This page intentionally left blank
G r a m m at ical Cat e gorie s
Grammatical categories (e.g. complementizer, nega...
131 downloads
1658 Views
2MB Size
Report
This content was uploaded by our users and we assume good faith they have the permission to share this book. If you own the copyright to this book and it is wrongfully on our website, we offer a simple DMCA procedure to remove your content from our site. Start by pressing the button below!
Report copyright / DMCA form
This page intentionally left blank
G r a m m at ical Cat e gorie s
Grammatical categories (e.g. complementizer, negation, auxiliary, case) are some of the most important building blocks of syntax and morphology. Categorization therefore poses fundamental questions about grammatical structures and about the lexicon from which they are built. Adopting a ‘lexicalist’ stance, the authors argue that lexical items are not epiphenomena, but really represent the mapping of sound to meaning (and vice versa) that classical conceptions imply. Their rule-governed combination creates words, phrases and sentences – structured by the ‘categories’ that are the object of the present inquiry. They argue that the distinction between functional and non-Â�functional categories, between content words and inflections, is not as deeply rooted in grammar as is often thought. In their argumentation they lay the emphasis on empirical evidence, drawn mainly from dialectal variation in the Romance languages, as well as from Albanian. m . r i ta m a n z in i a n d l e o na r d o m. s avoia are both Full Professors of General Linguistics at the University of Florence.
In this series 94. 95. 96. 97. 98. 99. 100.â•… 101. 102. 103. 104. 105. 106. 107. 108. 109. 110. 111. 112. 113. 114. 115. 116. 117. 118. 119. 120. 121. 122. 123. 124. 125. 126. 127. 128.
j oan b yb e e : Phonology and language use l aur i e b aue r : Morphological productivity t homas e r ns t : The syntax of adjuncts e l i zab e t h c l os s t r augot t and r i c h a r d b. d a sh e r : Regularity in semantic change maya hi c kmann: Children’s discourse: person, space and time across languages di ane b l ake mor e : Relevance and linguistic meaning: the semantics and pragmatics of discourse markers i an r ob e r t s and anna r ous s ou: Syntactic change: a minimalist approach to grammaticalization donka mi nkova: Alliteration and sound change in early English mar k c . b ake r : Lexical categories: verbs, nouns and adjectives c ar l ota s . s mi t h: Modes of discourse: the local structure of texts r oc he l l e l i e b e r : Morphology and lexical semantics hol ge r di e s s e l : The acquisition of complex sentences s har on i nke l as and c he r yl zo l l : Reduplication: doubling in morphology s us an e dwar ds : Fluent aphasia b ar b ar a dancygi e r and e ve s w e e t se r : Mental spaces in grammar: conditional constructions he w b ae r man, duns tan b r ow n a n d g r e v i l l e g . c o r b e t t : The syntax–morphology interface: a study of syncretism mar c us t omal i n: Linguistics and the formal sciences: the origins of generative grammar s amue l d. e p s t e i n and t. dani e l se e ly: Derivations in minimalism paul de l acy: Markedness: reduction and preservation in phonology ye huda n. fal k: Subjects and their properties p. h. mat t he w s : Syntactic relations: a critical survey mar k c . b ake r : The syntax of agreement and concord gi l l i an c at r i ona r amc hand: Verb meaning and the lexicon: a first phase syntax p i e t e r muys ke n: Functional categories j uan ur i age r e ka: Syntactic anchors: on semantic structuring d. r ob e r t l add: Intonational phonology second edition l e onar d h. b ab b y: The syntax of argument structure b. e l an dr e s he r : The contrastive hierarchy in phonology davi d adge r , dani e l har b our a n d l au r e l j. wat k i n s: Mirrors and microparameters: phrase structure beyond free word order ni i na ni ng zhang: Coordination in syntax ne i l s mi t h: Acquiring phonology ni na t op i nt zi : Onsets: suprasegmental and prosodic behaviour c e dr i c b oe c kx, nor b e r t hor ns t e i n a n d ja i r o n u ň e s: Control as movement mi c hae l i s r ae l : The grammar of polarity: pragmatics, sensitivity, and the logic of scales m. r i ta manzi ni and l e onar do m . savo i a : Grammatical Â�categories: variation in Romance languages Earlier issues not listed are also available
CAMBRIDGE STUDIES IN LINGUISTICS General Editors: p. austin, j. bresnan, b. comrie, s. crain, w.dressler, c. j. ewen, r. lass, d. lightfoot, k. rice, i. roberts, s. romaine, n. v. smith
Grammatical Categories: Variation in Romance Languages
Grammatical Categories Va r i at i on i n Ro m ance L anguag e s
M. Rita M a n zin i University of Florence
Leo na r d o M . S avo ia University of Florence
cambrid ge uni ve r s i t y p r e s s Cambridge, New York, Melbourne, Madrid, Cape Town, Singapore, São Paulo, Delhi, Tokyo, Mexico City Cambridge University Press The Edinburgh Building, Cambridge CB2 8RU, UK Published in the United States of America by Cambridge University Press, New York www.cambridge.org Information on this title: www.cambridge.org/9780521765190 © M. Rita Manzini and Leonardo M. Savoia 2011 This publication is in copyright. Subject to statutory exception and to the provisions of relevant collective licensing agreements, no reproduction of any part may take place without the written permission of Cambridge University Press. First published 2011 Printed in the United Kingdom at the University Press, Cambridge A catalogue record for this publication is available from the British Library Library of Congress Cataloguing in Publication data Manzini, Maria Rita Grammatical categories : variation in romance languages / M. Rita Manzini, Leonardo Maria Savoia. â•… p.â•… cm. – (Cambridge studies in linguistics ; 128) ISBN 978-0-521-76519-0 (hardback) 1.╇ Grammar, Comparative and general–Grammatical categories.â•… 2.╇Language and languages–Variation.â•…I.╇Savoia, Leonardo Maria, 1948–â•…II.╇Title. P240.5.M36 2011 415–dc22 2010052183 ISBN 978-0-521-76519-0 Hardback Cambridge University Press has no responsibility for the persistence or accuracy of URLs or external or third-party internet websites referred to in this publication, and does not guarantee that any content on such websites is, or will remain, accurate or appropriate.
Contents
List of tables Acknowledgements
page x xi
Introduction:€grammatical categories and the biolinguistic perspective
1
1
he structure and interpretation of (Romance) T complementizers
13
1.1
Romance complementizers are nominal and head their own noun phrase Structure of the complementizer phrase 1.2.1â•… Combining a left periphery in the complementizer phrase and in the embedded sentence; combining two complementizers 1.2.2â•… Some potential problems The left periphery beyond complementizers 1.3.1â•… Is order dictated by interpretation€– or interpretation by order? 1.3.2â•… Embedded contexts Conclusions
1.2
1.3
1.4
2
14 19
23 30 37 38 43 47
49
2.2 2.3
Variation in Romance k-complementizer systems Systems with two k-complementizers 2.1.1â•… Definite and indefinite complementizers€– and alternative analyses 2.1.2â•… Generalized wh–complementizers ‘If’ The interaction with (non-)finiteness
3
Sentential negation:€adverbs
80
3.1
Sentential negation adverbs are nominal and argumental
83
2.1
49 54 61 65 73
vii
viii╅ Contents 3.1.1╅ Further evidence 3.1.2╅ Sentential negation adverbs as nominal arguments 3.2 Ordering sentential negation with respect to other adverbs 3.2.1╅ The order of negation with respect to aspectual adverbs 3.2.2╅The order of negation with respect to quantificational and manner adverbs 3.2.3╅ General discussion 3.3 The interaction of adverbial and verbal positions:€the participle
88 92 94 106
4
128
Sentential negation:€clitics
111 118 120
4.1 Interactions of negation clitics and subject clitics 4.2 Interactions of negation clitics with object clitics 4.2.1â•… Non-negative n 4.3 Negative concord and negative doubling
131 138 145 152
5
The middle-passive voice:€evidence from Albanian Data 5.1.1â•… Middle-passive morphologies 5.1.2â•… The interpretation of the middle-passive morphologies 5.1.3â•… The Arbëresh varieties The u clitic Specialized inflections 5.3.1â•… Be–participle
159
The auxiliary:€have/be alternations in the perfect Evidence 6.1.1╅ Theoretical background Auxiliary selection independent of transitivity/voice 6.2.1╅ Auxiliary selection according to person Splits according to transitivity/voice 6.3.1╅ Auxiliary selection according to voice 6.3.2╅ Auxiliary selection according to transitivity 6.3.3╅ Irreversibility Finer parametrization 6.4.1╅ Interactions between auxiliary selection according to transitivity/voice and according to person 6.4.2╅ The third auxiliary Some conclusions
196
5.1
5.2 5.3
6 6.1 6.2 6.3
6.4
6.5
160 160 164 169 172 184 188
196 203 208 209 216 216 218 222 223 224 228 233
Contentsâ•… ix 7 7.1
7.2
7.3
7.4
8 8.1 8.2 8.3 8.4 8.5
he noun (phrase):€agreement, case and definiteness T in an Albanian variety Theoretical and empirical background 7.1.1â•… Nominal inflections in Albanian 7.1.2â•… Generative approaches to case Analysis of Albanian nominal inflections 7.2.1â•… Consonantal inflections 7.2.2â•… Vocalic case inflections and lack of inflections 7.2.3â•… Prepositional contexts 7.2.4â•… Summary The Albanian noun phrase 7.3.1â•… The genitive 7.3.2â•… The adjective 7.3.3â•… Adjectives as heads of the noun phrase Concluding remarks ( Definite) denotation and case in Romance:€history and variation The Latin case system Romance case systems:€Romanian Other Romance case systems€– and alternative accounts Loss of case in Romance:€Romansh –s Pronouns€– and some conclusions Notes References Index
236 237 239 244 246 250 255 259 261 262 262 266 272 275
276 277 286 295 302 308
312 331 345
Tables
6.1
6.2 7.1 7.2 8.1
x
istribution of be (E) and have (A) according to person D in the present perfect (in Central and Southern Italian varieties) page 212 Distribution of be, have and syncretic forms in the present perfect in Piedmontese and Lombard varieties 231 Distribution of nominal inflections in Albanian 244 Denotational properties of Albanian nominal inflections 262 Denotational properties of Latin nominal inflections 286
Acknowledgements
The research reported in this book has been financed largely through PRIN grants from the MURST/MIUR, namely Per una cartografia strutturale delle configurazioni sintattiche:€ microvariazione nei dialetti italiani (1997–1999), La cartografia strutturale delle configurazioni sintattiche e le sue interfacce con la fonologia e la semantica. Parametri morfosintattici e fonosintattici (1999–2001), Categorie linguistiche:€Categorie di flessione nominale e verbale (Accordo, Aspetto); Nome e Verbo (2001–2003), I sistemi linguistici ‘speciali’ (apprendimento, disturbi) e la variazione tra i sistemi linguistici ‘normali’. Categorie funzionali del nome e del verbo (2003–2005), Strutture ricorsive in sintassi, morfologia e fonologia. Studi sulle varietà romanze. slave e albanesi (2005–2007), Morfosintassi e lessico:€ Categorie della flessione nominale e€verbale (2007–2009). Special thanks go to all our informants, both Romance and Albanian, though space limitations prevent us from mentioning all of them here. Our debt to the friends and colleagues whose work inspired ours should be obvious from the references. However, we take this opportunity to thank Neil Smith, as a (former) general editor of the series, for helping our project along.
xi
Introduction:€grammatical categories and the biolinguistic perspective According to Chomsky (2000b:€ 119), ‘the human language faculty and the (I–)languages that are manifestations of it qualify as natural objects’. This approach€– which ‘regards the language faculty as an “organ of the body”’€– has been labelled the ‘biolinguistic perspective’ by Chomsky (2005:€1). Hauser, Chomsky and Fitch (2002:€1570) base their discussion of the key biological question of evolution on the ‘biologically and individually grounded’ use of the term language ‘to refer to an internal component of the mind/brain (sometimes called “internal language” or “I-language”)’. They distinguish two conceptions of the faculty of language, one broader (FLB) and one narrower (FLN): FLB includes FLN combined with at least two other organism-internal systems, which we call ‘sensory-motor’ and ‘conceptual-intentional’ … A key component of FLN is a computational system (narrow syntax) that generates internal representations and maps them into the sensory-motor interface by the phonological system and into the conceptual-intentional interface by the (formal) semantics system … Most, if not all, of FLB is based on mechanisms shared with nonhuman animals … FLN€ – the computational mechanism of recursion€ – is recently evolved and unique to our species.â•…â•… (Hauser, Chomsky and Fitch 2002:€1571)
The conception of the language faculty and of (I-)languages as ‘natural’, ‘biologically grounded’ objects corresponds to specific theories concerning their internal articulation: the I-language consists of a computational procedure and a lexicon. The lexicon is a collection of items, each a complex of properties (called ‘features’) … The computational procedure maps an array of lexical choices into a pair of symbolic objects, phonetic form and LF [logical form] … The elements of these symbolic objects can be called ‘phonetic’ and ‘semantic’ features, respectively, but we should bear in mind that all of this is pure syntax and completely internalist.â•…â•… (Chomsky 2000b:€120)
The internal articulation of the FLN is crucial to the biolinguistic programme, no less than its applications to domains such as language evolution, genetics 1
2â•… The biolinguistic perspective and neurology. Here we address some points concerning this; specifically, we concentrate on the issue of language variation, starting with the idea that ‘the diversity and complexity can be no more than superficial appearance … the search for explanatory adequacy requires that language structure must be invariant’ (Chomsky 2000b:€7), and ‘There is a reason to believe that the computational component is invariant, virtually … language variation appears to reside in the lexicon’ (Chomsky 2000b:€120). From this perspective, a central aim of our work is to provide empirical support for what we may call the lexical parametrization hypothesis (Manzini and Wexler 1987), and thus to make more precise the sense in which it holds. Without a doubt ‘one aspect is “Saussurean arbitrariness”, the arbitrary links between concepts and sounds … However, the possible sounds are narrowly constrained, and the concepts may be virtually fixed’ (Chomsky 2000b:€120). In the present study, we address the issue of how the linguistically relevant conceptual space yields different (I-)languages beyond the obvious aspect of ‘Saussurean arbitrariness’. Before proceeding to the empirical core of the argument, we briefly introduce some of the conceptual underpinnings of the framework we adopt, beginning with the thesis that language ‘is a system that is, as far as we know, essentially uniform. Nobody has found any genetic differences … since its emergence there has not been any significant evolution. It has stayed that way’ (Chomsky 2002:€147). This view is shared by much current work on human cognitive and linguistic evolution (Lieberman 1991; Jackendoff 2002). The conclusion holds both for living languages and for ancient ones (whether documented and no longer spoken or merely reconstructed); as argued by Labov (1994), the same mechanisms of (surface) variation and change affect all of them. To take a comparative typological perspective: no evidence of anything like speciation has been found … Languages from typologically very different areas have the same latent structural potential … this survey has uncovered no evidence that human language in general has changed since the earliest stage recoverable by the method used here. There is simply diversity, distributed geographically.â•…â•… (Nichols 1992:€227)
As for this geographically distributed diversity: a residual zone or a set of residual zones will contain a good deal of the world’s possible linguistic diversity in microcosm, and both the existence of internal diversity and its actual profile are stable and obviously very natural situations. Diversity of a particular kind may even be regarded as the state to which a group of languages will naturally revert if left undisturbed … Spread zones, in contrast, are typically highly divergent from one another, but each is internally quite homogeneous … Just which language spreads in a
The biolinguistic perspective╅ 3 spread zone is a matter of historical accident, and this historical accident can distort the statistical distribution of linguistic types in an area.╅╅ (Nichols 1992:€23)
The set of languages considered in this work presents the kind of variation that we expect in natural languages in the absence of external constraints. Because of the political and cultural factors which, for centuries, have kept the Italian peninsula in conditions of great administrative and social fragmentation, dialectal differentiation in Italy has been preserved for longer (i.e. up to the present day) than in other areas of Western Europe, including Romance-speaking ones. Thus Italian varieties provide a rich and articulated picture of language variation that contrasts with that of other intensively studied varieties such as those of English. The view we take is that it is linguistic situations such as those in Britain, for example, that represent a somewhat misleading picture of variation, reflecting not only the internal shaping forces of language development, but also external mechanisms of social and political standardization. The variation seen in Albanian, including the major Gheg vs. Tosk divide in mainland Albania, and Arbëresh varieties of Southern Italy, has the same general character as that observed in Romance varieties. In the internalist (i.e. ‘biologically, individually grounded’) perspective that we adopt, variation between two or more varieties (linguistic communities) is in fact not qualitatively different from variation within the same variety (community), or even within the production of a single speaker. For example, to the extent that a speaker alternates between stylistic levels according to the situation of use, s/he will have a ‘bilingual’ competence of sorts€ – which, given the lexical parametrization hypothesis adopted here, can be accounted for as the co-existence of different lexicons with a single computational component (MacSwan 2000). Suppose, then, that the lexicon is the locus of linguistic variation€ – in the form of a uniform (i.e. invariant) computational component, and of an invariant repertory of interface primitives, both phonological and conceptual. Non-trivial questions arise at this point:€how can the lexicon vary on the basis of a universal inventory of properties (or ‘features’), and why does that variation in the lexicon result in variation in order, agreement, selection, and other relations that are computationally determined? These questions are amply debated in current linguistic theory. Our empirical discussion aims to support certain positions emerging from the debate, as opposed to others which are in principle equally possible. In particular, the answer to the preceding questions is mediated for various scholars by the notion that there is a fundamental distinction between functional and non-functional elements. Thus, within the Distributed Morphology framework, Embick (2000:187) assumes a ‘distinction between the functional and lexical vocabularies of a language … functional categories merely
4â•… The biolinguistic perspective instantiate sets of abstract syntacticosemantic features’, on which the derivational component operates. The actual phonological terminals corresponding to these abstract categories are inserted only after a level of morphological structure, where readjustment rules apply (Late Insertion). It is evident that the overall architecture of the grammar implied by this model is considerably more complex than one in which ‘the formal role of lexical items is not that they are “inserted” into syntactic derivations, but rather that they establish the correspondence of certain syntactic constituents with phonological and conceptual structures’ (Jackendoff 2002:€131). Kayne’s (2006, 2008a) parametrization model, while avoiding recourse to Late Insertion, is close to Distributed Morphology in assuming that functional items correspond to a universal lexicon of sorts. Lexical and hence grammatical differences depend on whether the elements of this functional lexicon are overtly realized or ‘silent’. Interestingly, for Kayne (2006), even variation in the substantive lexicon can be reduced to variation in functional structure in the sense just defined, as can be seen in his construal of shallow as ‘LITTLE deep’, that is, essentially as the specialized lexicalization of deep in the context of the silent functional category ‘little’. Manzini and Savoia (2005, 2007, 2008a) pursue a model under which, again, there is a unified conception of lexical variation€– however, this is of the type traditionally associated with the substantive lexicon:€there is a conceptual and grammatical space to be lexicalized and variation results from the distinct partitioning of that space. There is no fixed functional lexicon which varies along the axis of overt vs. covert realization€– so-called functional space is just like all other conceptual space, and all lexical entries are overt. Thus, the distinction between functional (i.e. grammatical) contents and conceptual ones is an external one; as such it may very well be useless, and at worst it may obscure the real underlying linguistic generalizations. Our conception of variation within the so-called functional lexicon is consistent with current conclusions regarding the conceptual space and the different ways in which it surfaces in natural languages. Fodor (1983) and Jackendoff (1994), among others, develop the Chomskyan theme that concepts, like other aspects of language, must have an innate basis€– largely because of the poverty of stimulus argument. It has already been observed by Lenneberg (1967) that lexical items are the overt marks of a categorization process through which human beings carve out an ontological system from the perceptual continuum of the external world. This process of categorization is of course only indirectly connected with the objects of the external world. Jackendoff (1994:€195) notes that the lexical forms employed to express spatial location and motion (e.g. The messenger is in Istanbul; The messenger went from Paris to Istanbul;
The biolinguistic perspectiveâ•… 5 The gang kept the messenger in Istanbul) typically also express possession (e.g. The money is Fred’s; The inheritance finally went to Fred; Fred kept the money), the ascription of properties (e.g. The light is red; The light went from green to red; The cop kept the light red), etc. This suggests that thought has a set of precise underlying patterns that are applied to pretty much any semantic field we can think about. Such an underlying ‘grain’ to thought is just the kind of thing we should expect as part of the Universal Grammar of concepts; it’s the basic machinery that permits complex thought to be formulated at all.â•…â•… (Jackendoff 1994:€197)
Dehaene, Izard, Pica and Spelke (2006) study geometrical concepts in an isolated group of Amazonian people whose language, Mundurukú, ‘has few words dedicated to arithmetical, geometrical, or spatial concepts’. They conclude that geometrical knowledge arises in humans independently of instruction, Â�experience with maps or measurement devices, or mastery of a sophisticated geometrical language … There is little doubt that geometrical knowledge can be substantially enriched by cultural inventions such as maps, mathematical tools, or the geometrical terms of language … however, the spontaneous understanding of geometrical concepts and maps by this remote human community provides evidence that core geometrical knowledge, like basic arithmetic is a universal constituent of the human mind.â•…â•… (Dehaene, Izard, Pica and Spelke 2006:€385, our italics)
In a similar vein, Hespos and Spelke (2004) study the acquisition of the conceptual distinction between ‘tight’ and ‘loose’ fit of one object to another in English-speaking children, which is not lexicalized in English, though it is in other languages like Korean. Their conclusion is that ‘like adult Korean speakers but unlike adult English speakers, these infants detected this distinction … Language learning therefore seems to develop by linking linguistic forms to universal, preexisting representations of sound and meaning’ (Hespos and Spelke 2004:€453). In short, the building blocks that are combined to make up the potentially infinite variety of human lexicons are innate. The lexicons of different languages are formed on this universal basis, covering slightly different extensions of it and in slightly different ways. The view we advocate here is simply that ways of representing the event, such as transitivity or voice (chapters 5–6), ways of connecting arguments to predicates (or to one another), such as cases (chapters 7–8), and more, are to be thought of as part of this general system. There is no separate functional lexicon€ – and no separate way of accounting for its variation. We started with the general Chomskyan biolinguistic, or internalist, picture of language, and of its basic components, both broadly and narrowly construed. Variation is crucial to establishing this model for the obvious reason that the uniformity thesis, as laid out above, requires a suitably
6â•… The biolinguistic perspective restrictive account of observed cross-linguistic differences. But, even more fundamentally, the lexical parametrization hypothesis that we adopt means that questions of variation will inevitably bear on the form of the lexicon, as one of the crucial components of the I-language. The other main component of the I-language is ‘the computational procedure’, which ‘maps an array of lexical choices into a pair of symbolic objects, phonetic form and LF’ (Chomsky 2000b, quoted above). As for the latter, Culicover and Jackendoff (2005:€ 6) aptly characterize a particularly popular conception of the relation of LF to the syntax (i.e. the computation) as ‘Interface Uniformity’, which holds that ‘the syntax-semantics interface is maximally simple, in that meaning maps transparently into syntactic structure; and it is maximally uniform, so that the same meaning always maps onto the same syntactic structure’. This bias inherent in much current theorizing provides a standardized way of encoding the data, but does not appear to have any strong empirical motivation; nor is the encoding it provides a particularly elegant or transparent one. Conceptually it corresponds to a picture where syntax ‘includes’ interpretation, in the sense that all relevant semantic information finds itself translated into syntactic structure. In contrast, we agree with Culicover and Jackendoff (2006:€ 416) on the idea that interpretation is ‘the product of an autonomous combinatorial capacity independent of and richer than syntax’, ‘largely coextensive with thought’, which syntax simply restricts in crucial ways. Linguistic meanings are merely an input to general inferential processes; the linguistic categorization of the conceptual space encoded by lexical items does not correspond to ‘meaning’ itself but rather to a restriction of the inferential processes producing it. Sperber and Wilson (1986:€174) provide a particularly compelling discussion of the point that linguistic expressions only denote because of their inferential associations:€ ‘Linguistically encoded semantic representations are abstract mental structures which must be inferentially enriched’. In such a model, the well-known indeterminacy of linguistic meanings becomes a key property of successful communication: A linguistic device does not have as its direct proper function to make its Â�encoded meaning part of the meaning of the utterances in which it occurs. It has, rather, as its direct proper function to indicate a component of the speaker’s meaning that is best evoked by activating the encoded meaning of the linguistic device. It performs this direct function through each token of the device performing the derived proper function of indicating a contextually relevant meaning.â•…â•… (Origgi and Sperber 2000:€160)
Note that we disagree with Culicover and Jackendoff (2005) on the model of syntax to be adopted. Our analysis depends on a representational version of minimalism, roughly in the sense of Brody (2003). Crucially, the LF primitives
The biolinguistic perspectiveâ•… 7 we employ are independently available within a minimalist grammar as defined by Chomsky (1995), and in this sense the approach we take is compatible with Chomsky’s model. In fact, we would argue that our views on lexical variation and on interpretation are the simplest construal of Chomsky’s (2000b) proposals, as summarized above€– much simpler than other current approaches, and in this sense closer to the core of minimalism and of the biolinguistic programme. Therefore, any theory maintaining a functional/lexical divide must define the boundary between the two€– which is a far from trivial task. The domain of spatial relations and of events involving them is a case in point. Spatial relations are covered by prepositions (or particles in their intransitive use), among other items. In particular, prepositions/particles can combine with elementary verbs to lexicalize events with a spatial component; for instance, English has put down (the book), Northern regional Italian has mettere giù (il libro). At the same time, Tuscan and literary Italian has a verb posare ‘put down’, and the examples could be multiplied (go in and enter in English, etc.). Particles in Germanic languages (but also in Romance, for instance in Northern Italian varieties) also allow for aspectual interpretations. If, on the basis of these, of the role they play in case systems, etc., we treat prepositions/particles as part of the functional lexicon, what should we infer about spatial primitives? Are they functional? If so, how is their relation to posare, enter, etc. (i.e. canonical lexical verbs) expressed? As mentioned above, the answer envisaged by authors such as Kayne (2006) is that apparent variation in the substantive lexicon reduces to variation in the pronunciation of functional categories; hence the substrings lexicalized by what would traditionally be thought of as lexical categories consist in reality of a number of functional specifications€– which may surface in some languages and not in others, or surface to different extents in different languages. In this way, the functional lexicon effectively spreads over considerable portions of the substantive lexicon; taking this to the extreme, one may want to say that lexical categories are but an epiphenomenon of abstract functional structure. Since the proposal we are putting forward is that lexicons are merely ways of partitioning an abstract categorial space, we are in a way suggesting theories close to those we are taking issue with. At the same time, we consider it significant that we take the step of calling the lexical/functional divide into question, while they typically don’t. To begin with, the different approaches make different empirical predictions in the data domains they both address. Thus, we have specifically referred to Kayne (2006, 2008a, 2009) and Distributed Morphology, since we can directly compare our respective approaches with regard to such domains as fine variation in clitic structures, where we believe our model to be preferable on grounds of descriptive as well as explanatory adequacy (Manzini and Savoia 2009b, 2010).
8â•… The biolinguistic perspective The lexical/functional issue seems to us particularly noteworthy, because at heart it concerns the distinction between the narrow and broad language faculty (FLN and FLB). Let us assume that there is a universal inventory of concepts, and that the lexicon represents a way of realizing it. In theories in which there are in fact two inventories, one for functional categories and one for non-functional ones, it seems to us that the functional and non-functional lexicons are implicitly or explicitly apportioned to the language faculty narrowly construed and broadly construed, respectively. The reduction of the divide that we are proposing has implications not only for the more technical aspects of the theory of grammar, but also opens up the possibility that the universal conceptual repertory which is partitioned by language-particular lexicons is part of the broadly construed language faculty in its entirety. In fact, we see no reason why the grammatically relevant categories investigated here should not constitute categorizations in a domain of general cognition. In other words, what we are saying is that the existence of a functional lexicon associated with the FLN is not a matter of logical or factual necessity€– and as such it should be open to scrutiny. Given the position that we tentatively take on the matter€– namely that eliminating the divide does not imply any empirical problem, and on the contrary allows for a certain simplification of the architecture of language€– we may wonder why such a distinction is so prominent in linguistics. Neuropsychological literature provides much evidence, based both on recent brain imaging techniques and on more traditional language disorders and acquisition studies, that different brain areas are implicated by different conceptual clusters. The prediction is that manipulable objects such as tools are strongly linked to motor behaviour and therefore their representational networks should comprise a significant amount of neurons in motor contexts. Animals, which are most of the time (visually) perceived rather than manipulated, should be represented by networks that partly reside in the visual cortex.â•…â•… (Bastiaansen et al. 2008)
Conversely, ‘assemblies representing function words remain limited to the perisylvian cortex and strongly left-lateralized in typical right-handers’ (Pulvermüller 1999:€260–1). This appears to underlie, in particular, the differential treatment of different sublexicons by aphasic patients (anomics, agrammatics, etc.). Given such results, it does not seem to us to be necessary to draw the conclusion that there is a functional lexicon associated with the computational system of natural language and distinguished on these grounds from a contentive lexicon. Another possibility is that there is a continuum of meaning complexity between the ‘simple’ concrete content words that have clearly defined entities they can refer to … more abstract items that may or may not be used to refer to objects and actions
The biolinguistic perspectiveâ•… 9 and function words … According to the present proposal, the important criterion is the strength of the correlation between the occurrences of a given word form and a class of non-linguistic stimuli or actions.â•…â•… (Pulvermüller 1999:€261)
In other words, it is not so much the functional lexicon that has a special status within the architecture of the mind-brain, but rather certain concrete contents as opposed to more abstract ones. Once freed from the burden of highly articulated inventories and hierarchies of functional categories, we can entertain a simpler syntax, much in the sense of Culicover and Jackendoff (2005). As already mentioned, on the other hand, we do not believe that levels of representations of the type proposed by Culicover and Jackendoff (2005), including rich notions such as grammatical functions, linking rules etc., are required by such a simpler syntax. Rather, the grammar implemented here is a representational version of current minimalist theories (cf. Brody 2003). The relation of the syntax, and more precisely its LF component, to interpretation, as outlined above, is crucial in our view to understanding the role of language variation in the overall economy of the faculty of language. If our construal of syntax and its relation to interpretation is correct, the syntax restricts interpretation, but does not ‘contain’ it (Culicover and Jackendoff 2006). Thus the boundary between syntax and interpretation is a loose one, allowing for a number of different matchings of syntactic form to (inferentially determined) meaning. The looseness of this relation seems to be an essential design feature of the faculty of language, in the sense that it permits the invariant constructs of syntax to cover changing meanings. Lexical items are at the core of language variation simply because they represent the core unit of this interface between syntax and interpretation. In this sense, variation is not an accidental property of the faculty of language, and neither are the characteristics of variation that we try to outline in this study. Rather, they pretty much represent a by-product of the general design of the language faculty. The aspect of our work which provides the title for this book (‘grammatical categories’) has to do with the redefinition of the grammatically relevant classes (i.e. the ‘categories’) of natural language. In general, we take it that the lexicons of natural languages are learnable in that lexical entries individuate natural classes. We apply this logic in particular to Romance complementizers which have the same form as wh–items (Italian che and the like) and to Romance sentential negations which have the same form as negative polarity arguments, in particular ‘nothing’ (Piedmontese nen etc.). In both cases we conclude that lexical identity of form is not a matter of homophony but reveals the sharing of deeper categorizations. This calls into question, among
10â•… The biolinguistic perspective other things, the classical functional categories of C(OMP) (chapters 1–2) and NEG (chapters 3–4). Elsewhere in this book, we find no reason to entertain a functional category status for the so-called AUX(iliaries) have and be, which are argued just to be main verbs selecting a participial clause (chapter 6). In chapter 5 the cluster of meanings associated with Romance si and its Albanian counterpart u are reduced to a unified characterization which also holds of other morphological instantiations of middle-passive voice. Even syncretisms involving case morphology€ – and the functional category K(ASE) according to some (cf. Fillmore 1968; Giusti 1995), are analysed in chapters 7–8 as instances of ambiguous interpretation of the same underlying category, rather than as instances of default lexicalization. This, in turn, requires a revision of the categorizations provided by standard morphological feature systems. It should be kept in mind that the functional structure that this book calls into question (COMP, NEG, AUX, K) is quite independent of recent cartographic proposals (see Cinque and Rizzi (2008) for an overview) which aim to provide a fine-grained picture of functional categories and the way in which they map to syntactic hierarchies. The result is an increase in the number of functional categories, yielding hierarchies of considerable complexity, which have been objected to on the grounds that they enrich the grammar by introducing a great number of new categories and orderings. Yet the same concern regarding the expressive power of the theory could be voiced for standard approaches to functional structure, since the creation of a new functional category or a new feature annotation of an existing category is not subject to any formal or substantive constraints. In this book we propose a take on the problem which goes back to the very first models of exploded structures (Larson 1988), and even further to the very first approaches to ‘functional’ structure in generative grammar (Rosenbaum 1967 on complementation). We argue that structures are indeed atomized, in the sense that a wealth of differentiated head positions are projected under Merge. At the same time, our contention is that a considerable amount of this atomization (perhaps all) does not derive from the introduction of novel categories, but simply from the recursion of certain elementary, identical cells. Thus, the complementizer (chapters 1–2) is not introduced as a specialized head C(OMP); rather, the clearly nominal nature of the complementizer in Romance languages (as in Germanic ones) suggests that the complementizer is the N complement of the matrix verb; in turn, this N takes the embedded sentence as its complement. This structure is as internally articulated as that of Rizzi (1997), but its internal articulation does not depend on a functional hierarchy. Rather, it depends on the recursion of ordinary nominal and sentential
The biolinguistic perspectiveâ•… 11 embeddings. Similarly, negation (chapters 3–4) is one of the earliest functional categories proposed under an articulated view of phrase structure, dating back at least to Pollock (1989). Based on evidence from Romance varieties, we propose, however, that so-called negative adverbs and heads are negative polarity elements, and, even more radically, that they participate in the argumental structure of the verb, coinciding specifically with the individuation of the internal argument position. Thus, in Romance languages, there is neither evidence for a lexicalized negative operator nor for a functional position hosting it. The final case study to be introduced here concerns the internal structure of nouns and noun phrases (chapters 7–8). Following an established trend in generative grammar, we argue for the conclusion that noun phrases (as well as adjective phrases) have the same internal organization as sentences. From this perspective, we take up the classical proposal of Higginbotham (1985) that the D(eterminer) saturates the obligatory (internal) argument of the nominal predicate; in this sense, D properties yet again represent an instantiation not of functional structure, but of predicate–argument structure. Case, in turn, is not construed as a (functional) primitive of grammar, but rather as a label covering much more elementary properties, relating again to the saturation of predicate– argument structures. Throughout the discussion, the emphasis is very much on empirical evidence. We repeatedly argue that our model not only fares better with respect to fairly reasonable simplicity metrics, but also that it has descriptive advantages. In fact, and quite strikingly in our opinion, less powerful theories are better suited to capturing complex (micro)variation data of the type we consider than theories potentially capable of greater descriptive power. In particular, we subscribe to the simplicity argument in favour of representational grammars advanced by Brody (2003). This implies abandoning derivations, including the notions of a cycle (phases) and an asymmetric search space (feature checking). What we retain is representational relations:€chains, agreement, etc. Simplicity is paramount, to the extent that existing empirical evidence does not provide any support for the more complex grammar. In particular, complex data concerning agreement (and variation in agreement patterns) are accounted for in Manzini and Savoia (2005, 2007, 2008a) by abandoning phifeature checking in favour of identity (or better, compatibility) of referential properties; uninterpretable and unvalued features are also eliminated under this approach. At no point is there any evidence that a derivational approach would have empirical advantages€ – on the contrary, the complexity of the variation effectively requires the simpler representational approach. In general, representational grammars are simpler than derivational ones in that the latter postulate
12â•… The biolinguistic perspective purely computational processes whose results are LF-relevant and hence redundant with LF constructs. Movement and the LF-relevant notion of chain are the obvious cases, but this also holds of the computational operation of agreement and its LF reflexes relevant for coreference etc. By contrast, the representational model views LF-relevant relations as determined directly by the interpretive calculus at the LF interface (chains by the theta-calculus, and so on). This adoption of a representational model goes hand in hand with the present take on functional structure. In the minimalist grammar of Chomsky (1995), functional categories, or features, bear the burden of computation, since they crucially enter into feature checking. To the extent that functional categories are reduced to ordinary predicates and arguments, functional feature checking is also replaced by ordinary selectional restrictions. In turn, empty functional heads and/or Specs, needed only for the checking of functional features, can be eliminated. As for the highly articulated hierarchies proposed in cartographic research, if functional and lexical properties of lexical entries are not formally different, as argued here, the most natural and economical assumption is that their ordering is restricted entirely by interpretive principles, including closure requirements, scope, etc. Needless to say, in the absence of any theoretical distinction between functional and lexical elements, we do not expect that they will be inserted at two different points in the derivation, namely in syntactic and morphological structure respectively (as in Distributed Morphology). On the contrary, we predict that all syntactic structure will be projected directly from lexical entries, independently of their properties. In other words, we propose a view in which morphological-level structures and relations are entirely unified with syntax. Lexical entries, in turn, are entirely characterized in terms of positive properties. Another theme of general significance is the question of how the present approach to functional structure relates to the issue of ‘grammaticalization’. In the terms of, say, Roberts and Roussou (2003), grammaticalization is essentially the reanalysis of a lexical category as a functional category, often with the result that a lexical entry can have both a lexical and a functional construal. In such cases we typically propose that there is a single lexical item with a unified (lexical, not functional) characterization; the che ‘that’ complementizer/wh–phrase (chapters 1–2) is a case in point. It is evident that, to the extent that this latter treatment can be generalized, the distinction between functional and lexical dissolves; if so, the problem of why exactly lexical categories would turn into functional ones (i.e. the problem of ‘grammaticalization’) simply does not arise.
