syntactic position different from the position of ned, lead me to reject the ned-as-Neg° analysis. I thus conclude that an analysis of ned as a vP adjunct with an uninterpretable negative feature that has to be checked by a negative operator is superior.

3.2

The negative operator

If the negation is not interpreted at the locus of the negative marker and the negative indefinites but rather in a higher position as shown in (15), another element needs to contribute negative force to the sentence as a whole. I assume this element to be a negative operator (in line with Zeijlstra 2004, 2008a; Penka 2011 and others). The negative operator can be overt, that is, it can be expressed by the negative marker of non-Strict NC languages. See example (19) from Italian. (19) [OPØ Non] è arrivato nessuno. neg aux arrived nobody ‘Nobody has arrived.’

Alternatively, the operator may remain phonetically empty. This is observed especially in Strict-NC languages (as pointed out in the introduction). Here, not only negative indefinites but also the sentential negative marker have to be analysed as semantically non-negative elements in order to explain the fact that the negation is interpreted in a higher position in the structure than the negative marker. This was illustrated in example (15), for which a structural representation is given in (20).

166 (20)

Jakob Steixner

CP 3 C TP dass 3 TP [iNeg]OP 3 ich T' 3 vP T 3 brauch1 DP vP [uNeg]keine Rede2 3 vP [uNeg]nicht 6 t2 halten t1

The negative elements here have a formal, syntactic feature [uNeg], i. e. an uninterpretable feature that has to be checked in the syntax by an element carrying an interpretable [iNeg] 5 (Zeijlsta 2004; Penka 2011). In (20) this element is a covert negative operator which I am representing as an adjunct, here at the TP level. Note that the operator could in principle be inserted at different syntactic positions which are constrained semantically. The negative operator is of type , taking complements of type . It can therefore only be merged to complete predicates with all arguments satisfied. Under an event-semantic approach, this is minimally after the event argument has been quantified over (Penka 2011:10). The [uNeg] features of the negative elements are licensed under Agree with the higher [iNeg] feature on the negative operator, see Pesetsky & Torrego (2007). This Agree relation can apply multiply, checking several instances of [uNeg] with a single operator. For Zeijlstra (2004, 2008a), doubling phenomena such as negative concord are due to multiple Agree (Hiraiwa 2001). Once checked, uninterpretable features are invisible at LF (Chomsky 1995). The negative elements with uninterpretable negative features signal the presence of a negative operator to the parser, but are themselves semantically non-negative. Zeijlstra (2008b: (25)) adopts the following Phonological Economy for negative operators: (21) a. A phonologically empty negative operator may be assumed iff it prevents a derivation from crashing. (Given multiple agree no second OPØ may be assumed if the first one is able to check all present [uNEG] features.) b. The phonologically empty negative operator may not occupy a position that is higher in the structure than necessary.

5 Zeijlstra assumes upward Agree, thus departing from the standard analysis of Agree such as proposed by Pesetsky & Torrego (2007), among others. This is most explicitly defended in Zeijlstra (2012).

Focus Intervention and Double Negation in Bavarian

167

This assumption will be shown to be essential when deriving double negation readings.

3.3

Deriving double negation

If the negative operator can be inserted in more than one position, it should in principle be possible to find examples of more than one Op¬ within a single clause. This is the case in DN readings of NC languages, discussed in more detail for Bavarian in section 4. DN readings have been described in the literature, among others, for Hungarian (Puskás 2012; Bende-Farkas 2005) and most Romance standard varieties. Example (22) adapted from Isac (2004: (28)) illustrates DN in Romanian, a Strict NC language that allows an NC interpretation in the same string under a more neutral intonation. A very similar effect is shown in the Bavarian example (23): when neamd is stressed, the DN reading becomes available. For an account of the correlation between stress on the NI, and DN interpretation, see sections 4.2 and 4.3. (22) Nimeni/NIMENI nu iubeste pe NIMENI. nobody neg loves pe nobody DN reading: ‘Everybody loves somebody.’ NC reading: ‘Nobody loves anybody.’ (23) weil nia neamd/NEAMD ned kema is because never nobody/NOBODY neg come aux a. DN reading: ‘because never did nobody come on Friday’ b. NC reading: ‘because never does anybody come on Friday’

A Strict NC language in which DN readings are available must permit the insertion of a negative operator in more than one position: In example (23), the higher operator checks the [uNeg] feature on the negative adverb nia, while the lower operator checks the [uNeg] features of the subject NI and the negative marker. This is illustrated in (24). For the time being, I suggest adjunction of the lower negative operator to vP.

168 (24)

Jakob Steixner CP 3 weil TP 3 TP [iNeg]OP 3 TP [uNeg]nia 3 T' 3 vP T 3 is1 vP [iNeg]OP 3 vP [uNeg]NEAMD2

3 vP 6 t2 kema t1

[uNeg]ned

DN readings in Bavarian and their relation to prosodic and information structural cues are extensively discussed in section 4.

3.4

Agree vs. Absorption

Weiß’ (1999) account builds upon the Absorption and Factorization theory (henceforth AFT) of negative concord developed by Haegeman (1995); Haegeman & Zannutini (1996); Zannutini (1997). In the AFT, NIs are semantically negative elements with the semantic content of "¬. The AFT postulates a Neg-criterion analogous to Rizzi’s (1996) Wh-criterion: a negative operator must be in a Spec-Head configuration with a [+NEG] X°, and vice versa. To derive negative concord, it is postulated that different (phrasal) negative elements adjoin to each other at LF, forming a complex or polyadic quantifier, with the semantic NEG feature factored out. Since the AFT assumes a negative semantic content for all NIs, this theory is well suited to explain the interpretation of NIs in fragment answers as unambiguously negative. The AFT explains the availability of DN readings as follows: The negative features of NIs that are not in a Spec-Head relation with Neg° cannot undergo Absorption. Thus, such NIs retain their inherent negative meaning leading to DN readings. Additional assumptions are required to explain why such structures are not simply ungrammatical given that these NIs do not satisfy the Neg-Criterion. A further problem of the AFT is the case of expletive

Focus Intervention and Double Negation in Bavarian

169

negation. Thus, it is not immediately clear how the AFT derives the availability of the NIss nadie and nada in (25), adapted from (Espinal 2007: 50). (Spanish) (25) Antes de que nadie diga nada, dejen que les dé la bienvenida. before of that nobody say nothing let that you give the welcome ‘Before anybody says anything, let me welcome you.’

The theory furthermore predicts that Absorption can only occur between negative elements located in the specifier of NegP. This is problematic for cases with three negative elements and concord between the first two, which will be discussed in more detail in section 4.2.1, ex. (31) ff. If we assume with Weiß (1999) that ned is the head of a NegP in Bavarian, no Absorption should take place at all,6 since the negative elements are not in a Spec-Head relation with Neg°. Another proposal within the AFT worth considering is the ‘Split NegP’ or ‘Big NegP’ hypothesis (Poletto 2008, this volume). An early variant of this proposal was applied to Bavarian by Brugger & Poletto (1993). They assume recursion of NegPs such that negation is interpreted at the higher NegP1 which remains covert, while ned surfaces in the lower NegP2, with NC requiring government of NIs by the higher NegP. Of interest to the theory advanced in the present paper is the central hypothesis that negation can be expressed in several functional positions in the clause, which is at least to some degree independent of the discussion of how to derive DN readings. For Poletto (this volume), negation is realized either in a FocusP (as the highest projection for negation), or in a MinimizerP, a ScalarP, or an ExistentialP lower in the clause. Poletto assumes that the sentential negative markers are merged as a complex phrase, in a vP argument position. Subsequently, the negation markers move to one of the positions defined above for the purpose of feature checking, resulting in the different syntactic positions for negative markers across Northern Italian dialects. While I will not pursue this idea further in this paper, it would be interesting to compare Poletto’s positions with the possible loci for merging a negative operator under the Agreement framework defended here.

6 This is possibly a problem restricted to Weiß’ (1999) specific implementation. The observed interpretation is derivable when we assume that ned, like negative indefinites, is merged as a phrasal argument in the VP and moves to the specifier of a (null) NegP, as Haegeman (1995) does for West Flemish. In this scenario, both NIs of (31) are in the specifier of NegP while ned remains in the VP, unable to undergo Absorption.

170

Jakob Steixner

4.

Negative Concord and information structure

4.1

Background

Previous analyses of Bavarian NC have focussed on structural constraints. In this section, I will explore the role of focus structure making DN readings surface. Section 4.2 shows that narrow focus on the VP or on a constituent within the VP can lead to DN interpretations when a series of negative elements is interrupted by a focus boundary. In section 4.3, I propose that this can be analyzed as an intervention effect in the sense of Beck (2006). In section 4.4, I attempt to extend the intervention effect analysis to the case of intervening quantificational adverbs like oft (‘often’), a case that has been discussed, again solely from a structural perspective, by previous authors (Bayer 1990; Brugger & Poletto 1993; Weiß 1999). Analysing the exclusivity of certain DN readings in terms of focus intervention allows for a unified account of the two superficially distinct phenomena discussed in section 4.2 and 4.4. Although beyond the scope of the current paper, it may furthermore allow a direct comparison with NC languages in which DN readings have been discussed in relation with information structure, e. g. Hungarian (Bende-Farkas 2005; Puskás 2012) or Spanish and Catalan (Espinal & Prieto 2011; Tubau et al. 2015).

4.2

Contrastivity

4.2.1 Contrast on DPs preceding ned Contrastively stressed non-initial NIs require a contrastive interpretation leading to DN readings. Comparing (26)–(27), repeated from the introduction, with negative concord involving two NIs and ned, with their variants in (28)–(29), we find that NC is blocked by narrow focus on the second of two NIs. (26) dass i nia koa Zeidung ned kaft hob that I never no newspaper neg bought aux a. NC reading: ‘that I never bought a/any newspaper’ b. ?DN reading: ‘that I never bought no newspapers’ (27) dass am Freidog nia neamd (ned) kema is that on Friday never nobody neg come aux a. NC reading: ‘that nobody ever came on Friday’ b. ?DN reading: ‘that never nobody came on Friday’

Focus Intervention and Double Negation in Bavarian

171

(28) dass i nia KOA Zeidung (ned) kaft hob that I never NO newspaper neg bought aux a. *NC reading: ‘that I never bought any newspaper’ b. DN reading: ‘that I never bought no newspapers’ = ‘that I always bought some newspaper’ (29) weil am Freitag nia NEAMD ned kema is because on Friday never NOBODY neg come aux a. * NC reading: ‘because never does anybody come on Friday’ b. DN reading: ‘because it was never the case that nobody came on Friday’

The generalization is that DN readings become obligatory when a non-initial NI is contrastively stressed. In terms of an AFT account along the lines of Weiß (1999), this has to be explained by stipulating that only the lower, stressed NI occupies the specifier of NegP, while the higher one is situated in the TP domain where its negative semantic content escapes Absorption. This is illustrated in (30). (30)

CP 3 weil TP 3 nia TP 3 T' 3 NegP T 3 is1 NEAMD2 Neg' 3 Neg vP ned 6 t2 kema t1

The generalization that stress on an NI precludes any other NI further to its left from participating in NC still has to be stipulated, though. It does not follow from Weiß’ theory that an NI followed by another NI has to be situated in the TP if and only if that other NI is contrastively stressed. Even more problematic for a account á la Weiß are cases where the third item in a series of negative elements is contrastively stressed. Recall that under the AFT, NC readings require NIs to have their negative semantics absorbed in a Spec-Head configuration with the negative head. In a sequence of NI1-NI2-NED, if NI2 is too high in the clause structure to enter a Spec-Head relation with NegP and have its negative semantics absorbed, it follows that NI1 is also too high, leading to an interpretation where each NI contributes an independent negation interpretation, thus predicting a triple negation reading. This is not borne out as

172

Jakob Steixner

shown in (31). In this example narrow focus blocks a full concord reading with only one negation interpreted for the entire clause, i. e. ned is interpreted independently of the NIs. In other words, NC is blocked between stressed ned on the one hand and the NIs outside the focus on the other hand. Concord between the NIs which are both outside the focus domain, on the other hand, is allowed to proceed normally and the two NIs nia and neamd jointly contribute only one negation. I refer to this as a ‘partial NC reading’. The only reading predicted by Weiß’ theory is ‘triple negation reading’, with no concord between any two negative elements independent of focus domains. That is, not only the negative marker ned but also each NI is expected to receive an independent negative interpretation. This reading is hardly, if at all, available (an illustration of Horn’s 2010 principle of triplex negatio confundit). (31) weil am Freitag nia neamd NED kema is Because on Friday never nobody neg come aux Partial NC reading: ‘because on Friday nobody ever didn’t come’ (i. e. everyone always comes, ¬$o(ccasion),p(erson): ¬come(p at o)) *Triple negation reading: ‘because never does everybody come on Friday’ ¬$o(ccasion), ¬$p(person): ¬come(p at o)

The tree in (32) illustrates the structure of (31) in the framework of Weiß (1999). With neither NI in the SpecNeg, Absorption, which requires a Spec-Head configuration with Neg°, cannot take place. Both NIs thus contribute an independent negation, in addition to the clausal negation expressed by ned as the head of NegP. The expected interpretation is therefore the unattested triple negation reading. (32)

CP 3 weil TP 3 nia TP ¬ (o) 3 neamd2 T' ¬ (p) 3 NegP T 3 is1 Neg vP NED 6 ¬ t2 kema t1

The Agree-based account advocated for in the present paper derives the correct reading of (31) as illustrated in (33). The first OPØ is adjoined to the TP and licenses the [uNeg] features on both NIs. The second, vP-adjoined, OPØ licenses the negative marker ned. The reading is thus a double negation reading with one

Focus Intervention and Double Negation in Bavarian

173

negation narrowly negating the focussed vP, and another licensing both NIs above it. Partial NC is licensed among the elements outside the sentences’ focus while the focussed ned is interpreted independently. (33)

CP 3 weil TP 3 TP [iNeg]OP 3 TP [uNeg]nia 3 T' [uNeg]neamd2 3 vP T 3 is1 vP [iNeg]OP 3 vP [uNeg]NED 6 t2 kema t1

Analogously, the DN reading in (29) above is derived by positing an OPØ with scoping over and licensing both NEAMD and ned but not nia. To summarize this section, the existence of narrow focus on an element within the vP induces readings where the negative elements within the scope of the focus operator are interpreted independently of those outside of its scope. This generalization is best captured by positing that focus triggers or requires the insertion of a second OPØ in a vP adjoined position.