1 The structure and interpretation of (Romance) complementizers
In this chapter we start from the observation that in Romance languages, Â�complementizers are nominal, belonging to the same argumental series as wh–phrases, although current theories treat them as functional projections of verbs, filling the same positions as verbs do. We argue that the Romance chetype complementizer is not a functional category of the verb, but rather a nominal head, which satisfies an argument slot of the matrix verb and which takes the embedded sentence as its complement (section 1.1). Both as a wh–phrase and as a complementizer, Italian che introduces a variable. If it Â�introduces an individual variable, the wh–phrase reading arises; if it introduces a propositional variable (ranging over situations or possible worlds) it is read as a so-called complementizer. In section 1.2, we argue that our proposal is compatible with fine distributional evidence relating to the left periphery of the sentence, in particular with the fact that elements such as topics or foci can occur both below the complementizer and above it. We capture this distribution by allowing them to occur at the left periphery of the embedded sentence (below the complementizer) or at the left periphery of the complementizer phrase (above the complementizer). This structural hypothesis predicts that occurrences of the relevant material below and above the complementizer can combine€ – a fact that requires quite complex functional hierarchies in alternative accounts. It should be kept in mind that, insofar as this chapter proposes an alternative to the generally adopted view that the complementizer is a functional projection of the sentence, it is not aimed specifically at so-called cartographic models. These, however, are discussed in some detail in section 1.3 (especially Rizzi 1997, 2001, 2004), because they explicitly consider the same type of data as we do. In section 1.2 we also briefly consider possible general objections to our proposal, for instance concerning the distinction between complementizer phrases and conventional noun phrases:€how is this distinction made for the purposes of selection? Extraction facts are also relevant:€ how are sentences 13
14â•… Romance complementizers:€structure & interpretation introduced by complementizers different from complex NPs? For the purposes of selection, complementizer phrases can be distinguished from conventional noun phrases in that they introduce a propositional and an individual variable respectively. As for extraction, the question why complementizer phrases, in contrast to conventional noun phrases, are not islands is the same as the question why complementizer phrases have a left periphery that is an ‘escape hatch’ (Chomsky 1973), and conventional noun phrases do not. This remains very much an irreducible primitive of all generative approaches (up to Chomsky 1995, 2008), since it is not the case that noun phrases lack a left periphery altogether (Szabolcsi 1994). We construe this primitive not as a structural, but as an interpretive one, contrasting propositional denotations with individual denotations. A different question concerns the fact that, although Romance complementizers belong to the wh–series, they typically have a non-interrogative interpretation. We return to this question in chapter 2, where we show that this property is parametrized. In particular, we illustrate Romance systems where the ‘that’ complementizer€– or an element of the wh–series€– lexicalizes the ‘if’ (interrogative/ hypothetical) complementizer as well. In general, the argument in favour of the present theory is based on explanatory adequacy. First, it simplifies the lexicon, allowing for a unified lexical entry for elements like Italian che, the only possible alternative being homophony. It also solves the mystery of why C would host such disparate categories as verbs and complementizers, reserving what we might continue to call C for verbs. Perhaps most interestingly, it reduces what would otherwise be potentially complex functional hierarchies to the recursion of simple predicate–argument structure (i.e. the complementizer is an argument of the matrix predicate, taking the embedded sentence as its argument). Furthermore, the burden of proof is on alternative theories to prove that they can account in an explanatory way for the fine variation we observe in complementizer systems, a matter to which we return in chapter 2.
1.1
Romance complementizers are nominal and head their own noun phrase
Current theories hold that complementizers are functional projections of the verb, that is, their position is essentially one which could in other circumstances be filled by a verb. This is true independently of whether there is a set of C positions (Rizzi 1997), or a single C position with a rich specifier structure (Chomsky 1995). But if complementizers fit into the same type of position as
Romance complementizers are nominalâ•… 15 verbs, it is unclear why they so clearly coincide with functional projections of the noun:€for instance with demonstratives in Germanic, or with wh–elements in Romance. Even if we were to accept that the relation of complementizers to the wh–system of Romance languages, to the demonstrative system of English etc. is a purely historical one, the question would arise of how nominal, quantificational elements could come to fit verbal specifications. In other words, the form of the problem would change, but not its substance. Consider for instance Italian, in which che, like English that, introduces finite declaratives, as in (1); since Kayne (1976) this has been identified with the relative clause introducer in (2). The same element also introduces interrogatives with the meaning of ‘what’ as in (3). Furthermore, che can appear as the wh–determiner of complex interrogative NPs as in (4). (1)â•… So che fai questo I.know that you.do this ‘I know that you do this’ (2) Il lavoro che fai è noto the workâ•… thatâ•… you.doâ•… is â•… known ‘The work you do is well-known’ (3) Che fai? what you.do? ‘What are you doing?’ (4) Che lavoro fai? which job you.do ‘Which job do you do?’
One possibility that we can reject is that Italian che simply has two lexical entries, one of which corresponds to the ‘that’ complementizer and the other to the ‘what’ wh–element. This solution does not have any explanatory value, given that the pattern that it describes is not an accidental coincidence observed in one or even a few languages, but a systematic phenomenon in Romance, as can be gleaned from the data to follow. What is more, Caponigro and Polinsky (2008) find the same formal identity between the wh–system and the complementizer system in a completely unrelated language, Adyghe (a NW Caucasian, Abkhazo-Adyghean language), showing that in Adyghe the same syntactic structure can be mapped to four different meanings, namely relative clause, complement clause, wh–interrogative and yes–no interrogative. Uncontroversially, in sentences like (3) che heads its own noun phrase, which in turn fills a position in the C field of the sentence, as in (5a). From an interpretive point of view, che introduces a variable corresponding to the internal argument of the predicate fare ‘to do’, as in (5b).
16╅ Romance complementizers:€structure & interpretation (5) a. [che] [C b. che x, fai x
[I fai
In examples of the type in (4), the interrogative noun phrase che lavoro ‘what job’ is associated with a structure of the type in (6a), where che represents a quantificational position Q within the noun phrase. Assuming that the structure of the noun phrase is parallel to that of the sentence, the inflected head noun is positioned in I€ – to parallel the position of the inflected verb within the sentence. From an interpretive point of view, che again introduces a variable restricted by the predicate lavoro ‘job’, as in (6b). (6) a. [Q cheâ•… [I lavoro]] [Câ•… [I fai b.â•… che x:€x lavoro, fai x
Apart from the fact that both complementizer and wh–phrase take the morphophonological form che, from a syntactic point of view the complementizer che in (1) and the wh–phrase che in (3) both occupy a position at the left periphery of the sentence. Since under current theories, complementizer che fills precisely the kind of position that we questioned at the outset, namely C, in (7a) it is assigned to an unnamed position above C€– the latter independently known to be a possible position of the verb in the sentence. From an interpretive point of view, complementizer che could be an operator introducing a variable, again like wh–phrase che€– the main difference between them being the nature of the variable. This ranges over individuals for the wh–phrase che in (5)–(6), while for complementizer che it ranges over situations/possible worlds, as in (7b).1 (7) a. [che [Câ•… [I fai questo b. che x: x fai questo
The fact that the operator-variable structure is interpreted as a question in (5)–(6), but not in (7), need not stand in the way of their unification, as we know that wh–phrase che can also have non-interrogative interpretations (very much like its English counterpart what). Thus che can be the head of a free relative, as in (8), or an exclamative, as in (9). The most natural interpretation of this range of data is that question, declarative (relative) and exclamative values are not intrinsic to the wh–phrase, but rather contextually determined.2 (8) Fai che ti pare do what to.you pleases ‘Do what you like’ (9) Che non farei! what not I.would.do ‘What I wouldn’t do!’
Romance complementizers are nominalâ•… 17 Given the discussion so far, if complementizer che appeared in the appropriate context, there is no reason why it shouldn’t assume the interrogative value of, say, English if or whether. Now, in well-known Romance languages like Italian, ‘if’ has a specialized lexicalization, se in Italian, in which the hypothetical and interrogative values overlap. In chapter 2, however, we will show that this state of affairs does not necessarily hold€– in other words, that there are languages in which the lexicalizations of ‘that’, ‘if’ and ‘wh–’ overlap, as we expect. Similarly, recall that the Adyghe morphology discussed by Caponigro and Polinsky (2008) is ambiguous between a ‘yes–no interrogative’ and a ‘complement clause’ interpretation (among others). Let us assume, then, that complementizer che is to be identified with wh– –phrase che (as the morphology but also the interpretation suggest) and that, as anticipated in the discussion of (7a), it cannot be hosted in C, since C is a verbal position, while che is nominal. An alternative structure is suggested by what is perhaps the earliest approach to finite complementation in generative grammar (Rosenbaum 1967) as well as by analyses of special subsets of complement sentences, notably factives (Kiparsky and Kiparsky 1970) and unselected questions (Adger and Quer 2001). These have in common the postulation of a nominal structure of some sort for sentential complementation. What we propose here is that every sentence introduced by che has a nominal layer, represented by a noun phrase headed by che itself. Thus, in a sentence like (1) the verb heading the matrix sentence takes as its complement a noun phrase headed by the che complementizer, which in turn takes the embedded sentence as its complement, yielding a structure of the type in (10). Following the assumptions we have already introduced in (6) concerning the position of head nouns in noun phrases, che is in I.3 (10)
I che C I fai V
questo
18â•… Romance complementizers:€structure & interpretation It should also be noted that the theoretical literature does contain treatments in which sentential introducers are removed from the C projection(s) of the embedded sentence and are made to depend directly on the main sentential tree. Notably, Kayne (1994), in analysing the infinitival introducers of Romance languages, such as Italian di ‘of’ etc., treats them not as complementizers, but as functional projections of the main verb, triggering movement of the embedded sentence to their Spec position. What we propose here is different, in that we are rejecting entirely the idea that complementizers are functional heads (either of the embedded or of the main sentence) and we are arguing that they are true arguments of the main verb, in turn taking the embedded sentence as their argument. For reasons of space we will not be able to deal with di–type introducers here. Manzini (1982) and Manzini and Savoia (2005) treat them as prepositions. A more direct comparison can be established with the work of Arsenijevic (2009), according to whom ‘the variable denoted by a wh–element gets bound by a question operator in questions or by an appropriate head in relatives. Its presence in F[inite] C[omplement] C[lause]s in a significant number of languages signals that the denotation of these clauses involves a variable’. This conclusion converges with those in the text, yet the syntax proposed by Arsenijevic (2009) for English markedly differs from ours. Thus, that, which semantically introduces a lambda operator, syntactically occupies the conventional C position. The variable over which the lambda operator abstracts corresponds to the Spec of a Force head generated immediately under the C head. This position is occupied either by a nominal expression with Force content such as claim or by its abstract incorporated counterpart for verbs such as to claim (analysed as make claim). In these terms the overall structure assigned to claim that John kissed Mary is a relativization headed by claim:€[N claim [C that [Force claim [John kissed Mary]]]]. Similarly, Kayne (2010) revises his (1976) idea that the relative que of French is really the que complementizer, by proposing instead that ‘that isn’t [a complementizer]. The that that introduces sentential complements is really a relative pronoun … The claim that English sentential that is a relative pronoun must be taken to extend … for example, to Italian che, to French que … From the present perspective that is not a Force in Rizzi’s (1997) sense nor the head of a CP phase in Chomsky’s (2001) sense’. The actual implementation that Kayne provides for these ideas differs markedly from ours, while bearing a considerable resemblance to Arsenijevic’s. In particular, for Kayne ‘factive sentences … have a deleted or silent FACT. If so … factives too must involve relative clause structures.’ Assume, then, that sentential complements and
Structure of the complementizer phraseâ•… 19 sentential subjects are always accompanied by a head noun, even if they are not factive … Either it will be raised from within the relative in a way largely parallel to what happens with overt fact … or a silent noun will be so raised’. We have insisted more than once on the empirical reasons, but also on the general theoretical problems that lead us to reject the silent categories approach of Kayne, for instance in the domain of clitic phenomena (Manzini and Savoia 2008a, 2009a, 2010; Savoia and Manzini 2010). These effectively apply to Arsenijevic (2009) as well. On the other hand, it seems to us that the extra assumptions concerning the (silent) Force phrase of Arsenijevic, or equivalently the silent head nouns of Kayne, are unnecessary, since both the semantics for propositional embedding and the identity of so-called complementizers and wh–phrases follow from the simpler syntax proposed here. We will return to this comparison in chapter 2. In the next section, we argue that the structure in (10) is compatible with the distributional evidence concerning the so-called left periphery of Romance languages€– in fact, it provides a particularly economical way of dealing with it. 1.2
Structure of the complementizer phrase
Given a structure like (10), we predict that material related to the quantificational and informational structure of the sentence (wh–elements, topic, focus, etc.) should be hosted by the left periphery of the embedded sentence and hence should follow the complementizer. Indeed the presence of such material under the complementizer is familiar from English as well as from Romance languages. In (11) we provide some examples from Italian, where (11a) illustrates a topic and (11b) a focus in the left periphery of the embedded sentence. (11)â•… a.â•… b.
So che questo non l’hanno preso I.knowâ•… that â•… this not it they.have taken ‘I know that they haven’t taken this’ So che questo hanno preso (non quello) I.know that this they.haveâ•… taken (not that) ‘I know that they have taken this one, not that one’
As expected, wh–phrases can also appear under the complementizer. In fact, in our own judgement, this option is open in Italian, as in (12). In any event it is fully productive in Southern Italian varieties such as Arena in (13), where any wh–phrase can appear either in the left periphery of the matrix sentence€– i.e. in its scope position, as in (13a’) and (13b’)€– or in the left periphery of the embedded sentence, hence under the complementizer, as in (13a) and (13b).
20â•… Romance complementizers:€structure & interpretation (12)â•… Credi che con chi se la prenderanno you.think that with whom MP it they.will.take ‘Who do you think they will take it out on?’ (13) Arena (Calabria) a.â•… ti kriði ka ðuvi ðrmi you thinkâ•… thatâ•… whereâ•… he.sleeps ‘Where do you think he sleeps?’ a’. duvi ti kriði ka ðrmi where you think that he.sleeps ‘Where do you think he sleeps?’ b. ti kriði ka pEkki vEni you thinkâ•… thatâ•… why he.comes ‘Why do you think he is coming?’ b’. pEkkiâ•… ti kriði ka vEni why you think that he.comes ‘Why do you think he is coming?’
In turn, the verb can invert with the subject in sentences embedded under the complementizer, yielding instances of embedded V2. In particular, among the (present-day) Romance languages, many Romansch and Ladin varieties, like La Pli and Scuol in (14) and (15) respectively, exhibit V2 in embedded sentences, hence under a che-type complementizer. On the standard assumption that V2 depends on the positioning of the verb in C, this means that the left periphery of sentences embedded under the complementizer, i.e. its C field, can host a verb as well. (14)â•… La Pli de Mareo (Alto Adige/South Tyrol) a.â•… i te diZi ke le liber a-i lit I you tell that the book have-I read ‘I am telling you that I have read the book’ b. i te diZi â•… ke endomaN/ gonot/ magari ve-el I you â•… tell thatâ•… tomorrow/ often/ may be comes-he ‘I am telling you that he is coming tomorrow/ often/ perhaps’ (15)
Scuol (Grisons) i m an dit tSa frs drmaS-t/ drm-al they me have told that perhaps sleep-you/ sleeps-he ‘They told me that perhaps you are/ he is sleeping’
The data that have been presented are merely compatible with the structure in (10) and do not in themselves provide any argument in its favour. An adequate analysis of the same data is available within the articulated theory of the C field proposed by Rizzi (1997, 2001, 2004). In particular, the distribution of Italian che and other Romance che-like complementizers in (11)–(15)
Structure of the complementizer phraseâ•… 21 corresponds to that predicted by Rizzi (1997) for the highest complementizer.4 The embedded V2 facts follow, as already proposed by Schwartz and Vikner (1996), if the verb occupies a different C position (in fact the lowest C position for Rizzi (1997)).5 By contrast, the existence of embedded V2 is problematic in frameworks like Chomsky’s (1995), where only one C position is available. In such a framework, one possibility is to derive embedded V2 from the positioning of the verb in I (Santorini 1989). The clitic nature of the inverted subject in (14)–(15) tends to discount this possibility, on the assumption that subject clitics are associated with a high inflectional domain€– hence if the verb were in I, subject clitics should precede, rather than follow, the verb. The only alternative that we can see to Rizzi’s (1997) articulated view of the C field is precisely the one we are advocating here:€namely, treating the complementizer as the head of an independent noun-phrase-like projection. The discussion so far only addresses the conditions that make structures like (14)–(15) possible; a different question altogether is what makes them necessary. According to McCloskey (2004) there is a particular interpretive value associated with embedded V2, which is revealed by the selectional restrictions discussed in the literature on Germanic languages. Thus, the highest layer of an embedded V2 structure€– i.e. the one filled by the complementizer€– realizes illocutionary force, and allows for inversion of the verb in the lower layer. In languages/contexts that do not allow for embedded V2, a single layer is present, hosting the complementizer and excluding V2; this does not realize illocutionary force. We are not aware of any data on Romance embedded V2 that would allow us to decide whether it obeys semantic restrictions. But suppose it does. The general schema of explanation proposed by McCloskey (2004) need not be tied to the double CP structure that he adopts; rather, it can be implemented, as far as we can see, by the structure that is at the heart of the present proposal. Thus, properties of the matrix sentence select a particular set of properties on the embedded complementizer€– which in turn select for V2 in the embedded sentence. Languages/contexts without embedded V2 simply do not select for the relevant properties on the complementizer head, and so the V2 position of the verb is not selected in the embedded sentence.6 While in (12)–(13) we have seen some cases where the che-type complementizer precedes a wh– phrase, there are many Romance varieties in which the che-type complementizer follows a wh–phrase in both main and embedded interrogatives. This pattern is quite robust in Northern Italian varieties; in (16) we exemplify just one of them.
22â•… Romance complementizers:€structure & interpretation (16) â•… Castellazzo Bormida (Piedmont) a.â•… kwan k i madZi when that you.pl eat ‘When do you eat?’ b. dallwO k i d4wOmi where that you.pl sleep ‘Where do you sleep?’
Now, precisely because the complementizer is the head of an independent constituent, a left periphery can be postulated for it as well. In this schema of explanation, therefore, the wh–phrase that precedes che belongs to the left periphery of the complementizer itself, roughly as in (17). Note that the wh–phrase dallw ‘where’ is simply categorized according to its relation to the predicate, as Loc(ative); we will return to the theory of left periphery elements that this labelling implies. (17) Castellazzo Bormida Loc dallw
c C I k C D i I drw mi c
The analysis in (17) predicts that the entire set of focus and topic elements should be able to appear to the left of the complementizer. Examples like those in (18)–(19) show that the left periphery of the complementizer can host not only wh–elements (presumably foci) but also topicalized material, including both adverbs and the lexical subject, doubled by the subject clitic following the complementizer.
Structure of the complementizer phraseâ•… 23 (18)
Castellazzo Bormida marjo dal’lw k u dRwm Mario where that he sleeps ‘Where does Mario sleep?’
(19)
Cerano (Piedmont) lo:/ adme� inda k i ve� they/ tomorrow where that they go ‘Where are they going (tomorrow)?’
Our structural hypothesis therefore accounts both for data in which the complementizer has topic and focus material to its right and for data in which it has topic and focus material to its left, and thus matches the descriptive power of an articulated C field of the type in Rizzi (1997). At this point the important question arises of whether a left periphery of the type in (16)–(19) is restricted to complementizer-headed noun phrases or whether it can associate with noun phrases in general. It is evident that, to the extent that the structure of noun phrases parallels that of sentences, noun phrases are predicted to host leftÂ�peripheral material exactly as the sentence does. In fact, the left periphery of the noun phrase in (6) hosts the wh–item che, paralleling the left periphery of the sentence in (5). Similarly, the ability of the left periphery of Hungarian DPs to act as an escape hatch for a possessor phrase was the key to Szabolcsi’s (1994) proposal of a parallel structure for noun phrases and sentences. Nevertheless, an asymmetry between sentences/complementizer-headed noun phrases and other noun phrases does in fact exist€– namely, that operators hosted at the left periphery of ordinary noun phrases must bind variables within the noun phrase itself. By contrast, sentences and complementizerheaded phrases can host operators binding long-distance variables. Within the analysis that we are suggesting, this asymmetry can be captured by saying that only propositions and nouns introducing propositional variables€– i.e. complementizers€– can support the relevant set of operators. This distinction is stipulated€– i.e. it is an apparently irreducible primitive of natural languages. But exactly the same is true of Chomsky’s (1973) original proposal of a C(OMP) node providing an ‘escape hatch’ for Subjacency in S(entences) but not in NPs, and its successors up to the present (Chomsky 1995, 2001, 2008). 1.2.1
Combining a left periphery in the complementizer phrase and in the embedded sentence; combining two complementizers Suppose we accept that left-peripheral material occurring after the complementizer is associated with the embedded sentence, while left-peripheral material occurring before the complementizer is associated with the left periphery of
24â•… Romance complementizers:€structure & interpretation the complementizer itself. We then predict that the two left peripheries should combine. The simplest verification of this prediction is that topics can not only precede the cluster of wh–phrase and che, as in (18)–(19), but they can also follow it, as in (20). In present terms, this means that in (20) the wh–phrase occurs in the left periphery of the complementizer, while the topic appears in the left periphery of the embedded sentence.7 (20)â•…
Castellazzo Bormida dallw kə  mat  drm where that the girl she sleeps ‘Where is the girl sleeping?’
Furthermore, a left periphery associated with the complementizer, in the shape of a wh–phrase and/or topics preceding it, can be combined with a left periphery in the embedded sentence in the shape of a verb in C. In other words, it is possible to embed V2 (as revealed by subject–verb inversion) under a sequence of wh–phrase and che complementizer, as in (21)–(22). Note that in the Cantoira example (22b), the complementizer is preceded not only by the wh–phrase, but also by the topicalized subject. (21)â•… Mezzenile (Piedmont) a. əndua k u wnt-i where that they go-they ‘Where are they going?’ b. kaŋ k u viunt-i when that they come-they ‘When are they coming?’ (22) Cantoira (Piedmont) a. ənduə k u dyərt-e whereâ•… thatâ•… theyâ•… sleep-they ‘Where do they sleep?’ b. lu: andua k u vEnt-e they where that they go-they ‘Where are they going?’
Data of the type in (21)–(22) are associated by the present theory with structures of the type in (23), in which the left periphery of the complementizer combines with the left periphery of the sentence embedded under it. The position of the verb, although we keep the conventional C label for it, is now characterized by exclusively verbal properties. Note also that two copies of the subject clitic D are present in (23), on either side of the verb; we assume that while the inverted subject clitic is in the ordinary subject (clitic) position
Structure of the complementizer phraseâ•… 25 immediately above I, the higher copy is in the C domain.8 In keeping with the labelling introduced in connection with (17), the wh–phrase is categorized as Loc(ative) by its intrinsic content€– as is the topicalized subject, notated as D like its clitic counterparts. We shall return to the lack of Top (Focus etc.) labels in section 1.3. (23) Cantoira D lu Loc andua (C) I ke D u C ��nt D e
Poletto (2000), working essentially within Rizzi’s (1997) framework, argues that in sentences like (21)–(22), che is inserted in an intermediate position of the C field. This is distinct from the higher C position (Force for Rizzi (1997)) involved in examples like (11)–(15), and from the lower C position (finiteness for Rizzi (1997)) involved in (16)–(20). Thus, theories of CP recursion can match the descriptive power of the present theory (with a single complementizer position combined with two left peripheries) by postulating three separate complementizer positions. Despite this apparent descriptive equivalence, the two theories are clearly different. Consider the fact that in several varieties interrogatives introduced by a wh–phrase alternate with interrogatives introduced by a wh–phrase and che. In at least some of them, the former have subject clitic inversion€– i.e. V2,
26â•… Romance complementizers:€structure & interpretation as in (24a’ and b’)€– but the latter don’t, as in (24a and b). In many more varieties, for instance Castellazzo Bormida, wh–questions introduced by the wh–phrase and che without inversion, as in (17), alternate with yes–no questions with inversion, as in (25). Both facts lead to the conclusion that in the relevant languages the cluster of wh–phrase and che is in complementary distribution with V2, which is otherwise required by interrogatives. (24)â•… Viguzzolo (Piedmont) a. indE k u drmæ where that he sleeps ‘Where does he sleep?’ a’. ind u drmæ-l where he sleeps-he ‘Where does he sleep?’ b. kwænt k u drmæ how.much that he sleeps ‘How much does he sleep?’ b’. kwand u drmæ-l when he sleeps-he ‘When does he sleep?’ (25)
Castellazzo Bormida 4 d4wm-4 she sleep-she ‘Does she sleep?’
A theory such as Poletto (2000) predicts the data in (24) by assuming that the che complementizer sits in the lowest C position; if so, verb movement will not be able to target that position€ – nor, according to Poletto (2000), will it be able to target any higher position because of minimality. However, under the schema of explanation adopted here, the wh–phrase is in the left periphery of the complementizer che and the verb is in the C position of the embedded sentence, exactly as discussed for (23). Hence the two positions are different, and neither interferes with the movement paths of the other, so the fact that they cannot both be filled in (24) must be explained on other grounds. In present terms, in varieties like (24), when the wh–phrase is introduced in the left periphery of the sentence, interrogative modality is lexicalized by the verb in C, as in (24a’ and b’). By contrast, if the wh–phrase is introduced in the left periphery of the complementizer, V2 is no longer necessary and is in fact excluded, as in (24a and b); we assume that this is so because of a selectional constraint, whereby the complementizer selects the declarative
Structure of the complementizer phraseâ•… 27 modality (i.e. the V in I rather than in C) in the embedded sentence. In the (rarer) languages of the type in (23), interrogative modality is lexicalized by the verb in C even when the sentence is embedded under the complementizer. There is no need, and indeed no evidence, for a different categorization of the complementizer which ultimately yields its different position, as in Poletto (2000). Let us mention also that, along with languages which form interrogatives with a wh–phrase and che (no V2), as in (16)–(20), languages which form them with a wh–phrase and V2 (no che), as in (24a’ and b’), and languages that form them with both che and V2, as in (21)–(23), there are languages that form them with neither. Thus, in (26) the wh–phrase introduces an interrogative sentence alone, i.e. without che-type complementizer, while at the same time the verb does not realize interrogative modality in C, but remains in I, as can be seen by the lack of inversion with the subject (clitic). (26)
Filattiera (Tuscany) ke kamiza t yi what shirt you want ‘Which shirt do you want?’