4.2.2 ned-NI orders and contrastivity When an NI appears after clausal negation, only a DN interpretation is available, as illustrated in (34) and (35). (34) I ko da ned NIX gebm. I can you.dat not nothing give a. DN reading: ‘I can’t give you NOTHING.’/ ‘I have to give you something.’ ¬ > CAN > ¬ > $ b. * NC reading: ‘I can’t give you anything.’ (35) I hob ned NEAMD gseng. I have not nobody seen a. DN reading: ‘I didn’t see NOBODY.’/ ‘I did see someone.’ ¬ > CAN > ¬ > $ b. * NC reading: ‘I didn’t see anyone.’

174

Jakob Steixner

The AFT accounts for this observation by assuming that the NIs simply remain in their VP base position. By not moving them into SpecNeg, Absorption of their negative semantic content is unavailable. However, it is crucial to note that these NIs are obligatorily stressed when appearing after ned. Without such a contrastive intonation examples (34) and (35) do not receive a DN interpretation but are simply ungrammatical. In other words, the behaviour of NIs parallels that of regular DPs following ned, which also have to be stressed and interpreted contrastively. This is illustrated in examples (36) and (37). The interpretations available are contrastive focus either on the object itself or on the VP, that is, what is asserted is that the (presupposed) seeing-event involving Peter does not have a dog as its theme, or that the presupposed event involving Peter is not one of seeing a dog. (36) Da Peda hod ned an HUND gseng *(sondern a KOTZ (gsuacht)). the Peter has neg a dog seen but a cat searched ‘Peter didn’t see a [DOG]F, but a CAT.’ / ‘Peter didn’t [see a DOG]F, but searched for a CAT.’ (37) Da Peda hod ned in HUND gfüttad *(sondern d’ Kotz (gstreichelt)). the Peter has neg the dog fed but the cat stroked ‘Peter didn’t feed the [DOG]F, but the CAT.’ / ‘Peter didn’t [feed the DOG]F but stroke the CAT.’

In section 4.2.1, we saw that contrastive focus induces a DN reading even if NIs appear in their canonical position relative to the negative marker ned, that is, in the position in which we observe NC readings as the salient interpretation in the absence of special prosodic marking. Having seen that contrast drives DN reading, we can analyse the obligatory contrastive focus on the NIs in (34) and (35) as the driver for their DN readings. Reference to the NIs’ position as the immediate cause for their failure to participate in NC thus becomes unnecessary; see Brugger & Poletto (1993).

4.3

Focus and intervention effects

The unifying property of the DN configurations discussed in section 4.2 is that the sequence of negative elements is interrupted by a Focus operator scoping over one of the negative elements but not the other(s). Let us recall Zeijlstra’s (2008b) Phonological Economy Principle from section 3.1, repeated here as (38).

Focus Intervention and Double Negation in Bavarian

175

(38) a. A phonologically empty negative operator may be assumed iff it prevents a derivation from crashing. (Given multiple agree no second OPØ may be assumed if the first one is able to check all present [uNEG] features.) b. The phonologically empty negative operator may not occupy a position that is higher in the structure than necessary.

According to this principle, the observed DN readings show that the first OPØ is unable to check all present [uNEG] features in such a configuration. I propose to explain this observation in terms of intervention effects as described by Beck (1996; 2006), Tomioka (2007) and others. A classical example for such an intervention effect is illustrated in (39), adapted from Beck (2006: ex. (7)). Despite the absence of a superiority effect in (39a), (39b) is ungrammatical due to the presence of the NI niemand (‘nobody’) as an intervener, and this unacceptability is due to the position of the intervener, as (39a) is well-formed. (39) a. Wen hat Luise wo gesehen? whom has Luise where seen ‘Whom did Luise see where?’ b. *Wen hat niemand wo gesehen? whom has nobody where seen ‘Whom did nobody see where?’ c. Wen hat wo niemand gesehen? whom has where nobody seen ‘Whom did nobody see where?’

Data like (39) lead Beck (2006: 3) to the generalization that “a wh-phrase in situ may not be c-commanded by a focusing or quantificational element”, which is ultimately reduced to a General Minimality Effect, as given in (40). (40) General Minimality Effect (Beck 2006:17) The evaluation of alternatives introduced by an XP cannot skip an intervening ~ operator. [i. e. a focus operator in Rooth’s (1992) sense, informally known as ‘squiggle operator’, JS]

As is already visible from the example in (39) above, negation is (at least in German) an element which introduces an intervention effect. With these prerequisites, it is possible to interpret Bavarian DN readings under narrow focus as an instantiation of the General Minimality Effect. In a structure like (33) from section 4.2.1, focus intervention blocks licensing of the [uNeg] feature of the negative marker ned. A full NC reading becomes unavailable. The only way to avoid uninterpretability is by introducing an additional Op¬ at the vP level, which yields a DN reading. As for why focus blocks the licensing of Neg features, it is conceivable that NIs introduce a covert ‘even’ type operator as part of their meaning. This has been argued for NPIs by Lahiri (1998), among others, and used by Beck (2006) to explain intervention effects. NPIs, and in our theory NIs, thus introduce a set of

176

Jakob Steixner

alternatives into the representation. Under broad focus, these would be evaluated at the sentence level. An intervening focus operator OP~ is forced to evaluate those alternatives and thus bind the NIs. They thereby become unavailable for an Agree relation with a high OPØ outside the scope of OP~. With a single OPØ in the clause, this would mean that the [uNEG] feature on any NI in the scope of a focus operator remains unchecked. As a last resort, and in line with Zeijlstra’s (2008b) Phonological Economy Principle in (38), a second OPØ is inserted below the position of the focus operator OP~. Finally, note that quantitative adverbs like ‘often’ frequently behave as interveners in Beck style intervention effects, as illustrated in (41) from Beck (2006: (26)). (41) a.* Luise zählt auf, welche Uni oft welche Linguisten eingeladen hat. Luise enumerates which university often which linguists invited has b. Luise zählt auf, welche Uni welche Linguisten oft eingeladen hat. Luise enumerates which Uni which linguists often invited has ‘Luise enumerates, which university often invited which linguists.’

Section 4.4 explores how blocking of NC with such adverbs, previously discussed in e. g. Brugger & Poletto (1993), Weiß (1999), can be explained with an intervention effect analysis. 4.4.1 discusses the basic cases, while section 4.4.2 shows a rescuing effect under topicalization that remains mysterious under previous analyses.

4.4

Quantifying adverbs

4.4.1 Blocking of NC with ‘often’ Going back to Brugger & Poletto (1993); Weiß (1998, 1999), it has been observed that oft (‘often’) allows for DN readings when following an NI, whether it follows (42) or precedes (43) ned.7 In both of these approaches, this is explained structurally. In their cartographic approach to NC, Brugger & Poletto assume two negation related positions in the clausal spine, a lower one where ned is spelled out and a higher one corresponding to the negation operator which determines the scope of negation and licenses NIs. They use the above mentioned blocking effect on NC as a diagnostic for the position of their higher NegP which appears still below the position of frequentative adverbs. Weiß, on the other hand, takes the position of neamd in (42) as an indicator for scrambling of the NI

7 Weiß does not actually discuss the DN readings, only the ungrammaticality of the NC interpretation. While the DN readings are not easily accessible without context, they are definitely available and become more salient with appropriate intonation.

Focus Intervention and Double Negation in Bavarian

177

into the TP domain. As a consequence, neamd is unable to have its negation properties absorbed under Spec-Head agreement with ned. (42) weil neamd ned oft kema is because nobody neg often come aux a. *NC reading: ‘because nobody came often’ b. ?DN reading: ‘because nobody came rarely (i. e., not often)’ (43) weil neamd oft ned kema is because n-body often neg come aux a. *NC reading: ‘because nobody came often’ b. ?DN reading: ‘because nobody often didn’t come’

Such structural theories successfully account for the basic cases above. However, they make an empirical prediction that has not been commented upon so far: They predict that the unavailability of NC should persist when the intervening oft is overtly displaced. It should be impossible to reconstruct oft into a position within the scope of negation, since its only available base position is above NegP.8 Section 4.4.2 shows that this prediction is not borne out. 4.4.2 NC and ‘often’ with Topicalization Under a rise-fall intonation, or ‘hat contour’, which generally favors scope inverted readings in German (cf. Büring, 1997, 2003; Krifka, 1998), it is possible to have NC with the negation scoping over the adverbial oft. This is illustrated in (43). (44) OFT is NEAMD ned kema. often aux nobody not come ‘Nobody came OFTEN.’

The structure of example (44) is given in the tree in (45). I assume the position where oft reconstructs to be above the NI neamd but below OPØ. This allows us to derive the correct scope.9

8 In these theories, the availability of sentences with ned before oft has to be explained by treating ned oft (‘not often’, or ‘rarely’) as constituent negation. This is done explicitly in Brugger & Poletto (1993: 69). 9 Reconstructing ‘often’ below negation should be impossible if we assume a unique position for clausal negation. This is done in Weiß (1999) who equates the locus of semantic negation with the overt position of ned. Brugger & Poletto (1993) too assume that the (unique) base position of often is above the (unique) base position of clausal negation, even as if they assume that the latter is a covert category.

178 (45)

Jakob Steixner

CP 3 OFT3 C' 3 is1 TP 3 TP [iNeg]OP 3 t3 TP 3 [uNeg]NEAMD2 T' 3 vP t1 3 vP [uNeg]ned 6 t2 kema t1

The corresponding base string before movement of oft to SpecCP would thus have to be (46). With oft remaining in its middle-field position, only the surfacetrue scope where OFTEN scopes over negation is available, though. This reading is represented in (47). (46) weil oft neamd ned kema is because often nobody not come aux ‘because often nobody came.’ (Scope: OFTEN > Ø >$) (47)

CP 3 weil TP 3 oft TP 3 TP [iNeg]OP 3 T' [uNeg]neamd2 3 vP is 3 vP [uNeg]ned 6 t2 kema t1

That is, oft can reconstruct below the negative operator when it does not overtly intervene. When it does overtly intervene, as in (46), it blocks the licensing of

Focus Intervention and Double Negation in Bavarian

179

[uNeg] elements in its scope by a higher OPØ. The only available reading, unless oft moves out of the way, is thus one where OPØ is inserted below often, as in (47). This sensitivity to the overt position of interveners is typical of Beck-style intervention effects discussed in section 4.3, which can often be alleviated by scrambling over an intervener. This can be seen in (41), repeated here as (48). Here, just like in our examples, oft acts as an intervener in a multiple wh-context only when it is overtly realized between the two wh-phrases in (48a). In (48b), where it does not, no intervention occurs. (48) a. *Luise zählt auf, welche Uni oft welche Linguisten eingeladen hat. Luise enumerates which university often which linguists invited has b. Luise zählt auf, welche Uni welche Linguisten oft eingeladen hat. Luise enumerates which Uni which linguists often invited has ‘Luise enumerates, which university often invited which linguists.’

In this section, I have argued that the licensing relation of Bavarian NC shows sensitivity to intervention effects as discussed by Beck (2006), among others. Together with Zeijlstra’s (2008b) Phonological Economy Principle, the treatment of NC-blocking as a Beck-style intervention effect allows us to describe the distribution of DN readings in the data considered correctly. DN readings arise when an NC interpretation is impossible due to an intervention effect. In these cases, a second OPØ is inserted in a position below the intervener to prevent the derivation from failing. This analysis allows a uniform treatment of two seemingly distinct phenomena: Intervention of ‘often’ as described in the literature, and DN readings in narrow focus sentences.

5.

Summary and Outlook

I argued that DN readings of Bavarian sentences with multiple negative elements are best explained as intervention effects mediated by focus interpretation rather than locality violations of Spec-Head Agreement between NegP and NIs. The empirical motivation for this shift of explanatory burden is that the Absorption and Factorization approach as applied to Bavarian by Weiß (1999) turns out to be too restrictive. As a theoretical consequence, the clausal negative marker ned cannot be the unique nexus of NC. As in Zeijljstra’s (2004) theory, I analyse ned and NIs as semantically non-negative elements with an uninterpretable formal [uNEG] feature. Double Negation readings are derived by positing that a second negative operator is inserted as a last resort in precisely those cases where one or more [uNEG] feature would otherwise go unchecked due to Beck (2006) style intervention effects.