Our examples so far involve main sentences. Leaving aside embedded questions introduced by the ‘if’ complementizer (to which we return in chapter 2), in embedded wh–questions V2 is generally not found. This is the case in languages in which no inversion is found in main wh–questions, as in the Castellazzo Bormida example in (16’), in languages that do have inversion, as in the Cantoira example in (22’), and also in the Viguzzolo example in (24’) in the absence of a complementizer. In other words, V2 is generally limited to root contexts. We conclude that, contrary to the Romansh/Ladin varieties with embedded V2 in (14)–(15), matrix predicates in Northern Italian languages do not select for V in C in the embedded sentence€– nor for properties of the che-type complementizer that in turn select for embedded V2. (16’)â•…
Castellazzo Bormida di-m ki k i4 vi6N tell-me who that he comes ‘Tell me who is coming’
(22’)
Cantoira di-me ki (k) e vint tell-me whoâ•… thatâ•… heâ•… comes ‘Tell me who comes’
28â•… Romance complementizers:€structure & interpretation (24’)
Viguzzolo a ŋ s2 k a f@ŋ Iâ•… notâ•… knowâ•… whatâ•… theyâ•… do ‘I don’t know what they do’
At the same, there is no structural impossibility of having V2 in embedded questions€ – a fact which is duly reflected by the existence of data like (27) attesting the possibility of the relevant patterns. In this case we assume that the verb indeed lexicalizes interrogative modality in C. It is this interpretive property that discriminates between the pattern in (16’), (22’) and (24’), and the (much rarer) one in (27). From a strictly syntactic point of view, they are equally possible. (27) â•… Castiglione d’Adda (Lombardy) a. di-m sa tSam-ot tell-me who call-you ‘Tell me who you are calling’ b. di m sa sE -t a dre a fa tell me what are you in the process of doing ‘Tell me what you are doing’
A final set of examples which have been considered in the literature in the context of Rizzi’s (1997) proposals concerning an articulated left periphery involve the lexicalization of left-peripheral material, typically a topic, between two copies of the same che-type complementizer, as in (28). Examples like (28) are predicted within a theory of complementizers as functional heads, on the assumption that more than one complementizer position in the hierarchy is instantiated, along the lines of Paoli (2007). (28)â•… Castellazzo Bormida a. l E mii ke nuiâ•… k a l lavu itâ•… isâ•… betterâ•… thatâ•… we thatâ•… weâ•… itâ•… wash ‘It is better that we wash it’ b. l E mii ke vujautSâ•… k i m la dagi it is better that you.pl that you.pl me it give ‘It is better that you give it to me’
In the structures that we have laid out so far there is no room for two complementizers co-occurring. This is not to say that structures cannot be provided for (28). On the contrary, the relevant examples can be adequately described through recursion of the complementizer phrase. In other words, as shown in (29), the matrix predicate takes as a complement the higher complementizer, which in turn selects the lower complementizer, hosting the topic in its left periphery, and embedding the complement sentence.9
Structure of the complementizer phraseâ•… 29 (29) Castellazzo Bormida I ke D vujautʃ I ke D i
... I dagi
Needless to say, the question is not only whether a structure, say (29), can be assigned to a sentence like (28), but also what forces this structure. McCloskey (2004) entertains two alternatives. The first is that double complementizer structures semantically differ from simple complementizer structures; the other possibility is that the higher layer of complementizer structure ‘exists solely to facilitate the adjunction’ of the embedded topic material (McCloskey 2004:€ fn.€ 30). As it turns out, Paoli (2007) argues that both possibilities are instantiated in Romance. Thus, she finds that the double complementizer structures of Turinese (but ‘not the “mainstream” type spoke in Turin’) and of Ligurian are semantically restricted, in that the lower complementizer must select the subjunctive. By contrast, our data from Castellazzo in (28) display the double complementizer phenomenon both with the indicative in (28a) and the subjunctive in (28b). In this respect they seem entirely comparable to the data that Paoli (2007) quotes from Medieval Romance (Tuscan and other varieties). Since the evidence we have does not point to a semantic distinction between double and simple complementizer structures, they must be distinguished on structural grounds. Specifically, based on the evidence that we have, we conclude that matrix predicates in Castellazzo can immediately embed complementizers, as in (29), or wh–phrases as in (16’), but not topic/focus material. This state of affairs can be captured in terms of a selectional constraint imposed by the matrix verb on any complementizer phrase it embeds€– roughly to the effect that it can contain only clause-typing material (in the sense of Cheng (1991)), i.e. only declarative che or the interrogative cluster of wh–phrase and
30â•… Romance complementizers:€structure & interpretation che. Two alternatives are then open for the insertion of focus/topic material. First, one can position such material in the left periphery of the embedded sentence, as in (20). Alternatively, it can be positioned in the left periphery of the embedded complementizer€– but then a higher complementizer must be added, satisfying the selectional constraint just defined, as in (29). In this respect, therefore, we reject the view of Paoli (2007:€ 1075), according to whom the lower complementizer ‘is not a subordinating particle, but overt realization of Top°’ in medieval Romance.10 1.2.2 Some potential problems Before we proceed with our discussion, we will try to clear the ground of a number of quite general questions that may be raised against our approach. Perhaps the most basic such question involves selection of an embedded complement by a matrix predicate. If a complementizer head in Romance is nominal and its projection is a noun phrase, how can we state the distinction between selecting for a complementizer phrase, i.e. a sentence, and selecting for an ordinary noun phrase? The answer is that in terms of the interpretive categories adopted here, complementizers (like bare sentences) correspond to propositions, while conventional noun phrases correspond to individual terms. Therefore selection can adequately be stated on the basis of interpretive categories.11 Selection may even provide an argument in favour of the present articulation of the left periphery, as opposed to the cartographic one. Suppose that a higher predicate embeds a sentence whose leftmost and highest element is, say, a topic, as would be the case, for instance, in (18) and (19). Under Rizzi’s (1997) theory the predicate effectively embeds a topic phrase€– which, according to Newmeyer (2005), provides no clear grounds for selecting the interrogative force in lower functional projections. Whether this turns out to be a serious problem or not, it does not arise in the theory that we are building. A verb selecting for a complementizer-headed noun phrase selects for the properties of the complementizer head. Alternatively, if the higher predicate selects a bare sentence€– i.e. one without a complementizer€– it selects for properties of the embedded verbal head. Another property that has been consistently used to explain differences in distribution between noun phrases and sentences is Case. In the present framework, Case cannot be used to this end precisely because complementizers, which introduce sentences, are nominal€– and should therefore have the same Case properties as ordinary nouns. The evidence with which Stowell (1981) introduces his Case Resistance Principle concerns the fact that noun phrases, including gerunds, can be the object of a preposition, but sentences cannot, as in (30a) vs. (30b).
Structure of the complementizer phraseâ•… 31 (30) a. We talked about the Marines going to Iraq b. *We talked about that the Marines went to Iraq
Now, elements such as before, after and without would seem to be prepositions since they can select noun phrases. But their Italian counterparts also embed sentences introduced by che, as shown in (31). Hence there appears to be no general selectional constraint against sentences as objects of prepositions or against complementizers in such a position. To the extent that more specific constraints are real, they can be stated again in terms of the interpretive notions of individual vs. propositional variable. (31) a. b.
Sono arrivato prima/dopo che I.am arrived before/after that ‘I arrived before/after you left’ Me ne sono andato senza me away am gone without ‘I went without you noticing it’
sei partito you.are left che te ne accorgessi that you of.it noticed
A different kind of question has to do with the fact that theories in which a sentential complement is contained within a nominal layer have been proposed before€– but typically differentiate between various types of sentential complements. Specifically, Kiparsky and Kiparsky (1970) differentiate complements of factive verbs, which are endowed with a nominal layer, from complements of non-factive verbs, which do not have such a layer. Crucially, different syntactic behaviours are argued to follow from the different structures. Thus, it is suggested that factive verbs are islands for extractions, since their structure makes them into complex NPs. Our judgement for Italian is that there really is no difference between extraction from a non-factive context like (32a) and extraction from a factive one like (32b)€– and even extraction from ‘the fact that’ in (32c) is not severely degraded. In contrast, there is a strong contrast with extraction from a relative clause, as in (32d). (32) a. b. c.
Chi pensi che non abbiamo visto? Who you.think that not we.have seen ‘Who do you think that we didn’t see?’ Chi ti dispiace che non possiamo vedere? who you regrets that not we.can see ‘Who do you regret that we can’t see?’ Chi ti dispiace il fatto che non possiamo vedere? who you regrets the fact that not we.can see ‘Who do you regret the fact that we can’t see?’
32â•… Romance complementizers:€structure & interpretation d. d’.
*Chi ti dispiace il motivo perâ•… cui non possiamo vedere who you regrets the motiveâ•… for whichâ•… not we.can see ‘Who do you regret the reason why we can’t see?’ Chiâ•… ti chiediâ•… perchèâ•… nonâ•… abbiano potuto vedere? who youâ•… ask why not they.haveâ•… been.ableâ•… to.see ‘Who do you wonder why they haven’t been able to see?’
The fact that (32b) patterns with (32a) may simplify our task with respect to the distinction between factive and non-factive complements, since we may not need to worry about providing different structures for them. But the clear contrast between (32a) and (32d) raises the question of whether they shouldn’t both be blocked as instances of complex NP islands. In fact, this question can be reduced to the one we discussed at the end of section 1.2.1 concerning the admissibility of long-distance operator material in the left periphery of sentences vs. noun phrases. There we concluded that complementizers and sentences (because of their common propositional content) differ from other noun phrases in being able to host such material€ – i.e. in behaving like ‘escape hatches’. If so, we expect that ordinary noun phrases may give rise to a complex NP island effect, whereas complementizers do not. In other words, there is nothing in the present proposal contradicting standard accounts of islands in terms of subjacency/ phases€– though we remain strictly non-committal with respect to them.12 Italian does display some limited sensitivity to factive islands with adjuncts, which can have both matrix and embedded scope in sentences like (33a), while in (33b) embedded scope appears to be quite hard to obtain. This suggests that the adjunct cannot be extracted from the factive sentence€ – though this extraction is possible from the non-factive one. In other words, factives show an asymmetry between arguments, as in (32b) and adjuncts, as in (33b) of the type well known from Cinque (1990), and accounted for since Rizzi (1990) as a Minimality effect. (33)â•… a.â•… b.
Perchèâ•… pensi che siano venuti? why you.think that they.are come ‘Why do you think that they are coming?’ Perchè ti dispiace che siano venuti? why you regrets that they.are come ‘Why do you regret that they came?’
Now, we know that in some languages the factive vs. non-factive distinction involves the choice of different complementizers, one such language being Greek, as studied by Roussou (1994). Therefore it is natural to propose that the embedded complementizer in (33b) has some property (selected by the higher predicate) that the complementizer in (33a) does not have, for instance a
Structure of the complementizer phraseâ•… 33 definiteness property, which, as suggested by Roussou (1994), is consistent with the interpretation of factivity. It is reasonable to think that this property triggers a Minimality effect in (33b) which is absent from (33a). In other words, the present analysis of complementizers need not interfere with current theories of extraction (with respect to which we remain non-committal, as before). Rather, the two issues are orthogonal. Summing up so far, the structure that we propose for complementizers is motivated in section 1.1 on the basis of their nominal nature in Romance, as seen in the fact that they have the same form as wh–elements. Section 1.2.1 shows that this structure is compatible with the distribution of the complementizer with respect to other elements of the so-called left periphery. But note that this latter result depends on the complementizer being the head of its own projection€– and does not depend on it having the same form as the wh–operator or even a nominal nature. In other words, though we argued that the nominal, wh– nature of Romance complementizers requires the structure we propose for them, nothing in the structure we propose requires nominal, let alone wh– properties in the complementizer. Complementizers which do not coincide with wh–elements can easily be seen in Romance systems. A simple example is provided in (34), from a variety from Sardinia which distinguishes the declarative complementizer ki, as in (34a), from kiE ‘who’ in (34b) and kalE ‘which’ in (34d), as well as from ittE ‘what’ in (34c–d). Thus, though the complementizer can be assigned to the k-series to which ‘who’ and ‘which’ belong, it does not coincide with any wh–item. For languages like (34), we of course maintain the same analysis as for Italian che. In fact, given the morphological relatedness of ki to the ksystem of wh–elements, a language like Luras can be described as having a specialized wh–operator for propositional variables. (34)â•… Luras (Sardinia) a.â•… m ana naDuâ•… ki enis kraza to.meâ•… they.haveâ•… told thatâ•… you.comeâ•… tomorrow ‘They told me that you are coming tomorrow’ b. kiE eniDi whoâ•… comes ‘Who is coming?’ c. ittE znâ•… fattEN whatâ•… they.are doing ‘What are they doing?’ d. ittE/ kalE libbru t a llEaDu what/ whichâ•… book to.youâ•… he.hasâ•… brought ‘What/which book did he bring you?’
34â•… Romance complementizers:€structure & interpretation We know from Germanic languages that complementizers can belong to a non-wh– nominal series, namely that of demonstratives, like English that. We maintain the idea that the complementizer introduces a propositional variable for these languages as well. On the other hand, a wh–complementizer of the type generally instantiated in Romance (but also in Adyghe according to Caponigro and Polinsky (2008)) is very naturally construed as a lambda abstractor; the question of how best to construe the demonstrative complementizer of English is left open here. A related open question is whether the different nature of the complementizer has reflexes in the syntax and in the interpretation of Romance vs. Germanic complementation. In chapter 2 we introduce Romance complementizer systems that have no counterpart in Germanic languages known to us, possibly indicating that these are possibilities open to wh–complementation systems but not to demonstrative ones. Further afield, there are languages in which so-called complementizers are verbal, rather than nominal in nature. A case in point is the Buru language of Eastern Indonesia, as discussed by Roberts and Roussou (2003) on the basis of Klamer (2000). In Buru, fen can function as a main verb followed by a quotation, as in (35a), or can combine with another verb of saying introducing direct speech as in (35b) or indirect speech as in (35c); the data are from Klamer (2000). (35)â•… a.â•… b. c.
Nak ana-t feneâ•… “Ng-ina nau daholo” 3sgPossâ•… child-Nomâ•… say 1sgVoc-motherâ•… 1sgPossâ•… bunch-head ‘Her child said, “Mother, the hand (of bananas) at the top of the stalk is mine”’ Da prepaâ•… fenâ•… “Siraâ•… ruaâ•… kaduk” 3sgâ•… speak say 3pl two arrive ‘She said “The two of them came”’ Da prepa fene ringe mata haik 3sg speak say 3sg die Prf ‘He said that he was already dead’
Under the present approach, in all three examples in (35) fen can be treated as the I head of its own projection€– in this case not a noun-phrase-like projection, but a verb phrase-/sentence-like projection. In the absence of other verbal specifications, fen is read as a main verb, as in (35a)€– which we predict to be possible precisely on the basis of the fact that it heads a verb phrase-/sentencelike constituent. On the other hand, the combination of fen with another verb in (35b–c) recalls so-called serial verb constructions. In Manzini and Savoia (2005) we consider serial verbs in connection with constructions in Southern Italian varieties in which aspectual/modal/motion verbs with reduced or absent inflectional properties embed sentential complements, on condition that their
Structure of the complementizer phraseâ•… 35 temporal reference and their EPP (Extended Projection Principle) arguments coincide. We conclude that these constructions involve complex predicate formation; this implies the coincidence of temporal reference and argumental structure€ – and in turn these properties will typically result in a lack of inflection on one of the two verbs. The same general properties seem to hold in (35b–c), in which the so-called complementizer status of fen can then be described in terms of complex predicate formation€– again on the basis of the usual structure in which it heads a verb phrase-/sentence-like projection. Data of the type in (35) are worth bringing up not only for their intrinsic interest€– but also because of an issue that we raised in passing in section 1.1 and then abandoned in subsequent discussion:€namely, that (35) and the like are routinely described in terms of processes of historical change. Thus, according to Klamer (2000:80), ‘we can explain the synchronic distributional restrictions on fen if we assume that historically … the report verb fen(e) has developed an alternative interpretation as a quote marker fen … In contexts where fen is preceded by another verb which reports words, thoughts, or perceptions, it has developed a complementizer interpretation’. To quote just one similar case, for Whitman (2000:€222) in Ewe ‘the categorial feature of bé “say” changes from V to C’. A closer term of comparison with the present theory is provided by Roberts and Roussou’s (2003) discussion of English that€– an element which we have cited more than once as providing evidence for the connection of complementizers with bona fide nominal heads. In a nutshell, when it comes to that ‘we are dealing with one and the same lexical item which can surface as either D or C’ (Roberts and Roussou 2003:€115). This analysis is compared to Davidson’s (1997 [1968]:€828–9) idea that ‘sentences in indirect discourse, as it happens, wear their logical form on their sleeves … They consist of an expression referring to a speaker, the two place predicate “said”, and a demonstrative referring to an utterance’. According to Roberts and Roussou (2003:€113–14), in Davidson’s analysis complementizer that ‘is actually the demonstrative’; by contrast, ‘it is possible to argue that that in terms of its position in the sentence has been grammaticalized as a C element’. In present terms, the explanation for the range of interpretations of, say, Italian che or English that has to do with conditions internal to a single grammatical competence system; in other words, there is no necessity for invoking several competence systems in a relation of historical change to one another. Specifically, no grammaticalization is implied, understood roughly as reanalysis from lexical to functional (Roberts and Roussou 2003). In this respect the present analysis is closer to the syntactically naive one of Davidson (1997
36â•… Romance complementizers:€structure & interpretation [1968]). The gist of our proposal is that Italian che and its Romance counterparts are nominal quantificational heads in all cases€ – with a syntax typical of nominal heads, i.e. that of serving as arguments of predicates, and with a semantic content typical of quantificational heads, i.e. that of introducing variables. Similarly, we surmise that both demonstrative that and complementizer that are nominal heads and arguments of verbs; the only difference is that complementizer that has a propositional restrictor and demonstrative that a nominal one. Before proceeding, we will consider just one more issue. Nothing that we have said so far leads us to expect that complementizers appear only in embedded contexts. In fact we have already seen a complementizer occurring in matrix sentences:€ namely, the che-type element following the wh–phrase in (16)–(24).13 The che complementizer also introduces yes–no matrix questions in a variety like Florence in (36), though in the same variety wh–questions are introduced simply by a wh–phrase (as in standard Italian). We surmise that the insertion of ke in yes–no questions corresponds to the presence of a focalization bearing on the main verb, paralleling the focalization on the wh–constituent in wh–questions, but consistent with the yes–no interpretation. If so, we can maintain that the yes–no complementizer proper (i.e. what we will call the polarity complementizer in chapter 2) is the se ‘if’ element that occurs in embedded yes–no questions (in Florence as in standard Italian). (36)
Firenze (Tuscany) ke lla viEne la maria that she comes the Mary ‘Is Mary coming?’
Finally, matrix instantiations of the complementizer are also possible in standard Italian, subject to a modal split, since matrix subjunctives can be introduced by the complementizer, as in (37), while matrix indicatives are not. In other Romance varieties, the che-type complementizer can introduce ordinary indicative sentences, as reported in the literature for Provençal (Ronjat 1937:€536 ff.) and Guascon (Rohlfs 1977:€205); an example from Sardinian is provided in (38). (37) Che entrino that they.enter ‘Let them enter’ (38)
Àllai (Sardinia) ka dZai ɖɖ a ffattu that alreadyâ•… it he.hasâ•… done ‘He has already done it’
The left periphery beyond complementizersâ•… 37 1.3
The left periphery beyond complementizers
In the discussion of distributional evidence in section 1.2, we argued that this evidence is also accounted for by our analysis when tested against the alternative view of complementizers as functional categories of the verb. The argument in favour of our hypothesis does not have to do with empirical coverage per se, but rather with the way in which this empirical coverage is achieved. In our view, there are features of the present proposal that make it more conceptually perspicuous (more explanatory) than available alternatives. In particular, the cartographic model, which aims at an empirical coverage comparable to ours, is associated with a trend towards finer-grained phrase structure. In a sense, such an approach is simply what the empirical evidence requires€– and in this sense we share it. At the same time, in the specific implementation of the cartographic programme presented, say, by Rizzi (1997, 2001, 2004), the fine articulation of syntactic structure derives from the postulation of specialized hierarchies of functional categories. The model of the left periphery that we have defined here is clearly different. Thus, where Rizzi (1997) has a hierarchy of functional projections for complementizers, we have a single complementizer position€– and this position is not functional. Rather, on the evidence of its morpholexical form as well as its interpretation, the complementizer is the head of a noun phrase that selects the embedded sentence. In other words, a certain amount of the articulation of the so-called left periphery is achieved through the recursion of predicate–argument structures€– the superordinate verb selecting the complementizer noun phrase as its argument, and this in turn selecting the embedded sentence as its argument. Critics of cartography note that functional hierarchies are potentially unrestricted devices, since a new position or set of positions in the hierarchy can always be introduced to meet new empirical evidence. The present approach exempts at least complementizer structures from this potential problem. On the other hand, the present chapter so far only deals with the complementizer itself. Nothing that we have said touches on the independent issue of whether there is a single (conventionally C) position to which the verb can move or more than one. In previous work we have sided in favour of more than one verb position above I€– based notably on the comparison of finite verbs in questions and V2 with imperatives and infinitives. Evidence concerning their distribution with respect to clitics (enclisis vs. proclisis), negation and other material leads us to the conclusion that at least two different (conventionally C) positions of the verb are involved. In adopting multiple head positions above I, we therefore follow Rizzi (1997) and the related cartographic literature.
38â•… Romance complementizers:€structure & interpretation At the same time, in the model of Rizzi (1997, 2001, 2004), the possible head positions above I are also used to order topic, focus and other phrasal material, for it is one of the postulates of standard cartography, based on Kayne’s (1994) anti-symmetry, that there is a single Spec position for each head; therefore a head hierarchy automatically defines a hierarchy of phrasal material€ – and in fact hierarchies of phrasal material are stated as hierarchical orderings of the heads to which they attach (cf. Cinque 1999 on adverbs). The issue of the ordering of phrasal material in the left periphery is logically independent of the status of complementizers. It is evident, however, that the present reanalysis of complementizers leads us to expect that a considerable amount of currently postulated functional structure may be dispensed with in that respect as well. In this respect, potentially more promising views are held by Chomsky (1995, 2000a, 2008), according to whom each head supports any number of Specs (as for Brody (2003)), while topic, focus and similar notions correspond to interface interpretations€– not to features (or categories) entering syntactic computations. 1.3.1 Is order dictated by interpretation€– or interpretation by order? In what follows, we concentrate on the respective position of wh–phrases and of lexical subjects, specifically in Northern Italian varieties with subject clitics. The topic(-like) nature of the lexical subject in Romance languages is supported precisely by their position in questions. Thus, in subject-clitic languages the subject clitic follows finite verbs in questions, including the auxiliary, as in (39). This corresponds to the position of the lexical subject in Germanic languages, and can be analysed accordingly, as a result of the clitic remaining in the same position as in declarative sentences€– while the verb alternates between I in declaratives and C in questions. By contrast, lexical subjects in Romance do not appear between the auxiliary and the participle, but only after the participle or before the auxiliary, i.e. in the right or left periphery of the sentence, as in (39) again. (39)â•… Oviglio (Piedmont) a.â•… E -l amni marju is he come Mario ‘Has Mario come?’ b. marju E -l amni Mario is he come ‘Has Mario come?’
When wh–phrases are brought into the picture, it remains true that lexical subjects are generally positioned in the right or left periphery of the sentence;
The left periphery beyond complementizersâ•… 39 however, in this latter case they are found before the wh–phrase, as in (40a), and not between the wh–phrase and the verb, as in (40b). The examples in (40c-d) illustrate the parallelism of other topics with lexical subject in this respect. Both are normally excluded from the position between the wh–phrase and the verb in C. (40)â•… Modena (Emilia) a. lo ki tSam -el he who calls he ‘Who does he call?’ b. *ki lo tSam -el whoâ•… he calls he c. la torta indo l E -t the cake where it have you ‘Where have you put the cake?’ d. *indo la torta l E -t where the cake it have you
mesa put mesa put
According to Rizzi (1997:€ 299), the positioning of the lexical subject is determined essentially by the Wh–Criterion of Rizzi (1996), whose satisfaction requires the verb and the wh–phrase to be in a head-Spec configuration in C(Focus)P, which forces their adjacency and hence the impossibility of a lexical subject (or other material) intervening between them. At the same time, even for Rizzi (1996:€87), the unacceptability of a sentence like (41a) in Italian contrasts with the grammaticality of (41b), in which the lexical subject intervenes between the wh–phrase perchè ‘why’ and the verb. (41)â•… a.â•… b.
*Dove Gianni è andato? where G. is gone ‘Where has Gianni gone?’ Perché Gianni è partito? why G. is left ‘Why has Gianni left?’
Working in a model with a single C position, Rizzi (1996) proposes that in sentences like (41b), it is the wh–phrase perchè that occupies this position, so that the verb is in the ordinary I position and the subject precedes it. However, in Northern Italian varieties the position of the lexical subject between ‘why’ and the verb combines with inversion of the subject clitic and the verb, as in (42), indicating that the verb is in C. (42)â•…
Modena perkE al putEin e l parti why the child is he left ‘Why has the child left?’
40â•… Romance complementizers:€structure & interpretation In fact, it is not just ‘why’ that allows for the lexical subject (or other topical material) to appear between the wh–phrase and the verb, as Rizzi (1996:€87) also remarks. In his terms, the obligatoriness of verb movement to C in interrogatives is weakened when a D-linked wh–phrase is involved, for reasons that remain unclear, so that the verb is (again) in I. Data from Northern Italian varieties confirm Rizzi’s (1996) intuition about the possibility of positioning the subject after a D-linked wh–phrase. At the same time, this can combine with inversion of the subject clitic after the verb, as for instance in (43), leading us to conclude that the verb is in C. Hence the phenomenon is not connected with the verb remaining in I. (43)â•… Corte (Veneto) a. kEl de kis marjo ljezara-lo which of these Mario will.read he ‘Which of these will Mario read?’ b. *ula marjo va-lo where Mario goes he ‘Where does Mario go?’
Incidentally, note that if we combine the previous generalizations about the relative order of wh–phrases and lexical subjects/topics with the proposal that the complementizer is a specialized nominal head with its own left periphery, we obtain a straightforward prediction concerning interrogatives introduced by wh–phrase and complementizer. Namely, we predict that we will find not only the order in (23), in which the topic precedes the wh–phrase and complementizer sequence, but also the order in which the topic appears after the wh–phrase and hence before the complementizer. This prediction is confirmed by data like (44a); similar evidence is noted by Poletto and Vanelli (1995:€153) for a variety from the Turin area. Crucially, the order in (44a) appears to be restricted by the same factors that we considered for questions introduced by a simple wh–phrase€– so that a D-linked wh–phrase favours the order in (44a), while a non D-linked wh–phrase tends to exclude it, as in (44b, b1). (44)â•… S.Maria Maggiore (Piedmont) a. kwal ad kwi guit ul dZua k u tSama which of those children the John that he calls ‘Which of those children does John call?’ b. ki k a tSama lu:r who that they call they ‘Who do they call?’ b’. *ki lu:r k a tSama who they that they call
The left periphery beyond complementizersâ•… 41 Proceeding now with the theoretical discussion, a new take on the ‘why’ problem is offered by Rizzi (2001). In addition to the three C positions (finiteness, Focus and Force, cf. fn. 3–5) proposed by Rizzi (1997), Rizzi (2001) introduces a further C position, namely Int(errogative), located between the highest C(Force) position and the C(Focus) one. This C(Interrogative) position is meant to host elements like ‘why’ as well as interrogative ‘if’, se in Italian. Crucially, in Rizzi’s (1997) model, Top phrases can be freely interleaved between any C projections€– therefore topicalized material is predicted to occur between ‘why’ in C(Interrogative) and the verb, even if the latter is in C(Focus). Note that an eventual unification of ‘why’ with D-linked wh–phrases requires the latter to be moved to (or inserted in) Spec of C(Interrogative) as well. Abstracting away from the theoretical postulates of the cartographic approach€– essentially the idea that all phrasal and head movement must be supported by the presence of a matching functional head€– the empirical generalization seems to be that wh–phrases can move into two different positions, namely a slightly higher one that can precede a topic and a lower one that cannot. The question is whether the codification of this generalization in terms of functional structure yields predictions or insights that are otherwise unavailable. Incidentally, the labels ‘interrogative’ and ‘focus’ proposed by Rizzi (2001) for the relevant positions, while clearly interpretive in nature, are not sufficient to characterize the interpretation in full, for there is obviously no sense in which the wh–phrase is a Focus in one case but not in the other (or interrogative in one case but not in the other). As already mentioned, the alternative is essentially the one suggested by Chomsky (1995, 2000a, 2008), namely that Topic, Focus and the like simply name interpretations of left-peripheral material. Suppose that the left periphery considered in (40)–(43) is freely ordered above the verb€– in the absence of any functional hierarchy (of categories or features) forcing it. If so, sentences of the type in (41a), which represent the crucial case of ill-formedness, cannot be excluded on formal, computational grounds; on the contrary, their illformedness must be interpretive in nature. It seems to us that a potential argument in favour of this theoretical stance comes from the fact that the unacceptability of sentences like (41a) is far from a matter of absolute judgement in particular, there is no connection between the intrinsic lexical shape of the wh–phrase and the range of positions that the lexical subject or other topic material can take with respect to it. Thus, Benincà (2001) notes the acceptability of sentences like (45), in which the wh–phrase does not appear to be D-linked. Benincà (2001) further connects the well-formedness of (45) to a ‘rhetorical question’ interpretation, implying
42â•… Romance complementizers:€structure & interpretation ‘a negative answer’. We are not sure that this is the correct characterization of rhetorical questions, which seem to correspond more generally to questions that presuppose a certain answer, not necessarily a negative one (‘Who wants more money? Everybody does!’). Even so, it seems to us that (45) need not be interpreted as a rhetorical question at all€– rather, it can be a genuine request for information. (45)â•… Chi la fisica la sa bene? Who the physics it knows well ‘Who knows physics well?’
Independently of what its correct characterization may be, we may agree that the interpretation of (45) differs from that of (41a). This leads us to predict that in the out of the blue context implied by (41a), the utterance in (45) is equally ill-formed, while (41a) becomes grammatical if the range of interpretations that make (45) grammatical is forced. In a theory where the relative position of wh–phrases and topics/lexical subjects is governed by functional hierarchies of categories/features, we must assume that the very same wh–phrases can be associated with one or the other of these functional specifications and can be placed according to them. The other way of thinking about the relevant data is that the computational component allows for any positioning of the relevant elements (in the left periphery); however, crucially, different orderings yield different interpretations. To the extent that both views can be used to characterize contrasts like those considered here, they are notational variants. However, one of them is arguably simpler€– in more than one respect. To begin with, notions such as topic and focus are encoded only once (as interpretations) in the theory we are upholding here€– while they are (redundantly) encoded twice (as interpretations and as grammatical properties) in the alternative theory. This simplicity argument has a counterpart when it comes to the single lexical entries. Assuming the maximally restrictive Inclusiveness principle of Chomsky (1995), according to which only intrinsic properties of lexical items (i.e. properties associated with that item in the lexicon) enter syntactic computation, we would have to admit that topic, focus and similar properties can be (optionally) associated with any nominal head. The alternative is to consider that these notions, like other notions that are relational in nature, do not correspond to features/categories at all, but rather to configurations (cf. Chomsky 2000a on theta-roles)€ – defined in this case at the LF interface. These simplicity considerations, in the absence of empirical evidence, prove decisive in our view.14
The left periphery beyond complementizersâ•… 43 1.3.2 Embedded contexts In the preceding section, we considered the relative order of wh–phrases and lexical subjects in matrix interrogatives. A further element of complexity is introduced by embedded interrogatives. Here the judgement of Rizzi (1997:€289) is that in Italian the topic can precede the wh–phrase, as in (46a), while the configuration in (46b), in which the topic follows the wh–phrase, is ‘slightly marginal’. For us, both examples are equally well-formed. In fact, Rizzi’s (1997) structural schemas allow the topic to either precede or follow the wh–phrase, assuming its position to be the same in (46a) and in (46b), namely C(Focus). (46)â•… a.â•… b.