180

Jakob Steixner

I leave open the extent to which the analysis presented here can be generalised for other NC languages, at least those that readily allow for Double Negation readings. It is an empirical fact that not all Negative Concord languages allow Double Negation readings with equal ease. This is not entirely unexpected. There is well known cross-linguistic variation with respect to the triggers of intervention effects. In order to explain the total absence of DN readings in some NC languages, it might however be necessary to depart from a uniform analysis of the formal and semantic makeup of Negative Indefinites across languages. At any rate, it would be highly interesting to take a closer look at DN and NC readings in languages such as Hungarian (Bende-Farkas, 2005; Puskás, 2012) or Spanish and Catalan (Espinal & Prieto, 2011; Tubau et al., 2015) which have also been described in information structural terms. The existence of DN readings in NC languages poses an interesting problem. If it is possible to insert more than one negative operator in the clause, why does this possibility seem to be absent in the basic cases? Why are not all NC structures highly ambiguous with respect to a DN reading? This problem is not even specific to the Agreement theory of NC. In the context of the Absorption and Factorization Theory, it also remains unclear why NIs cannot simply always escape Absorption – as DN readings show they can in principle. I would like to speculate that this is ultimately due to the universal pragmatic markedness of negative utterances. Negative utterances function as rejections of a proposition that needs to be given or easily accommodated. They require their affirmative counterpart to be accessible through the previous discourse or world knowledge (Larrivée, 2015; Horn, 1991, 2010; Israel, 2004; Givón, 1978): In order to felicitously utter a sentence like “John did not come”, the proposition that he did must be present in the common ground, as an explicit claim or plausible assumption. DN readings are thus only felicitous where this holds of a proposition that is already negative. DN readings are thus pragmatically felicitous only when one negative operator can be construed as backgrounded. The correlation between information structure and DN interpretations is thus twofold: On the one hand, the binding of [uNEG] features fails through focus intervention for formal reasons, leading to the unavailability of NC readings in these constructions. On the other hand, explicit backgrounding of a negative element creates the prerequisites for the felicitous DN readings the presuppositions of which would otherwise be difficult to accommodate.

Focus Intervention and Double Negation in Bavarian

181

References Bayer, Josef (1990). What Bavarian Negative Concord Reveals about the Syntactic Structure of German. In Joan Mascaró & Marina Nespor (eds.), Grammar in Progress. Glow Essays for Henk van Riemsdijk, pp. 11–23. Dordrecht: Foris. Beck, Sigrid (1996). Quantified Structures as Barriers for LF Movement. Natural Language Semantics 4, 1–56. Beck, Sigrid (2006). Intervention Effects Follow from Focus Interpretation. Natural Language Semantics 14 (1), 1–56. Bende-Farkas, Ágnes (2005). Focus and Negative Concord in Hungarian. In Paul Dekker & Michael Franke (eds.), Proceedings of the Fifteenth Amsterdam Colloquium, pp. 47–52. ILLC/Department of Philosophy University of Amsterdam. Brugger, Gerhard & Cecilia Poletto (1993). On Negation in German and Bavarian. University of Venice Working Papers in Linguistics 2 (3), 41–79. Büring, Daniel (1997). The Meaning of Topic and Focus. The 59th Street Bridge accent. London: Routledge. Büring, Daniel (2003). On D-trees, beans, and B-accents. Linguistics and Philosophy 26 (5), 511–545. Chomsky, Noam (1995). The Minimalist Program. Cambridge: Cambridge University Press. Espinal, M. Teresa (2007). Licensing Expletive Negation and Negative Concord in Catalan and Spanish. In Franck Floricic (ed.), La negation dans les langues romanes, pp. 49–74. Amsterdam: John Benjamins. Espinal, M. Teresa & Pilar Prieto (2011). Intonational encoding of double negation in Catalan. Journal of Pragmatics 43 (9), 2392–2410. Giannakidou, Anastasia (2000). Negative . . . Concord? Natural Language and Linguistic Theory 18, 457–523. Givón, Thomas (1978). Negation in Language: Pragmatics, Function, Ontology. In P. Cole (ed.), Pragmatics (Syntax and Semantics 9), pp. 69–112. New York: Academic Press. Haegeman, Liliane (1995): The Syntax of Negation. Cambridge: Cambridge University Press. Haegeman, Liliane & Raffaella Zanuttini (1996). Negative Concord in West Flamish. In Adriana Belletti & Luigi Rizzi (eds.), Parameters and Functional Heads. Essays in Comparative Syntax, pp. 117–197. New York: Oxford University Press. Heim, Irene (1982). The Semantics of Definite and Indefinte Noun Phrases. Ph. D. thesis, University of Massachusetts. Hiraiwa, Ken (2001). Multiple agree and the defective intervention constraint in Japanese. MIT Working Papers in Linguistics 40, pp. 67–80. Horn, Lawrence R. (1991). Duplex negatio affirmat…: the economy of double negation. In L. Dobrin (ed.), CLS 27, Part Two: The Parasession on Negation, pp. 88–106. Horn, Lawrence R. (2010). Multiple negation in English and other languages. In Lawrence R. Horn (ed.), The Expression of Negation, pp. 111–148. Berlin: Mouton De Gruyter.

182

Jakob Steixner

Isac, Daniela (2004). Focus on Negative Concord. In Reineke Bok-Bennema et al. (eds.), Romance Languages and Linguistic Theory 2002. Selected Papers from “Going Romance”, Groningen 2002, pp. 119–140. Amsterdam: John Benjamins. Israel, Michael (2004). The pragmatics of polarity. In L. Horn & G. Ward (eds.), The Handbook of Pragmatics, pp. 701–723. Wiley-Blackwell. Krifka, M. (1998). Scope Inversion under the Rise-Fall Contour in German. Linguistic Inquiry 29 (1), 75–112. Lahiri, Utpal (1998). Focus and Negative Polarity in Hindi. Natural Language Semantics 6, 57–123. Laka, Itziar (1990). Negation in Syntax: On the Nature of Functional Categories and Projections. Ph. D. thesis, MIT. Larrivée, Pierre (2016). The Markedness of Double Negation. In Pierre Larrivée & Chungmin Lee (eds.), Negation and Polarity. Experimental and cognitive perspectives, pp. 177–198. Dordrecht: Springer. Penka, Doris (2011). Negative Indefinites. Oxford: Oxford University Press. Penka, Doris & Arnim von Stechow (2001). Negative Indefinita unter Modalverben. Linguistische Berichte (Sonderheft) 9, 263–286. Pesetsky, David & Esther Torrego (2007). The syntax of valuation and the interpretability of features. In Karimi, Simin et al. (eds.): Phrasal and Clausal Architecture. Syntactic derivation and interpretation, pp. 262–294. Amsterdam: John Benjamins. Poletto, Cecilia (2008). On negative doubling. Quaderni di lavoro ASIT 8: 57–84. Puskás, Genoveva (2012). Licensing Double Negation in NC and non-NC languages. Natural Language and Linguistic Theory 30 (2), 611–649. Rizzi, Luigi (1996). Residual verb second and the wh-criterion. In A. Belletti & L. Rizzi (eds.), Parameters and Functional Heads. Essays in Comparative Syntax, pp. 63–90. New York: Oxford University Press. Rooth, Matts (1992). A theory of focus interpretation. Natural language semantics 1(1), 75–116. Schwarz, Bernhard & Rajesh Bhatt (2006). Light Negation and Polarity. In R. Zanuttini et al., Crosslinguistic Research in Syntax and Semantics: Negation, Tense and Clausal Architecture, pp. 175–197. Washington D.C.: Georgetown University Press. Tubau, Susagna, Joan Borràs-Comes, Pilar Prieto, & M. Teresa Espinal (2015). Prosody and Metalinguistic Double Negation in Catalan and Spanish. In Pierre Larrivée & Chungmin Lee (eds.), Negation and Polarity. Experimental and cognitive perspectives, pp. 145–176. Dordrecht: Springer. Tomiaka, Satoshi (2007). Pragmatics of LF intervention effects: Japanese and Korean Wh-interrogatives. Journal of Pragmatics 39 (9), 1570–1590. Weiß, Helmut (1998). Syntax des Bairischen. Studien zur Grammatik einer natürlichen Sprache. Number 391 in Linguistische Arbeiten. Tübingen: Niemeyer. Weiß, Helmut (1999). Duplex negatio non semper affirmat: A theory of double negation in Bavarian. Linguistics 37, 819–846. Zanuttini, Raffaella (1997): Negation and Clausal Structure: A Comparative Study of Romance Languages. Oxford/New York: Oxford University Press. Zeijlstra, Hedde (2004). Sentential Negation and Negative Concord. Ph. D. thesis, Universiteit van Amsterdam.

Focus Intervention and Double Negation in Bavarian

183

Zeijlstra, Hedde (2008a). Negative Concord is Syntactic Agreement. Ms., available online at: http://ling.auf.net/lingBuzz/000645. Zeijlstra, Hedde (2008b). On the Syntactic Flexibility of Formal Features. In T. Biberauer (ed.) The Limits of Syntactic Variation. Amsterdam: John Benjamins. Zeijlstra, Hedde (2012). There is only one way to agree. The Linguistic Review 29 (3), 491–539.

Doris Penka

Splitting at most*

Abstract: Negative quantifiers are known to give rise to split scope readings, where some other operator, in particular a modal, takes scope in between the negative and the indefinite meaning component. The fact that split scope readings are not only restricted to negative quantifiers, but arise with downward monotonic quantifiers more generally, has been taken as an argument against a decompositional analysis. This paper discusses split scope readings that arise in connection with the superlative modifier at most, which shows a particularly interesting pattern of interaction with modals, as it naturally combines with possibility modals to yield split scope readings, but not with necessity modals. While recent pragmatic approaches to the meaning of superlative modifiers successfully account for the interaction of at least with modals, they fail for downward monotonic at most. I argue that the interaction of at most with modals can be explained if the pragmatic approach is amended with the assumption that at most is decomposed into at least plus an antonymizing operator defined in terms of degree negation. Keywords: split scope, superlative modifiers, superlative quantifiers, ignorance inferences, decomposition of antonyms

1.

Introduction

Negative quantifiers are known to give rise to split scope readings, where another operator, in particular a modal, takes scope in between the negative and the existential meaning component. As illustrated in the paraphrases of the following examples from German and English, the negation takes wide scope over * For valuable comments and suggestions, I would like to thank Irene Heim, Maribel Romero, Arnim von Stechow, and Daniel Büring, as well as the editors of this volume. This work was supported by DFG grant PE 1765/2–1 and the Zukunftskolleg at the University of Konstanz.

186

Doris Penka

the modal while the existential is interpreted with narrow scope under the de dicto reading.1 (1) Bei der Operation muss kein Anästhesist anwesend sein. at the surgery has-to no anesthetist present be ‘It is not required that an anesthetist be present during the surgery.’ (2) The company need fire no employees. ‘The company is not obligated to fire any employees.’

[German]

[from Potts 2000]

The most prominent reading of sentence (1), for instance, is the one paraphrased saying that the presence of an anaesthetist is not obligatorily required. It is not about a particular anaesthetist, but rather about the presence of some anaesthetist or other, corresponding to a de dicto reading of the indefinite. At the same time, the necessity modal is interpreted in the scope of negation, expressing the absence of an obligation. This split reading cannot be derived under the standard analysis of negative indefinites as negative quantifiers where the negation and the existential quantifier form a lexical unit. Several analyses have been proposed, some maintaining the assumption that negative quantifiers are lexical units (de Swart 2000; Abels & Martí 2010) others decomposing them into a negation and an indefinite (among others Jacobs 1980; Rullmann 1995; Penka 2011). As has been observed by de Swart (2000) and Abels & Martí (2010), split scope readings arise not only with negative quantifiers, but also with other downward monotonic quantifiers like few and numerals modified by fewer than or at most. This is illustrated in the following examples. (3) The inmates of this prison are allowed to write few letters. ‘The inmates of this prison are not allowed to write more than a small number of letters.’ (4) At MIT one needs to publish fewer than three books in order to get [Hackl 2000] tenure. ‘At MIT one doesn’t need to publish more than two books in order to get tenure.’ (5) A muslim can marry at most four women. ‘A muslim cannot marry more than four women.’

[de Swart 2000]

The fact that downward monotonic quantifiers in general give rise to split scope readings has been taken as an argument against approaches that derive split scope readings of negative indefinites by decomposing the quantifiers into a negation and a quantificational part. Supposing that split scope is a unified phenomenon, de Swart (2000) argues that the decomposition required is not always morphologically 1 Since the focus of the paper is on split scope readings, only these readings are paraphrased here and in the following examples although other readings may be available as well.

Splitting at most

187

transparent (fewer than three for instance would have to be decomposed into negation and more than two), which makes the analysis implausible. This paper focuses on split readings arising with at most. Considering at most is particularly interesting and instructive with respect to the question whether split scope should receive a unified analysis because at most gives rise to a split scope reading in combination with possibility modals, but not with necessity modals. This contrast in the availability of split readings, which does not arise for any other downward entailing quantifier, is illustrated in (6) and (7). Under its most salient reading, sentence (6) expresses prohibition of checking out more than ten books. This corresponds to the split reading: to express prohibition, the possibility modal is interpreted in the scope of negation (‘It is not allowed’ ‘It is prohibited’), while ten books is interpreted with narrow scope under the modal to express a de dicto reading (the sentence is not about particular books, but rather expresses that taking out any set of more than ten books is prohibited). (6) You can check out at most ten books from the library. ‘You are not allowed to check out more than ten books from the library.’ (7) ?You have to read at most ten books for this class. You are not required to read more than ten books for this class.

In contrast, the split reading is not available for (7) with a necessity modal.2 If it were, the sentence could be understood as expressing permission of reading ten or fewer books: Interpreting the necessity modal in the scope of negation would express absence of an obligation corresponding to permission of doing the opposite (‘You are not required to read more than ten books’ ‘You are allowed to read ten or fewer books’). But intuitively, this is not a possible reading of (7). What the sentence actually means is not easy to say either, because sentences where at most occurs under a necessity modal in general are odd and are harder to interpret than combinations of at most with possibility modals (see experimental studies by McNabb & Penka 2014a,b). In order to explain this asymmetry in the availability of split readings arising with at most, I build on a recent approach to the semantics of superlative modifiers like at least and at most (Büring 2008; Cummins & Katsos 2010; Schwarz 2011 and 2013; Kennedy 2015). It has been proposed to account for a characteristic of at least and at most that makes them a particularly interesting object of study in semantics and pragmatics, namely the fact that they give rise to 2 The sentence has a reading − the so called speaker insecurity reading − which might at first glance seem to express negation of the obligation of reading more than ten books. However, there is a crucial difference between this reading and the split scope reading, as will become clearer in the discussion in section 2 and 3: whereas sentence (6) under the split reading can be used to state a rule, (7) cannot.