Mi domando il premio Nobel a chi lo potrebbero dare myself I.ask the prize Nobel to whom it they.could give ‘I wonder to whom they could give the Nobel prize’ Mi domando a chi il premio Nobel lo potrebbero dare myself I.ask to whom the prize Nobel it they.could give
Recall from the previous section that for Rizzi (1997), movement of the verb to C(Focus) normally excludes the wh–phrase–topic order in matrix questions. Therefore he automatically predicts that the latter resurfaces in embedded questions, as in (46b), because of the absence of verb movement. Data from Northern Italian varieties, however, call his explanation into question. In a null-subject language like Italian, it is hard to detect the position of the verb in the sentence on independent grounds, but in Northern Italian varieties the position of the subject clitic provides a reliable independent test. In many of these languages, inversion of the verb with the subject clitic does not take place in either matrix or embedded questions, providing evidence that the verb maintains its I position in all cases. Yet in the same languages, lexical subjects and other topics may appear in either right-peripheral or left-peripheral positions in matrix interrogatives, while the occurrence of the topic or lexical subject between the wh–phrase and the verb is not attested. This contrasts with embedded questions, in which the lexical subject/topic can either precede or follow the wh–phrase, as shown in (47). (47)â•… Fontanigordaâ•… (Liguria) a.â•… k2lu li duve u druome that.oneâ•… thereâ•… whereâ•… he sleeps ‘Where does he sleep?’ b. ne suo k2lu li duve u druome not I.know that.one there where he sleeps ‘I don’t know where he sleeps’
44â•… Romance complementizers:€structure & interpretation b’.â•… ne suo duve k2lu li u druome not I.know where that.one there he sleeps ‘I don’t know where he sleeps’
In other words, the contrast between matrix and embedded questions in a language without subject-clitic inversion is exactly the same as the one observed in a language with subject-clitic inversion, as in (48) vs. (40). (48)â•… Modena a. a n sO briza to fradEl indo al I not know not your brother where he ‘I don’t know where your brother goes’ b. a n sO briza ki lo l a I not know not who he him has ‘I don’t know who he has called’
va goes tSamE called
The conclusion that the ordering of the lexical subject/topic with respect to the wh–phrase in matrix questions is not determined by the position of the verb with respect to the wh–phrase is supported by at least one independent argument, from subject-clitic languages with interrogative inversion. In several relevant Northern Italian varieties, the subject clitic is not simply inverted after the verb but doubled before and after the verb. The relevant examples are (23) for questions introduced by a combination of wh–phrase and complementizer and (24a’, b’) for questions introduced by a simple wh–phrase. Therefore the supposed Spec-head adjacency of the wh–phrase and the verb is systematically disrupted by the preverbal subject clitic. In terms of the structure already provided in (23), in (24a’, b’) a D position intervenes between the wh–phrase in the left periphery of the sentence and the verb in C. If one wanted to claim that the clitic represented adjoined material, one would at the very least have to clarify how the double adjunction (of a proclitic and an enclitic) comes about. In short, for the various reasons reviewed, the contrasts relating to the positioning of the lexical subject in matrix and embedded questions cannot be due to the Wh–Criterion (or its variants). In other words, the relative ordering of wh–phrases and topics/lexical subjects seems to be determined by their intrinsic properties, without the position of the verb playing any role. If so, then it is no longer clear that the general acceptability of wh–phrases in front of lexical subjects/topics in embedded questions, and their acceptability in certain matrix questions (introduced by ‘why’, D-linked wh–phrases etc.) should not be accounted for in the same way. Answering the question of what would unify these various contexts is beyond the scope of the present chapter, which aims simply at establishing a
The left periphery beyond complementizersâ•… 45 theory of complementizers€– including of course its compatibility with a more general model of the left periphery. Nevertheless, the solution envisaged by Manzini and Savoia (2005) is worth mentioning, since it represents an extension of sorts of the approach to complementizers themselves. The idea is that embedded interrogatives, in which the wh–phrase can precede topic material, are effectively (free) relatives, of which the wh–phrase represents the nominal head. In other words, just as the complementizer is not in the left periphery of the sentence but rather projects its own noun phrase, the wh–phrase can project its own noun phrase, which embeds a sentence, eventually including topical material. This proposal can equally be applied to matrix questions, as Manzini and Savoia (2005) effectively do for matrix questions introduced by ‘why’. In other words, ‘why’ is not in the left periphery of interrogative questions; rather, it systematically introduces them as a sort of propositional operator (complementizer). In any event, we may assume that ‘why’ projects a nominal constituent€– which could then be taken to be more akin to a (free) relative. The latter proposal could further be extended to D-linked wh–phrases, and more generally to all wh–phrases that embed a topic. Other extant proposals in the literature seem to rest on intuitions compatible with the analysis just sketched. We have already mentioned Rizzi’s (1997) treatment of ‘why’ as a complementizer head. Kayne and Pollock’s (2001) discussion of pourquoi ‘why’ in French is also relevant. They consider the fact that ‘why’ and other wh–phrases like en quel sens ‘in which sense’ do not trigger inversion of the verb with the subject clitic in French matrix questions, in contrast to other wh–phrases. What they suggest is that questions like en quel sens les fleurs parlent ‘in which sense flowers speak?’ ‘include an abstract verb corresponding to say’, as if one were to say ‘In which sense are you saying that flowers speak?’. This proposal shares with ours the intuition that ‘why’ and other items of the same class belong to a different sentential(-like) constituent with respect to the sentence they introduce. At the same time, for reasons of restrictiveness of the theory, but also for strictly empirical reasons, we reject what Kayne (2006, 2008a) calls silent categories (Manzini and Savoia 2008a, 2009a, 2010; Savoia and Manzini 2010), including the ‘abstract verb’ of Kayne and Pollock (2001). In our intuition, the right interpretation for a sentence like (45), with the order wh–phrase–topic, is more likely to be simply ‘who (is it that) knows physics?’. Similarly, the interpretation of en quel sens les fleurs parlent? could be reconstructed as ‘in which sense (is it that) flowers speak?’. In other words, in both cases the embedded sentence is predicated of the wh–phrase€– more or less as we expect for (free) relatives, or clefts.15
46â•… Romance complementizers:€structure & interpretation For the sake of completeness, we note that in the discussion that precedes, the position of lexical subjects is assimilated to that of topics. However, this does not necessarily hold in all languages (witness the case of English), nor even in all Romance languages. Thus, in French, the lexical subject appears between the wh–phrase and the verb in so-called complex inversion examples, as in (49), independently of interpretive factors. The fact that the verb is inverted with respect to the subject clitic means that it is in a relatively high (conventionally C) position, so that the lexical subject must be in a higher position still€– yet not a topic position. The present analysis already provides a non-topic position for subjects above C, namely the position taken by the preverbal subject clitic in the structure in (23). The ability of the lexical subject in French to appear in a position which in Northern Italian varieties is only available for subject clitics correlates with the fact that while lexical subjects are obligatorily doubled by subject clitics in Northern Italian varieties, they lexicalize the subject alone in French€– and determine complementary distribution with subject clitics. (49) Quand Jean vient-il? When J. comes-he ‘When is John coming?’
At the same time, French, like Northern Italian varieties, reserves the lowest subject position, internal to the I domain, for clitics; therefore only subject clitics can be found inverted after the verb in questions. This contrasts with a language like English, in which the lexical subject ordinarily appears after the verb in questions€– and hence presumably occupies the same position as subject clitics in Romance languages. The English-type position of the subject can also be seen in some Romance languages, namely Romansh varieties in which the lexical subject can appear between the auxiliary and the participle, as in (50).16 (50)â•… Trunâ•… (Grisons) a. ain ilts taLO:ʀ kuʀdai pEʀ tiaʀa are the dishes fallen to ground ‘Have the dishes fallen to the ground?’ b. an iLts afOnts duʀmiu have the children slept ‘Have the children slept?’
There is a final set of data from the work of Rizzi (1997) that goes potentially unpredicted by the present approach. According to the data presented there, a topic can precede a wh–phrase not only in matrix questions, but also in embedded questions. However, a che-type complementizer cannot be preceded by a topic. The relevant data for embedded questions are of the type in (46);
Conclusionsâ•… 47 by contrast, Rizzi (1997) judges (51) and the like to be ill-formed. Benincà (2001), however, does not exclude topics in front of declarative complementizers altogether, since according to her an example like (51b) is well-formed. However, (51b) involves a so-called ‘hanging topic’; for Benincà (2001), Â�so-called clitic left dislocation, as exemplified in (51c), remains ill-formed. For us, the two sentences in (51b–c) have the same status; i.e. they are both acceptable. The same holds for a sentence like (51a), which is ambiguous between the descriptive categories of hanging topic and clitic left dislocation. (51)â•… a.â•… b. c.
Credo il tuo libro che loro lo apprezzerebbero molto I-believe your book that they it would.appreciate a.lot ‘I believe that they would appreciate your book a lot’ Sono certa questo libro che non ne ha mai parlato nessuno I.am certain this book that not of.it has ever spoken anybody ‘I am certain that nobody has ever talked about this book’ Sono certa su questo tavolo che non ci hanno messo niente I.am certain on this table that not there they.have put anything ‘I am certain that nobody put anything on this table’
The theory of Rizzi (1997) is constructed in such a way as to exclude examples of the type in (51). In particular, the che complementizer in these examples is identified with the highest C position, namely C(Force), which closes off the C field (cf. fn. 4, 6); therefore the prediction is that no leftperipheral material can precede it. In the present approach, however, in which che is the head of an independent nominal projection, the possibility must be open for its left periphery to host topical material. In fact, this is the structure we have postulated for sentences like (23) in section 1.2. Therefore examples like (51) are predicted to be grammatical. 1.4
Conclusions
In a nutshell, the present proposal represents a viable alternative to theories of C as a functional projection of the verb€– and possibly a better one in that it allows us to simplify functional architectures, treating complementizers in terms of the recursion of predicate argument structures. Another advantage resides in the simplification of the lexicon, to the extent that the complementizer can be given a unified lexical entry with the wh–phrases with which it is (often) homophonous. Though this would appear to be an even smaller gain than the previous one, consider that in a minimalist model the lexicon is all there is to language variation€– therefore simplifications in the lexicon are simplifications of the only learning task children have in front of them.
48â•… Romance complementizers:€structure & interpretation In the next chapter, we will pursue this second theme, providing some case studies in language variation and illustrating the account available for them under the present model. We argue that the model we are proposing is capable of accounting for the intricate parametrization of Romance varieties in a transparent way, effectively predicting that certain parametric values should be instantiated. In our view, these predictions depend on treating the complementizer as an argument, and specifically as an argument belonging to the same wh–series as classical wh–phrases. The burden of proof is on other theories to show that the same results can be mimicked if complementizers are functional projections of the verb. This may very well be possible€– but almost certainly through additional assumptions.
2 Variation in Romance k-complementizer systems
As observed in chapter 1, finite complementizers in Germanic and Romance are clearly nominal, belonging to argumental series such as demonstratives and wh–elements respectively. Therefore we propose that the complementizer is not a functional category of the verb, but rather an independent nominal head, which satisfies an argument slot of the matrix verb and which takes the embedded sentence as its complement. The aim of the present chapter is to investigate the range of variation in the lexicalization of complementizers in Romance varieties. We argue that only the treatment of the complementizer sketched in chapter 1 allows for a transparent account of the observed variation. First, the overlapping of Romance complementizers with the wh–system follows precise patterns, which excludes the possibility that we are merely dealing with homophony. Second, an account of the variation internal to the complementizer system requires the complementizer to be interpreted as introducing a propositional variable€– which is natural if it is a nominal head, but not if it is a functional projection of the verb. In section 2.3, we also consider the finiteness restriction that complementizers are subject to, arguing that they do not select for embedded temporal/modal properties (this being a potential argument in favour of their status as a functional projection of the verb), rather they select for the properties of the embedded EPP argument€– hence of the proposition as a whole.
2.1
Systems with two k-complementizers
The system of standard Italian, which we have considered in chapter 1, is characterized by a single complementizer introducing finite declarative clauses€– which is also the distribution familiar for English that. However, there are many Central and Southern Italian varieties which have two finite declarative complementizers, as in (1)–(2). One of the two complementizers, generally ka, is systematically found to introduce complements to verbs 49
50â•… Variation in Romance k-complementizer systems of ‘saying’ and ‘knowing’. The other complementizer, generally ke/ki (k), is found to introduce complements to verbs of ‘believing’ and ‘wanting’, though in some varieties (Guglionesi) some predicates alternate between the two complementizers. In other varieties, where all these predicates select the same complementizer ka, a second complementizer is found to introduce complements to ‘before’ and ‘after’ (Montenerodomo). Another important common property of these systems is that they involve some overlap between the complementizer system just described and the wh–quantifier system. Crucially, this overlap never affects the ka complementizer; instead it is the ke/ki complementizer that quite generally overlaps with the wh–quantifier for ‘what’. (1)â•… Guglionesi (Molise) a.â•… i. m ɔnnə dəttə ka vE krE to.me they.have said that he.comes tomorrow ‘They told me that he will come tomorrow’ a. ii. pEndzə ka vvE krE I.think that he.comes tomorrow ‘I think that he will come tomorrow’ b. i. vujjə kə vi krE I.want that you.come tomorrow ‘I want you to come tomorrow’ b. ii. pEndzə kə vvE krE I think that he.comes tomorrow ‘I think that he will come tomorrow’ b. iii. sNg aSSeutə prəmə kə tteu mənəssə I.am gone.out before that you would.come ‘I went before you came’ b. iv. sNg aSSeutə dppə kə ssi məneutə I.am gone.out after that you.are come ‘I went after you came’ c. kə ffi what you.do ‘What are you doing?’ (2) Montenerodomoâ•… (Abruzzi) a. i. m om ditt ka vi dum:n to.me they.have said that you.come tomorrow ‘They told me that you will come tomorrow’ a. ii. pEndz ka iss ve dum:n I.think that he comes tomorrow ‘I think that he will come tomorrow’ a. iii. vuless ka mniSS I.would.want that he.would.come ‘I would want him to come’
Systems with two k-complementizersâ•… 51 b. i. b. ii.â•… c.
so SSeut preim k mneiv I.am gone.out before that he.came ‘I went before he came’ so SSeut dop k tu avi mneut I.am gone.out after that you had come ‘I went after you came’ k ffi what you.do ‘What are you doing?’
The existence of double complementizer systems in varieties of Central and Southern Italy has been noted in the literature. However, Rohlfs (1969 [1954]:€190) puts it in the same bracket as another form of split lexicalization for sentential introducers, found in so-called Balkan languages as well as in some Southern Italian and Sicilian varieties. The relevant systems are characterized either by the lack of morphological infinitives or by their very reduced presence. Therefore the embedded infinitival clauses of English or Italian, associated with control and raising interpretations, are rendered in these Balkan (-like) languages by the embedding of finite clauses introduced by a specialized particle, which in Calabrian varieties, for instance, is mu. Thus, for Rohlfs, Neapolitan pèns ca vèn ‘I think that he come’ vs. vògli k mmang ‘I want that he eats’ and Calabrian pensu ca vèni ‘I think that he comes’ vs. vogghiu mu (mi) mangia ‘I want that he eats’ represent strictly comparable systems. The strongest argument for the independence of the two phenomena is that they actually combine. Thus, the Calabrian variety from Arena in (3), which has control and raising complements introduced by mu, also independently presents two complementizers of the k-series, roughly with the distribution described above for Montenerodomo, as in (2a–b). Of the two complementizers, it is the ki one, selected notably by ‘before’ and ‘after’, that has the same form as the wh–quantifier for ‘what’, as in (3c). (3)â•… Arena (Calabria) a. i. mi Dissiru ka vEni dmani to.meâ•… they.said thatâ•… he.comes tomorrow ‘They told me that he will come tomorrow’ a. ii. kriju ka vEni I.believe that he.comes ‘I think that he will come’ a. iii. E mmiu ka vini it.is better that you.come ‘It is better for you to come’ b. nESSivi duppu ki vinni I.went.out after that he.came ‘I went after he came’
52â•… Variation in Romance k-complementizer systems c. d.
ki ffatSi ijju what does he ‘What is he doing?’ vuu mu niï†fiï†fiu I.want Prt I.go.out ‘I want to go out’
We assume that introducers of the type of mu in Calabrian (or të in Albanian) are not complementizers in the sense defined in the preceding paragraph€– i.e. they are not (nominal) heads taking the embedded proposition as their (sentential) complement. Rather, they are internal to the embedded clause, to which they contribute modal properties (Rivero 1994; Roberts and Roussou 2003) and/or EPP properties involved in control and raising interpretations. Thus, we will put mu and the like aside without further discussion. Ledgeway (2003a, 2005, 2009) also argues against the identification of double complementizer systems with systems including a complementizer and a so-called subjunctive particle. It is possible that the three-complementizer split (ca, cu and che) described by Ledgeway (2005:€367 ff.) on the basis of Sgrilli (1983) for Early Salentino represents a system comparable to Arena in (3)€– with cu representing the ‘subjunctive particle’, very much like ku in contemporary Salentino varieties. Let us go back, then, to (1). The distribution of ka and k in varieties such as Guglionesi is reminiscent of the fact that standard Italian employs the indicative in complements of ‘to say’ and the subjunctive in complements of ‘to think’ or ‘to want’, with regional variants allowing for the indicative at least under ‘to think’, as in (4). (4) a. b. c.
Dico che viene I.say that he.comes ‘I say that he will come’ Penso che venga/ viene I.think that he.come/ he.comes ‘I think that he will come’ Voglio che venga I.want that he.come ‘I want him to come’
This similar distribution (and the somewhat similar distribution of indicatives vs. infinitives in the English translations) seems to suggest that some common property lies at the core of the complementizer split and the modality split. However, it should be stressed that it is not the case that one phenomenon depends on the other. In particular, in the examples in (1)–(2) the two
Systems with two k-complementizersâ•… 53 complementizers can embed the same indicative verbal forms, while ka in the Montenerodomo example in (2aiii) can embed the one form of subjunctive that appears to be present in the language, which can be roughly characterized as the counterfactual. Further examples of this pattern in Guglionesi are provided below in (9). In other words, there is no obligatory selection of the verb modality by the complementizer or vice versa. The evidence at our disposal also allows us to exclude the possibility that the complementizer split can be linked to the relatively impoverished modal system of varieties like (2); in other words, not only is it the case that there is no selection relation between complementizers and subjunctives, but there is no complementary distribution (functional equivalence) either. A case in point is represented by Sardinian varieties such as Paulilàtino in (5), in which a full-blown indicative€– subjunctive system of the standard Italian type combines with a double complementizer system of the type exemplified in (2) with Abruzzese varieties. (5) Paulilàtino (Sardinia) a. i. m anta nau ka bbeni kkraza to.me they.have said that he.comes tomorrow ‘They told me that he will come tomorrow’ a. ii. pEnts ka bbenizi I.thinkâ•… that you.come ‘I think that you will come’ b. i. pEnts ki bbEndzEDE I.think that he.come ‘I think he will come’ b. ii. kErdz ki bbEndzEzE I.want that you.come ‘I want you to come’ b. iii. sE bissiu appustisâ•… ki ze Benniu I.am gone.outâ•… after that you.areâ•… come ‘I went after you came’ b. iv sE bissia primma ki EssE Benniu DuE I.am gone.out before that be come you ‘I went before you came’ c. ittE faEzE what you.do ‘What are you doing?’
It will be noted that the ka complementizer combines with morphological indicatives in (a) while the ki complementizer combines with morphological subjunctives. However, the correlation between complementizer and verb modality breaks down in connection with complements to ‘before’ and ‘after’
54â•… Variation in Romance k-complementizer systems in (4biii) and (4biv) respectively. Following the pattern of standard Italian, the complement of ‘before’ is a subjunctive (a counterfactual) while the complement of ‘after’ is an indicative. Nevertheless, in both cases the complement is introduced by the ‘modal’ complementizer ki. This should of course be taken together with the existence of languages, like Montenerodomo in (2) or Arena in (3), in which complements to ‘before’ and ‘after’ maintain a specialized complementizer while other declarative sentential complements are uniformly introduced by ka. The lack of a one-to-one correlation between complementizers and modality is discussed at length by Ledgeway (2003a, 2005), who studies double complementizer systems in Early Neapolitan and other Southern Italian texts. He concludes that indicative clauses are frequently introduced by ‘CHE’. He also notices that selection for complementizers cannot be directly imputed to matrix predicates, since ‘it is not difficult to find minimal pairs … where the same main clause predicate selects in one case for an indicative clause headed by CA and in another for an indicative clause headed by CHE’ (Ledgeway 2005:€348). In other words, the historical varieties studied by Ledgeway have a complementizer system strictly comparable to that illustrated here by Guglionesi in (1) and in fact also by Paulilàtino in (5); the similarity between Ledgeway’s varieties and Sardinian ones is independently noted by Damonte (2006), who extends Ledgeway’s analysis to them. Nevertheless, Ledgeway subscribes to Formentin’s (1998:€ 432) conclusion that ‘the conjunction ca never introduces subjunctive clauses’, thus establishing a one-way implication between subjunctive and CHE complementizers. This oneway implication is not upheld by our data. Recall that we have already reviewed systems, such as Montenerodomo in (2), in which ka introduces a subjunctive (2aiii). More to the point, Sardinian varieties, which have fully productive subjunctives, show that both complementizers combine with both moods (indicative and subjunctive). Relevant data are found below in (11) for the variety from Làconi. Similarly the free alternation of ka and k with subjunctives is exemplified for Guglionesi in (9). Evidence that the subjunctive does not imply chi independently emerges in the Baunei corpus of Damonte (2006:€92), who quotes examples of ca with the subjunctive such as Mi pare ca custas cadirasa siente meda comodasa ‘it seems to me that these chairs are (lit:€be) very comfortable’. 2.1.1
Definite and indefinite complementizers€– and alternative analyses According to the conclusions of chapter 1, the finite k-type complementizer in Romance languages is an independent nominal head introducing a propositional
Systems with two k-complementizers╅ 55 variable, whose content is restricted by the embedded sentence. If so, the difference between the two complementizers in (1) is most naturally construed as a difference between two types of variable/quantification. We could assume, for instance, that the ka complementizer is a definiteness element, effectively the counterpart to a definite determiner, so that the LF of the Guglionesi example in (1a.ii) would be of the type in (6a). By contrast, the LF of a sentence like the Guglionesi example in (1b.ii) could include an indefinite quantification, corresponding to a free variable bound by existential closure, as sketched in (6b). (6) (Guglionesi) a. I think the x:€x he comes tomorrow b. I think for some x:€x he comes tomorrow
Because the evidence and the results discussed in this paper are mostly morphosyntactic in nature, we must content ourselves with a sketchy account of the interpretation. The evidence presented above is, however, sufficient to yield the generalization that of the two complementizers present in languages like Guglionesi in (1), it is always the indefinite complementizer that overlaps with a wh–quantifier. This is because the other complementizer effectively introduces a definite description, which is hardly compatible with wh–quantification€– while the indefinite complementizer introduces a propositional variable subject to existential closure, more or less like the argumental wh–variable. Manzini and Savoia (2005) provide examples from a considerable number of languages which behave like Guglionesi, and in all cases the same generalization holds, as shown by the summary table reproduced in (A) in the Appendix. Of course it is also logically possible to have double complementizer systems of the type in (6) in which no overlap with the wh–system is found, as summarized in table (A’) in the Appendix; the Paulilàtino sentence in (5) exemplifies this type of system.1 The preceding results are important for the present discussion for more than one reason. A preliminary point is that the identity of form shown by complementizers and wh–elements covers grammars that are not just different in general terms, but are specifically different in terms of the finite complementizer system€– having either a single complementizer or a split between complementizers. In other words, it cannot be objected that the formal identity of the complementizer with the wh–system is in a sense the property of just one grammar, and therefore to be treated as accidental. More importantly, the fact that in two complementizer systems it is the indefinite complementizer that coincides with the wh–system supports the semantics for complementizers sketched here, and indirectly the syntax in chapter 1.
56â•… Variation in Romance k-complementizer systems Now the question is:€under the alternative view, which has complementizers as C functional heads, could one provide characterizations for the two different complementizers, and capture the overlap with wh–quantifiers (or lack thereof)? Ledgeway (2003a, 2005), dealing with data closely comparable with ours, does provide an answer to the first part of the question. Before evaluating it, let us note that the two complementizer systems of languages like those in (1) may have a reflex even in a language like Italian in complementizer deletion phenomena; we will therefore consider first what the existing literature says about those. In fact there appears to be a good match between matrix predicates that require the indefinite complementizer in (6b) and those which, according to the literature, allow for the deletion of the che complementizer in Italian, as in (7b–d). By contrast, contexts which require the definite complementizer, as in (6a), seem to coincide with those that do not admit complementizer deletion in Italian, as illustrated in (7a). (7)â•… a.â•… b. c. d.
So *(che) viene I.know that he.comes ‘I know that he is coming’ Penso (che) venga I.think that he.come ‘I think that is coming’ Vorrei (che) venisse I.would.want that he.came ‘I would like him to come’ E’ meglio (che) venga It.is better that he.come ‘It is better for him to come’
The issue is made more complex by the existence of at least one variety in which effectively any matrix predicate admits complementizer deletion. The latter is identified by Cocchi and Poletto (2002) with the variety spoken in Florence. As in many cases involving Italian, the question can legitimately be asked whether, on the contrary, the judgements attributed to the standard are not unduly constrained by normative considerations; in fact even the judgements in Cocchi and Poletto (2002) appear to be unnecessarily restrictive for the Italian spoken in Florence.2 Be that as it may, the reason to address the (potential) pattern in (7) is that phenomena of complementizer dropping have been prominent in the theoretical literature since the study of the that-t filter in English by Chomsky and Lasnik (1977). Therefore distributions of the type in (7) have been discussed more than once (Poletto 2001; Giorgi and Pianesi 2004), providing us with a possible alternative to the analysis advocated here.
Systems with two k-complementizers â•… 57 On the basis of the preceding discussion of double complementizer systems, we would be led to characterize the alternation in (7) by saying that in order to introduce a propositional definite description the complementizer is necessary, as in (7a); while other types of propositional complementation admit the bare embedding of a sentence under the matrix verb, as in (7b–d). As for the question of why contexts obligatorily introduced by the complementizer feature the indicative, while subjunctive contexts admit complementizer dropping, we can assume that the indicative vs. subjunctive split is itself connected with the definite or indefinite binding of a propositional variable. In fact, Manzini (2000) notices that mood selection is not exclusively determined by the matrix predicate. Thus, ‘to know’ in (8a) normally requires the indicative. The latter remains possible when the matrix predicate is questioned, as in (8b), in which case the reading is still a factive one presupposing the truth of the embedded proposition. Questioning the matrix sentence also renders the subjunctive possible, in which case, however, the embedded proposition is in the scope of the question operator, yielding a non-presuppositional reading. (8)â•… a.â•… b. c.
So che è/ *sia venuto I.know that he.is/ he.be come ‘I know he came’ Sai che è venuto? you.knowâ•… that he.is come ‘Do you know that he came?’ Sai che sia venuto? you.know that he.is come ‘Do you know if he came?’
Alternations like (8b–c) can be described by saying that the indicative (like ka–type complementizers) introduces a propositional definite description, read outside the scope of polarity operators like the question operator. By contrast, the subjunctive (like ke–type complementizers) introduces an indefinite propositional variable, interpreted within the scope of polarity operators. This accounts for the parallelism between the complementizer split and the indicative vs. subjunctive split apparently present in (7). Let us now consider Giorgi and Pianesi (2004) and Poletto (2001). They assume, as we do, that so-called complementizer deletion does not correspond to the actual deletion of lexical material nor to the zero instantiation of the C position. For Giorgi and Pianesi (2004), while the sequence of che and a subjunctive provides separate instantiations for modality (through the che complementizer) and for agreement (through the I position of the verb), in complementizer deletion contexts, the Mood and Agr properties are conflated in the
58â•… Variation in Romance k-complementizer systems I head, and simply lexicalized by the verb. The reason why che cannot delete in indicative contexts has to do with the fact that in indicative contexts T and Agr properties cannot conflate, but must be independently lexicalized. To be more precise, ‘the T-features are duplicated in che/C in such a way that the indexical component can be evaluated outside the clause’ (Giorgi and Pianesi 2004:€200). According to Giorgi and Pianesi (2004) the satisfaction of this latter condition is independently required in order to obtain the so-called Double Access Reading, under which the situation denoted by the embedded sentence is taken to hold both at the time of the matrix event and at the time of utterance. What is relevant here is that Giorgi and Pianesi’s (2004) framework could in principle account for double complementizer systems, as the result of a split between exponents of C endowed with Tense (ka) and exponents of C endowed with Mood (ke/ki). But since there is no (direct) connection between the whâ•fi system and subjunctive Mood (e.g. wh–questions do not necessarily select the subjunctive) the overlapping of modal complementizers and the wh–system (to the exclusion of the non-modal complementizers) would not be predicted. The present discussion provides the missing link; but crucially it does so by extricating the complementizer from the Mood and Tense system to which Giorgi and Pianesi (2004) confine it€– and by assuming that it is an argumental element. Correspondingly it is removed from the C position and construed as an independent (nominal) head. Let us now turn to Poletto (2001), who draws a parallel between complementizer deletion and Germanic V-to-C movement; the latter, at least in some languages, is in complementary distribution with the lexicalization of the complementizer (and thus seems to result in its deletion). The idea is that ‘the class of verbs selecting C[omplementizer] D[eletion] complements is exactly the same as in Germanic embedded V2 contexts’ (Poletto 2001:€267). Thus both phenomena can be captured by the movement of the verb to the C position. On the assumption that ‘CD is possible only when the embedded verb is a subjunctive, a future or a conditional form’ and that ‘these forms all have a modal quality’, Poletto (2001:€278) proposes that ‘a [–realis] feature … is realized on the head of the complement and attracts the verb into the CP domain’. Let us leave aside the question of whether the verb does or does not move to the C position in Italian complementizer deletion contexts, what is relevant here is Poletto’s (2001) characterization of these contexts in terms of a [–realis] feature in C. In these terms, it is not difficult to recognize the common conceptual core of Poletto’s (2001) analysis and Giorgi and Pianesi’s (2004), namely that complementizer deletion involves modal properties, absent from indicative contexts. The point on which we differ from both analyses is once again made clearer if we try to extend Poletto’s (2001) account to the double
Systems with two k-complementizers â•… 59 complementizer systems considered here. One could assume, for instance, that what we have called the indefinite complementizer lexicalizes the€–realis feature, while the other complementizer satisfies the +realis feature in C. But if so, there is no reason why the former should overlap with the wh–system (to the exclusion of the latter), since wh–questions do not appear to have any (immediate) connection with€–realis (not being restricted, say, to the subjunctive). As before, the present theory has the advantage that it can make direct predictions on this point€– but these crucially depend on complementizers being extricated from the C position, and treated as argumental, nominal heads. We consider next the proposals put forth by Ledgeway (2003a, 2005). The empirical generalization that Ledgeway proposes concerning the distribution of the two complementizers depends on a further set of data concerning the distribution of CHE and CA with respect to embedded left-peripheral material. In his corpus of eleven early texts, he finds that out of a total of 327 examples of CA-clauses … a mere 10.1% were found to contain one or more elements in the left periphery, whilst from a total of 1,061 examples of indicative clauses introduced by CHE … 41.8% were found to host one or more elements in the left periphery … Secondly even if there do occur some examples of the complementizer CA preceding one or more elements of the left periphery … with very few exceptions, all such examples involve elements of the Focus field, namely … foci … and fronted indefinite quantifiers.â•…â•… (2005:€360)
Ledgeway’s characterization of CA and CHE is in fact partially based on mood selection, since he assumes that ‘CA and CHE are invariably merged in … Fin° … as an overt reflex of the different modal specifications (indicative vs. subjunctive) they check in Fin°, from which automatically follows the claim that CA only occurs in indicative clauses and CHE only in subjunctive clauses’ (2005:€372). However, ‘the appearance of topics and foci forces the complementizers CA and CHE to move out of Fin° to target Force°’ (2005:€376); crucially ‘this movement operation is spelt out morphologically only in the case of the indicative complementizer CA which invariably surfaces as CHE’ (2005:€374–5). The empirical generalization, namely that the presence of an embedded left periphery forces the complementizer to be CHE, deserves to be investigated for contemporary varieties as well. What we have found is that our varieties do not observe such a restriction, since both foci (e.g. (9c)) and topics (including clitic left dislocation in (9b)) appear under ka in Guglionesi in (9) and in the Sardinian varieties in (10)–(11).3 We also exemplify topics and foci under k and ki respectively, in order to establish that our varieties do not retain the asymmetry between the two complementizers, simply reversing their respective positions (i.e. ka higher and k/ki lower).4
60â•… Variation in Romance k-complementizer systems (9) Guglionesi a.â•… m ann  ka krE to.me they.have said that tomorrow you.came ‘They told me that tomorrow you would come’ b. m ann  ka u kES ts l annâ•…  to.me they.have said that the cheese Refl it they.have taken the boys ‘They told me that the cheese was taken by the boys’ c. m ann  ka ï†fi ts ann tt i ajjEun to.me they.have said that the cheese Refl they.have taken the boys ‘They told me that it was the cheese that the boys took’ d. vujj ka/k pur tEu mnss I.want that also you come(subj) ‘I want you too to come’ e. vujj ka/kâ•… krE pur lrâ•… mnssn I.wantâ•… that tomorrowâ•… also they come(subj) ‘I want them as well to come tomorrow’ (10)â•… Paulilàtino a. m anta nau ka raza bbenizi to.meâ•… they.haveâ•… saidâ•… thatâ•… tomorrowâ•… you.come ‘They told me that you will come tomorrow’ a’. m anta nau ka u libru ɖ aza leddzju to.me they.have said that the book it you.have read ‘They told me that the book you read’ b. EstE mmendzus ki fintsaza juanni bEndzEDEâ•… kraza it.is better that even John comes tomorrow ‘It is better that John as well comes tomorrow’ (11)â•… Làconi (Sardinia) a. dEɔ krEɔ ka/tɔi issu Buru/ kraza I believe that he too/ tomorrow ‘I believe that he will come as well/tomorrow’ b. dεɔâ•… krEɔ tSi/kaâ•… issu Buru/â•… kraza I believeâ•… that he too/ tomorrowâ•… ‘I believe that he will come as well/tomorrow’
eniDi he.comes EdZaDa he.comes
Let us now consider Ledgeway’s (2003a, 2005) analysis, which, as we have seen, consists of a more conservative assumption, namely that complementizer choice depends on mood, and an innovative proposal, namely that movement of CA results in it being spelled out as CHE. The first part of the proposal is in essence the same as that reviewed above for complementizer deletion. Thus our generalization that the ‘subjunctive’ complementizer is the one that overlaps
Systems with two k-complementizersâ•… 61 with a wh–element (if such an overlap is present in the system), remains as inaccessible to Ledgeway as it does to Giorgi and Pianesi and Poletto, and for much the same reason. Indeed, connecting complementizers directly to modality (as is natural in a treatment where both belong to the functional spine of the verb) leaves us without any basis for associating the complementizer with the argumental, nominal system of wh–phrases. The novel part of Ledgeway’s (2003a, 2005) proposal is that movement of CA from the Fin to the Force position results in its morphological spellout as CHE. Now, there is no syntactic and/or morphological theory that can literally change a lexical item into another (because of Chomsky’s (1995) Inclusiveness, among others). Nor is it possible to have a lexical entry specified for a context defined by movement (since only intrinsic properties of lexical items belong in the lexicon). Therefore, Ledgeway’s analysis seems to imply that double-complementizer languages have a spell-out for indicative complementizers in Fin€ – and another spell-out for subjunctive complementizers as well as for complementizers in Force. This means that the unification achieved by Ledgeway’s analysis is only apparent€– as it requires a disjunctive lexical entry for the CHE complementizer. Alternatively, one could try to unify the entry for CHE by recourse to underspecification€ – i.e. to the idea that CHE is simply the default complementizer of the relevant languages, inserted as the Elsewhere case. But if so, the generalization noted here concerning its overlapping with the wh–system would become impossible to explain, in the absence even of modal properties on the complementizer. 2.1.2 Generalized wh–complementizers Section 2.1.1 discusses systems with a split between definite and indefinite complementizers of the type of Guglionesi in (1) or in Paulilàtino in (5). A different grammar, however, seems to be exemplified by Montenerodomo in (2) or by Arena in (3), in which the k/ki complementizer overlapping with a wh–question word is restricted to a few contexts, which in our data coincide with complements to ‘after’ and ‘before’. These contexts do not correlate with the verb modality€– which can be subjunctive under ‘before’ but is consistently indicative under ‘after’. Interestingly, all of the languages tabulated in (A’) of the Appendix, in which there is no overlap between the wh–system and the complementizer system, show the definite vs. indefinite complementizer split of the type in section 2.1.1. By contrast, the distribution of k/ki complementizers in languages like Arena in (3) or Montenerodomo in (2) appears to be necessarily connected with their wh–nature.