188

Doris Penka

ignorance inferences. (That is, by using at least and at most, a speaker generally conveys that she is unsure about the precise value under discussion; see section 2.) But in its basic version this approach does not account for the interaction of at most with modals and cannot derive the spilt reading of at most under possibility modals. I propose a modified analysis of at most where at most is decomposed into its positive antonym at least and an antonymizing operator defined in terms of degree negation (Büring 2007, Heim 2008). This analysis provides a principled account of split readings arising with at most as well as ignorance inferences and their (non-)obviation under modals. The paper is organised as follows: Section 2 sets the stage for the analysis by discussing superlative modifiers and the ignorance inferences they give rise to. It also introduces a pragmatic approach to ignorance inferences analysing them as quantity implicatures and shows that it successfully accounts for the interaction of at least with modals, but not for at most. In section 3, I propose to decompose at most into an antonymizing operator and at least and show that this successfully accounts for the interaction of at most with modals under the pragmatic approach. Section 4 addresses the question whether split scope of different downward entailing quantifiers should receive a unified analysis. Section 5 summarizes and concludes the chapter.

2.

Superlative modifiers and ignorance inferences

2.1

Speaker insecurity and authoritative readings of at least and at most

The superlative modifiers at least and at most have recently received a lot of attention in the semantics and pragmatics literature. What makes them particularly interesting is the fact that in most contexts, they imply speaker ignorance, i. e. they convey that the speaker is not sure about the precise value under discussion (see Geurts & Nouwen 2007, Nouwen 2010). Sentence (8), for instance, conveys that the speaker is not sure how many beers exactly John had last night. The only thing she is sure about is that the number is not less than three. But for all she knows, John might have had four or more beers. (8) John had at least three beers last night.

In certain environments, however, the implication of speaker uncertainty is absent. In particular, it has been observed that ignorance inferences can be suppressed in certain combinations of at least and at most with modals (see Geurts & Nouwen 2007). Sentence (9), where at least occurs under a necessity modal, has a reading which Büring (2008) calls authoritative. Under this read-

Splitting at most

189

ing, (9) does not convey speaker ignorance, but rather expresses that 10pp is the required minimum length of the paper. This reading is graphically illustrated in (9a), where ‘––’ signifies the range of permissible paper lengths – which I will also refer to as deontic range – and 10pp is its lower bound. (9) The paper has to be at least 10 pages long. a. ‘10 pages is the required minimum length of the paper’ [-––––– 10pp

authoritative reading

b. ‘According to what the speaker knows, the required minimum length might be 10 pages or it might be more.’ [/////////–––––speaker insecurity reading 10pp

Sentence (9) has another reading conveying speaker ignorance, which can be brought out by prefixing the utterance with “I don’t know exactly. But I think …”. Under this reading, which Büring (2008) calls speaker insecurity reading, the speaker is unsure about the required minimum length of the paper. For all she knows, the lower bound of permissible paper lengths could be 10pp or more. This reading is graphically illustrated in (9b), where ‘/////’ signifies the epistemic range, i. e. the values that for all the speaker knows might or might not be permissible. To bring out the difference between the two readings, consider a situation in which the regulations specify that only papers that are 15pp or longer will be accepted. In this situation, the sentence is judged false under the authoritative reading (9a), whereas under the speaker insecurity reading (9b) the speaker cannot be blamed for making a false statement. Not all combinations of superlative modifiers and modals allow for both the authoritative and the speaker insecurity reading. Geurts & Nouwen (2007) observe that at least in combination with necessity modals and at most in combination with possibility modals have both the authoritative and the speaker insecurity reading, whereas possibility modals plus at least and necessity modals plus at most allow for the speaker insecurity reading only. (The question which readings are available for the combination of necessity modal and at most is actually more complex and will be discussed in more detail in section 3.4.) The readings that Geurts & Nouwen claim to be available for the different combinations are summarized in (9) to (12). (10) The paper can be at least 10 pages long. ‘According to what the speaker knows, the maximally allowed length might be 10 pages or more.’ –––––[///////// speaker insecurity reading only 10pp

190

Doris Penka

(11) The paper has to be at most 10 pages long. ‘According to what the speaker knows, the required minimum length might be 10 pages or less.’ /////////] ––––– speaker insecurity reading only 10pp (12) The paper can be at most 10 pages long. a. ‘10 pages is the maximally allowed length of the paper.’ ––––-] 10pp

authoritative reading

b. ‘According to what the speaker knows, the maximally allowed length might be 10 pages or less.’ –––––/////////] speaker insecurity reading 10pp

The ignorance implications of at least and at most and their interaction with modals are currently subject to a lot of work in semantics and pragmatics (amongst others Geurts & Nouwen 2007; Büring 2008; Cummins & Katsos 2010; Nouwen 2010 and 2015; Schwarz 2011 and 2013; Cohen & Krifka 2014; Coppock & Brochhagen 2013; Kennedy 2015). But none of the analyses proposed so far fully accounts for the interaction of at least and at most with modals. In this paper, I elaborate an approach that seems particularly attractive and derives ignorance as quantity implicatures. As we will see, this approach in its basic version does not account for the interaction of at most with modals and cannot derive the spilt reading of at most under possibility modals.

2.2

Ignorance inferences as quantity implicatures

One approach, pioneered by Büring (2008) and further developed by Cummins & Katsos (2010), Schwarz (2011, 2013) and Kennedy (2015), derives ignorance inferences of superlative modifiers in the pragmatic component. Building on a parallel to ignorance inferences arising with disjunction, which are generally taken to be derived via Gricean reasoning, the ignorance inferences triggered by at least and at most are analysed as quantity implicatures. Schwarz (2011, 2013) spells out an analysis of superlative modifiers as degree operators and shows that ignorance inferences of at least and at most can be derived in the same way as the ignorance implications of or in a neo-Gricean framework (Sauerland 2004).3 3 Mayr (2013) and Schwarz (2013) note that Sauerland’s algorithm needs to be revised and based on the notion of Innocent Exclusion (Fox 2007) in order to prevent the generation of unattested scalar implicatures for scalar modifiers. I neglect this issue for the purpose of this paper and circumvent the problem for Sauerland’s basic algorithm by considering just those scalar alternatives that asymmetrically entail the assertion and where the number is closest to the modified numeral.

Splitting at most

191

The essential ingredients of Schwarz’ analysis are the following: In the semantics, at least and at most are analysed as degree operators. As shown in (13) at least and at most take a degree (of type d) and a degree property (of type dt) as arguments and express non-strict comparison between the degree and the maximal degree of which the degree property holds. (13) a. [[at least]]= λdd. λDdt. max(D) ≥ d b. [[at most]]= λdd. λDdt. max(D) ≤ d

In the pragmatics, utterances with at least or at most trigger scalar alternatives which are the result of substituting (i) the modified number by other numerals or measure phrases and (ii) at least and at most by each other or exactly. (14) [[exactly]]= λdd. λDdt. max(D) = d

With these assumptions ignorance inferences are generated for unembedded occurrences of at least and at most in Sauerland’s (2004) system, where scalar implicatures and ignorance inferences are two sides of the same coin. Scalar implicatures arise if primary implicatures of the form “the speaker is not certain that ϕ”, where ϕ is a stronger scalar alternative, can be strengthened to secondary or scalar implicatures of the form “the speaker is certain that not ϕ”. Ignorance inferences arise if the stronger alternatives cannot simultaneously be false while the assertion is true, or putting it differently, the assertion is equivalent to the disjunction of the stronger alternatives. In this case the alternatives are called ‘symmetric’. The derivation of ignorance inferences is illustrated in the following for example (15). Since according to the meaning rule in (13) at least 10pp is a quantifier over degrees, it has to undergo QR resulting in the LF (16a). Under the standard semantics of gradable adjectives (Heim 2000), shown in (16b) for long, the truth conditions in (16c) are derived, according to which the length of the paper is 10pp or more. (15) The paper is at least 10 pages long. (16) a. [at least 10pp] [ λd [the paper is d-long]] b. [[long]]= λdd. λxe. length(x) ≥ d c. max{d: length(the paper) ≥ d} ≥ 10pp length(the paper) ≥ 10pp henceforth abbreviated as max{d: long(p,d)} ≥ 10pp

In a neo-Gricean pragmatic framework, this meaning is considered against alternative assertions the speaker could have made instead. The scalar alternatives of (15) correspond to all the sentences where the numeral modifier at least is substituted by exactly or at most and/or the measure phrase 10pp is substituted by other paper lengths, cf. (17). Out of these, the alternatives that are more

192

Doris Penka

informative, i. e. asymmetrically entail the assertion, are the ones formed by substituting either at least by exactly or 10pp by 11pp, as shown in (18). (17) Scalar alternatives to (15): The paper is NumMod n pages long.

where NumMod ∈ {at least, exactly, at most} n ∈ { …, 9, 10, 11, …}

(18) Stronger scalar alternatives: a. The paper is exactly 10 pages long. max{d: long(p,d)} = 10pp b. The paper is at least 11 pages long. max{d: long(p,d)} ≥ 11pp *4 max{d: long(p,d)} > 10pp

at least º exactly 10pp º 11pp

Because the stronger alternatives in (18) are symmetric, i. e. one of them has to be true for the assertion to be true, none of the primary implicatures can be strengthened to secondary/scalar implicatures, because this would contradict the conjunction of the assertion and all the primary implicatures. Instead, the assertion A and the primary implicatures PI1 and PI2 taken together entail possibility implicatures. This is shown in (19), where using Gazdar’s (1979) notation, Kϕ corresponds to ‘the speaker knows/believes ϕ’ and Pϕ to ‘the speaker considers ϕ possible’. (19) a. Assertion: b. Primary implicatures:

K max{d: long(p,d)} ≥ 10pp ØK max{d: long(p,d)} = 10pp ØK max{d: long(p,d)} > 10pp c. Possibility implicatures: P max{d: long(p,d)} = 10pp P max{d: long(p,d)} > 10pp

A PI1 PI2 (follows from A + PI2) (follows from A + PI1)

The primary and possibility implicatures together correspond to ignorance inferences, which are of the form Pϕ & PØϕ (note that ØKϕ is equivalent to PØϕ). According to them, the speaker does not know whether the paper is exactly 10pp long or whether the paper is more than 10pp long. Together with the assertion, this correctly reflects the meaning of sentence (15). (20) Ignorance implicatures generated: a. P max{d: long(p,d)} = 10pp & P Ømax{d: long(p,d)} = 10pp b. P max{d: long(p,d)} > 10pp & P Ømax{d: long(p,d)} > 10pp

Ignorance inferences for unembedded occurrences of at most are generated in the same way, the only difference being that now the alternative with a lower numeral is symmetric to the alternative where at most is substituted by exactly. 4 The symbol * is used when the equivalence is based on the simplifying assumption that the relevant scale is discrete, i. e. that only full-page lengths are considered. I make this assumption for expository reasons; the analysis works in the same way when dense scales are considered.

193

Splitting at most

(21) The paper is at most 10 pages long. max{d: long(p,d)} ≤ 10pp (22) Stronger scalar alternatives: a. The paper is exactly 10 pages long. max{d: long(p,d)} = 10pp b. The paper is at most 9 pages long. max{d: long(p,d)} ≤ 9pp * max{d: long(p,d)} < 10pp

at most º exactly 10pp º 9pp

(23) Ignorance implicatures generated: a. P max{d: long(p,d)} = 10pp & P Ømax{d: long(p,d)} = 10pp b. P max{d: long(p,d)} < 10pp & P Ømax{d: long(p,d)} < 10pp ‘The speaker doesn’t know whether the paper is exactly 10pp long or whether the paper is less than 10pp long.’

The neo-Gricean analysis thus accounts for the fact that unembedded occurrences of superlative modifiers give rise to ignorance inferences. It also makes certain predictions for the interaction of superlative modifiers with modals, which are discussed in the following subsections.

2.3

Interaction of superlative modifiers with necessity modals

Since superlative modifiers are analysed as degree operators, two different scope orders are possible when they interact with modals. If a superlative modifier is interpreted in the scope of a necessity modal, the stronger scalar alternatives in (26) are not symmetric. That is, they can simultaneously be false while the assertion is true. For example (24) this is the case if the permissible paper length corresponds to a range including 10pp and more. Because the alternatives are not symmetric, primary implicatures get strengthened to scalar implicatures in Sauerland’s (2004) system, as shown in (27). (24) The paper has to be at least 10 pages long. (25) a. has to [[at least 10pp] λd [the paper be d-long]] b. □ max{d: long(p,d)} ≥ 10pp ‘In all the acceptable worlds, the length of the paper is 10pp or more.’

□ > at least

(26) Stronger scalar alternatives: a. □ max{d: long(p,d)} = 10pp at least º exactly ‘In all the acceptable worlds, the length of the paper is exactly 10pp.’ b. □ max{d: long(p,d)} ≥ 11pp * 10pp º 11pp □ max{d: long(p,d)} > 10pp ‘In all the acceptable worlds, the paper is longer than 10pp.’

194

Doris Penka

(27) a. Assertion: K □max{d: long(p,d)} ≥ 10 b. Primary implicatures: Ø K □max{d: long(p,d)} = 10pp Ø K □max{d: long(p,d)} > 10pp c. Scalar implicatures: K Ø□max{d: long(p,d)} = 10pp K Ø□max{d: long(p,d)} > 10pp

According to the scalar implicatures generated, the speaker is sure that the paper does not have to be exactly 10pp long and that the paper does not have to be more than 10pp long. Together with the asserted content this is true iff the permissible paper lengths correspond to a range of values whose lower bound is 10pp. This corresponds to the authoritative reading illustrated in (28). (28)

[-––––– 10pp

authoritative reading

If a superlative modifier takes wide scope over a necessity modal, the speaker insecurity reading results. Although the scope order at least > □ is truth-conditionally equivalent to □ > at least (see Heim 2000), the pragmatic reasoning is different. Because wide scope of at least and exactly in the alternatives leads to symmetric alternatives – just as in the case of unembedded occurrences – ignorance inferences rather than scalar implicatures are generated. (29) a. [at least 10pp] λd [ has to [the paper be d-long]] b. max{d: □ long(p,d)} ≥ 10pp ‘The required minimum length of the paper is 10pp or it is more.’

at least > □

(30) Stronger scalar alternatives:5 a. max{d: □ long(p,d)} = 10pp at least º exactly ‘The required minimum length of the paper is exactly 10pp.’ b. max{d: □ long(p,d)} ≥ 11pp * max{d: □ long(p,d)} > 10pp 10pp º 11pp ‘The required minimum length of the paper is more than 10pp.’ (31) Ignorance implicatures generated: a. P max{d: □ long(p,d)} = 10pp & P Ø max{d: □ long(p,d)} = 10pp b. P max{d: □ long(p,d)} > 10pp & P Ø max{d: □ long(p,d)} > 10pp

These ignorance implicatures express that the speaker is unsure about the required minimum length of the paper; she does not know whether it is exactly 10pp or more than 10pp. Together with the asserted content, this corresponds to the speaker insecurity reading. (32)

[/////////–––––10pp

speaker insecurity reading

5 For reasons of space, from now on I only state the truth-conditions of the alternatives along with the substitutions by which they are generated from the LF under consideration.

195

Splitting at most

Under the neo-Gricean account the two readings of sentences with a necessity modal and at least come down to a difference in scope: The authoritative reading arises if the superlative modifier is interpreted in the scope of the necessity modal, and the speaker insecurity reading results from wide scope of the superlative modifier. As evidence for this scopal ambiguity, Büring (2008) observes that only the authoritative reading is available if movement of at least over the modal is blocked for independent reasons, in particular if at least is contained within a finite clause. Sentence (33) only has the authoritative reading (28), and the speaker insecurity reading (32) is absent. It thus contrasts with the minimally different (24), where at least occurs in an infinitival and both readings are available. (33) It is required that the paper be at least 10 pages long.

The derivations for at most are again parallel to those for at least. If at most takes narrow scope under the necessity modal as in (34) the stronger scalar alternatives are not symmetric and the authoritative reading is derived, as shown in (35) to (38). (34) The paper has to be at most 10 pages long. (35) a. has to [[at most 10pp] λd [the paper be d-long]] □ > at most b. □ max{d: long(p,d)} ≤ 10pp ‘In all the acceptable worlds, the length of the paper is 10pp or less.’ (36) Stronger scalar alternatives: a. □ max{d: long(p,d)} = 10pp at most º exactly ‘In all the acceptable worlds, the length of the paper is exactly 10pp.’ b. □ max{d: long(p,d)} ≤ 9pp * 10pp º 9pp □ max{d: long(p,d)} < 10pp ‘In all the acceptable worlds, the paper is shorter than 10pp.’ (37) Scalar implicatures generated: a. KØ□max{d: long(p,d)} = 10pp b. KØ□max{d: long(p,d)} < 10pp ‘The speaker is sure that the paper doesn’t have to be exactly 10pp long and that the paper doesn’t have to be less than 10pp long.’ (38)

–––––-] 10pp

authoritative reading

If at most takes wide scope, the stronger scalar alternatives are symmetric and the speaker insecurity reading is derived, as shown in (39) to (42). (39) a. [at most 10pp] [ λd [ has to [the paper be d-long]]] b. max{d: □ long(p,d)} ≤ 10pp ‘The required minimum length of the paper is 10pp or less.’

at most > □

196

Doris Penka

(40) Stronger scalar alternatives: a. max{d: □ long(p,d)} = 10pp at most º exactly ‘The required minimum length of the paper is exactly 10pp.’ b. max{d: □ long(p,d)} ≤ 9pp * 10pp º 9pp max{d: □ long(p,d)} < 10pp ‘The required minimum length of the paper is less than 10pp.’ (41) Ignorance implicatures generated: a. P max{d: □ long(p,d)} = 10pp & P Ø max{d: □ long(p,d)} = 10pp b. P max{d: □ long(p,d)} < 10pp & P Ø max{d: □ long(p,d)} < 10pp ‘The speaker doesn’t know whether the required minimum length of the paper is exactly 10pp or whether the required minimum length of the paper is less than 10pp.’ (42)

/////////]–––––10pp

speaker insecurity reading

In general, ignorance inferences are obviated if a superlative modifier is interpreted in the scope of an operator that breaks symmetry. This arguably also accounts for other cases of ignorance obviation, e. g. under universal quantifiers (Schwarz 2011) and generics (Nouwen 2010). However, we will see in the next section that a possibility modal does not break symmetry.

2.4

Interaction of superlative modifiers with possibility modals

While we just saw that a necessity modal breaks symmetry if a superlative modifier is interpreted in its scope and scalar rather than ignorance implicatures arise, a possibility modal does not break symmetry. Even if the possibility modal takes wide scope over a superlative modifier, the stronger scalar alternatives cannot simultaneously be false while the assertion is true. Therefore ignorance inferences are generated. This is shown in the following for at most (the derivations for at least are again parallel). If at most is interpreted in the scope of the modal as in (44), a reading results which is weak for several reasons. For one thing the truth conditions merely say that there is an acceptable world where the length of the paper is 10pp or less. This allows for other lengths also being permissible. (43) The paper can be at most 10 pages long. (44) a. can [[at most 10pp] [λd [the paper be d-long]]] ^ > at most b. ^ max{d: long(p,d)} ≤ 10pp ‘There is an acceptable world where the length of the paper is 10pp or less.’

197

Splitting at most

In the pragmatic component, strong ignorance inferences are generated: (45) Stronger scalar alternatives: a. ^max{d: long(p,d)} = 10pp at most º exactly ‘There is an acceptable world where the length of the paper is exactly 10pp.’ b. ^max{d: long(p,d)} ≤ 9 pp * 10pp º 9pp ^max{d: long(p,d)} < 10pp ‘There is an acceptable world where the length of the paper is less than 10pp.’ (46) Ignorance inferences generated: a. P ^max{d: long(p,d)} = 10pp & P Ø^max{d: long(p,d)} = 10pp b. P ^max{d: long(p,d)} < 10pp & P Ø^ max{d: long(p,d)} < 10pp

According to these ignorance inferences the speaker does not know whether the paper can be exactly 10pp long or whether the paper can be less than 10pp long. Thus, for all the speaker knows, the maximally allowed length might be 5pp or the required minimum length might be 10pp. Since this does not exclude a lot of epistemic alternatives, this reading is very weak. This weak reading might not be detectable because there is another reading with stronger truth conditions and sensible ignorance inferences, derived from an LF where the superlative modifier takes wide scope as in (47): (47) a. [at most 10pp] λd [allowed [the paper be d-long]] b. max{d: ^long(p,d)} ≤ 10pp ‘The maximally allowed length of the paper is 10pp or less.’

at most > ^

(48) Stronger scalar alternatives: a. max{d: ^long(p,d)} = 10pp at most º exactly ‘The maximally allowed length of the paper is exactly 10pp.’ b. max{d: ^long(p,d)} ≤ 9pp * 10pp º 9pp max{d: ^long(p,d)} < 10pp ‘The maximally allowed length of the paper is less than 10pp.’ (49) Ignorance implicatures generated: a. P max{d: ^long(p,d)} = 10pp & P Ømax{d: ^long(p,d)} = 10pp b. P max{d: ^long(p,d)} < 10pp & P Ømax{d: ^long(p,d)} < 10pp

According to these ignorance inferences the speaker is not sure whether the maximally allowed length of the paper is exactly 10pp or whether the maximally allowed length is less than 10pp. Together with the asserted content, this corresponds to the attested speaker insecurity reading illustrated in (50). (50)

–––––/////////] 10pp

speaker insecurity reading

198 2.5

Doris Penka

Summary of predictions of the neo-Gricean account

In sum, the neo-Gricean account of ignorance inferences makes the following predictions regarding the interaction of superlative modifiers with modals: If at least or at most are interpreted with wide scope over a necessity or possibility modal, the speaker insecurity reading conveying speaker ignorance is derived. The authoritative reading results if at least or at most are interpreted in the scope of a necessity modal. Narrow scope under a possibility modal leads to a reading with strong ignorance inferences. This correctly accounts for the readings observed for at least. As discussed in section 2.1 at least gives rise to an authoritative reading in combination with a necessity modal (cf. (9)), but not in combination with a possibility modal (cf. (10)). At most, in contrast, gives rise to an authoritative split reading in combination with a possibility modal (cf. (12)). While the truth-conditions derived from an LF where at most takes wide scope correspond to the split reading excluding worlds where the paper is longer than 10pp, the reading is not authoritative because ignorance inferences are generated in the pragmatic component. Therefore the neo-Gricean analysis, while successful for at least, does not account for the interaction of at most with modals.

3.

A decompositional analysis of at most

3.1

Decomposing at most

In order to account for the split reading of at most and the particular pattern of interaction with modals, I propose that at most is morpho-syntactically complex. Building on the idea that marked members of antonym pairs like short (vs. long) or little (vs. much) are generally decomposed in the syntax into an antonymizing operator and the corresponding positive antonym (Büring 2007; Heim 2008; Alxatib 2013), I propose that at most n is decomposed into an antonymizing operator ant and at least n: (51) at most n = [[ant-n]d(dt)-at least](dt)t

As for the meaning of at least I adopt the degree operator semantics proposed by Schwarz (2011) and Kennedy (2015), repeated as (52) from (13a) above. (52) [[at least]]= λdd. λDdt. max(D) ≥ d

For the semantics of the antonymizing operator ant, an obvious candidate would be Heim’s (2006) degree operator little, which is defined in (53) and

Splitting at most

199

expresses degree negation (meaning that a degree property D does not hold to degree d) (see also Alxatib 2013). (53) [[little]]= λdd. λDdt. ØD(d)

[Heim 2006]

Heim uses this meaning of little to derive the split reading of sentences like (54) from the LF where little takes inverse scope over the modal. (This arguably also works for split scope of few as in example (3).) (54) We can grow very little before we run out of space. [Heim 2006] ‘It is not possible for us to grow more than very little before we run out of space.’

With this definition of little, however, at most n cannot be decomposed into little plus at least n. Instead, at most n would have to be decomposed into little plus more than n, which is not in line with the idea that negative antonyms involve their positive counterparts. But there are in fact independent reasons why Heim’s definition of little needs to be revised (see also Beck 2012). The definition in (53) does not derive the intended meaning of sentences like (55), where that serves as direct degree argument of weigh and anaphorically picks up the measure phrase 40 kg. With the definition of little in (53) and the meaning of weigh in (56b) in analogy to the standard semantics of gradable adjectives, the truth conditions in (56c) are derived, according to which Sue weighs less than 40 kg. But this does not correctly render the meaning of sentence (55). Intuitively, (55) is perfectly compatible with Sue weighing exactly 40 kg, and in order for the presupposition of too to be fulfilled, Sue needs to have the same weight as Mary, i. e. 40 kg. (55) [Mary only weighs 40 kg.] Sue weighs that little too. (56) a. [that little] [λd [ Sue d-weighs]] b. [[weigh]]= λdd. λxe. weight(x) ≥ d c. Ø[weight(s) ≥ 40 kg] weight(s) < 40 kg

To derive the correct meaning for sentence (55), we need a definition of little in which only higher degrees are negated, but not the degree contributed by the first argument: (57) [[little]]= λdd. λDdt. "d’ > d: ØD(d’)

With this revised definition we get the truth conditions (58), according to which Sue does not weigh more than 40 kg. After strengthening by scalar implicature the

200

Doris Penka

meaning is that Sue’s weight is exactly 40 kg, which correctly captures the meaning of sentence (55).6 (58) "d’ > 40 kg: Ø[weight(s) ≥ d’] Ø[weight(s) > 40 kg]

I thus conclude that (57) is the correct definition of little (it also works for the split scope cases Heim 2006 discusses) and for the antonymizing operator ant more generally.7 Note that this renders the antonymizing operator ant equivalent to the non-complex meaning Schwarz (2011, 2013) assigns to at most, cf. (13b) from above (see also Beck 2012). (59) [[ant]]= λdd. λDdt. "d’>d: ØD(d’) (13b) [[at most]]= λdd. λDdt. max(D) ≤ d

Since in the analysis that I propose at most consists of the antonymizing operator ant (corresponding to Schwarz’ at most) and at least, more complex LFs are generated than under the analysis of Schwarz. This has important consequences for the interaction with modals, as we will see shortly.

3.2

Alternatives and ignorance inferences of at most

Before we can turn to the interaction with modals, we first need to address the question what consequences the decompositional analysis has for the pragmatics. We need to know in particular what the scalar alternatives are for an utterance with at most. In this respect I follow Katzir (2007) and Fox & Katzir (2011), who argue that alternatives are structurally defined and generated by substitution of lexical categories and deletion. In particular, I assume that the scalar alternatives for an utterance with at most are generated by (i) substituting numerals or measure phrases by each other; (ii) substituting at least by exactly (see Schwarz 2011) and (iii) deleting ant (see Alxatib 2013). In addition, I adopt the common assumption that modals are substituted in the alternatives. With these assumptions about the meaning of at most and scalar alternatives, ignorance inferences for unembedded occurrences of at most are generated in the same way as above: the stronger scalar alternatives, which are the same as 6 See Rett (2007) on how to derive the fact that negative antonyms are generally evaluative, e. g. that (55) conveys that 40 kg falls below the standard weight of women comparable to Sue. Incidentally, for Rett’s analysis to work, the antonymizing operator also needs to be defined as in (57) rather than as in (53). 7 Under the decompositional approach to antonyms, little would be ant plus much. Since much is semantically vacuous, the meaning of ant corresponds to the meaning of non-decomposed little.