62â•… Variation in Romance k-complementizer systems In our view the key to the distribution observed resides in the fact that prepositions like ‘before’ or ‘after’ can easily embed complements introduced by wh–arguments/adjuncts, as in the comparative structures in Italian (12). In turn, the comparatives in (12) appear to be interpreted in relation to a temporal argument. Thus (12a) corresponds to an LF roughly of the form ‘I arrived at a time t before/earlier than a time t such that I thought I would arrive at t’, where the most embedded (italicized) proposition undergoes ellipsis.5 (12)â•… a.â•… b.
Sono arrivato prima di I.am arrived before of ‘I arrived earlier than I thought’ Sono arrivato dopo di I.am arrived after than ‘I arrived later than planned’
quanto how.much
pensassi I.thought
quanto stabilito how.much planned
On the basis of (12) we propose that languages that embed a wh–complementizer under ‘after’ and ‘before’ effectively select a (comparative-like) wh–structure. As noted before, the evidence we are considering is essentially morphosyntactic in nature and therefore hardly insightful when it comes to the semantics; hence semantic analyses are sketched here essentially for the sake of falsifiability. We suggest that the wh–complementizer acts in this case as a sort of generalized quantifier, binding the propositional variable, restricted by the following sentence and, at the same time, the temporal variable. We then expect that the generalized quantifier that we have hypothesized also turns up as a whâ•fi binder of arguments; in other words that it is not specialized as a propositional introducer, but will overlap with wh–arguments€– as is indeed the case. It is worth noting that Ledgeway (2009) notices the existence of doublecomplementizer systems that cannot be assimilated to those in section 2.1.1 in relation to the Calabrian variety of Cosenza. He reports that even in the speech of younger generations chi is obligatory in optative expressions such as chi ti vò affucà ‘(that you) go drown’; in a more conservative variety he finds chi to occur in complement sentences when the embedded verb is in the subjunctive (counterfactual). His generalization is that ‘ca is a passe–partout complementizer compatible both with indicative and with subjunctive, while chi is a modally marked complementizer’. Needless to say, this characterization cannot be extended to varieties like Arena in (3), in which ki is selected by ‘after’.6 Let us nevertheless consider his analysis€– namely that ca realizes the Force head while chi lexicalizes the Fin head in Rizzi’s (1997) schema, i.e. the head ‘responsible for marking modal distinctions’. In this schema, therefore, the Cosenza variety would involve the opposite distribution to the varieties of section 2.1.1, in which che is assigned to the Force position, while ca is restricted to Fin. Now, the point being made here is that the k of languages like Guglionesi
Systems with two k-complementizersâ•… 63 and the ki of languages like Arena have in common the fact that they overlap with the wh–system. From this perspective, Ledgeway’s (2003a, 2005, 2009) analysis would be saying that in double-complementizer systems, the wh–complementizer is sometimes a specialized Fin and sometimes a default. It seems to us that a theory that captured the continuity between the various double-complementizer varieties would have a clear explanatory advantage.7 This leads us to the next point to be examined, namely whether our theory can in fact capture this continuity. Crucially, in all languages of the same type as Guglionesi in (1), the indefinite complementizer, in the sense of (6b), turns up in ‘before’/‘after’ contexts as well. Now, if the wh–complementizer in languages like Arena in (3) or Montenerodomo in (2) is to be able to function as some sort of generalized binder of propositional and wh–variables, it cannot be definite (which would exclude wh–quantification). Therefore, in languages with definite vs. indefinite complementizers like Guglionesi, the contexts relevant for its lexicalization will be picked up by the indefinite complementizer, acting as a generalized wh–quantifier. This conclusion only depends on the indefinite nature of the complementizer and not on its actual formal identity with a wh–argument. Therefore we predict that the indefinite complementizer will also pick up ‘before’/‘after’ contexts in systems where it does not overlap with a wh–argument, such as Paulilàtino in (5).8 The final fact worth bringing up is that in all of the languages tabulated in (A–A) in the Appendix, what we have called the indefinite and generalized wh–complementizers also introduce relative clauses. Relevant examples are given in (13)–(16). These involve that relatives, and not who relatives, since, although the relevant complementizers overlap with ‘what’, as illustrated in (1)–(3) and (5), ‘who’ has a distinct lexicalization, as illustrated in (13b) and in (14c)–(16c). We can extend to (13)–(16) the analysis that we have already proposed for (12), treating ki/k as some sort of generalized binder for the propositional variable and the argumental one (actually a resumptive pronoun in the examples in (14b)–(16b). The treatment of ‘after’ and ‘before’ complements in what precedes also implies that we should find the indefinite/generalized wh–complementizer introducing comparatives. The data in Paulilàtino’s (16d) confirm this prediction. (13)â•… Guglionesi a.â•… E kkull k vvad sEmbr he.is that that I.see always ‘He is the one that I see all the time’ b. ki vE who comes ‘Who comes?’
64â•… Variation in Romance k-complementizer systems (14)â•… Montenerodomo a.â•… e kkwillâ•… k vve:tâ•… sEmbr it.isâ•… that thatâ•… I.see always ‘It is the one I see all the time’ b. e kkwill k i so dat l sld it.is that that to.him I.am given the money ‘It is the one I gave the money to’ c. ki ve whoâ•… comes ‘Who comes?’ (15) Arena a. su kkijjiâ•… ki vviju sEmpri they.areâ•… those thatâ•… I.see always ‘They are the ones that I see all the time’ b. su kkijji ki tSi dEttsi li srdi they.areâ•… thoseâ•… thatâ•… thereâ•… I.gave the money ‘They are the ones I gave the money to’ c. ku vni whoâ•… comes ‘Who comes?’ (16) Paulilàtino a. Es kussu ki bbi zEmpErE he.isâ•… that thatâ•… I.see always ‘He is the one that I see all the time’ b. Es kussu ki appɔ jau z inari he.is that that him I.have given the money ‘He is the one that I gave the money to’ a’. juanniâ•… ki Estâ•… ammiu meu John thatâ•… is friend mine ‘John, who is my friend …’ c. kiE bbeniDi who comes ‘Who comes?’ d. tEndzâ•… pru llibrsâ•… ki (n)â•… inarE I.have moreâ•… books thanâ•… not money ‘I have more books than money’
In a language like Ardaùli in (17), of the Paulilàtino general type, the indefinite complementizer ki introduces restrictive relatives and the definite ka complementizer introduces appositive relatives. We suggest that in this language the choice of the complementizer is sensitive to the nature of the embedded variable. In particular, ki is incompatible with appositive relatives, in that they contain an individual variable; thus, the language resorts to the definite complementizer ka.9
‘If’â•… 65 (17)â•… Ardaùli (Sardinia) a.â•… Es kussu ki bbi sEmpErE he.isâ•… that that I.see always ‘He is the one that I see all the time’ b. Es kussu ki app jau zu inarE he.is that that himâ•… I.have given the money ‘He is the one that I gave the money to’ a’. dzuanni ka Estiâ•… ammiuâ•… meu John, thatâ•… is friend mine ‘John, who is my friend …’ c. kiE eniDi whoâ•… comes ‘Who comes?’ d. tEndzâ•… pru llibrɔsâ•… ki (n)â•… inarE I.have moreâ•… books thanâ•… not money ‘I have more books than money’
As discussed in chapter 1, Arsenijevic (2009) and Kayne (2010) propose theories of complementation as relativization. We may wonder how these fare with respect to the relativization facts in (13)–(17). Consider for instance Paulilàtino€ – i.e. a language that has a two-complementizer system, but in which relative clauses are formed without exception by the indefinite complementizer. If the definite ka complementizer is nothing but a relative pronoun, why doesn’t it ever overtly appear in relatives? In fact, the preliminary question arises of how Kayne or Arsenijevic would deal with double-complementizer systems. In Arsenijevic’s terms, different complementizers could select for different types of ‘Force’ arguments. But even so, it is difficult to see why what we have characterized as the indefinite complementizer has the same form as the relative complementizer, but the definite one does not. For it does not seem possible to say that relative clauses associate with a particular type of Force. Kayne’s proposal would have to deal with the facts in a similar way; therefore the same critique applies (cf. also the discussion on conditionals in the next section).
2.2
‘If’
The complementizer system of Italian, as discussed in chapter 1, and that of the varieties considered so far, have a specialized lexicalization for the interrogative/hypothetical complementizer, se in Italian, corresponding (more or less) to English ‘if’. Thus the Italian data in (1)–(4) of chapter 1 are completed by (18).
66â•… Variation in Romance k-complementizer systems (18)â•… a.â•… b.
Se piove (spesso) esco if it.rains often I.go.out ‘If it rains I (often) go out’ Non so se viene not I.know if he.comes ‘I don’t know if he’ll come’
As it turns out, this state of affairs does not hold of necessity, but only as one of several possible parametric options. In some Sardinian varieties with a double declarative complementizer system, the indefinite complementizer also lexicalizes hypothetical and interrogative ‘if’, as exemplified by Làconi in (19). Note that the indefinite complementizer tʃi belongs to the wh–lexical series, exemplified by tʃinni ‘who’ in (19ci). Note also that the same distinction holds between the complementizer for restrictive and appositive relatives as we discussed for Ardaùli in (17); as before, it is the complementizer introducing restrictive relatives that overlaps with the indefinite/‘if’ complementizer. (19)â•… Làconi a. b. i. b. ii b. iiiâ•… c. i. c. ii. d. i. d. ii e. i
m anti nau ka ennis kraza to.me they.have said that you.come tomorrow ‘They told me that you come tomorrow’ bZ tSi EdZas kraza I.wantâ•… that you.comeâ•… tomorrow ‘I want you to come tomorrow’ E mmedZuzu tSi bbEdZaza it.is better that you.come ‘It is better for you to come’ sE essia prima Zi fRssaz arribbau I.am gone.out before that you.were come ‘I went out before you came’ a ttSinni tserriaza to whom you.call ‘Who are you calling?’ itta faizi whatâ•… you.do ‘What are you doing?’ funti gussuzu tSi tserriu zEmprE they.were those that I.call always ‘They were the ones that I always call’ srrE Dua ka EstE ammia mia sister yours that is friend mine ‘Your sister, who is a friend of mine …’ tSi llEzE BEdï†ı that you.want I.come ‘If you want, I will come’
‘If’â•… 67 e. iiâ•… f.
tSi rEDE n bbEss thatâ•… it.rains not I.go.out ‘If it rains, I won’t go out’ n iï†fiï†fiiu tSi u tserriu not I.know that him I.call ‘I don’t know if I shall call him’
Another major pattern of lexicalization of ‘that’ and ‘if’ is exemplified in (20) by Miglionico. This has a single declarative complementizer and a separate complementizer for hypotheticals and questions€– reproducing the conditions of a language like Italian. However, the hypothetical/question complementizer actually belongs to the wh–system, in fact overlapping with the wh–element for ‘who’, namely tʃi. Because of this, the embedded sentence in (20e) is ambiguous between the wh–question reading ‘who’ and the yes–no question reading ‘if’. (20)â•… Miglionico (Lucania) a. m vnn ditt ka vin kra to.me they.have said that he.comes tomorrow ‘They told me that he comes tomorrow’ b. s assut prim ka tu vniss I.am gone.out before that you came ‘I went out before you came’ c. i. tSi ve:n whoâ•… comes ‘Who is coming?’ c.â•… ii.â•… tSe ffa ï†fin what they.do ‘What are they doing?’ d. tSiâ•… vvuo ve if you.wantâ•… I.come ‘If you want, I come’ e. nanâ•… sattS tSi ven not I.know who/if he.comes ‘I don’t know who comes/if he comes’
In short, in the Làconi examples in (19) there is a single lexicalization for the indefinite declarative complementizer and for the question/hypothetical complementizer. Table (B) of the Appendix summarizes varieties for which this distribution holds; all of them are Sardinian, of a type in which there is no overlap between the complementizer and the wh–system. Crucially, from the table in (B) it can be seen that the hypothetical/interrogative complementizer never has the same form as the ka definite complementizer. In turn, in a language like Miglionico in (20) the lexicalization for the hypothetical/interrogative
68â•… Variation in Romance k-complementizer systems complementizer overlaps with that of a wh–element, while the (single) declarative complementizer has a different lexicalization. The data summarized in table (C) of the Appendix show that it is always the hypothetical complementizer that coincides with the wh–system (as opposed to the declarative one). The complementizer system of languages like (19)–(20) closes a potential gap in the discussion so far. As discussed in chapter 1, we know that wh–items are not intrinsically interrogative; in Italian, for instance, argument che can head a free relative as in (21a) or an exclamative as in (21b). We interpreted these data as showing that question, declarative or exclamative values are not intrinsic to the wh–element, but rather contributed by contextual operators. This is far from an isolated case in natural languages. For instance, so-called N-words in Romance are not intrinsically negative, but rather introduce a variable which can receive a negative interpretation in the scope of a negative operator€– and can equally well be licensed in non-negative contexts (cf. chapters 3–4). (21)â•… a.â•… b.
Faiâ•… che ti pare do whatâ•… youâ•… please ‘Do as you please’ Che non farebbe! what not he.would.do ‘What wouldn’t he do!’
On these grounds we expect that a complementizer belonging to the wh–series will not require a question interpretation. What does represent a potential problem for the present theory, however, is that none of the complementizers considered in section 2.1 allows for such an interpretation. Nothing in what we have said so far blocks the possibility that a k–complementizer could be interpreted as interrogative in the scope of an appropriate question operator, in a similar way to specialized interrogative wh–complementizers like English whether. From this perspective, systems like (19)– (20) are interesting in the first instance because they instantiate precisely the possibility we predict. Let us consider, then, what the characterization of the hypothetical/interrogative complementizer may be, beginning with a language like Italian (or indeed English) in which it has a specialized lexicalization. Hypothetical sentences are interpreted much in the same way proposed previously for complement sentences, that is, as the restriction of a propositional variable. According to Lewis’s (1975) classical treatment, the latter is bound by an adverb of quantification, or by a generic quantifier in the absence of other overt quantifications. Thus the sentence in (18a) has an LF of the type in (22a), if the adverb of
‘If’â•… 69 quantification is computed. Otherwise the propositional variable is closed by a universal quantification, as in (22b). (22)â•… a.â•… For many situations/possible worlds x:€x it rains, I go out (in x) b. For all situations/possible worlds x:€x it rains, I go out (in x)
An element which introduces a propositional variable (restricted by the embedded propositional content) is associated in the present grammar with structures of the type argued for in chapter 1 for k-type complementizers€ – which we will then extend to Italian se ‘if’.10 The latter can in turn be interpreted not only in a hypothetical context as in (15), but also in an interrogative context€– exactly like English ‘if’. The embedded interrogative interpretation simply follows if the variable introduced by se and the like (and restricted by the proposition following it) can be licensed in the scope of a question operator. We provisionally assume that this is what the interpretation of an interrogative complementizer amounts to, as in (23).11 (23)â•… I wonder for which situations/possible worlds x:€x John comes
At this point the characterization of the se complementizer is fundamentally the same as for the che complementizer, and more specifically for the indefinite complementizer in (6b), where the variable that the complementizer introduces is closed quantificationally. Therefore we fully expect that ‘if’ and the indefinite declarative complementizer can be lexicalized by the same item, yielding the Làconi system in (19). In view of the proposals reviewed in section 2.1.2 to the effect that complement sentences are relatives, it is particularly worth remarking on the independent literature treating conditionals as relatives€– specifically as free relatives. Thus, for Bhatt and Pancheva (2006), ‘turning to conditionals, our proposal that they are interpreted as free relatives amounts to the claim that they are definite descriptions of possible worlds’. As for the syntax, ‘the null operator in Spec, CP of if-clauses and likely the when itself in e.g. German conditionals, is a definite binder of the possible world variable’. As Bhatt and Pancheva note, the proposal that there is a covert operator in the Spec, CP of conditional if clauses goes back to Larson (1985), where the focus of the discussion is interrogative if clauses. For the latter, Larson posits the presence of a covert whether, which he extends to conditional if clauses. For Bhatt and Pancheva, ‘the fact that if functions in many languages as both a conditional and an interrogative complementizer makes sense within the general proposal that conditionals are free relative clauses. In English and in many other languages, this syncretism would be part of a more general structural parallelism between questions and free
70â•… Variation in Romance k-complementizer systems relatives’. However, if ‘free relatives are interpreted as definite descriptions, i.e. with the variable abstracted over being bound by a definite operator’ (Bhatt and Pancheva 2006), the lexicalization patterns observed in Làconi, in which ‘if’ overlaps with the indefinite complementizer, and not with the definite one, seem to be difficult to account for. More generally, the present account does not require us to postulate zero world operators in the syntax. Quite simply, the so-called complementizer is itself the element that introduces a (situation/ possible world) variable in the syntax of LF.12 If anything, our problem is how to provide a sharper characterization for languages like Italian that have two distinct complementizers, or languages like Paulilàtino in (5) that end up having three. Let us begin by considering a Â�language with three distinct forms, including the definite and indefinite declarative complementizers, as well as the hypothetical/interrogative complementizer, as exemplified for Paulilàtino in (5) and (24). Following the discussion in section 2.1, the definite complementizer introduces a propositional complement akin to a definite description, while the indefinite complementizer introduces an indefinite variable. The latter is also true of the proposition introduced by ‘if’, which therefore needs to be further differentiated. (24)â•…
Paulilàtino si ‘prEDEâ•… n Ess ifâ•… it.rains notâ•… I.go.out ‘If it rains, I won’t go out’
In introducing the contexts triggering subjunctive in Italian in section 2.1.1, we mentioned that they are not necessarily lexically selected; rather, subjunctive can be triggered in the scope of the question operator in (8c). In the scope of the question operator, the declarative complementizer, ordinarily embedded by a verb like to know, alternates with the ‘if’ complementizer, as in (25)€– a phenomenon discussed in the literature under the name of ‘unselected questions’ (Adger and Quer 2001). Needless to say, the reading of the embedded sentence in (25) is always non-factive, i.e. within the scope of the matrix question operator, independently of whether the indicative or the subjunctive is selected. As shown in (26), a matrix negation has very much the same effect as a question operator, both with respect to the triggering of the subjunctive in (26a) and to the possibility of the ‘if’ complementizer in (26b). In (26) only the co-occurrence of the che complementizer with the embedded indicative in (26a) yields a presupposed (factive) reading of the embedded sentence; both the subjunctive and the ‘if’ complementizer force the sentence to be read within the scope of the negation operator.
‘If’â•… 71 (25)â•… Sai se è/ siaâ•… venuto? you.knowâ•… ifâ•… he.is/â•… be come ‘Do you know if he has arrived?’ (26) a.â•… Nonâ•… sa che sono/siaâ•… guarito not he.knowsâ•… thatâ•… I.am healed ‘He doesn’t know that I am healthy again’ b. Nonâ•… sa se sono/siaâ•… guarito not he.knowsâ•… that/ifâ•… he.be healed ‘I don’t know if he is healthy again’
The fact that ‘if’ is triggered in the scope of a matrix question or negative operator with an otherwise declarative (factive) verb, as in (25)–(26), points the way to a solution for the problem we are concerned with, i.e. the characterization of the contexts specialized for ‘if’. Negations and questions are two of the fundamental contexts triggering (negative) polarity items€– the third one being the hypothetical one. Thus, for instance, in Italian nessuno ‘anybody’, niente ‘anything’ etc. are licensed in the scope of negation (yielding the equivalent of the negative quantifiers nobody, nothing etc. in English), as well as in the scope of a question or of a hypothetical. We suggest that the complementizer system of languages like Italian or Paulilàtino in (24) (or, for that matter, English) is sensitive to polarity, so that complementizers, i.e. nominal heads responsible for introducing propositional variables, are lexicalized by two different items in non-polarity and polarity contexts, surfacing in the latter as se in Italian, if in English etc. Correspondingly, we shall refer to these as the polarity complementizers. In section 2.1 we concluded that it is to be expected that the complementizer system, i.e. the system of propositional variable introducers, and the wh–system may coincide. More to the point, in cases of split complementizer systems the coincidence is predicted to involve the indefinite complementizer, i.e. the one introducing an indefinite variable, rather than the definite complementizer, which introduces a (propositional) definite description. This mode of reasoning can now be extended to the ‘if’ complementizer. In languages of the type of Làconi in (19) there is no overlap between the complementizer and wh–systems. However, languages like Miglionico in (20) verify this prediction for the systems with a polarity complementizer ‘if’ and a nonpolarity complementizer ‘that’. For in Miglionico, as in the other varieties tabulated in (C) of the Appendix, it is the ‘if’ complementizer that belongs to the wh–system€– and not the (single) declarative complementizer. A further parameter brought to the fore by varieties like Miglionico has to do with which element of the wh–system the complementizer identifies with. In
72â•… Variation in Romance k-complementizer systems Italian, as reviewed in chapter 1, the overlap is between the declarative complementizer and the wh–element for ‘what’. By contrast, in Miglionico in (20) the overlap is between the ‘if’ complementizer and the wh–item for ‘who’. This parameter is independent of the others considered. Thus, in Ruvo di Puglia in (27) the ‘if’ complementizer has the same form as ‘what’, and not ‘who’ as in Miglionico. By contrast, in a variety like Secinaro, otherwise entirely comparable to Guglionesi in (1), the modal complementizer has the same form as ‘who’, and not ‘what’, as in (28). (27)â•… Ruvo di Puglia (Apulia) a.â•… tS cɔ:v nn iss that it.rainsâ•… not I.go.out ‘If it rains I won’t go out’ a’. nn tsattï†fi tï†fi fratt m cɔm not I.know if brother-yours me calls ‘I don’t know if your brother is calling me’ b tSi vEn whoâ•… comes ‘Who comes?’ c. tSe ffɔS whatâ•… you.do ‘What are you doing?’ (28) Secinaro (Abruzzi) a. vujj k vvi I.want that you.come ‘I want you to come’ b. k vvE who comes ‘Who comes?’ c. ku ffE what you.do ‘What are you doing?’
In terms of the discussion in chapter 1, the wh–quantifier for ‘what’, such as Italian che, is characterized by the absence of any lexical restriction. Thus che questions can be answered by an animate or inanimate noun phrase or by a proposition, as in (29). Of course a lexical restriction can be added by a noun phrase, when che ‘what’ appears as its specifier€– or by a sentence, when che appears as the ‘that’ complementizer. (29)â•… a.â•… Che c’ è? what there is ‘What is it?’
Interaction with (non-)finitenessâ•… 73 b. Il campanello ‘the doorbell’ b’. Il postino ‘the postman’ b’’. Suona il postino rings the postman ‘The postman is ringing’
By contrast, the wh–quantifier for ‘who’ is characterized by a restriction to humans, which poses an interesting question as to its ability to appear as a complementizer at all. In fact, the occurrence of, say, Miglionico tʃi in (20) as a complementizer excludes the possibility that the restriction to humans is encoded in the lexicon. What we suggest instead is that it is a lexical property of tʃi that it requires a restriction. In complementizer contexts the latter is provided by the proposition that tʃi introduces. In wh–contexts, on the other hand, the restriction to humans may represent an interpretive closure in the absence of lexical restrictions. If this line of reasoning is correct, we should be able to find languages in which the wh–quantifier for ‘who’ also appears as the wh–determiner, restricted by a following noun phrase. A case in point is Fontanigorda in (30), which, like Italian, has a single declarative complementizer overlapping with a wh–quantifier€– except that the overlap involves ‘who’, rather than ‘what’. As expected, the same element also appears as the wh–determiner, as shown in (30d). (30)â•… Fontanigorda (Liguria) a.â•… m aŋ ittu ke te vie dɔpu to.meâ•… they.haveâ•… saidâ•… thatâ•… youâ•… comeâ•… afterwards ‘They told me that you are coming afterwards’ b. ke te tSammi whoâ•… youâ•… call ‘Who do you call?’ c. kuɔseâ•… te fE what youâ•… do ‘What are you doing?’ d. ke kamiï†ıaâ•… te te bEtti which shirt youâ•… yourselfâ•… put.on ‘Which shirt are you putting on?’
2.3
The interaction with (non-)finiteness
The last question we shall consider in this chapter concerns yet another asymmetry between wh–quantifiers and complementizers, observed even in
74â•… Variation in Romance k-complementizer systems languages in which they have the same lexicalization. Thus, in standard Italian che ‘what’ introduces both finite and infinitival questions, as in (31a). In contrast, complementizer che is restricted to finite contexts, and infinitival embeddings involve, if anything, prepositional introducers like di ‘of’ in (31b). (31)â•… a.â•… b.