201

Splitting at most

under Schwarz’ (2011) analysis, are symmetric and thus ignorance inferences are generated. This is illustrated in the following for sentence (60). (60) The paper is at most 10 pages long. (61) a. ant-10pp [λd2 [at least-d2 [λd1 [the paper is d1-long]]]] b. "d’ > 10pp: Ø [max{d: long(p,d)} ≥ d] Ø max{d: long(p,d)} > 10pp (62) Scalar alternatives: The paper is Polarity NumMod n pages long.

(63) Stronger scalar alternatives: a. The paper is exactly 10 pages long max{d: long(p,d)} = 10pp b. The paper is at most 9 pages long. Ø max{d: long(p,d)} > 9pp * max{d: long(p,d)} < 10pp

where Polarity ∈ { ant, ∅ } NumMod ∈ { at least, exactly} n ∈ { …, 9, 10, 11, …} ant, at least º exactly 10pp º 9pp

(64) Ignorance inferences generated: a. P max{d: long(p,d)} = 10pp & P Ømax{d: long(p,d)} = 10pp b. P max{d: long(p,d)} < 10pp & P Ømax{d: long(p,d)} < 10pp ‘The speaker isn’t sure whether the paper is exactly 10pp long or whether the paper is less than 10pp long.’

3.3

Interaction of at most with possibility modals

In the discussion of the interaction of at most with modals, let us start with possibility modals, for which the basic neo-Gricean account cannot derive the attested split reading. Recall that the split reading, which is the most prominent reading of sentence (65), is authoritative and expresses prohibition of the paper being longer than 10pp. Under the decompostional analysis of at most, the three different scope orders in (66) are possible for this sentence. (65) The paper can be at most 10 pages long. (66) a. can [ant-10pp [λd2 [at least-d2 [λd1 [the paper be d1-long]]]]] ^ > ant > at least b. ant-10pp [λd2 [at least-d2 [λd1 [can [the paper be d1-long]]]]] ant > at least > ^ c. ant-10pp [λd2 [can [at least-d2 [λd1 [the paper be d1-long]]]]] ant > ^ > at least

Crucially, the decompositional analysis makes available the LF (66c) where ant takes wide and at least takes narrow scope with respect to the modal. Under this scope order the alternatives are not symmetric and thus scalar implicatures are generated resulting in the authoritative reading, as shown in detail in the following.

202

Doris Penka

(67) a. ant-10pp [λd2 [can [at least-d2 [λd1 [the paper be d1-long]]]]] ant > ^ > at least b. "d’ > 10pp: Ø^ max{d: long(p,d)} ≥ d Ø^ max{d: long(p,d)} > 10pp ‘There is no acceptable world where the length of the paper is more than 10pp.’

This LF already looks promising in terms of its truth conditions, according to which the paper is not allowed to be longer than 10pp. This is definitely part of the meaning intuitively conveyed by sentence (65) under the split reading. In addition, pragmatic inferences arise by considering the scalar alternatives of the following form: (68) Scalar alternatives: The paper is Polarity Modal NumMod n pages long.

where Polarity ∈ {ant, ∅} Modal ∈ {allowed, required} NumMod ∈ {at least, exactly} n ∈ {…, 9, 10, 11, …}

We now have to consider eight scalar alternatives. It turns out that out of these, only the two shown in (69) asymmetrically entail the assertion.8 (69) Stronger scalar alternatives: a. Ø^ max{d: long(p,d)} > 9pp * □ max{d: long(p,d)} < 10pp b. □ max{d: long(p,d)} = 10pp

10pp º 9pp ant, ^ º □, at least º exactly

Crucially, (69a) generated by substituting the numeral with a lower value does not have a symmetric counterpart, i. e. there is no alternative which together with (69a) exhausts the assertion. This is due to the fact that ant, which has the semantics attributed by Schwarz (2011) to at most, can be deleted in the alternatives but not substituted by exactly. The alternative (69a) thus leads to the scalar implicature (70a), according to which the speaker is sure that the paper does not have to be shorter than 10pp. In addition, the alternative (69b) also leads to a scalar implicature (70b), according to which the speaker is sure that the paper does not have to be exactly 10pp long. (70) Scalar implicatures generated: a. K Ø□ max{d: long(p,d)} < 10pp b. K Ø□ max{d: long(p,d)} = 10pp

Taken together, the assertion and the scalar implicatures express that the permissible paper lengths correspond to a range of values whose upper bound is

8 The truth conditions in (69a) can be derived by two different substitutions: either by substituting 10pp by 9pp (as shown) or by substituting at least by exactly and 10pp by 9pp. In case of equivalent alternatives, I only consider the one requiring the fewest substitutions.

203

Splitting at most

10pp. In other words, we derive the authoritative reading (71), which corresponds to the split reading of (65). (71)

––––--] 10pp

authoritative reading

This shows that the decompositional analysis can derive the split (authoritative) reading for the combination of at most with a possibility modal, which other analyses fail to account for. In the analysis I propose, the split reading is derived from an LF where ant takes wide scope. This leads to the prediction that the split (authoritative) reading should not be available if movement out of the scope of the modal is blocked for independent reasons. Evidence that this prediction is borne out comes from sentences like (72), where at most is embedded in a finite clause. While the sentence is less than perfect and hard to interpret, it seems clear that it does not have the split (authoritative) reading, according to which 10pp is the maximally allowed length of the paper.9 (72) It is permitted that the paper is at most 10 pages long.

In addition to the split authoritative reading, the other two readings that the basic neo-Gricean approach discussed in section 2 derives are also generated from the other two available LFs (66a) and (66b). In general, if ant and at least take adjacent scope, the same pragmatic inferences and readings are derived as for Schwarz’ (2011) non-decomposed at most. If both ant and at least are interpreted in the scope of the possibility modal, we get the same symmetric stronger alternatives (74a) and (74b) as in (45). These lead to the ignorance inferences (75). Because we now also consider alternatives which are generated by replacing the possibility modal with a necessity modal, we also derive the scalar implicatures (76) on the basis of the nonsymmetric stronger alternatives (74c) and (74d). (73) a. can [ant-10pp [λd2 [ at least-d2 [λd1 [the paper be d1-long]]]]] ^ > ant > at least b. ^"d’ > 10pp: Ø[max{d: long(p,d)} ≥ d’] ^ max{d: long(p,d)} ≤ 10pp ‘There is an acceptable world where the length of the paper is 10pp or less.’

9 The absence of the authoritative reading if wide scope of the modal is enforced also provides an argument against attributing obviation of ignorance inferences under possibility modals to a Free Choice effect (see Coppock & Brochhagen 2013 and Nouwen 2015 for proposals in this direction). Current analyses of Free Choice effects arising with disjunction and indefinites (Fox 2007 among others) derive Free Choice permission readings from an LF where the possibility modal takes wide scope. If the authoritative reading of at most was due to a Free Choice effect, then the fact that (72) does not have the authoritative reading is unexpected.

204

Doris Penka

(74) Stronger scalar alternatives: a. ^max{d: long(p,d)} ≤ 9pp * ^max{d: long(p,d)} < 10pp b. ^max{d: long(p,d)} = 10pp c. □ max{d: long(p,d)} ≤ 10pp d. □ max{d: long(p,d)} = 10pp

10pp º 9pp ant, at least º exactly ^º□ ant, ^ º □, at least º exactly

(75) Ignorance Inferences generated: a. P ^ max{d: long(p,d)} < 10pp & P Ø^ max{d: long(p,d)} < 10pp b. P ^ max{d: long(p,d)} = 10pp & P Ø^ max{d: long(p,d)} = 10pp ‘The speaker does not know whether the paper is allowed to be exactly 10pp long and whether the paper is allowed to be less than 10pp long.’ (76) Scalar implicatures generated: a. K Ø□ max{d: long(p,d)} ≤ 10pp b. K Ø□ max{d: long(p,d)} = 10pp ‘The speaker is sure that the paper does not have to be exactly 10pp long and that the paper does not have to be 10pp long or shorter.’

As discussed for Schwarz’ analysis above, the derived ignorance inferences are strong. Although the truth conditions and scalar implicatures taken together now entail that the paper does not have to be one specific length, the resulting reading is still weak. In any case, it does not express an upper or lower bound to the lengths that are permissible according to the knowledge of the speaker. Because this reading is not informative, I assume, following Schwarz, that it is not detectable. The speaker insecurity reading is derived from the LF (66b), where both ant and at least take scope over the possibility modal. In this case, the derivation is completely parallel to (47)−(49) above: There are two symmetric stronger alternatives, which lead to ignorance implicatures. (77) a. ant-10pp [λd2 [at least-d2 [λd1 [can [the paper be d1-long]]]]] ant > at least > ^ b. "d’ > 10pp: Ø[max{d: ^long(p,d)} ≥ d’] Ø[max{d: ^long(p,d)} > 10pp] max{d: ^long(p,d)} ≤ 10pp ‘The maximally allowed length of the paper is 10pp or less.’ (78) Stronger scalar alternatives: a. max{d: ^long(p,d)} ≤ 9pp * max{d: ^long(p,d)} < 10pp b. max{d: ^long(p,d)} = 10pp

10pp º 9pp ant, at least º exactly

(79) Ignorance inferences generated: a. P max{d: ^long(p,d)} = 10pp & P Ømax{d: ^long(p,d)} = 10pp b. P max{d: ^long(p,d)} < 10pp & P Ømax{d: ^long(p,d)} < 10pp ‘The speaker is not sure whether the maximally allowed length is exactly 10pp or less than 10pp.’ (80)

–––––/////////] 10pp

speaker insecurity reading

205

Splitting at most

3.4

Interaction of at most with necessity modals

To complete the discussion of the readings the decompositional analysis makes available, we also need to re-consider the interaction of at most with a necessity modal, as in (81). Crucially, we need to make sure that we do not derive an unattested split-scope reading granting permission for the paper to be 10pp or shorter. The three scope orders in (82) have to be considered. (81) The paper has to be at most 10 pages long. (82) a. has to [ant-10 [λd2 [at least-d2 [λd1 [the paper be d1-long]]]]] □ > ant > at least b. ant-10 [λd2 [at least-d2 [λd1 [has to [the paper be d1-long]]]]] ant > at least > □ c. ant-10 [λd2 [has to [ at least-d2 [λd1 [the paper be d1-long]]]]] ant > □ > at least

The LF (82c) would seem to be the basis for the split scope reading, according to which the paper does not have to be longer than 10pp and which is in fact not attested. But while the truth-conditions (83b) express permission for the paper to be no longer than 10pp, strong ignorance inferences based on the symmetric alternatives (84a) and (84b) are generated in the pragmatic component. As before, I assume that the strong ignorance reading is masked by the existence of the speaker insecurity reading with sensible ignorance inferences. (83) a. ant-10pp [λd2 [has to [at least-d2 [λd1 [the paper be d1long]]]]] b. "d’ > 10pp: Ø□ max{d: long(p,d)} ≥ d Ø□ max{d: long(p,d)} > 10pp ‘The paper does not have to be more than 10pp long.’ (84) Stronger scalar alternatives: a. Ø□ max{d: long(p,d)} > 9pp b. ^ max{d: long(p,d)} = 10pp c. Ø^ [max{d: long(p,d)} > 10pp]

ant > □ > at least

10pp º 9pp ant, □ º ^, at least º exactly □º^

(85) Ignorance inferences generated: a. P Ø□ max{d: long(p,d)} > 9pp & P □ max{d: long(p,d)} > 9pp b. P ^ max{d: long(p,d)} = 10pp & P Ø^ max{d: long(p,d)} = 10pp ‘The speaker is not sure whether the paper is required to be longer than 9pp and she is not sure whether the paper is allowed to be exactly 10pp long.’ (86) Scalar implicature generated: K ^ max{d: long(p,d)} > 10pp ‘The speaker is sure that the paper is allowed to be longer than 10pp.’

The speaker insecurity reading is derived from the LF where ant and at least take adjacent scope over the modal (which makes the derivation again equivalent to the one for non-decomposed at most):

206

Doris Penka

(87) a. ant-10pp [λd2 [at least-d2 [λd1 [has to [the paper be d1-long]]]]] ant > at least > □ b. "d’ > 10pp: Ø[max{d: □long(p,d)} ≥ d’] Ø[max{d: □ long(p,d)} > 10pp] max{d: □ long(p,d)} ≤ 10pp ‘The required minimum length of the paper is 10pp or less.’ (88) Relevant scalar alternatives: a. max{d: □ long(p,d)} ≤ 9pp * max{d: □ long(p,d)} < 10pp b. max{d: □ long(p,d)} = 10pp

10pp º 9pp ant, at least º exactly

(89) Ignorance inferences generated: a. P max{d: □ long(p,d)} < 10pp & P Ømax{d: □ long(p,d)} < 10pp b. P max{d: □ long(p,d)} = 10pp & P Ømax{d: □ long(p,d)} = 10pp ‘The speaker is not sure whether the required minimum length is exactly 10pp or less than 10pp.’ (90)

/////////]––––– 10pp

speaker insecurity reading

Finally, if both ant and at least are interpreted in the scope of the necessity modal, the authoritative reading results: (91) a. has to [ant-10pp [λd2 [at least-d2 [λd1 [the paper be d1-long]]]]] □ > ant > at least b. □ "d’ > 10pp: Ø[max{d: long(p,d)} ≥ d’] □ Ø[max{d: long(p,d)} > 10pp] □ max{d: long(p,d)} ≤ 10pp ‘In every acceptable world, the length of the paper is 10pp or less.’ (92) Stronger scalar alternatives: a. □ max{d: long(p,d)} ≤ 9pp * □ max{d: long(p,d)} < 10pp b. □ max{d: long(p,d)} = 10pp

10pp º 9pp ant, at least º exactly

(93) Scalar implicatures generated: a. K Ø□ max{d: long(p,d)} < 10pp b. K Ø□ max{d: long(p,d)} = 10pp ‘The speaker is sure that the paper does not have to be less than 10pp long and that the paper does not have to be exactly 10pp long.’ (94)

––––--] 10pp

authoritative reading

The decompositional analysis thus inherits from the neo-Gricean account the prediction that for the combination of at most with a necessity modal both the speaker insecurity and the authoritative reading are available. But intuitively only the speaker insecurity reading seems to be available for sentence (81) and the authoritative reading (94), according to which 10pp is the maximally allowed length is not possible. Thus, the approach seems to overgenerate readings. But on closer inspection, the unavailability of the authoritative reading of (81) might in

Splitting at most

207

fact be due to independent factors. It is instructive to observe that the authoritative reading is readily available if at most is embedded in a finite clause under a necessity modal. This is illustrated by the sentences in (95) gathered from the internet. Intuitively, they do not express speaker ignorance, but rather report or set the upper bound of the range of permissible values: (95) a. [I am looking for suggestions for a dorm room microwave for my son.] The college requires that it be at most 1 cu feet in volume and at most 800 Watts.10 ‘1 cu feet is the maximally allowed volume and 800 Watts is the maximally allowed power.’ b. This algorithm requires that variables be used at most once.11 ‘The maximally allowed number of variable uses is one.’

Data like (95) suggest that the authoritative reading is available, and in fact the only reading possible, if there is no choice but to interpret the modal with widest scope. Thus, I take it to be a welcome prediction that wide scope of a necessity modal results in the authoritative reading. The question remains, however, why the authoritative reading does not seem to be available when at most is contained within an infinitival complement of a necessity modal as in (81). A way to explain this could be to relate the scopal interaction of modals with the antonymizing operator to the well-known scope preferences modals show vis-à-vis sentential negation. Since the modal verb have to is known to take narrow scope with respect to negation (see e. g. Iatridou & Zeijlstra 2013), this would predict that have to also takes narrow scope with respect to ant, making the authoritative reading unavailable. A further prediction would be that the authoritative reading is readily available if at most is combined with must, which scopes over negation. In fact, the authoritative reading seems to be more easily available when at at most is combined with must than when it is combined with have to. But still, the combination does not seem to be particularly natural. This might have to do with the fact that the authoritative reading expressed by at most plus necessity modal is the same as the one expressed by the combination of at most with a possibility modal. Since both possibility modals and at most are naturally used to express an upper bound, this combination might seem more natural than combining at most with necessity modals, which naturally express lower bounds (see Breheny 2008). In order to see whether the parallel between sentential negation and the antonymizing operator in their interaction with modals can be upheld, more modals would have to be considered as well as other constructions arguably involving ant or little (in particular the constructions Heim 2006 and Büring 2007 discuss).

10 http://www.bogleheads.org/forum/viewtopic.php?f=11&t=100975 (accessed 30 June 2014). 11 http://homepages.inf.ed.ac.uk/wadler/papers/oncetech/oncetech.ps (accessed 30 June 2014).

208

Doris Penka

To summarize the discussion of the readings when at most is combined with a necessity modal, the decompositional analysis derives the speaker insecurity and the authoritative reading. The fact that the authoritative reading in many cases does not show up is likely to be attributable to independent reasons. As desired, it is not possible to derive a split (authoritative) reading, which would express permission for the paper being 10pp or shorter.

4.

A unified analysis of split scope?

This paper started from the observation that a split scope reading arises if at most is combined with a possibility modal, but not if it is combined with a necessity modal. This sets at most apart from other downward monotonic quantifiers like negative indefinites and numerals modified by fewer than, which show no asymmetry in terms of split readings across the two types of modals. An investigation of at most can therefore provide important evidence for the adequate analysis of split scope, and in particular contribute to the debate whether split scope of all downward monotonic quantifiers should receive the same analysis. The asymmetry between possibility and necessity modals is unexpected under unified accounts of split scope readings (de Swart 2000 and Abels & Martí 2010) since they apply independently of the type of modal. But the asymmetry can be explained under the analysis I propose, where at most is decomposed into an antonymizing operator and at least. This provides an argument for deriving split readings arising with different quantifiers in a case-by-case fashion. Decomposing e. g. negative indefinites into a sentential negation and an indefinite (Penka 2011 among others), and comparative quantifiers into a comparative operator and a quantificational part (Hackl 2000) allows for accounting for the particular patterns of split readings these quantifiers exhibit. In principle, it would be possible to extend the analysis in terms of an antonymizing operator to other quantifiers exhibiting split scope readings. The negative determiner no, for instance, could be analysed as meaning ‘no more than zero’ and consisting of the antonymizing operator and Hackl’s (2000) gradable determiner many, as shown in (96). Sentence (97a) would then receive the truth conditions (97b) saying that there does not exist a set consisting of students who failed with a cardinality greater than zero. (96) a. no = [[ant-0]d(dt)- manyd((et)(et))] (97) a. No student failed. b. "d > 0: Ø$X [ |X|=d & students(X) & failed(X)]

Splitting at most

209

But although theoretically possible, such an analysis for negative indefinites does not seem adequate. In contrast to numerals modified by at most or fewer than, negative indefinites do not display overt degree morphology. Hence analysing them in terms of degree semantics does not seem justified if the morpho-syntactic composition of quantifiers is taken seriously. Moreover, negative indefinites in many languages show an interaction with sentential negation and give rise to negative concord. Thus, relating them to sentential negation seems more adequate than relating them to the antonymizing operator. These considerations point to the conclusion that split scope of different types of downward monotonic quantifiers should not be subsumed under a unified analysis but rather be analysed in a case-by-case fashion.

5.

Summary and conclusion

The aim of this paper was to account for the specific pattern of interaction with modals that at most exhibits in terms of split scope readings. Split scope readings are available for at most in combination with possibility modals, but not necessity modals. Embedding the discussion of split scope readings of at most in the recent semantic and pragmatic literature on superlative modifiers, it was noted that the split reading of at most under possibility modals is authoritative where the ignorance inferences usually triggered by superlative modifiers are obviated. Building on the neo-Gricean approach to ignorance inferences, which successfully accounts for the interaction of at least with modals, but in its basic version fails for at most, I showed that the neo-Gricean approach successfully accounts for the interaction of at most with modals if it is supplemented with the idea that at most is decomposed into an antonymizing operator, defined in terms of degree negation, and at least. The analysis explains the fact that at most gives rise to ignorance inferences when it occurs unembedded or in combination with a necessity modal, but can obviate ignorance inferences in combination with a possibility modal to yield an authoritative split reading. Several consequences emerge from this analysis bearing on issues that are currently under debate in the semantics and pragmatics literature. First, it shows that split readings arising with at most can be derived under an approach where at most is decomposed into a negation and another meaning component. This decomposition also seems to be on a principled basis, as the second meaning component is the corresponding positive antonym at least. But crucially, in contrast to negative indefinites like no doctor, the negation involved is not sentential negation but rather degree negation that figures in the formation of antonyms of gradable adjectives. This suggests that instead of a unified analysis, split scope of downward monotonic quantifiers should be analysed in a case-by-

210

Doris Penka

case fashion taking the morphological make-up of the respective quantifier seriously. Second, the particular pattern of interaction with modals which at most exhibits lends support to the idea that negative antonyms are generally decomposed in the syntax into an antonymizing operator, defined in terms of degree negation, and the corresponding positive antonym. This pattern of interaction also has consequences for the question what triggers this decomposition, or, putting it differently, which element of a pair of antonyms should be analysed as the negative one. For the antonym pair at least – at most it seems that semantic rather than morphological properties are decisive. The evidence coming from the interaction with modals suggests that it is at most which is more complex and composed of the antonymizing operator, rather than at least, although the former is morphologically based on the positive form much and the latter on the negative form little. It might seem that at least is the more likely candidate for decomposition, since it already contains little, which is semantically equivalent to the antonymizing operator. But the readings that are available in combination with modals suggest that it is the downward monotonic modifier at most that involves the antonymizing operator. In sum, the investigation of split readings of at most in this paper contributes to current debates in the semantics and pragmatics literature in several respects. First, it lends support to the idea that negative antonyms are decomposed in the syntax into an antonymizing operator, defined in terms of degree negation, and the corresponding positive antonym. Second, it shows that a pragmatic account of the ignorance inferences of superlative modifiers can explain the prima facie puzzling interaction of at least and at most with modals. Finally, it provides evidence against a unified analysis of split scope readings and in favour of a caseby-case analysis for different types of downward monotonic quantifiers.

References Abels, Klaus & Luisa Martí (2010): A unified approach to split scope. Natural Language Semantics 18: 435–470. Alxatib, Sam (2013): ‘Only’ and association with negative quantifiers. PhD thesis, MIT. Beck, Sigrid (2012): DegP Scope Revisited. Natural Language Semantics 20: 227–272. Breheny, Richard (2008): A new look at the semantics and pragmatics of numerically quantified noun phrases. Journal of Semantics 25: 93–139. Büring, Daniel (2007): More or less. In Elliott, Malcolm, James Kirby, Osamu Sawada, Eleni Staraki & Suwon Yoon (eds.): Proceedings from the annual meeting of the Chicago Linguistic Society 43, Vol. 2. Chicago, IL: University of Chicago, Chicago Linguistic Society, 3–17.

Splitting at most

211

Büring, Daniel (2008): The least at least can do. In Chang, C. B. & H. J. Haynie (eds.): Proceedings of the 26th West Coast Conference on Formal Linguistics. Somerville, MA: Cascadilla Press, 114–120. Cohen, Ariel & Manfred Krifka (2014): Superlative quantifiers and meta-speech acts. Linguistics and Philosophy 37: 41–90. Coppock, Elizabeth & Thomas Brochhagen (2013): Raising and resolving issues with scalar modifiers. Semantics & Pragmatics 6: 1–57. Cummins, Chris & Napoleon Katsos (2010): Comparative and superlative quantifiers: Pragmatic effects of comparison type. Journal of Semantics 27: 271–305. de Swart, Henriette (2000): Scope ambiguities with negative quantifiers. In Heusinger, Klaus von & Urs Egli (eds.): Reference and Anaphoric Relations. Dordrecht: Kluwer Academic Publishers, 109–132. Fox, Danny (2007): Free choice disjunction and the theory of scalar implicatures. In Sauerland, Uli & Penka Stateva (eds.): Presupposition and implicature in compositional semantics. Palgrave: MacMillan, 71–120. Fox, Dannny & Roni Katzir (2011): On the characterization of alternatives. Natural Language Semantics 19: 87–107. Gazdar, Gerald (1979): Pragmatics: Implicature, Presupposition, and Logical Form. New York: Academic Press. Geurts, Bart & Rick Nouwen (2007): ‘At least’ et al.: The semantics of scalar modifiers. Language 83 (3): 533–559. Hackl, Martin (2000): Comparative Quantifiers. PhD thesis, MIT. Heim, Irene (2000): Degree Operators and Scope. In Jackson, Brendan & Tanya Matthews (eds.): Proceedings of SALT X. Ithaca, NY: CLC Publications, 40–64. Heim, Irene (2006): Little. In Gibson, M. & J. Howell (eds.): Proceedings of SALT 16. Ithaca, NY: CLC Publications, 38–58. Heim, Irene (2008): Decomposing Antonyms? In Grønn, Atle (ed.): Proceedings of Sinn und Bedeutung 12.Oslo: ILOS, 212–225. Iatridou, Sabine & Hedde Zeijlstra (2013): Negation, polarity and deontic modals. Linguistic Inquiry 44(4), 529–568. Jacobs, Joachim (1980): Lexical decomposition in Montague grammar. Theoretical Linguistics 7: 121–136. Katzir, Roni (2007): Structurally-defined alternatives. Linguistics and Philosophy 30: 669– 690. Kennedy, Christopher (2015): A “de-Fregean” semantics (and neo-Gricean pragmatics) for modified and unmodified numerals. Semantics & Pragmatics 8(10), 1–44. Mayr, Clemens (2013): Implicatures of modified numerals. In Caponigro, Ivano & Carlo Cecchetto (eds.): From Grammar to Meaning: The Spontaneous Logicality of Language. Cambridge: Cambridge University Press, 139–171. McNabb, Yaron & Doris Penka (2014a): The interpretation of superlative modifiers and deontic modals: An experimental investigation. In Etxeberria, U., A. Fa˘la˘us,, A. Irurtzun & B. Leferman (eds.): Proceedings of Sinn und Bedeutung 18, 271–288. McNabb, Yaron & Doris Penka (2014b): The processing cost of interpreting superlative modifiers and modals. In Degen, J., M. Franke & N. Goodman (eds.): Proceedings of the Formal & Experimental Pragmatics Workshop, Tübingen, 29–35. Nouwen, Rick (2010): Two kinds of modified numerals. Semantics and Pragmatics 3: 1–41.