Che faccio/â•… fare? whatâ•… I.do/ to.do ‘What should one/I do?’ Mi hanno dettoâ•… di/*cheâ•… essereâ•… venuti to.meâ•… they.haveâ•… said to/that be come ‘They told me that they had come’
The problem is not a descriptive one, since complementizer che can be distinguished from wh–che on the basis of both the syntactic context of insertions and its semantic characterization. From an explanatory point of view, however, the objection may legitimately be raised that the connection between complementizers and finiteness crucially links them to the functional projections of the verb. Now, finiteness has two components€– one relating to tense, and another relating to agreement. As is well known, some Romance languages allow these two components to be distinguished, since they feature so-called inflected infinitives. A case in point is provided by Sardinian varieties, including Paulilàtino in (32). Both inflected and non-inflected infinitives can be introduced by a prepositional complementizer, as in (32b)€ – but inflected infinitives also allow for the finite complementizer, as independently observed by Jones (1993). The finite complementizer, then, is sensitive not to the temporal/modal/aspectual properties of the verb, but to the presence of an agreement inflection. As we have seen, Paulilàtino, like many Sardinian varieties, has two declarative complementizers, a modal and a non-modal one€– and it is the modal complementizer that co-Â�occurs with inflected infinitives. However, Manzini and Savoia (2005) also provide examples from varieties which have only one declarative complementizer, which also appears in front of the inflected infinitive (for instance, Siniscola). (32)â•… Paulilàtino a.â•… l appu attu innantis dE/ ki TrrarE- s tuE it I.have done before to/ that come.back-2sg you ‘I did it before you came back’ b. l an fattu innantis dE EnnErE-â•… (nE) /ki nnErE-nE i it they. have done before to come- 3pl that come-3pl they ‘They did it before they came’
In our theory (cf. in particular chapter 5) the so-called agreement inflection is a lexicalization of the EPP argument of the sentence internal to the verb. If
Interaction with (non-)finitenessâ•… 75 no lexicalization of the EPP argument is present, either by an inflection or by a lexical subject, as is the case for non-inflected infinitives, we assume that the EPP argument is introduced as a variable at the interpretive interface. This variable can then be given a bound reading (control/raising) or a generic reading (arbitrary control). In such terms, the restriction of the che-type complementizer to finite sentences€– including inflected infinitivals€– can be restated as an incompatibility between it and the EPP variable. The question then becomes why the finite complementizers would have this crucial property. We may surmise that the presence of an EPP variable within the sentence defines an open predicate, rather than a proposition. Since we have proposed that the finite complementizer of Romance, and presumably that of English as well, takes a proposition as its complement, we may conclude that what the complementizer is actually incompatible with is the open expression resulting from the presence of the EPP variable. Similarly, for Roussou (2010), the fact that in English that requires a propositional complement ultimately yields the that-t filter, precisely because it excludes the possibility that the EPP argument is a variable. In Roussou’s terms, the absence both of that and of the EPP gives rise to a case of predicate embedding€ – which allows the base-generated wh–phrase in the matrix left periphery to bind a variable in the argument structure of the embedded predicate. In fact, Roussou (2010) analyses both the *for-t and the *for–to filters as results of the same restriction, to the effect that for, like that, embeds a proposition, barring EPP variables. Though a discussion of the that-t filter is beyond the scope of the present work, it is worth recalling that, according to the classical conclusion of Taraldsen (1978), languages like Italian lack that-t filter effects (whence the apparent lack of correlation with the finiteness constraint on the complementizer) in that they are null-subject languages. We propose that what is at stake is simply the fact that while in non-null-subject languages the EPP must be satisfied by a syntactic-level argument (the subject), in null-subject languages a morphological-level argument (the inflection) suffices. The latter therefore closes off the proposition even in the case of wh–extraction (effectively acting as a resumptive pronoun). Similarly, non-null-subject Romance languages, in contrast to English, generally have subject clitics and resolve the propositionality requirement on chetype complementizers by lexicalizing one of them. On the basis of a comparison with Romansh varieties, Taraldsen (2002) and Manzini and Savoia (2005) in particular conclude that the€–i of French qui is a subject clitic of this type. A further twist on the finiteness problem is introduced by the ‘if’ complementizer. In a language like standard Italian, hypothetical se ‘if’ is restricted to
76â•… Variation in Romance k-complementizer systems finite sentences, as in (33a), while interrogative se normally occurs in front of infinitival questions, as in (33b). Now, recall that in Làconi in (19) the same tʃi form lexicalizes both the modal declarative complementizer and the polarity (i.e. hypothetical/question) one. As it turns out, while declarative tʃi is restricted to finite sentences, interrogative tʃi combines with infinitivals, as in (34). In the discussion surrounding (23), we interpreted the interrogative complementizer as a wh–element ranging over propositions in the scope of a question operator. We must conclude that this interpretation makes it compatible with embedded EPP variables. In this respect, ‘if’ in the wh–construal behaves exactly like other wh–elements, including the English wh–complementizer whether. (33)â•… a.â•… b. (34)
*Seâ•… piovere, … if to.rain Non so se uscire not I.know if to.go.out ‘I don’t know whether to go out’
Làconi n iSSiu tSi u tsErriai not I.know if him to.call ‘I don’t know whether to call him’
The issue is further complicated by the fact that while Italian se or Làconi tʃi can introduce infinitival questions, as in (33)–(34), neither English if nor French si can do so. In present terms, this means that French si and English if must embed complete propositions, very much like che or that, even if they are construed in the scope of an interrogative operator. The present analysis can be usefully compared with Kayne’s (1991). For Kayne, the finiteness constraint on complementizers reflects their incompatibility with the PRO subject of infinitivals; in other words, he substantially agrees with the view advocated here that the finiteness requirement relates to the EPP rather than to the temporal properties of the sentence. However, for him, what is responsible for this constraint is the structural government relation holding between the complementizer and PRO. The lack of such a constraint on Italian se correlates with a further structural property, i.e. with the relatively high position of the infinitive, which protects PRO from government by the complementizer. That the position of the infinitive is higher in Italian than in French is shown, according to Kayne, by the fact that the clitic precedes the infinitive in French, while it follows the infinitive in Italian. However, he acknowledges that his theory ends up not accounting for the incompatibility of Italian che with the infinitive, since on the evidence of cliticization the infinitive must be in the same high position in declaratives as in questions (1991:€95). In Làconi
Interaction with (non-)finitenessâ•… 77 the problem is even starker, since the same tʃi complementizer cannot combine with infinitival declaratives€– as predicted by Kayne on the basis of the preverbal position of the clitic seen in (19)€– but can combine with infinitival questions. Furthermore, in the account of Kayne (1991), whether is oblivious to the finiteness of the sentence it embeds because it is not a complementizer at all, but a wh–phrase€– where these two notions are given a structural characterization once more, as a C head and a C Specifier respectively. However, a structural characterization as a C Specifier cannot be applied to Làconi tʃi€– on pain of not being able to predict that (as a C head) it excludes non-finite sentences in declarative contexts.13 A different type of evidence concerning definiteness restrictions on Â�complementizers of the k-series comes from Northern Italian vÂ�arieties Â�(including some that display the standard Italian identity of form between the wh–quantifier for ‘what’ and the complementizer), which form questions by combining whâ•fi quantifier and complementizer, as discussed in detail in chapter 1 (especially section 1.2). Most of the relevant languages alternate between wh– che in finite sentences and the wh–element alone in infinitivals, as shown in (35) for Zoldo Alto. However, there are (a few) otherwise entirely comparable languages in which the wh– che cluster introduces both finite and infinitival questions, like Civate in (36)€– though the che-type complementizer is otherwise restricted to finite complements. (35)â•… Zoldo Alto (Veneto) a.â•… di-me ke ke te faTe tell-meâ•… whatâ•… thatâ•… youâ•… do ‘Tell me what you are doing’ b. no sai ke fa notâ•… I.knowâ•… whatâ•… to.do ‘I don’t know what to do’ (36) Civate (Lombardy) a. tSe ke te tSamet whoâ•… thatâ•… youâ•… call ‘Who are you calling?’ b. so mia tSe ke tSa’ma I.know not who that to.call ‘I don’t know who to call’
The pattern of Zoldo Alto is what we expect on the basis of the assumption that in the wh–che sequence, the che-type element is the ordinary complementizer; as such we predict it to be sensitive to the finiteness of the embedded clause and in particular to exclude an infinitival sentence, whose EPP
78â•… Variation in Romance k-complementizer systems argument is a variable under present assumptions. Let us consider, then, the case of Civate. Evidently the presence of the wh–quantifier in the left periphery of the ke complementizer is sufficient for the latter to also be interpreted in the scope of a question operator€– that is, as a wh–element of sorts ranging over propositions. If so, we expect it to have the property that we have independently reviewed for such elements above, that is, that of being insensitive to the finiteness restriction. Appendix (A)
Comp def
Comp indef/wh–
rel
who
what
Sonnino Pontecorvo Colledimacine Montenerodomo Torricella Peligna Secinaro Civitaluparella Vastogirardi Capracotta Guardiaregia Guglionesi Frigento Nocara Albidona Terranova Pollino Morano Orsomarso Conflenti Platania Gizzeria Sorbo S.Basile S.Pietro a Maida Iacurso Arena Umbriatico Gerace S.Agata del Bianco
ka ka ka ka ka ka ka ka ka ka ka ka ka ka ka ka ka ka ka ka ka ka ka ka ka ka ka
ke k k k k k k k k k k ke k k k ki ki ki ki ki ki ki ki ki ki ki ki
ke k k k k k k k k k/ka k ki k k k ki ki ki ki ki/ka ki ki ki ki ki ki ki
ki ki ki ki k k ki kia kia ki ki ki k kwE ku ku ku kinE kinE kina kinE ku ku ku kini ku ku
ke k k k k ku k k k k k ke k k k ki ki ki ki ki ki ki ki ki ki ki ki
(A’) Dorgali Aritzo Ardaùli Paulilàtino Gallo
ka ka ka ka ka
ki tSi ki ki ku
ki tSi ki€– ka ki ku
kiE (e)tSinE kie kiE ki
ittE ittE (e)itte ittE kwe
Interaction with (non-)finitenessâ•… 79 Tufillo
ka
k
k
kia
Canosa Sannita
ka
k
k
ki
(k) dEkk ki
(B) Làconi Orroli Settimo S.Pietro Sìliqua
comp ka ka ka ka
comp/ if tSi ki ki ki
rel tSi€– ka ki€– ka ki ki
who tSinni (e)kini kini (ak)kini
what itta itta itta itta
(C) Miglionico Minervino Murge Gravina in Puglia Bitetto Ruvo di Puglia Canosa Bisceglie Martina Franca Molfetta Mesagne Putignano Brindisi Grottaglie Carmiano Uggiano Copertino Melissano
comp ka ka ka ka ka ka kE ka ka ka ka ka ka ka ka ka ka
if tSi tS tSi tS tS tS tS tS tSi tSi tS tSi tSi Si tSi tSi tSi
rel ka ka ka ka ka ka kE ka ka ka ka ka ka ka ka ka ka
who tSi tS tSi tS tSi tS tS tS tSi tSi tS tSi tSi Si tSi tSi tSi
what tSe tSk attSok tS tS tSE tS tS tSE tSe tSE tSE tSe tSE tSE tSe tSi
3 Sentential negation:€adverbs
The currently standard approach to the structure of negation (in Romance) was inaugurated by Pollock (1989), who proposed that negation adverbs such as pas in French fill the Spec position of a NegP projection generated below the I position targeted by the verb. The head of NegP can in turn be filled by a negative clitic like ne in French, whose higher inflectional position depends on movement, i.e. cliticization. In other languages, which include colloquial French, no negative head is present. Belletti (1990) applies the same theory to Italian, which only has a negative head; the latter originates in the Neg position and moves higher as a result of cliticization. A more complex set of data, involving Northern Italian varieties, is considered by Zanuttini (1997), who proposes that there are several Neg positions. Specifically, a Neg position is generated above I, while below I there are three Neg positions. The inflectional Neg position hosts negative clitics in languages like Italian which do not require a sentential negation adverb. In contrast, languages which require a sentential negation adverb generate it in one of the lower Neg positions; if a clitic combines with the adverb, it is generated in the head of the relevant Neg position and moves to the inflectional domain via cliticization. In other words, for Zanuttini (1997) preverbal clitic negations are associated with two different structures, according to whether they negate alone or combine with a negative adverb. In turn, the lower Neg positions are defined in relation to the general hierarchy of adverbs proposed by Cinque (1999); according to Zanuttini (1997), her Neg2, Neg3 and Neg4 positions occur within the aspectual adverbial series, while they do not interact in any significant way with either the temporal or the modal series. Neg1 corresponds to the negative clitic position in the inflectional domain. According to Cinque (1999:€106), the aspectual adverbial series is ordered according to the hierarchy partially reproduced in (1a). The three adverbial negation positions proposed by Zanuttini (1997:€99) are ordered with respect to this hierarchy as in (1b). The three different Neg positions correspond to three different types of sentential negation adverbs, which Zanuttini individuates in 80
Sentential negation:€adverbsâ•… 81 Northern Italian varieties on the basis of their distribution. The higher adverbial Neg position Neg2 corresponds to sentential negation adverbs such as pa in Piedmontese varieties or the optional mica of standard Italian. In the same Piedmontese varieties in which pa occurs as Neg2, the Neg3 position is filled by sentential negation adverbs of the nen type. In turn, the lowest adverbial negation position Neg4 corresponds to the Lombard type no. (1)â•… a.â•… [Tanterior already [AspTerminative no longer [AspContinuative still [AspPerfective always…[Voice well b. [Neg2 [already [Neg3 [no longer [still [always … [well [Neg4
In closely related literature, the functional category Neg also plays a role in explaining interpretive facts, specifically ‘negative concord’, whereby two negative elements in the sentence do not give rise to a double negation, but rather are interpreted as instantiations of a single logical negation. Consider, for instance, a Romance language like Italian which has a negative clitic; the latter combines with a negative argument or adverb (niente ‘nothing’, mai ‘never’, etc.) to yield a single logical negation. Haegeman and Zanuttini (1991) explain this phenomenon on the basis of a Neg Criterion, modelled on the Wh–Criterion of Rizzi (1996). On the basis of the Neg Criterion, the negative clitic in the head position of a NegP requires a Neg operator in its Spec and vice versa. The head-Spec configuration is in turn read as an agreement relation, yielding a single negation interpretation. This analysis presupposes that niente, mai and the like€– i.e. n-words€– are negative quantifiers. A consistent body of literature on Romance languages argues for a different conclusion, namely that Romance n-words are negative polarity items (Rizzi 1982; Laka 1990; Longobardi 1992; Acquaviva 1994). In other words, n-words have no intrinsically negative properties, but are simply existentials, or free variables in the terms of Heim (1982), which are interpreted in the scope of the negation or another polarity operator. The most immediate argument in favour of the polarity status of n-words in Romance comes from the fact that they occur in modal environments (questions, hypotheticals) without any implication of negative meaning. The argument that is often advanced in favour of a negative quantifier status for n-words, namely that they appear in fragments (Zanuttini 1997), depends on the fact that fragments are the result of ellipsis€– and that they interpretively correspond to a full sentential structure, obviously capable of hosting an abstract negative operator. In what follows we assume that n-words are negative polarity items (and not negative quantifiers). This entitles us to (provisionally) disregard the role played by NegP in negative concord. We shall return to negative concord in chapter 4, where we shall also consider the sentential negation clitics which enter into it.
82â•… Sentential negation:€adverbs In this chapter we focus on sentential negation adverbs. Specifically, in Â� section 3.1 we note that sentential negation adverbs either take the same form as negative arguments such as ‘nothing’ or are bare nouns. We argue that treating them as nominal elements, specifically connected to the internal argument of the predicate, allows for an explanation of data such as the fact that they trigger the partitive, or their sensitivity to the person (1st/2nd vs. 3rd) of the internal argument itself. Correspondingly, we abandon the Neg categorization in favour of a reassignment of sentential negation adverbs to a nominal, argument-related category, tentatively labelled Q.1 In section 3.2 we examine the distributional phenomena targeted by the hierarchy in (1) – keeping in mind also the fact that recent literature (Ernst 2002; Svenonius 2002; Nilsen 2003) argues that they are sensitive not to syntactic hierarchies but to semantic constraints on selection and/or scope. In particular, we reassess the sub-hierarchy in (1) in the light of our treatment of sentential negation adverbs as both nominal and argumental. We conclude that the adverbs of the sub-hierarchy in (1) are attached to the eventive domain of the sentence in accordance with LF selection requirements and that they are ordered within this domain according to the same macrocategories encountered in the argumental and inflectional domains€– i.e. roughly by quantificational (Q), deictic (Loc) and nominal/argumental (N) properties. In section 3.3 we briefly reconsider another major distributional pattern concerning the position of adverbs with respect to verbs. Cinque (1999) and Zanuttini (1997) predict that any verbal head appearing internally to the hierarchy of adverbs in (1) merely cuts it into two segments, without any reordering of the adverbs relative to one another. We argue that if the ability of non-finite verbs to appear in the middle of the sequence in (1) is construed as an indication of their low position, then the fact that they notably precede pronominal clitics (i.e. inflectional material) remains unexplained. We conclude that the non-finite verbs are quite high in the sentence (participles also define independent sentences as in Kayne (1993)). The adverbs in (1) attach even higher because the lack of tense specifications allows for an eventive treatment of the higher projections of the sentence (along the lines of the flexible matching of syntactic and interpretive projections proposed by Svenonius (2002) and Ernst (2002)). Overall, then, we devote this chapter to providing a recategorization of sentential negation adverbs; the alternative account of their distributional properties is meant to establish the compatibility of this recategorization with known facts. As anticipated, we leave the recategorization of negation clitics and the reappraisal of negative concord for chapter 4.
Sentential negation adverbs are nominalâ•… 83 3.1
Sentential negation adverbs are nominal and argumental
In several Northern Italian varieties, notably Piedmontese ones of the type also studied by Zanuttini (1997), the sentential negation adverb has the same form as the negative argument for ‘nothing’. This is true of nEŋ as in (2), nEint(a) as in (3), nota as in (4), and rEŋ as in (5), as well as nia in Ladin varieties, as in (6). As a result of the variable valency of the verb, the (b) examples are ambiguous between the argumental and sentential negation interpretations; specifically, since the verb can be construed both transitively and intransitively, the negative item can fill the internal argument slot, or it can have a reading equivalent to a sentential negation. Conversely, it is only the valency of the verb that disambiguates the sentential negation readings in (a) and the argumental readings in (b). (2) Montaldo (Piedmont) a.â•… i u tSam nEŋ Iâ•… himâ•… callâ•… not ‘I don’t call him’ b. i mdZâ•… nEŋ I eat nothing/not ‘I don’t eat (anything)’ (3) Oviglio (Piedmont) a. a n El vig nEinta I notâ•… himâ•… seeâ•… not ‘I don’t see him’ b. u n maJdZaâ•… nEinta heâ•… notâ•… eats nothing/not ‘He doesn’t eat (anything)’ (4) Quarna Sotto (Piedmont) a. tSamuâ•… not -t I.call notâ•… you ‘I am not calling you’ b. j vOg nota I see no/nothing ‘I don’t see (anything)’ (5) Stroppo/Macra (Piedmont) a. lu tSamuâ•… rEŋ himâ•… I.call not ‘I don’t call him’ b. al fai rEŋ he does nothing ‘He does nothing’
84â•… Sentential negation:€adverbs (6)â•… La Pli de Mareo (Alto Adige/South Tyrol) a.â•… i ne le kErdiâ•… nia Iâ•… notâ•… himâ•… call not ‘I don’t call him’ b. ju ne li nia I notâ•… readâ•… not/nothing ‘I don’t read (anything)’
The common lexicalization of the sentential negation adverb and negative arguments is recognized in the historical literature, specifically in connection with what is taken to be a ‘grammaticalization’ process changing the negative argument into a functional category Neg. Thus, Roberts and Roussou (2003) mention ‘negative quantifiers/pronouns’ as one of the ‘principal sources for clausal negators’ (English not is an example of this). In fact, there are many Northern Italian varieties in which the types nen, neinta, etc. for the sentential negation do not have the same form as ‘nothing’, so that the connection between the former and negative arguments appears to be purely etymological. However, varieties of the type in (2)–(6) seem to point to a bona fide lexical identity of negation and negative argument. Therefore we may wonder whether the best theory is one which posits two lexical entries for them, with the negation associated with the category Neg and the argument with the category N/Q. In fact, only a subset of the negative elements reviewed so far contains the negative morphology n-; thus, the type rEŋ is etymologically connected to a bare noun (
Trun (Grisons) Elts dOrmŋâ•… bk(a) theyâ•… sleep not ‘They don’t sleep’ Pramolloâ•… (Piedmont) a drm pa he sleeps not ‘He doesn’t sleep’ Premanaâ•… (Lombardy) elâ•… fo: miE it I.do not ‘I don’t do it’ Finale Emiliaâ•… (Emilia) i n dramâ•… briʑa theyâ•… notâ•… sleep not ‘They don’t sleep’
Sentential negation adverbs are nominalâ•… 85 Interestingly, in relation to sentential negations of the type in (7) MeyerLübke (1899:€§693–4) proposes that what we describe as sentential negation adverbs originate in a partitive construction. In support of his proposal he quotes Old French examples such as (8), in which the ‘negative adverb’ mie, a bare noun ‘minimizer’, overtly co-occurs with the partitive. (8)â•… de s’espee ne volt mieâ•… guerpir of his swordâ•… notâ•… he.wanted not to abandon ‘He didn’t want to abandon his sword’
(Chanson de Roland 465)
This type of data recalls the phenomenon described by Pesetsky (1982) for Russian whereby the accusative object in non-negative contexts alternates in negative contexts with an object morphologized either in the accusative or the partitive, as in (9). (9)â•… a.â•… b.
Ja ne polucal pis’ma I not received letters(acc.pl.) ‘I didn’t receive letters’ Ja ne polucal pisem I not received letters(gen.pl.) ‘I didn’t receive letters’
Considerable interaction between the negation and case assignment to the internal argument of the verb is also found in Northern Italian varieties. Given its potential relevance for the status of the negation (purely functional or still nominal), we shall review it next. In one group of languages, which includes the Piedmontese varieties in (10), the negation triggers the partitive even in the presence of a definite interpretation. Thus, the partitive clitic in (10) doubles the proper name in topic position, and indeed can normally alternate with an accusative, as in (10a). It will be noted that the languages exemplified are characterized by enclitics in declarative finite sentences (cf. Tortora 2002). However, the variety from Quarna Sopra exemplified in (15) combines the same phenomenon with proclisis on the finite declarative verb, showing that it cannot depend on enclisis. (10)â•… a.â•… b.
Trecate (Piedmont) (a mmarju) tSamumâ•… -ru/â•… -na mija the Mario we.call him/of.himâ•… not ‘We are not calling Mario’ Cerano (Piedmont) Marjoâ•… tSamaâ•… -n mea Mario call of.him not ‘Don’t call Mario!’
The data in (10) shed a new light on a different, but fairly obviously connected phenomenon that has long been known for French. Thus, as discussed
86â•… Sentential negation:€adverbs by Kayne (1984), in negative environments French allows for indefinite noun phrases (bare plurals and bare mass singulars) introduced by de; these same forms are not allowed in the absence of negation, as illustrated in (11). The same phenomenon is fairly widespread in Northern Italian varieties, as shown in (12). (11)â•… Jeâ•… *(ne)â•… veuxâ•… pas de cadeaux I not want notâ•… ofâ•… gifts ‘I (don’t) want gifts’ (12) a. b. c. d.
Casorezzo (Lombardy) al maJdʒaâ•… miŋgaâ•… da biskOti heâ•… eats not of biscuits ‘He doesn’t eat biscuits’ Casei Gerola (Lombardy) u bev mi d vi he drinks not of wine ‘He doesn’t drink wine’ Cravagliana (Piedmont) al maJɟa mia da buskuti heâ•… eats not of biscuits ‘He doesn’t eat biscuits’ Stroppo/ Macra al beu rEŋ de viŋ heâ•… drinksâ•… notâ•… of wine ‘He doesn’t drink wine’
For the object noun phrase in (11), Kayne (1984) proposes a structure including a non-lexicalized negative quantifier Q followed by the partitive de cadeaux, yielding a structure of the type je ne veux pas [ Q [de cadeaux]]. The presence of the empty Q quantifier means that the noun phrase as a whole is subject to the Empty Category Principle (ECP) of Chomsky (1981); this in turn predicts a generalization that appears to hold at least of French, namely that noun phrases of the type under consideration are restricted to the object position. This proposal of Kayne (1984) can be considered a predecessor of his recent ideas about ‘silent’ categories (Kayne 2006, 2008a). Manzini and Savoia (2009a, 2010) and Savoia and Manzini (2010) argue in some detail against this conception on grounds of restrictiveness of the theory, as well as on specific empirical grounds. Here we note only that Chomsky’s (1981) ECP is formulated for empty categories whose content is determined in that framework by the computation, not by the lexicon€– so that the extension of the ECP to empty Qs of the type in Kayne (1984) in fact represents a substantial revision of the theory.
Sentential negation adverbs are nominalâ•… 87 In at least one of the languages in (12), namely Stroppo in (12d), the sentential negation adverb rEŋ has the same form as the negative argument, as illustrated in (5). This suggests an alternative analysis of the partitive in (12). Suppose that we take the fact that ‘not’ and ‘nothing’ in languages like Stroppo have the same form not to be just a matter of homophony (eventually to be explained on diachronic grounds), but rather as an indication of the fact that a single lexical entry (with non-disjunctive properties) is involved. If we take the negative argument content to characterize the lexical entry of rEŋ as a whole, we obtain the Logical Form in (13) for (12d); this is endowed with the interpretive properties indicated by Kayne (1984), but without any recourse to silent Q elements. Quite simply, we can assume that the role of the negative quantifier licensing the partitive is played by rEŋ itself. To be more precise, rEŋ introduces the polarity variable existentially closed within the scope of the polarity (here negation) operator. (13)â•… Stroppo/ Macra al beuâ•… [Q rEŋ]â•… [de viŋ]
For the data in (10), we envisage an analysis along the lines of that provided for (12). In fact (10) represents a generalization of the structures already proposed for (12), where no definiteness restriction applies. On the contrary, we must assume that the elements that can enter into the partitive construction in (12) are sensitive to the same constraints that generally restrict the occurrence of (object) bare nouns in Romance to plurals and mass singulars. The proposal in (13) raises many questions, some of which can be settled on the basis of fairly standard assumptions. For instance, the fact that rEŋ de viŋ in (13) is not a constituent can be viewed as a result of the fact that rEŋ as a sentential negation is in a Q-floated position. Other questions require more articulated answers. In particular, one may wonder what the difference is between ‘He doesn’t drink’ and ‘He drinks nothing’ if ‘not’ and ‘nothing’ coincide in the lexicon; we shall return to this question in section 3.1.2. Genitive of negation in Russian (9) arises with a negative clitic, rather than with a negative adverb; the reason will become apparent when we discuss negation clitics in chapter 4. In short, we have argued that the lexical formal identity between negative arguments and sentential negation adverbs in (2)–(6) supports a single lexical entry for these elements. We have further argued that this analysis provides an explanation of why sentential negations license partitives, as in (10)–(12). More generally, our idea is that the same analysis applies to all sentential negation adverbs€ – and not only to those of which the relevant phenomena hold.
88â•… Sentential negation:€adverbs If negations of the type of nEn, nEinta, nuta illustrated in (2)–(4) have the referential content of the negative argument ‘nothing’ (with which they may lexically coincide), one may wonder what the referential content may be for mia in (12a–c), (10) and for similar minimizers (briza, pa, etc). In fact, they have the only reference independently known to be compatible with bare singular count Ns, i.e. again that of negative (polarity) items. To illustrate, bare singular Ns, though generally excluded in Romance, become possible in the scope of a negative operator, as illustrated in (14) for standard Italian. Note that what is crucial for the argument in the text is that bare count singulars are allowed only in negative (polarity) contexts; it is not crucial that all negative (polarity) contexts should license all bare count singulars.2 (14)â•… a.â•… b. c.
*(Non)â•… si muoveâ•… foglia not M/Pâ•… moves leaf ‘Not a leaf stirs’ *(Non)â•… alzò ciglio not he.raisedâ•… eyebrow ‘He didn’t raise an eyebrow’ *(Non) proferì parola not he.said word ‘He didn’t say a word’
3.1.1 Further evidence Before we proceed with a discussion of the proposal we are putting forward, and the various issues it raises, we will review some more complex evidence relating to the interaction of sentential negation adverbs with the argument structure of the predicate. The varieties of Quarna Sopra and Quarna Sotto both alternate between a ‘nothing’-type adverb, i.e. nota, and a bare-N adverb, namely mia, as illustrated in (15) and (16) respectively. Consider first Quarna Sopra in (15). In this language, nota appears in intransitive (unergative) contexts, as in (15a), as well as in the middle voice (reflexive), as in (15b). In transitive contexts, nota appears when the internal argument is a 1st or 2nd person pronoun, as illustrated by (15c). With 3rd person internal arguments, nota Â�co-occurs with accusative lexical objects and clitics, as in (15d–d’), whilst with partitive clitics and with lexical objects introduced by the partitive preposition di ‘of’, we find mia, as in (15e). Note that the partitive (15e) has an interpretation that makes it equivalent to the accusative (15d). (15)â•… Quarna Sopra (Piedmont) a. i drm/ t drum/ G drm/ i drumma/ i drumi/ i drmn I sleep/ you sleep/ he sleeps/ we sleep/ you sleep/ they sleep ‘I don’t sleep’ etc.
nt not
Sentential negation adverbs are nominalâ•… 89 b. m lau nt/ t lavt nt/ s lava nt me I.wash not/ you you.wash not/ himself he.washes not ‘I don’t wash myself’ etc. c. m/t vgn nt me/you they.see not ‘They don’t see me/you’ d. i camnâ•… ntâ•… u te fril theyâ•… call not the your brother ‘They don’t call your brother’ d’. u vgn nt him they.see not ‘They don’t see him’ e. n camn miaâ•… d u te fril of.himâ•… they.callâ•… not ofâ•… theâ•… yourâ•… brother ‘They don’t call your brother’ e’. n vgn mi of.it/of.them they.see not ‘They don’t see (of) it/them’
In the variety of Quarna Sotto in (16), nota appears with intransitive (unergative) verbs, as in (16a), and in the middle (reflexive), as in (16b). Interestingly, unaccusatives, as in (16c), can split between nota and mia according to person, so that 1st and 2nd person co-occur with nota, while 3rd person co-occurs with mia. In transitive contexts, mia and nota again split according to person. Thus, nota occurs with 1st and 2nd person objects, as in (16d), while mia co-occurs with 3rd person objects, either in the form of a partitive, as in (16e) and (16f’), or of an accusative lexical object, as in (16f). Accusative clitics do not occur in negative contexts. For enclisis of the clitic to declarative finite verbs, we refer the reader to the brief discussion of this phenomenon in relation to the examples in (10). (16)â•… Quarna Sotto a.â•… j drmma/ d drumma/ G drmma/ j drumoma/ j drumi/ drmu nota I sleep/ you sleep/ he sleeps/ we sleep/you sleep/â•… they.sleep not ‘I don’t sleep’ etc. b. jâ•… lau not-m/ d lEu not-t/ G lava not-s I washâ•… not-me/ you wash not-you/ he washes not-himself ‘I don’t wash myself/ you don’t wash yourself/ he doesn’t wash himself’ c. sum/ t i/ soma/si Ju nota I.am/you are/â•… we.are/you.are come not ‘I haven’t come’ etc. c’. i mia Ju/ Ju nota he.is not come/ come not ‘He hasn’t come’
90â•… Sentential negation:€adverbs d. â•… vgâ•… notɐ -m/â•… -t/â•… -u he sees not me/ you/ you(pl) ‘He doesn’t see me/you’ e. vg mi-n he sees not-of.it/them ‘He doesn’t see it/ them’ f. lavu mia kamiz they.washâ•… not the shirts ‘They don’t wash the shirts’ f’.╇ beu mi d viŋ he drinks not of wine ‘He doesn’t drink wine’
Person split phenomena are pervasive in the Romance languages. For instance, in the clitic system, 1st and 2nd person forms distinguish at most between nominative and non-nominative; 3rd person forms are the only ones to register any difference between accusative and oblique, both in the morphology and in the position of the clitic.3 Similarly, it will be seen from (15b) and (16b) that only the 3rd person has the specialized middle (reflexive) marker si. We surmise that, although a 1st or 2nd person element and a 3rd person one can equally serve as arguments of a predicate, they do so through different syntactic means. In particular, the position and the morphology of 1st and 2nd person arguments are not necessarily sensitive to their anchoring in the event structure (i.e. whether they are the first or second internal argument of a ditransitive, etc.). In this sense, we speak of ‘discourse-anchored’ elements, i.e. elements whose position and morphology are sensitive only to their denotational content. By contrast, the morphosyntactic properties of 3rd person elements reflect their argumental role (accusative vs. dative marking, specialized middle marking by si, etc.). In this sense, we speak of ‘event-anchored’ elements.4 Let us consider Quarna Sotto in (16). In present terms, mia selects for event-anchored internal arguments. The latter include 3rd person accusatives and partitives, as well as the 3rd person nominative arguments of unaccusatives. In turn, nota appears in the complementary set of environments, i.e. those where there isn’t an event-anchored internal argument. Therefore it combines with discourse-anchored (i.e. 1st and 2nd person) internal arguments and with predicates, i.e. unergatives, that either have no internal argument or, as we assume here, have an incorporated one (Hale and Keyser 1993). Reflexives are of particular interest; if the distribution of mia and nota was simply sensitive to a person split, we should expect the reflexive to have nota in the 1st and 2nd and mia in the 3rd€– whether reflexives are unaccusative structures (Marantz 1984) or transitive structures, with the reflexive clitic playing the role
Sentential negation adverbs are nominalâ•… 91 of the internal argument (Burzio 1986) (cf. chapters 5–6). On the other hand, many independent (for instance distributional) facts point to the conclusion that si patterns with 1st and 2nd person object clitics rather than with 3rd person accusatives. The notion of discourse anchoring is meant to capture this fact as well; specifically, si is characterized as the free variable of the system (as in chapters 5–6 here) and is dissociated from the event-anchored set on the basis of its indefinite/quantificational nature. Therefore we account for the fact that 3rd person reflexives will pattern with 1st or 2nd person internal arguments in being selected by nota. The interaction of the negation adverb with the person split is hardly expected if the adverb corresponds to a functional category Neg€– whose content presumably is that of the logical connective of negation. In other words, it is hard to find a reason why the logical operator of negation would be lexicalized in two different ways according to the argument structure of the verb. We are not saying that the interaction of the negation adverb with the argument structure of the verb cannot be stated in terms of a conventional theory treating the adverb as an instance of logical negation. What we are saying is that any such connection would remain a pure stipulation. Suppose, however, that, as outlined in section 3.1, the sentential negation adverb is in reality a nominal element bearing a particularly close relation to the internal argument of the verb (with which it can lexically coincide). Suppose, furthermore, as also suggested in section 3.1, that the negative adverb can be interpreted to the extent that it forms a partitive structure with the internal argument€– eventually triggering an overt partitive case on it, as seen in Quarna Sotto itself. From this perspective, connecting the so-called sentential negation to the nominal, argumental set of categories, its interaction with the argumental structure of the predicate is not surprising and in fact expected. Consider Quarna Sopra in (15), then. In this language, mia is found in structures with 3rd person partitive internal arguments, while nota occurs in all other structures, including those with 3rd person accusatives, 1st/2nd person objects, intransitives (unergatives) and middles (reflexives). Because partitive structures are restricted to 3rd persons, it is evident that the distribution found in Quarna Sopra presupposes a person split analogous to the one examined for Quarna Sotto; in other words, Quarna Sopra further restricts the distribution of mia observed in Quarna Sotto to event-anchored (i.e. 3rd person) partitives. The distribution of mia in Quarna Sopra again argues in favour of the nominal, argumental status of the sentential negation. Thus, although it can be stipulated that two different Neg functional categories, whose content is presumably that of logical negation, are inserted depending on the presence of a partitive, 3rd
92â•… Sentential negation:€adverbs person internal argument or on its absence, there is no connection between the two sets of facts, and hence no explanation why such a state of affairs holds. By contrast, an explanation of this state of affairs follows from the view of negation that we are developing here, which connects the so-called sentential negation to the nominal, argumental set of categories. 3.1.2 Sentential negation adverbs as nominal arguments On the basis of the preceding evidence, so-called negation adverbs are nominal categories.5 Furthermore, the interactions of the sentential negation with the internal argument of the verb reviewed above (the fact that the negation is lexicalized differently according to the person of the internal argument, the partitive under negation, the ambiguity between adverbial and argumental reading of ‘nothing’) point to the conclusion that the negation is connected to the internal argument slot. This conclusion is supported by the fact that from a purely truth-functional perspective, negating the internal argument, as in, for example, I ate nothing, is equivalent to negating the sentence, e.g. I didn’t eat. At the same time, negating the internal argument and negating the sentence as a whole cannot simply be identified. Consider the simple case in which the sentential negation co-occurs with lexicalization of the internal argument by a noun phrase, as in I didn’t eat the apple. The analysis that we propose is based on languages in which the negation selects a partitive rather than an accusative internal argument. In this case it is evident that the sentential negation can be construed as introducing a quantification over the internal argument€ – which correspondingly is lexicalized as a partitive. As already sketched in section 3.1, in connection with (13), we extend this analysis to all cases where the sentential negation co-occurs with an overt lexicalization of the internal argument. Thus we take it that examples such as (16d–e) from Quarna Sotto, independently of the actual presence of a partitive, have the same Logical Form, as schematized in (17a) and (17b). In (17), N labels the partitive and P labels 1st /2nd person. In accordance with reasonable assumptions about the nature of elements selecting the partitive, we impute to the negation the (nominal) categorization Q, which characterizes weak quantifiers in the structure of DPs, as we have already done in (13). The logical negation operator in whose scope the polarity elements not and mi are interpreted is notated by ⌐.. (17)â•… Quarna Sotto a.â•… [⌐â•… [D [I vg [Q not [1/2Pm]] b. [⌐ [D â•… [I vg [Q mi[N n]]
Sentential negation adverbs are nominalâ•… 93 On the basis of the analysis in (17), we also expect that the sentential negation interpretively combines with the internal argument of unaccusatives, despite the fact that this is lexicalized as the EPP argument. As for as the contrast mentioned above between I ate nothing and I didn’t eat, the crucial observation is that two different argumental frames of to eat are involved. The nothing example implies a transitive argumental frame, where nothing satisfies the internal argument. On the other hand, the not example involves unergative to eat, which, following Hale and Keyser (1993), we construe as a concealed transitive with an incorporated internal argument. From this perspective, not again introduces a quantification over the internal argument, which in this case is incorporated. We can make this more precise if we assume, along the lines of Hale and Keyser (1993) and Chomsky (1995), that an unergative predicate and more generally a transitive predicate (of which the unergative is a subtype) consists of two layers. The most embedded layer (headed by V in Chomsky (1995)) corresponds to an elementary event, while the outer layer (headed by v in Chomsky (1995)) corresponds to a causation imputed to the external argument. Thus the incorporated internal argument of unergatives is the elementary event to which the external argument applies. But if so, transitive structures also include an elementary event to which the external argument applies, which can be construed as an internal argument as well. The only difference with respect to unergatives is that the internal argument of the elementary predicate becomes the internal argument of the transitive predicate as a whole, much as in overt incorporation structures such as Romance (or better Bantu) causatives. This allows us to refine the account proposed for sentences like (17) or English I didn’t eat the apple€ – which we now predict to be ambiguous. If the negation quantifies over the internal argument of the elementary event, we obtain the reading ‘It was not the apple that I ate’ or ‘No situation of me eating anything was a situation of me eating the apple’. If the negation quantifies over the elementary event, we obtain the reading ‘It was not eating the apple that I did’ or ‘No situation of me doing anything was a situation of me eating the apple’. As we shall see in section 3.2.2, this type of ambiguity characterizes quantificational adverbs in general, and, following de Swart (1993), quoted by Ernst (2002:€347), it is connected to the focus Â�structure of the sentence. Since unaccusatives reduce to elementary events, according to Hale and Keyser (1993) and Chomsky (1995), the latter is of course what is quantified over€– or rather, its internal argument. In the present approach, as a rule, we avoid building interpretive structure into the syntax in the absence of independent morphosyntactic evidence. Now, no such evidence seems to us to be forthcoming in the case of the v-V structure
94â•… Sentential negation:€adverbs of Chomsky (1995) or the incorporation of V into v. For instance, all evidence pertaining to the postverbal subject in Italian, i.e. the in situ subject under the VP-internal subject hypothesis, seems to indicate that it shares the same predicative low domain of the sentence with the internal (and other) arguments. In other words, we object to the functional category v very much on the same grounds as we object to (much) functional structure in the rest of this work€– namely that it simply provides a syntactic encoding for what is in reality an interpretation. In this sense it is at best redundant€– and at worst it potentially obscures the real nature of the phenomena it seeks to capture. Therefore we maintain a syntax in which the elementary event corresponds to the verb, or the verb plus its internal argument€– while the complex predicate (causative) reading simply arises by composition of the external argument with the elementary predicate. 3.2
Ordering sentential negation with respect to other adverbs
Suppose that sentential negations are nominal in nature and attached to the internal argument position of the verb€– as argued in section 3.1. The question still arises of how the present theory can accommodate the evidence presented by Zanuttini (1997) and Cinque (1999) concerning the relative order of several types of adverbs and other adverbial material. As mentioned at the outset, Zanuttini and Cinque seek to account for the data by postulating three Neg positions ordered within the aspectual adverb (sub)hierarchy as in (1). The evidence is somewhat complicated by the fact that many varieties have at least two different sentential negation adverbs, which Zanuttini characterizes as presuppositional vs. non-presuppositional. In general, it seems to us that such a distinction is difficult to maintain, given that all instances of negations, and more generally of focus, must have a presupposition as part of their interpretation. In what follows we will simply refer to what Zanuttini calls the presuppositional adverb as the mica-type adverb, since mica lexicalizes this type of adverb in standard Italian. We begin by reviewing the data€– the reader interested in the analysis can skip directly to the table in (38) and to section 3.2.1. In some varieties, negation adverbs precede the aspectual series including ‘already’, ‘any longer’, ‘still/yet’ and ‘always’. A case in point are varieties like Castellazzo in (18), in which this position characterizes the mica-type adverb mEj in (18b) as well as the ordinary sentential negation adverb nEit in (18a). Other varieties have a single sentential negation adverb to cover the two contexts€– and once again this occurs before the entire aspectual series of adverbs.
Ordering sentential negation and other adverbsâ•… 95 This is illustrated in (19) for a variety in which the negation adverb is mia and in (20) for a variety in which it is pa. Quite straightforwardly, all of these adverbs would correspond to the Neg2 position of Zanuttini (1997) and Cinque (1999). (18)â•… Castellazzo Bormida (Piedmont) a.â•… a ŋ l u nEitQ zQ/ pi/ aŋkurQ/ d lNk I not it have not already/ any longer/ yet/ always ‘I haven’t done it already/ any longer/ yet/ always’ b. a ŋ l u mEjQ zQ/ pi/ aŋkurQ/ d lNk I not it have not already/ any longer/ yet/ always ‘I haven’t done it already/ any longer/ yet/ always’
fa:tS done fa:tS done
(19) Pozzaglio (Lombardy) a. El drma mia p/ Jam/ sEmper/ bE: he sleeps not any longer/ yet/ always/ well ‘He doesn’t sleep any longer/ yet/ always/ well’ b. El g a mia bEle durmi:t he Loc has not already slept ‘He hasn’t already slept’ (20) Pomaretto (Piedmont) a. a drm pa bŋ/ pi/ ŋkar/ sampre he sleeps not well/ any longer/ yet/ always ‘He doesn’t sleep well/ any longer/ yet/ always’ b. al a pa dZO dyrmi he has not already slept ‘He hasn’t already slept’
The evidence becomes less straightforward when we consider sentential negation adverbs that appear inside the aspectual series. The hierarchy in (1) leads us to expect that, insofar as they instantiate Neg3, they will follow ‘already’ and precede the lower adverbs. In reality, in a language like Cantoira in (21), in which the sentential negation adverb follows ‘already’, we observe that it also follows ‘any longer’ in (21a) and (21c), and ‘still/ yet’ in (21a’) and (21c), while preceding ‘always’ and ‘well’. The mica-type adverb is unproblematic, since it has the distribution already observed in (18)–(20); i.e. it could correspond to Neg2 in Zanuttini’s (1997) terms. The relevant examples are given in (22). (21)â•… Cantoira (Piedmont) a. u dyrt dï†ı pi Jiŋ he sleeps already any longer not ‘He already doesn’t sleep any longer’ a’. u miJdZunt aŋku Jiŋ they eat yet not ‘They don’t eat yet’
96╅ Sentential negation:€adverbs b.╅ c. d.
u miJdZunt Jiŋ biŋ/ sEmp they eat not well/ always ‘They don’t eat well/ always’ u l ont pi/ aï†−ku Jiŋ tSama: they him have any longer/ yet not called ‘They haven’t called him yet/ any longer’ u l ont Jiŋ sEmpe tSama: they him have not always called ‘They haven’t always called him’
(22)â•… Cantoira a.â•… u l ont pa pjy/ Jku/ sEmpe tSama: they him have not any longer/â•… yet/ alwaysâ•… called ‘They haven’t called him yet/ any longer/ always’ a’. u l ont pa dï†ı miï†−dZa they it have not already eaten ‘They haven’t already eaten it’
Zanuttini (1997:€74) considers the ordering of nen ‘not’ after pi ‘any longer’, arguing that pi nen ‘no longer’ is a single constituent, placed in the position otherwise occupied by ‘any longer’ alone. However, even if Zanuttini (1997) was right about pi nen, we would be led to conclude that the sentential negation adverb is lower in the hierarchy in (1) than Neg3€– on the basis of the fact that it also follows (a)nku ‘still/yet’. This conclusion somewhat diminishes the interest of discussing the issue of whether pi nen is in fact a constituent, because, on the basis of a lower ranking of nen in the hierarchy, we would automatically predict that it follows pi. However, Zanuttini (1997) deploys a number of arguments in favour of the single constituent status of pi nen, which deserve some discussion. A classical constituency test applied by Zanuttini involves the impossibility of inserting any lexical material between pi and nen. But in infinitival contexts, where the negation adverb is in preverbal position, pi and the like can actually follow the verb, as seen in (23) for Cantoira itself, as well as other varieties. Of these, Margarita and Mezzenile have the same pattern as Cantoira in finite sentences.6 As for Montaldo, a comprehensive set of data is provided in (36)–(37) below. Note that our argument cannot simply be dismissed by noting that the string in (23) is nen … pi rather than pi … nen, as we are arguing for the possibility that these two elements represent separate constituents, irrespective of their relative order. (23)â•… a.â•…
Cantoira dZi t e dit t Jiŋ tSama-lu I to.youâ•… have said to not call-him ‘I told you not to call him any longer’
pjy any longer
Ordering sentential negation and other adverbsâ•… 97 b. c. d.
Mezzenile (Piedmont) t Ei di t Jint tSamaâ•… pjy Jy:n to.youâ•… I.haveâ•… saidâ•… to not call any longer anybody ‘I told you not to call anybody any longer’ Margarita (Piedmont) j di-te t nEŋ tSamE-u pi I I.have told-you to not call-him any longer ‘I told you not to call him any longer’ Montaldo i sJ kuntEnta yd nEŋ avEi pi palQ I am happy to not have any longer spoken ‘I am happy not to have spoken any longer’
Furthermore, it is true that some varieties feature an incorporated form of pi nen; Zanuttini cites pin for Bollengo, while our data include, for instance, pjiŋ in Piverone (Manzini and Savoia (2005)). However, the argument could be turned against Zanuttini’s proposal. Thus, only a pin-type form can be rightly considered as a single lexicalization of ‘no longer’. By contrast, a language which has the combination pi nen can be argued to lexicalize the negative adverb ‘no longer (no more)’ literally by combining two constituents, namely ‘more’ (pi or più in standard Italian) and the negative adverb. Working within the framework of Zanuttini (1997), it is also possible to describe the fact that the negative adverb precedes anku ‘still/yet’ by saying that anku is incorporated into the negation. The proposal is no less plausible than the idea that pi incorporates, since in many Northern Italian varieties there exists a negative counterpart of ‘still/yet’, represented here by am in Pozzaglio, as in (19) (to be compared with non-negative am in S. Angelo, as in (27d)). Yet one may wonder why Zanuttini (1997) doesn’t simply reformulate the hierarchy in (1) by positioning Neg3 further down, below ‘still/yet’. The answer has largely to do with the distribution she observes in infinitives. In her data, the negation adverb of the nen type can only precede the infinitive, while pi nen can either precede or follow. She takes this to indicate that nen is higher in the adverb hierarchy than pi nen (treated as a single constituent). Thus, if the infinitive is positioned between Neg3 and the ‘any longer’ Asp position, the order nen€– infinitive€– pi nen is derived. The fact that pi nen can also precede the infinitive simply means that the verb can move slightly lower. In proposing this analysis, Zanuttini (1997) extends to Piedmontese varieties the schema of explanation proposed by Pollock (1989) for French infinitives; for, according to Pollock, the fact that French infinitives follow pas means that they move to a relatively low position, lower in any event than the position of the finite verb (which precedes pas). One difficulty with this extension, as
98â•… Sentential negation:€adverbs already noted by Kayne (1991), lies in the fact that while French has proclisis on the infinitive, Piedmontese varieties have enclisis, as illustrated in (23) for Cantoira and Margarita. The simplest account for this fact is that the infinitive (like other modal forms of the verb, e.g. the imperative) fills a relatively high, C-related position. But if so, the ordering of adverbs with respect to non-finite verbs cannot be explained by the low position of the verb within the adverb hierarchy. We return to these matters in section 3.3. What is also directly relevant here is that in our data, for instance the Margarita examples, nen can occur not only before the infinitive, as noticed by Zanuttini (1997), and illustrated here in (24a), but also after the infinitive, as in (24b). Since pi nen can also occur before the infinitive, again as noticed by Zanuttini and illustrated here in (24c), this means that even if we adopt Zanuttini’s (1997) analysis of the position of the infinitive, the language provides no evidence in favour of pi nen occurring in a lower position than nen. (24)â•… Margarita a.â•… j di-te t nEŋ tSamE-u I have told-you to not call-him ‘I told you not to call him’ b. j di-te t tSamE-u nEŋ I have told-you to call-him not ‘I told you not to call him’ c. j di-te t pi nEŋ tSamE-u I have told-you to any longer not call-him ‘I told you not to call him any longer’
In short, the evidence reviewed seems to provide no support for Zanuttini’s (1997) conclusion that the position of nen inside the aspectual string corresponds to the Neg3 position in (1). As far as we can see, it can be made compatible with such a hypothesis, given, for instance, a high enough degree of optionality in the movement of the non-finite verb, in order to account for the data in (24). But, of course, making the movement of the non-finite verb highly optional reduces the predictive power of the model, eventually turning it into a series of descriptive statements. An equally (or more) problematic case is represented by the third class of negations individuated by the hierarchy in (1)€– which, according to Zanuttini (1997), corresponds to no adverbs found in the Milan and Pavia varieties. In particular, in varieties like Milan or Casorezzo, the latter of which is illustrated in (25), the sentential negation no precedes relatively high aspectual adverbs such as ‘always’ and ‘still/yet’ in (25a–c’), while it follows an adverb like ‘well’ in (25a), which is much lower in the hierarchy in (1). The logic of the
Ordering sentential negation and other adverbsâ•… 99 hierarchy, of course, is that an element like no which follows ‘well’ should follow other aspectual adverbs. This unexpected distribution combines with another property not observed before, namely that no is not positioned between the auxiliary and the participle as in the varieties considered so far (cf. for instance (18), (22)). Rather, it follows the participle, as in (25c). (25)â•… Casorezzo a. al drmi ben n:€ se:mpr he sleeps well not always ‘He doesn’t always sleep well’ b. ly la vedi ben n ŋkamo he it sees well not yet ‘He doesn’t see it well yet’ c. l a drmiâ•… ben n: he has slept well not ‘He hasn’t slept well’ c’. l a drmiâ•… n ŋkamo he has slept not yet ‘He hasn’t slept yet’
Zanuttini (1997) argues that the overall pattern of Milanese (or Casorezzo, as illustrated here) can be explained by assigning no to the Neg4 position, lower than ‘well’. An argument in favour of this conclusion is provided by the Â�participle. According to Zanuttini, this moves to a position immediately higher than ‘well’, leaving the low Neg4 negation to its right. Other adverbs, which are higher in the hierarchy, appear to the left of the participle, i.e. between the participle and the auxiliary. Zanuttini acknowledges the problem represented by the positioning of no in front of ‘always’ and ‘still/ yet’ as in (25a–b). In these examples, she proposes that no actually fills a different position, namely the Neg2 position associated with the negation adverbs in (18)–(20) and with the mica-type adverb in varieties like (22). Interestingly, she argues that a no appearing in Neg2 is interpreted as a mica-type adverb. Yet this analysis would lead us to expect that when no precedes ‘always’ or the aspectual adverbs, it also precedes ‘well’. Data of this type are indeed attested, as in (26). But the data already reported in (25a–b) show that the orderings of no after ‘well’ and before ‘still/yet’/‘always’ can also be combined. (26) Casorezzo a. al drmi n se:mpr be: he sleeps not always well ‘He doesn’t always sleep well’ b. ly la vedi n ŋkamoâ•… be: he it sees not yet well ‘He doesn’t see it well yet’
100â•… Sentential negation:€adverbs Furthermore, we note that in at least some of the varieties with a no negation of the relevant type, a separate mica-type negation is also present. A case in point is S. Angelo in (27), in which the negation follows ‘well’ as in (27a) and the participle as in (27c), but a mia negation is also present, as in (d). This tends to discount Zanuttini’s (1997) idea that the no negation may play the role of the mica-type negation in contexts like (27b), where it is ordered in front of other low adverbs, as opposed to preceding them. (27)â•… S. Angelo Lodigiano (Lombardy) a.â•… El drme bEŋ n he sleeps well not ‘He doesn’t sleep well’ b. El drme n sEmper he sleeps not always ‘He doesn’t always sleep’ c. i m aŋ tSamade n theyâ•… meâ•… haveâ•… called not ‘They haven’t (yet) called me’ d. El tSami mia nn am himâ•… I.call not yet ‘I am not calling him yet’
A further set of varieties which display a no negation is exemplified here by Viguzzolo in (28). The distribution of no in these varieties overlaps with that described here for nen, mia, etc. in the languages in (18)–(20). Thus, no precedes all adverbs in the aspectual series, as in (28a), and it normally appears together with these between the auxiliary and the participle, as in (28b–b’). The language of Viguzzolo also has a mia negation which has the same distribution as no, as illustrated by (29). (28)â•… Viguzzolo (Piedmont) a.â•… a nn al tSameŋâ•… n ŋku/ empr/ be theyâ•… notâ•… him call not yet/ always/ well ‘They don’t call him yet/ always/ well’ b. a l aŋ n (semper) tSama theyâ•… himâ•… haveâ•… not always called ‘They haven’t always called him’ b’. aŋ n ŋku maï†−dʒa they.haveâ•… not yet eaten ‘They have not yet eaten’ c. u nn a n drumi be heâ•… notâ•… hasâ•… notâ•… slept well ‘He hasn’t slept well’
Ordering sentential negation and other adverbsâ•… 101 (29)â•… Viguzzolo a.â•… a nn al tSameŋâ•… mej py they not him call not any longer ‘They don’t call him any longer’ a’. u n drm mej ŋku he not sleeps not yet ‘He doesn’t sleep yet’ b. u n l a mej py det he not it has not any longer said ‘He hasn’t said it any longer’ b’. a nn aŋ mej ŋku/ʑa maï†−dZa they not have not yet/already eaten ‘They have not already/ yet eaten’
Zanuttini (1997) is aware that in some varieties, the no-type negation appears before the participle, and illustrates this distribution for Pavia. In her analysis, however, no in Pavia still occupies the Neg4 position. Her independent evidence in favour of this low position resides in the fact that in Pavia, no can also appear after the participle. The fact that no precedes the aspectual series is analysed, as before, as an effect of its optional appearance in the mica-type negative position, i.e. Neg2. As for sentences in which no precedes rather than follows the participle, this is analysed as a consequence of the fact that the participle moves to a lower head position. A variety present in our data which illustrates the variable order of the no negation with respect to the participle is Castiglione in (30b–c), in which (as in Viguzzolo) the negation also precedes ‘well’, as in (30a–b). Furthermore, Castiglione has a separate mica-type negation, illustrated in (30d). (30)â•… Castiglione d’Adda (Lombardy) a.â•… el drm n beŋ/ sEmper he sleeps not well/ always ‘He doesn’t always sleep/ sleep well’ b. i aŋ durmi:d nâ•… (beN) they have slept not well ‘They haven’t slept (well)’ c. i aŋ n (semper) durmi:d they have not always slept ‘They haven’t (always) slept’ d. el drm miGa beŋ he sleeps not well ‘He doesn’t sleep well’
102â•… Sentential negation:€adverbs Since in both Viguzzolo and Castiglione the negation precedes ‘well’, as it does other adverbs, the simplest analysis for these languages is one that assimilates their no negation to the high negations in (18)–(20). But this, in turn, implies that the post-participial position of the negation in Castiglione is entirely unconnected to its supposedly low position in the adverb hierarchy. Indeed, the order participle€– no€– ‘well’ in (30b) is not predicted by Zanuttini’s (1997) analysis, since, if the post-participial position of no depends on its filling Neg4, then it should precede ‘well’ and not follow it. Similarly, the order no€– participle€– ‘well’ in (28c) cannot be explained by the claim that no is pre-participial because the participle is even lower than no€– for in that case no ought to follow ‘well’. Zanuttini’s (1997) analysis of the post-participial position of no proves equally problematic for Milanese-type varieties, represented here by Casorezzo. In Zanuttini’s own terms, though no is supposed to be quite low when it follows ‘well’, it is also supposed to raise to a higher position when preceding aspectual/quantificational adverbs. But now observe that no follows the participle independently of which adverbs it combines with and in which orders, as in (25c–c’). Thus, by Zanuttini’s own logic it must be the case that the participle is high enough to precede the whole adverbial string in (1). But if so, of course, the post-participial position of the negation does not tell us anything about whether it in fact appears low in the adverbial string€– which is, quite strikingly, exactly the same conclusion we reached for Margarita in (24). We provide our own analysis of the position of adverbs with respect to verbal heads in section 3.3, and we disregard this issue altogether in the rest of this section. More questions concerning Zanuttini’s (1997) model arise when we consider that the patterns reviewed so far are not the only ones created by the interaction of negation with the aspectual series. The nen pattern in our data that provides the best match with Zanuttini (1997) is illustrated in (31). This has ‘already’ and ‘any longer’ preceding the negation, and the other relevant adverbs following it. The pattern is predicted by Zanuttini on the basis of the incorporation analysis of pi nen. The data in (32) provide a comparison with the mica-type negation in the same language. (31)â•… Mombercelli (Piedmont) a.â•… u d42m nEŋ aŋkua/â•… bE he sleeps not yet/ well ‘He doesn’t sleep yet/ well’ b. u d42m ʑa/ pi nEŋ he sleeps already/â•… any longerâ•… not ‘He doesn’t sleep already/ any longer’
Ordering sentential negation and other adverbsâ•… 103 c. i lu tSmu nEŋâ•… sEmp theyâ•… himâ•… call not always ‘They don’t always call him’ (32)â•… Mombercelli a. i d42mu pa py they sleep not any longer ‘They don’t sleep any longer’ b. i aŋ pa ŋku/ sEmp/ ʑa dumi they have not yet/ always/ alreadyâ•… slept ‘They haven’t slept yet/ already/ always’
But now consider Pamparato in (33), in which ‘already’ precedes the nen(t)type negation, which is therefore a candidate for Neg3, as in (33d). As expected under the hierarchy in (1), nen(t) precedes ‘any longer’, as in (33a) and (33c). What is not expected is that nen(t) follows ‘still/yet’, as in (33b) and (33d). The data in (34) provide a comparison with the mica-type negation. (33)â•… Pamparato (Piedmont) a.â•… i d24muâ•… naint tSy they sleep not any longer ‘They don’t sleep any longer’ a’. u drm naint saimp he sleeps not always ‘He doesn’t always sleep’ b. i d24muâ•… ŋku naint they sleep yet not ‘They don’t yet sleep’ c. i aŋ naint tSy d24mi they have not any longerâ•… slept ‘They haven’t slept any longer’ d. i aŋ ŋku/ ʑQ naint d24mi they have yet/ already not slept ‘They haven’t yet/already slept’ (34) Pamparato (Piedmont) a. i d24muâ•… pQ tSy / ŋku theyâ•… sleep notâ•… any longer/â•… yet ‘They don’t sleep yet/any longer’ b. i aN pQ Nku/ ʑQ d24mi they have not yet/ already slept ‘They haven’t slept yet/ any longer’
Other languages that share with Pamparato in (33) the property of having a negation that precedes ‘no longer’ while following ‘still/yet’ include Romansh varieties. A case in point is Mustér in (35), in which the negation is of the
104â•… Sentential negation:€adverbs bu(ka), bare-N type. In other words, the ordering pattern is not tied to the morphological ‘nothing’-type negation. As shown in (35a, c), bo in Mustér precedes ‘any longer’€– but, as shown in (35b, d), it follows ‘still/yet’. (35)â•… Mustér (Grisons) a.â•… jauâ•… drmlâ•… boâ•… pli/ adina/ bain I sleep not any longer/â•… always/â•… well ‘I don’t sleep any longer/ always/ well’ b. jau drml aun bo I sleep yet not ‘I don’t sleep yet’ c. i aŋ bo duʀmiu pli they have not slept any longer ‘They haven’t slept any longer’ d. i aŋ auŋ bo duʀmiu they have yet not slept ‘They haven’t slept yet’
A further pattern attested in our data is illustrated in (36) (the data in (37) provide a comparison with the mica-type negation). In (36), nEŋ precedes ‘already’€– which, in terms of the hierarchy in (1), would have to mean that we are in the presence of an exponent of Neg2. Yet again nEŋ follows both ‘any longer’ and ‘still/yet’, which would once again have to constitute cases of incorporation into negation. (36)â•… Montaldo a.â•… i u tSam pi nEŋ I him call any longer not ‘I don’t call him any longer’ b i u vEg nEŋ sEmp I him see not always ‘I don’t always see him’ c. i ŋku nEŋ fQ-u I have yet not done-it ‘I haven’t yet done it’ d. i nEŋ dʒQ fQ-u I have not already done-it ‘I haven’t already done it’ e. i nEŋ drmi bEŋ I have not slept well ‘I haven’t slept well’ (37)â•… Montaldo a.â•… i u vEg pæ pi/ sEmp I him see not any longer/ always ‘I don’t see him any longer/ always’
Ordering sentential negation and other adverbsâ•… 105 b.â•… i pQ Nku/ dZa fQ-u I have not yet/ already done-it ‘I haven’t done it already/ yet’
In the table in (38), we summarize the data we have provided concerning the relative position of the negation and the other adverbs in the sub-hierarchy in (1). It seems to us, from looking just at the distribution of the negation adverb with respect to these other adverbs, that there are two major patterns, rather than three. Thus, in varieties of the type in (38a), the negation always precedes the relevant subset of adverbs. On the other hand, in varieties of the type in (38b), their relative order is somewhat variable.7 It will be noticed that when it comes to the patterns in (38b), practically any relative order of the negation adverb with respect to ‘already’, ‘still/yet’ and ‘any longer’ is attested€– with one striking exception. What is not attested in our data is the order which should be the basic one given the hierarchy in (1), with ‘already’ preceding the adverb, and all other members of the hierarchy following it. This absence is significant, and would be so even if we wanted to follow Zanuttini’s (1997) line of explanation with respect to the possible permutations€– i.e. that the reordered substring roughly reflects cases of incorporation. Indeed, we notice that sometimes ‘still/yet’ is reordered, and sometimes ‘any longer’€– in other words, in neither case is reordering necessary, which means that there is no independent reason why the supposedly basic order should be excluded. Therefore the fact that it is not found should at least cause some perplexity. (38)
Relative position of negation and other adverbs
a. Castellazzo B.â•› Pozzaglioâ•› Pomarettoâ•› Viguzzoloâ•› Castiglione d’A.â•› b. Cantoiraâ•› Mombercelliâ•› Pamparatoâ•›
already no longer still
always
nɛ̃â•›itQâ•› (18a) mEâ•›jQâ•› (18b) mia (19) pa (20) nâ•› (28) mejâ•› (29) nâ•› (30) mIGâ•›a
– zQâ•› – zQâ•› – bEâ•›le – dʒ╛╛
– pi – pi – pâ•› – pi
– ʑa
– py
– aŋ↜kurQâ•› – aŋ↜kurQâ•› – Jâ•›amâ•› – ŋ↜karâ•› – ŋ↜ku – ŋ↜ku
– d lâ•›ŋ↜kâ•› – d lâ•›ŋ↜kâ•› – sEâ•›mper – bɛ̃â•›: – sampre – bâ•›ŋ↜ – ɕâ•›empâ•›r – be – sEmper
– beâ•›ŋ↜ – beŋ↜
Jiŋ↜(21) pa (22) nŋ↜(31) pa (32) naint (33)
dZ╛╛
aŋ↜ku€– – ŋ↜ku aNku4aâ•› – Nkuâ•› ŋ↜ku€–
– sEâ•›mp – sEâ•›mp – sEmpâ•› – sEmpâ•› – saimp
– biŋ↜
– py
ʑâ•›a€– – ʑâ•›Qâ•› ʑâ•›Qâ•›–
pi€– – pjy pi – py – tSâ•›y
well
– bEâ•›
106â•… Sentential negation:€adverbs already no longer still Mustérâ•› Montaldoâ•› Casorezzoâ•› S.Angelo L.â•›
pQâ•› (34) –zQâ•› bo(35) nEŋ↜(36) –dZQâ•› pQâ•› (37) – dZQâ•› nâ•› (25)–(26) nâ•› (27) mia
– tSyâ•› – pli pi€– – pi
– Nkuâ•› aun€– Nkâ•›u€– – Nkâ•›u – ŋ↜kamâ•›
always
well
– adina – sEmpâ•› – sEmpâ•› – sEâ•›mper – sEâ•›mper
– ain –bENâ•› beâ•›(ŋ)↜€– bEâ•›ŋ↜–
3.2.1 The order of negation with respect to aspectual adverbs Summarizing so far, the functional hierarchies for negation adverbs, and adverbs more generally, proposed by Zanuttini (1997) and Cinque (1999) do not seem to be sufficient to account for the spread of variation observed in Italian varieties. A separate question is whether they are necessary. The alternative generally suggested is that ‘adverb attachment is driven by interpretation’ (Svenonius 2002:€209), essentially as in the earliest treatments of adverb placement in generative grammar (Jackendoff 1972). Specifically, Ernst (2002:€92) claims that ‘the most important determinant of adjunct licensing is an adjunct’s scope (and other selectional) requirements, encoded as lexical requirements and verified at LF, rather than syntactic feature licensing, as in Cinque (1999) and other current work’. As Nilsen (2003:€13) observes, according to Ernst (2002) ‘adverbs can attach freely to any functional projection as long as the semantic requirements of the adverbs (FEO-calculus) are respected. Hence Ernst has it that, while adverbs are ordered by his FEO-calculus, the categories that the adverbs attach to are ordered by an orthogonal relation’, i.e. the basic functional sequence. Nilsen (2003:14) argues against this approach on the grounds that (among other things) ‘this set-up seems to force us to abandon the view that T has semantic import’. Indeed, according to Ernst (2002:€97), ‘events and propositions are not necessarily mapped to any one particular projection … A relatively high projection like TP may represent an event as long as nothing forces it to be converted to a proposition (such as a modal auxiliary or adverb); likewise a low projection like PredP can denote a proposition if no element above it requires an event’. What is more, Nilsen (2003:€18) objects to Ernst (2002), as well as Svenonius (2002), on the grounds that ‘although … the facts discussed by Cinque (1999) should ultimately be derived from more fundamental considerations, … the primary tool to do so can[not] be enrichment and manipulation of ontological
Ordering sentential negation and other adverbsâ•… 107 categories’€– i.e. the categories (fact, event, etc.) in terms of which the selectional properties of adverbs are stated. While it is evident that a limited repertory of conceptual entities is generally desirable, it seems to us that the flexible interpretation of syntactic categories required by Ernst (2002) is not in itself worse than a rigid interpretation of them. In fact, it is one of the overarching themes of the present work that the interface between syntax and interpretation is not rigid€– in the sense both that ambiguity is not necessarily resolved syntactically and that the same interpretation can be conveyed by different syntaxes. From this perspective, we are not surprised to find that a given syntactic domain€– here the I domain€– potentially maps to different semantic entities, or that the opposite may happen:€the same semantic entity may map to different syntactic domains. Let us consider, then, the relative orders illustrated above in which sentential negation adverbs appear internally to the substring of aspectual/manner adverbs individuated by Cinque’s (1999) and Zanuttini’s (1997) work. We assume, as is standard, that the lowest verbal domain of the sentence is characterized by the closure of thematic roles by arguments. The fact that negation adverbs, and adverbs in general, precede the argumental string suggests that they are inserted outside this lower verbal domain. At the same time, several adverbs, and in particular those in the substring in (1) that directly interests us here, typically follow the finite verb, which appears in the I head of the inflectional domain. Therefore, between the lower thematic domain and the higher inflectional one there must be at least an intermediate domain, which in Manzini and Savoia (2005) we name E (suggesting Event). It is in this domain that we find the adverbial substring in (1). We begin with the aspectual adverbs già ‘already’, ancora ‘still/yet’ and più ‘any longer’. Ernst (2002:€341–7) construes the interpretation of ‘already’ in the following terms (based on work by Michaelis (1996)):€‘the immediate scope of already denotes a state S, located at reference time, and whose inception precedes the time of an expected possible state S’ of the same type as S’. Thus a sentence like Karen has already performed is interpreted as saying that ‘S, the result state of Karen’s performing … precedes another interval at which she was expected to finish performing’. As for still, ‘the state scoped by still holds at reference time, held at a previous time and its expected end e was before reference time’. Thus, in a sentence like They still were doing it yesterday, ‘yesterday identifies the reference time, and still indicates that the process was ongoing at an earlier event-time as at reference-time’. As for ‘no … longer’, a widely held analysis takes it to be the negative counterpart of ‘still’ (Löbner 1989).