212

Doris Penka

Nouwen, Rick (2015): Modified Numerals: The Epistemic Effect. In Alonso-Ovalle, Luis & Paula Menendez-Benito (eds.): Epistemic Indefinites. Oxford/New York: Oxford University Press, 244–266. Penka, Doris (2011): Negative Indefinites. Oxford/ New York: Oxford University Press. Rett, Jessica (2007): Antonymy and Evaluativity. In Friedman, T. & M. Gibson (eds.): Proceedings of SALT 17. Ithaca, NY: Cornell University, 210–227. Rullmann, Hotze (1995): Geen einheid. Tabu 25: 194–197. Sauerland, Uli (2004): Scalar implicatures in complex sentences. Linguistics and Philosophy 27: 367–391. Schwarz, Bernhard (2011): Remarks on Class B numeral modifiers. Handout of a talk at the workshop Indefinites and Beyond, Universität Göttingen, November 19, 2011. Schwarz, Bernhard (2013): At least and Quantity Implicature: Choices and Consequences. In Aloni, Maria, Michael Franke & Floris Roelofsen. Proceedings of the 19th Amsterdam Colloquium, 187–194.

Index

Absorption 41, 168–169, 171–172, 174, 179–180 adverb 9, 22, 26, 33–36, 38, 50, 83–84, 87– 89, 106, 112, 116, 123–126, 160, 170, 176 – adverbial negator, see also negative adverb 25–26, 31, 36, 38 – higher adverb space 111–112, 122–125 – lower adverb space 111–113, 122, 125, 127 – quantifying adverbs 158, 160, 170, 176 agreement 9–10, 21–22, 32–33, 44, 53, 64, 70, 91, 135, 146, 157, 162, 169, 177, 179– 180 – Agree 28, 41, 51, 64, 66, 70, 72, 92, 135– 137, 146, 166, 168, 172, 176 – multiple agree 137, 166, 175 antonyms 199–200, 209–210 – antonymizing operator 185, 188, 198, 200, 207–210 – decomposition of antonyms 185 assertion 119–120, 190–194, 196, 202 authoritative (reading) 188–190, 194–195, 198, 201, 203, 206–209 Bantu 100 Bavarian 16, 31, 39–42, 92, 157–164, 167– 170, 175, 179 Big DP 14 Big NegP, see also negation phrase 14–15, 81, 90, 97–98, 169 Calabrian 15, 105, 109–113, 118–120, 123, 125–127 Catalan 106, 170, 180

c-command 16, 31, 41, 135–136, 140, 142, 147, 159, 175 clause boundness 94 cleft 15, 66, 70, 107, 127 composition 8, 10, 13–14, 120–121, 127, 209 contrastivity, see also contrastive focus 170, 173 Czech 31–32, 36, 55, 58–59, 75, 132, 136, 140–142, 144 decomposition 17, 42, 185–188, 198–201, 203, 205–206, 208–210 de dicto 186–187 degree 31, 35, 49, 51, 57–58, 113, 169, 191, 199, 209 – degree negation 185, 188, 199, 209–210 – degree operator 190–191, 193, 198 – DegreeP 66 – degree property 191, 199 diachronic change 7, 12, 39, 51, 74, 76 disjunction 190–191, 203 Distributed Morphology 60 Double Negation 8, 10–11, 13, 15–16, 40, 50–51, 53, 82, 92, 94, 135, 150–151, 153, 157–162, 167–177, 179–180 doubling 15, 30, 82, 90–92, 94–96, 98, 101, 166 – clitic doubling 82, 91, 93–94 – DP doubling 81, 92–95, 101 – negative doubling (NegP doubling) 14– 15, 24–25, 29, 81–82, 91–95, 98, 159 emphasis

31, 34, 54, 134

214 English 9–12, 16–17, 25, 30, 35–36, 53, 55, 57–59, 62–64, 66, 75, 83, 94, 98–99, 106, 132, 185 EPP 66, 70, 72 existential, see also indefinite, quantifier 98–99, 137, 165, 185–186 ExistentialP 100, 169 Factorization 41, 168, 179–180 Feature Economy 13–14, 21–22, 27, 31, 44 features 7, 10, 14, 26–28, 30–31, 33–34, 38, 43, 49, 51, 60, 62, 64–66, 68–70, 72, 74– 76, 87–88, 91, 93, 98, 100, 102, 131, 135– 137, 141–143, 146–150, 152–153, 163, 166–169, 172, 175, 180 – feature valuation 135, 146–147 – formal features 14 – interpretable features 16, 26–29, 31, 37–39, 41, 53, 64, 131, 135–137, 140–143, 146–150, 152–153 – negative features 16, 37, 40, 44, 54, 65, 67, 71, 75–76, 134, 137, 165–166, 168 – uninterpretable features 16, 27–29, 31, 37–39, 41, 43, 53, 64, 66, 72, 131, 135–138, 140–143, 145–150, 152–153, 165–167, 172, 175–176, 179–180 Florentine 15, 83, 89, 99, 121–123, 125–126 focus 9, 12, 16, 23, 35, 37–38, 42–44, 49, 51, 58, 70, 89, 96–100, 106, 110, 135, 142, 157–159, 170, 172–176, 179–180 – contrastive focus 36, 164, 174 – FocP (FocusP) 64, 66, 69–70, 72, 75, 100, 169 Free Choice 203 French 9–10, 12, 14, 25–26, 29, 32, 35, 49– 54, 58–59, 62, 65–66, 68–70, 72–76, 81, 83, 90–92, 96, 99, 101, 124 functional hierarchy 32, 74 functional projections 14, 30, 37, 81 German, see also Low German 8, 10–12, 16, 23, 31, 33–34, 36, 42–43, 83, 90, 99, 151, 157–158, 160, 163–164, 175, 177, 185–186

Index

grammaticalization 15, 21–22, 25–28, 31– 32, 34–35, 37, 39, 44, 98, 105, 107, 113– 117, 127 Greek 15–16, 58–59, 131–135, 137–139, 141–146, 148, 151–153 Head Preference Principle 27 Hungarian 58–59, 131, 138–139, 146, 148, 152–153, 167, 170, 180 ignorance inferences 185, 188, 190–194, 196–198, 200–201, 203–206, 209–210 implicature 97, 188, 190, 192–194, 196– 197, 204 – conversational implicature 95, 101, 118 – possibility implicature 192 – primary implicature 191–194 – quantity implicature 188, 190 – scalar implicature 190–195, 199, 201– 206 – secondary implicature 191–192 – sentence implicature 82, 96, 100–101 incorporation 123–124, 126 indefinite 10, 22, 24–25, 28–29, 31, 38–44, 75, 115, 120, 134, 146, 151, 158, 161, 163– 165, 185–186, 203, 208 – existential indefinite 164–165 – negative indefinite 13, 24–25, 29–30, 38–43, 50, 60, 75, 90, 136, 147, 157–158, 160–161, 165, 169, 180, 186, 208–209 information structure 11–13, 16, 37, 162, 170, 180 Innocent Exclusion 190 intervention effects 158, 160, 174–176, 179–180 intonation 135, 142, 164, 167, 174, 176–177 Italian 11–12, 16, 31, 33, 37, 65–66, 69, 82– 84, 87–95, 99, 101, 105–110, 113–114, 116, 118, 120, 122–126, 132, 134, 136, 141, 143–144, 159, 165 Italian dialects, see also Calabrian, Salentino, Florentine 37, 81, 99, 113 – Central Italian dialects 54, 87, 99 – North Eastern Italian dialects 83–84, 91, 99

Index

– Northern Italian dialects 14–15, 81–84, 87, 89–92, 95–96, 98, 107, 169 – Southern Italian dialects 15, 54, 91, 98– 99, 107, 113 Jespersen’s Cycle 7, 10, 13, 21–22, 25, 30– 32, 38, 49, 51–54, 68, 72–76, 96, 100, 117 Ladin 99 Late Merge Principle 27 lexical 7–9, 12, 14–16, 27, 30, 33, 39, 49, 51, 58, 62–63, 65–66, 68–69, 71–72, 74–76, 105–107, 110–111, 113–116, 118, 120, 124, 126–127, 186, 200 lexicalization 26, 99–100, 105, 112–113, 120–123, 125, 127 Logical Form (LF) 28, 35, 40, 147, 157, 166, 168, 191, 194, 197–199, 200–205 logical operator 13, 17 Low German 13, 21–25, 34, 38–39, 44 minimizer 10, 12, 15,31, 83, 88–89, 96, 98– 101, 105, 113–114, 127 MinimizerP 26, 100, 169 Minimize Structure 13, 21–22, 31, 33–34, 44, 68 modal 16–17, 94, 151, 163, 185–190, 193, 195–196, 198–203, 205, 207–210 – modal particles 21–22, 33–34, 37, 44 – necessity modal 17, 185–189, 193–196, 198, 203, 205–209 – possibility modal 17, 35, 185, 187–190, 196, 198, 201, 203–204, 207–209 movement 13, 15, 41, 43, 63–65, 69, 72, 75, 81, 86–90, 94, 101, 105, 113, 127, 163– 165, 178, 195, 203 – movement of negative markers 87–88 – verb movement 9, 105, 112, 125–127, 163 nanosyntax 14, 49–51, 54, 59–60, 62–63, 68, 75–76 negation 7–17, 21–32, 35, 37–44, 49–54, 56–60, 63, 65–68, 70–72, 75–76, 81–83, 86, 88, 90, 92–101, 105–106, 113–114,

215 118, 120, 126–127, 131, 133, 136–137, 140–144, 146–153, 157–165, 169, 171– 173, 175–178, 185–187, 207–209 – completive negator 12, 105, 122–123, 125–127 – constituent negation 35, 43, 58, 70, 89, 94, 108, 115, 177 – discontinuous negation 9, 13–14, 50, 81–82, 96, 101, 117 – emphatic negation 15, 54, 73, 75–76, 105–107, 109, 113, 117, 119–120, 127 – intensive negator 12, 105, 127 – metalinguistic negation 11 – plain negation 108 – presuppositional negation 12, 15, 96, 101, 105, 107, 110–111, 113, 115, 117, 119, 123, 125–127 – sentence negation 49–50, 72, 93 negation phrase (NegP) 9–10, 14–16, 21, 25–32, 37, 39–41, 43–44, 60, 62, 64–66, 69–70, 72, 75, 81–90, 92–93, 96–101, 122–127, 136–137, 151–152, 161, 163, 165, 169, 171–172, 176–177, 179 negation typology 30–31 negative adverb, see also adverbial negator 12, 44, 52, 54, 107, 113–114, 119, 167 Negative Concord 8, 10–13, 15–16, 24–26, 28–32, 39–42, 44, 81, 85–86, 91–92, 110, 131–150, 152–153, 157–177, 179–180, 209 – non-strict negative concord 110, 131, 165 – strict negative concord 165, 167 negative inflection 9 negative marker 8–16, 21–22, 24, 26, 28, 30–32, 34–35, 37–38, 44, 49–62, 65–69, 72–76, 81–102, 113, 115, 131–132, 142, 159–165, 167, 169, 172, 174–175, 179 negative morpheme 55, 60–61, 64, 106 negative operator 10–11, 13, 16, 41–44, 137, 140, 143, 146, 157, 162, 165–169, 175, 178–180 negative polarity item (NPI) 7–8, 11, 26, 29, 39, 50–51, 115–116, 132, 135, 163, 175 negative spread 25, 29, 38, 92, 134, 159

216

Index

negator, see also negation 12, 14, 16, 26, 28–29, 31–32, 35–37, 39, 52–56, 65, 73– 76, 107–110, 114–117, 120, 127 NegP hypothesis, see also negation phrase 10, 14, 21, 30, 44, 169 neo-Gricean 190–191, 193, 195, 198, 201, 203, 206, 209 numeral 186, 190–192, 200, 202, 208–209 n-word 11–12, 15–16, 28–29, 42–44, 87, 91–92, 98–99, 116, 131–140, 145–150, 153, 158 Old Saxon

21–22

polarity 10, 26–27, 31, 34, 49, 51, 53, 58, 62, 64–65, 68–72, 75, 92–93, 98, 157, 201–202 PolarityP 62, 64, 66, 69–71, 75 pragmatic component 190, 197–198, 205 presupposition, see also presuppositional negation 107, 110–111, 113, 117–120, 123, 126–127, 180, 199 probe 41, 64, 66, 70, 72, 90, 136 pronoun 13, 21–22, 32–35, 42, 44, 68, 91, 161 quantifier 41–44, 91–92, 96, 98, 101, 115, 135, 139–144, 146–147, 166, 168, 170, 175, 185–188, 191, 208–210 – downward monotonic quantifier 185– 186, 208–210 – existential quantifier 17, 186 – negative quantifier 16–17, 42–43, 92, 124, 135, 185–186 – QuantifierP 66 – universal quantifier 196

reinforcer 22, 38, 106, 115, 119 Rhaetoromance 83, 88, 96 Romanian 141, 143–144, 167 Salentino 15, 105, 113, 115, 117–121, 123– 125, 127 scalar alternative 190–197, 200–202, 204– 206 scope 7–9, 12–13, 16, 22, 24–25, 29–32, 35, 39–44, 55–59, 61, 65–67, 69–70, 74–75, 120, 137, 140–144, 151, 163–165, 170, 173, 176–179, 185–187, 193–196, 198– 199, 201, 203–207 – narrow scope 38, 59, 143, 186–187, 195, 198, 201, 207 – split scope 13, 16–17, 42, 163–164, 185– 188, 199–200, 205, 208–210 – wide scope 59, 141, 143–144, 185, 194– 198, 203, 207 Spanish 11, 16, 37, 91, 94, 141, 144, 159, 169–170, 180 spell out 54, 60, 62–64, 68–69, 71–76, 150, 190 split IP hypothesis 9, 25 superlative modifiers 185, 187–190, 193, 196, 198, 209–210 syncretism 55, 58–62, 65, 75–76 third-factor principles 21–22, 27, 44 topicalization 161–162, 176–177 truth condition 191, 194, 196–199, 202, 204–205, 208 valuation 26, 147 verb second 22, 36, 40, 158, 161