108â•… Sentential negation:€adverbs In the varieties of Montaldo, Pamparato and Mombercelli in (38b), the negation adverb precedes only one of the aspectual adverbs, respectively ‘already’, ‘any longer’ and ‘still/yet’. The rigid extrinsic ordering that characterizes Cinque’s (1999) hierarchies is a clear liability here, since there is no way of interspersing Neg positions and adverb positions so as to make the different possible orders emerge. The incorporation of single adverbs (in particular pi ‘any longer’) into the negation, suggested by Zanuttini (1997) as a way of making Cinque’s (1999) hierarchy compatible with the data, hints at a different type of explanation, whereby the properties of the single lexical item become relevant; in other words, the relative ordering of adverbs is conditioned by lexical idiosyncrasies. This also seems unsatisfactory to us; as we indicated when presenting the data, there is no clear evidence that the sequences that Zanuttini (1997) treats as single constituents are in fact constituents. Our idea is that the syntactic grids determining the relative order of adverbs are much coarser grained than the functional hierarchies of Cinque (1999), and refer to the same nominal categories that are relevant for the ordering of argumental material. It is in this connection that the conclusion of section 3.1€– that the so-called negation adverb is in reality a nominal category connected to the internal argument of the predicate€– becomes relevant for ordering as well. Let us begin with structures of the type in (13) or (17), where the negation adverb occupies a Q position. As we have just seen, this can be taken to be internal to an intermediate domain between V and I, i.e. E. We model the nominal arrays of verbal domains on inflectional-level clitic structures, where we independently argue for Q elements (represented in the clitic domain by the si variable, cf. chapters 5–6). In the inflectional clitic domain there is a restricted range of elements above Q, which includes, in particular, what we may take to be the left edge position of the theory, notated R (with a nod to Referentiality). Evidently, this left edge position could be imputed to the adverbs that precede the negation in Q, as illustrated in (39) for the Piedmontese varieties in (38b). (39)â•… a.â•… b. c. d.
Cantoira [R dZ6/ pi/ aŋkuâ•… Mombercelli [R ʑa/ pi Pamparato [R ʑQ/ Nku Montaldo [R pi/ Nku
[Q Jiŋ [Q nEN [Q naint [Q nEŋ
If the Q position of the negation remains constant, the aspectual adverbs ordered after it occur in a lower position. Our general schema for nominal
Ordering sentential negation and other adverbsâ•… 109 positions recognizes a series of positions below the left edge R position and the quantificational Q position, which are essentially linked to deixis, including 1st and 2nd person arguments and locatives. We may want to assign the aspectual adverbs that follow the negation in Q to this set of positions, adopting, for instance, the label Loc in (40)€– but keeping in mind that deictic relations in time, rather than in space, are being denoted. (40)â•… a.â•… b. c.
Mombercelli [Q nEŋ [Loc aNku4a Pamparato [Q naint [Loc tSy Montaldo [Q nEŋ [Loc dZQ
In a nutshell, this analysis amounts to saying that negations of the type being investigated generally have a Q position. There they are mostly followed by adverbs of the aspectual series in a deictic position (conventionally Loc). In some varieties, some aspectual adverbs are shifted to the left edge of the domain (here R) on a lexical basis. In a sense this is the analysis that Zanuttini (1997) gives for the pi nen order:€namely, that the relatively high position of the negation is masked by ‘any longer’ incorporating with it. The difference is that in the analysis that we are now developing, the reordering of the aspectual series around the negation requires no incorporation. At the same time, the existence of truly (i.e. morphologically) incorporated forms of ‘no longer’ and ‘not yet’ supports the idea that selection for particular lexical items is at stake in the reordering. At this point we can bring into the picture another major type of language, namely the type in (38a), in which the negation adverb systematically precedes the aspectual series. In terms of the analysis being considered, they can be described as languages with the negation adverb relatively high in the R or Q position, with other aspectual adverbs following it in the deictic reference positions (i.e. Loc), as shown for Castellazzo in (41). Manzini and Savoia (2005), who do not entertain the possibility of a deictic (Loc) position for the aspectual adverbs, assign them to the Q position, on the basis that they imply a quantificational closure over events. At least as far as the schema in (41) is concerned, the Q and Loc alternatives yield equivalent results. (41)â•… Castellazzo Bormida [R/Q mEjQ/ n˜itQâ•… [Q/Loc zQ/ pi/ aŋkurQ
That two different positions, presumably R and Q as in (41), are needed for the negation is a conclusion motivated by independent evidence. In many
110╅ Sentential negation:€adverbs varieties the mica-type negation can actually combine with the ordinary sentential negation adverb, as in (42). In this case, the mica-type negation generally takes the higher position, presumably R, and the ordinary negation takes the lower position, presumably Q. (42)╅
Margarita u Q pa nEŋ tSama-u he has not not called-him ‘He has not called him’
In cases where two aspectual adverbs combine with the negation, both can precede it, as illustrated in (43) for varieties (Cantoira and Mombercelli) in which we have shown that each of the aspectual adverbs involved can separately precede the negation. (43)â•… a.â•… b.
Cantoira u dyrt dZ pi Jiŋ he sleeps already any.longer not ‘He doesn’t already any longer sleep’ Mombercelli u d42m ʑa pi nEŋ he sleeps already any.longer not ‘He doesn’t already sleep any longer’
A straightforward application of the structural schemas already proposed leads us to assume that while the negation regularly appears in Q, the left edge position R of the system can be iterated for aspectual adverbs. Alternatively, adopting Manzini and Savoia’s (2005) idea that aspectual adverbs can appear in Q, it is Q that turns out to be iterated, equivalently for present purposes, as shown in (44). (44)â•… Cantoira [R d [R/Q pi [Q Jiŋ
If we put together the pattern in (42), where the mica-type negation precedes the ordinary negation, with the fact that aspectual adverbs precede the ordinary negation, we may expect orders like (45a), where the mica-type negation is followed by an aspectual adverb and then by the ordinary negation. Similarly, since aspectual adverbs can also follow the ordinary negation, we may expect the pattern in (45b), where the mica-type negation is highest and the aspectual adverbs are on either side of the ordinary negation. (45)â•… a.â•…
Margarita u Q pa pi/ ŋku nEŋ tSama-u he has not any.longer/ yet not called-him ‘He hasn’t called him yet/ any longer’
Ordering sentential negation and other adverbsâ•… 111 b.
Montaldo i pQ ŋku nEŋ dZa fQ-u I have not yet not always done-it ‘I haven’t yet done it already’
The distribution of the two aspectual adverbs on either side of the ordinary negation in (45b) can be predicted for Montaldo simply on the basis of the union of the structures (39d) and (40c). Assuming that the position of the micatype negation, in turn, is R, we obtain the structure in (46). (46)â•… Montaldo [R pQâ•… [R/Q ŋkuâ•… [Q nEŋâ•… [Loc/Q dZa
Before considering in more detail what degree of mobility the aspectual adverbs as well as the negation have, in the next section we will analyse the order of the negation with respect to a quantificational adverb like ‘always’, as well as with respect to the low manner adverb ‘well’. 3.2.2
The order of negation with respect to quantificational and manner adverbs Accounting for the ordering of negation with respect to ‘always’ and ‘well’ implies some preliminary understanding of these adverbs. Let us first consider ‘always’€ – and quantificational adverbs more generally. According to Ernst (2002:€ 347), who quotes work by de Swart (1993), ‘these adjuncts quantify over subsets of events within the set denoted by their sister constituent; the precise delimitation of these sets depends on the focus structure of the sentence’. So a sentence like John always eats pasta can mean ‘In all situations in which John eats something, he eats pasta’ or ‘In all situations in which John does something, he eats pasta’. When it comes to order, according to Ernst (2002:€363) ‘frequency adverbs … have fewer restrictions than aspectual adverbs’. Yet the fact to be explained is that ‘always’ is found obligatorily after the negation adverb in all varieties studied here or that we are aware of. It seems evident that this obligatory ordering is dictated by semantic requirements. In fact ‘always’ is obligatorily read in the scope of the logical negation. The Romance languages, like English, have a separate lexical item for a universal ‘always’ taking scope over the logical negation, i.e. never (standard Italian mai etc.). Obviously the reading in the scope of the logical negation requires ‘always’ to be lower than the negation adverb, i.e. the element which implies the logical negation itself. Given the ordering of aspectual adverbs with respect to the negation adverb sketched in section 3.2.1, and the rigid ordering of the quantificational adverb with
112╅ Sentential negation:€adverbs respect to negation which we have just analysed, we predict that the three combine in a fixed order, which is what can be seen in the examples in (47). (47)╅ a.╅ b. c.
Cantoira uj t pi Jiŋâ•… sEmp dyrmi he has any.longer not always slept ‘He hasn’t always any longer slept’ Margarita j pa pi nEŋâ•… sEmpre vist-je them I.haveâ•… not any.â•… longer not always seen-them ‘I haven’t always seen them any longer’ Pamparato i aŋ ʑQ naint saimp d24mi they have already not always slept ‘They haven’t always already slept’
If we keep to the idea that the negation lexicalizes Q properties in the E domain, quantificational adverbs like ‘always’, which follow it, must appear in a lower position in the same domain. However, the deictic position(s) that we have assigned to aspectual adverbs are not consistent with the interpretive properties of quantificational adverbs, which simply denote relations between sets of events very much independent of temporal deixis. This then leaves the lowest position of the nominal sequence, namely N, which we connect to the internal argument of the predicate (cf. chapters 7–8). In fact, we have seen at least one way in which ‘always’ is related to the internal argument€– namely by implying a focalization on it, in one of the possible interpretations of John always eats pizza, namely ‘whenever John eats something, he eats pizza’. Alternatively the entire elementary predicate can be in the scope of the focus operator as in the interpretation ‘all situations of John doing anything are situations of John eating pizza’. As discussed in section 3.1.2, the same ambiguity is present with the quantificational adverb we are mainly concerned with here, i.e. the negation. Thus John doesn’t eat pizza can be understood as ‘It is not pizza that John eats’ or as ‘It is not eating pizza that John does’. In section 3.1.2 we concluded that in the latter reading the negation quantifies over the elementary event that can be construed as the internal argument of the causation (or other) property introduced by the application of the external argument. In the former reading the negation quantifies over the internal argument of the elementary predicate or equivalently of the whole transitive predicate, namely pizza. If a parallel analysis is applied to non-negative quantificational adverbs, it is natural to conclude that these adverbs have at their disposal the N position connected to the internal argument. This in turn yields a straightforward structure for
Ordering sentential negation and other adverbsâ•… 113 the sequences in (47), as illustrated in (48) for (47c). Remember at the same time that another possibility is suggested by structures like (46) in which the negation in R or Q precedes elements (potentially) in the same types of positions. On this basis we may assume that the negation in Q precedes the quantificational adverb also in Q, where it can attach because of its quantificational content. (48)â•… Pamparato [R ʑQâ•… [Q naintâ•… [Q/N saimp
Before we proceed further, let us introduce the last major type of adverbials included in the hierarchy in (1), namely manner adverbs like ‘well’. According to Ernst (2002) manner adverbs are part of a larger class of predicational adverbs; ‘the manner adverbial restricts the denotation to a set of events of V-ing characterized by their property of (manifesting) rudeness, dimness, strangeness and the like’ (Ernst 2002:€ 259), for instance in She left rudely or The bulb shone dimly. The low point of attachment of manner adverbials reflects the fact that what they are predicated of is events (as opposed to situations/sentences). However, Ernst (2002) does not consider the strict relation that an adverb like ‘well’ has not only with the event, but also with the internal argument of that event. For instance, in several Southern Italian varieties ‘well’ (bene in standard Italian) translates as the adjective for ‘good’ (buono in standard Italian). In (49a–c) we report examples in which this adjective can be seen to overtly agree with the accusative internal argument. In the same varieties, the internal argument of unaccusatives also agrees with ‘well’ as shown in (49a’–c’). As for unergatives, either ‘well’ takes an invariant form corresponding to the masculine singular, as in the Marzano example in (49a’’), or it agrees with the external argument, as in the Gizzeria example in (49b’’). (49)â•… Marzano Appio (Campania) a. u/a veru bbuonu/ bbna him/her I.see well.m/ well.f ‘I see him/her well’ a’. issu/ essa se lava sEmbe bbuonu/ bbna he/sheâ•… him/herself washesâ•… alwaysâ•… well.m/well.f ‘S/he always washes well’ a’’. issu/essaâ•… a rurmitu bbuonu he/she has slept well ‘S/he slept well’ Gizzeria (Calabria) b. u/â•… a/ i lavu bbnu/â•… bbna/bbni him/her/themâ•… I.wash well.m/well.f/well.pl ‘I wash him/her/them well’
114â•… Sentential negation:€adverbs b’. b’’.
iɭɭu/ iɭɭa s a he/sheâ•… him/herselfâ•… hasâ•… ‘S/he has washed well’ iɭɭu/ iɭɭa a ddrmutu He/sheâ•… hasâ•… slept ‘S/he slept well’
ɭɭavatu bbnu/ bbna washedâ•… well-m/well.f bbnu/ bbna well/ well.f
Orsomarsoâ•… (Calabria) c. (a kammisa/ i kavutsuni) aɟɟu lavta/ lavtiâ•… bbna/â•… bbuni the shirt.f/ the hose.pl I.have washed.f/pl. well.f/ well.pl ‘The shirt/ the hose I washed well’ c’. arâ•… arrivtu/arrivta bbunu/ bbna s/he.has arrived.m/f. well.m/well.f ‘S/he has arrived well’
The same connection between manner adverbs and the internal argument can be seen in English sentences like I ate a quick pizza, meaning I quickly ate a pizza/I ate a pizza quick(ly). Similarly, it is possible to say that John has done a job well, or equivalently that John has done a good job. Therefore the manner adverb is connected to the internal argument position insofar as it is predicated either of an elementary event or of its internal argument€– which is also adopted as the internal argument of the transitive predicate as a whole. This is also what makes it possible for an adjective agreeing with (or embedded under) an accusative argument to modify an entire (elementary) event. The preceding discussion leads not only to the conclusion, shared with Ernst (2002), that manner adverbs attach to a relatively low domain in the sentence (namely the E domain here), but also that they are connected in particular to the N position of the domain. The nominal nature of manner adverbs is evident in languages like (49), but also in standard Italian and Northern Italian varieties in which bene is a noun, as in il bene ‘the good’ (e.g. il bene comune ‘the common good’) and so is male ‘badly’, as in il male ‘the evil’ (cf. il bene e il male ‘good and evil’). In the light of our hypotheses it is particularly interesting to consider data that show the position of ‘well’ not only with respect to negation, but also with respect to aspectual and quantificational adverbs, as in (50). (50)â•… a.â•… b.
Cantoira u dyrt Jiŋ sEmp biŋ heâ•… sleepsâ•… notâ•… alwaysâ•… well ‘He doesn’t always sleep well’ Margarita u parla pi nEN bEN he speaksâ•… any.longer not well ‘He doesn’t speak well any longer’
Ordering sentential negation and other adverbsâ•… 115 c.
Piverone (Piedmont) al drm ŋku Jiŋ sEmpe bEŋ he sleeps yet not always well ‘He doesn’t always speak well yet’
In each case ‘well’ closes the adverbial string, which is what we expect if it takes the lowest N position in the domain. The other adverbs will then take the positions that we have suggested for them in the preceding discussion, yielding the structures in (51) for the examples in (50). Recall that in the discussion of quantificational adverbs, in particular ‘always’, we suggested that the Q position may be recursive, while aspectual adverbs that appear before negation are inserted in the left-edge position R. (51)â•… a.â•… b. c.
Cantoira [Q Jiŋâ•… [Q sEmpâ•… [N biŋ Margarita [R pi [Q nEŋ [N bEŋ Piverone [R ŋku [QJiN [Q sEmpe [N bEŋ
With this background, we can consider the final question implicit in the hierarchy in (1), namely the position of the no-type negation that follows manner adverbs. Recall that Zanuttini (1997) is led to propose the low Neg4 position in (1) for these elements on the basis of two pieces of evidence:€ first, their post-participial position, and second, the fact that they follow low manner adverbs like ‘well’. In reality, the evidence concerning the relative ordering of adverbs also shows that the no-type negation precedes quantificational and aspectual adverbs. Crucially, when negation is combined both with ‘well’ and ‘always’/‘yet’, two orders are attested in our data. One of them has the negation preceding both the quantificational/aspectual adverb and the manner adverb, as in the Casorezzo example in (26a); this is consistent with Zanuttini’s (1997) proposal that the relative ordering of Neg4 in front of other adverbs depends on its taking a higher position. What is not consistent with her proposal is the alternative order, which has the manner adverb before the negation and the other adverbs following it, as in the Casorezzo example in (25b). In reviewing the data, we further indicated that the existence of a dedicated mica-type negation in varieties like S.Angelo makes it implausible that no should take on the role of the mica-type negation when it is positioned higher. The alternative we propose is therefore simply that the no-type adverb has the very same position in the eventive domain string as the other negation adverbs reviewed so far, namely Q, as indicated in (52). This position is
116â•… Sentential negation:€adverbs compatible with quantificational, aspectual and manner adverbs following it, assuming that they take their canonical quantificational/deictic/nominal positions as in (52b). It is also compatible with the alternative order in (52a), in which only the manner adverb precedes no, on the assumption that in this case ‘well’ occurs in the left-edge R position of the system. Under this account, there is no difference in the positioning of Milanese-type no and Piedmontese nen; rather, there is a difference as to which types of adverbs are found in the ‘left-edge’ of the domain (here R), namely the manner adverb in Milanese varieties and aspectual adverbs in Piedmontese ones. (52)â•… Casorezzo a.â•… [R ben [Q nâ•… [Q/ Loc se:mpr/ ŋkamo b. [Q n [Q/ Loc se:mpr/ ŋkamo [N ben
Note that it does not follow from (52) that the relative ordering with negation higher than ‘well’ is restricted to cases where there is an adverb of another type. For some varieties this is as it should be, since there is evidence of variation in order even in simpler strings, as in (53). (53)â•…
Arconateâ•… (Lombardy) a maJdʒu be:n no/ no beŋ I eat well not/ not well ‘I don’t eat well’
For varieties like Casorezzo itself in which no optional order is attested in simple adverbial pairs, we of course predict the complex order in (52a)€– i.e. precisely the one that is problematic under other analyses. As for the order in (52b), we note that at least the quantificational adverb ‘always’ systematically precedes ‘well’ in the absence of negation, as in (54). We conclude that the optionality in (52) depends on whether the relative order ‘well’€– negation or ‘always’€– ‘well’ dictates the overall pattern; we shall elaborate on this alternative directly below. (54)â•… Casorezzo a.â•… Eŋ sempr maï†−dZa:â•… they.haveâ•… alwaysâ•… eaten ‘They have always eaten well’ b. Eŋ maï†−dZa:â•… semprâ•… they.haveâ•… eaten always ‘They have always eaten well’
be: well be: well
Assuming that the account in (52) is on the right track, it raises the question of why there should be variation of the kind observed, with some languages putting ‘well’ before negation (to the exclusion of aspectual adverbs) and other
Ordering sentential negation and other adverbsâ•… 117 languages (i.e. the Piedmontese varieties of section 3.2.1) showing the reverse distribution. An obvious observation is that all languages that allow the order in (52a) have a negation adverb of the no type. This belongs to the n- negative polarity series but does not have the same form as any negative polarity argument. Rather, it has the same form as the deictic negation. On the other hand, the varieties that allow aspectual adverbs before negation have negation adverbs either of the ‘nothing’ type or of the bare N type reviewed in section 3.1. From this perspective, the Piedmontese ordering in (39)–(40) can be described by saying that nominal negation adverbs (‘nothing’ or bare N type) only allow deictic adverbs out of their domain (i.e. aspectual adverbs under the characterization provided here). Since ‘well’ appears to be lexicalized by clearly nominal elements, such as ‘good’ in the examples in (50) or bene in standard Italian and Northern varieties, we may equivalently describe Piedmontese varieties by saying that nominal adverbs must remain in the domain of the nominal-type negation. By contrast, deictic-type, i.e. no-type, negations in Milanese varieties, only allow nominal adverbs, such as low manner adverbs, out of their domain. Equivalently, we may describe Milanese varieties by saying that deictic (i.e. aspectual) adverbs must remain within the domain of the deictic negation. The generalizations just stated suggest that the parameter differentiating between Milanese-type and Piedmontese-type varieties may not need to be (entirely) stipulated. Rather, the different inherent properties of the negations involved may determine their placement relative to other adverbial subseries. As for the general principles under which the parallelism indicated may fall, they could perhaps be identified with a requirement for the negation (of a certain type) to close off certain types of adverbial subdomains. This may interact with other requirements on the adverbial sequence, such as the one in (54), whereby the manner adverb ‘well’, which is in the logical scope of the quantificational adverb ‘always’, is also in its syntactic domain. In particular, the complex order in (52b) takes this latter requirement into account. In contrast, in (52a) the composition of the negation with the quantificational adverb ‘always’ (read in its logical scope) evidently means that the negation determines the overall order, in particular with respect to the manner adverb ‘well’. In any event, the principal aim of this discussion has been to indicate how an account of the variation between Piedmontese and Milanese-type varieties is possible if we abandon the account provided by the various Neg positions in the hierarchy in (1)€– which we have shown to be descriptively inadequate. Finally, in analysing Milanese-type varieties we disregarded the fact that no appears after the participle, though Zanuttini (1997) and Cinque (1999) regard this as crucial. The reason why we did not discuss this is that the post-participial
118â•… Sentential negation:€adverbs position of no is also found in languages like Castiglione in (30) in which no precedes ‘well’ and the other adverbs considered here. In other words the two phenomena are (at least partially) independent of one another. Yet the issue deserves to be considered in more detail, since it still seems to be true that there are no negations that both follow ‘well’ and precede the participle. Before we consider the issue in section 3.3, however, we will try to draw some general conclusions concerning the present discussion of the relative ordering of adverbs. 3.2.3 General discussion Cinque (1999), Ernst (2002) and Nilsen (2003), from opposing points of view, argue that a theory of adverb ordering that uses both semantic and syntactic requirements is less economical than one restricting itself to a single component€ – though in semantically based models like Ernst (2002) or Svenonius (2002), a certain degree of freedom is observed in adverb placement because it is sensitive to semantic primitives (events, facts, etc.) that are not necessarily isomorphic with syntactic ones (TP, VP, etc). More clearly, for Nilsen (2003) the positioning of sentential-level adverbs is determined entirely by the fact that they are positive polarity items that must be outside the scope of non-veridical operators. However, Nilsen (2003) does not consider the event/ predicate-level adverbs studied here, while Ernst (2002) explicitly excludes negation adverbs from his discussion. We do not doubt that notions of positive and negative polarity (Nilsen 2003) or selection for events vs. propositions (Ernst 2002) play a role in adverb placement. Yet it seems to us that Cinque (1999) is correct in concluding that purely semantic constraints are not sufficient to explain the distribution of adverbs. This is particularly clear in the case of the negation adverbs which represent the focus of the present discussion, because the scope of the logical negation is entirely independent of their surface position with respect to aspectual and manner adverbs. On the other hand, Cinque’s (1999) conclusion that syntactic constraints are sufficient for adverb ordering works (to the extent that it does) only because interpretive facts are encoded in the syntax in the form of functional hierarchies. This encoding introduces a degree of rigidity absent from semantic accounts. In this respect, the discussion in this section can be read as an elaborate argument that the variation in the position of the negative adverb cannot be captured by (1)€– not even allowing for the various readjustments proposed by Zanuttini (1997). Our general diagnosis of the kind of difficulties encountered by previous syntactic analyses is the same as that presented in chapters 1–2 for complementizers, namely that categories lifted from formal semantics accounts or
Ordering sentential negation and other adverbsâ•… 119 descriptive (functional, typological, etc.) accounts do not necessarily yield the correct primitives on which to build the syntax. Our proposal differs from those put forward in previous literature in assuming that the positioning of adverbs in what we have called the eventive E domain of the sentence is sensitive to much the same categorizations (Definiteness, Quantification, deixis, etc.) as the placement of arguments in the inflectional domain of the sentence (clitics) or of the noun phrase. From the present perspective, manner adverbs close the adverbial string in that they are connected to the N-internal argument€– understood as the elementary event that the higher-level specifications of the sentence select, or as the internal argument of this elementary event, which is construed as the internal argument of the sentence as a whole. Next, we have argued that the aspectual series of adverbs is essentially connected to deixis, in the same way as demonstratives in the structure of noun phrases, or locatives as arguments of the elementary predicate€– though in the case of aspectual adverbs the connection is to temporal rather than spatial deixis. On the basis of what we know about deictic categories in noun phrases and in arguments, we expect aspectual adverbs to appear between quantificational adverbs and manner/nominal adverbs, which they do. Alternatively or complementarily, aspectual adverbs may be construed as belonging to a macrocategory of quantificational adverbs (Manzini and Savoia 2005), in which case they are predicted to commute in position with quantificational adverbs (narrowly construed). Incidentally, ‘already’, ‘still’ and ‘any longer’ typically have verbal counterparts, such as ‘to continue’ in the case of ‘still’, ‘to stop’ in the case of ‘no longer’ etc., which for Cinque (1999, 2006) also instantiate the same functional categories as their adverb counterparts. Yet we have independent reasons to believe that so-called auxiliaries in general are just ordinary verbs selecting sentences, as argued in chapters 5–6. It is also worth noting that in standard Italian there is an alternative lexicalization of già ‘already’, namely bell(o/i/ etc.), which is clearly adjectival/nominal in nature, as in (55); in many Northern varieties this is the only lexicalization of ‘already’. This adjectival lexicalization connects the aspectual adverb to the manner adverb (if anything)€– and through it in turn to the elementary event/internal argument specifications. (55)â•… L’ ho (sempre) bell’ e fatto it I.have always nice and done I have (always) already done it’
Proceeding from bottom to top, quantificational adverbs close (or may close) the aspectual–manner range; by and large, all that needs to be said about the
120â•… Sentential negation:€adverbs sentential negation adverb is that it belongs to the quantificational series. At the same time, the (re)positioning of certain adverbs to its left, namely aspectual adverbs in Piedmontese varieties and manner adverbs in Milanese-type varieties, provides evidence for a left-edge position of sorts within the adverbial field. This (re)positioning seems to depend (in part) on the lexical shape of the negation adverb; thus, deictic negation adverbs of the Milanese type co-occur with left-edge positioning of manner (i.e. nominal) adverbs, while the nominal-type negations of Piedmontese varieties co-occur with the repositioning of aspectual (i.e. deictic) adverbs. In other words, the interplay of lexical terminals, such as adverbs in the case at hand, depends on their actual lexical make-up and cannot be predicted solely on the basis of abstract arrays of features making them up. To take other examples, it matters for the purposes of ordering whether the content of ‘already’ is carried by già, which appears at the beginning of the sequence in (1), or by bell(o) as in (55), which instead appears at the end of the sequence; it matters whether ‘well’ is the nominal element bene or the adjectival element buono, and so on. The theories of adverb ordering reviewed in this section, even when they ostensibly aim to account for a variation spread, as does Zanuttini (1997), ignore this dimension of the problem. In a nutshell, the distribution of adverbs depends on the interaction of interpretation with the syntactic and lexical structures that restrict it in natural language. Shaping the adverbial space purely through interpretive notions of scope (Nilsen 2003) or selection for events, facts, etc. (Ernst 2002) seems to be insufficient to account for the fine variation in the positioning of the verbal range which is the focus of the discussion in Zanuttini (1997) and partly also in Cinque (1999). At the same time, a syntactic theory along the lines of Cinque (1999) or Zanuttini (1997) contains elements of rigidity that prevent it from accounting for various data€– except perhaps by allowing for languageparticular variations in the hierarchy whose universal nature represents the core of the theory, and therefore defeating its very purpose. 3.3
The interaction of adverbial and verbal positions:€the participle
In the discussion of the relative positioning of adverbs in section 3.2, we omitted the evidence that according to Cinque (1999) and Zanuttini (1997) comes from the positions of adverbs with respect to non-finite verbal heads, including the infinitive and the participle. One reason that led us to factor out this evidence is that no correlation holds, say, between the placement of adverbs after the participle and their placement with respect to manner or other low adverbs,
Interaction of adverbial and verbal positionsâ•… 121 as discussed for Castiglione in (30). Similarly, the conclusion of Cinque (1999) and Zanuttini (1997) that verbal heads (participles and infinitives) preceded by negation and other adverbs are particularly low in the structure is contradicted by evidence from the placement of non-adverbial material, notably (en)clitics, as seen in (23)–(24). At the same time, it appears that constraints are observed in the placement of adverbial material with respect to verbal heads€– subject to cross-linguistic variation. Therefore, in this section we briefly address the nature of the observed constraints and variation, concentrating on the participle. In section 3.2 we saw that negation adverbs are generally positioned between the auxiliary and the perfect participle in perfect tenses; yet there are languages in which the negation appears after the participle. Zanuttini (1997) and Cinque (1999) account for this pattern by assuming that the negation occupies a very low position, namely Neg4, so that the participle, moving towards the left periphery of the low sentential domains, systematically leaves it to its right. However, we have already shown that the post-participial negation of Casorezzo, although it follows ‘well’ as in (25c), precedes quantificational/ aspectual adverbs, as in (25c’), so that the latter are in turn preceded by the participle. This means that by Zanuttini’s (1997) and Cinque’s (1999) own reasoning there are no empirical grounds for claiming that the post-participial position of the negation is due to its low position in the adverb hierarchy; it could be due instead to the relatively high position of the participle. If so, we expect to find that negations that Zanuttini (1997) considers to occupy a (relatively) high position can also follow the participle. Indeed, recall that Zanuttini (1997) attributes the no–aspectual adverb order seen in Milanese-type varieties, as in (25c’), to a repositioning of the negation in Neg2; yet in (25c’) the negation follows the participle. Another case in point is Quarna Sotto, which was introduced in section 3.1.1. Both nota and mia can follow the participle, as in (56a), though they are typical Piedmontese negations of the ‘nothing’ and bare N types assigned by Zanuttini to Neg3. Furthermore, at least mia also appears before the participle as in (56b); and the relative order ‘still/yet’–mia is possible with either placement of the participle. Following Zanuttini’s (1997) and Cinque’s (1999) own reasoning, alternations like those in (56) would prove that two slightly different positions are available for the participle, a higher one in (56a) and a lower one in (56b). (56)â•… Quarna Sotto a.â•… i ŋ ku tSamaâ•… mi- n/ not- m theyâ•… haveâ•… yet called notâ•… of.him/â•… not-â•… me ‘They haven’t yet called him/ me’
122â•… Sentential negation:€adverbs b. i n ku mi/ mi ku Ju they are yet not/ not yet come ‘They haven’t come yet’
In general, Cinque (1999) must invoke relatively high or low positions for the participle to account for the fact that adverbs which (unlike Neg) do not have several positions at their disposal can occur to the left or right of the participle. Thus, in the examples involving auxiliary–participle structures provided so far, the adverbs in (1) mostly appear between the auxiliary and the participle, with the exception of ‘well’, which appears after the participle, as do the negation adverbs just discussed. Yet in Southern Italian varieties the preferred order has the adverbs in (1) following the participle, as illustrated by one example among many in (57); note that in (57a) the lexicalization of the negation is carried entirely by the clitic, on which more will be said in chapter 4. The logic of Cinque’s (1999) explanation would be that in (most) Southern Italian varieties the participle occupies a higher position than in (most) Northern Italian ones. (57)â•… Celle di Bulgheria (Campania) a.â•… nunn addZu vistu ccu/ aŋkra not I.have seen any.longer/ yet ‘I haven’t seen him yet/ any longer’ b. l addZu vistu ddZa him I.have seen already ‘I have already seen him’
Yet, as with the infinitive briefly mentioned in section 3.2 (cf. (24)), the analysis of the relative order of adverbs and participles proposed by Cinque (1999) is contradicted by evidence concerning the order of the participle with respect to other clausal material. There is evidence (cf. here chapter 5) that perfect participles define independent sentences, as is the case for the so-called absolute (i.e. adjunct) participles of standard Italian (Belletti 1990). These participles seem to be in quite a high position in their own sentence, since enclitics can attach to them. This also shows that the sentences they define are provided with a full inflectional layer€– in fact, if the high position of the participle is identified with the modal C position, they are also endowed with the C layer. In the Piedmontese varieties studied by Zanuttini (1997) and in this chapter, clitics in auxiliary–participle constructions do not typically procliticize to the auxiliary, as in standard Italian and most other Italian (or Romance) varieties, but rather appear post-participially. The data in (45) and (47c), introduced for independent reasons and reproduced below, illustrate this state of affairs. If, following Cinque (1999, 2006), we adopt the traditional point of view that perfect tenses are monoclausal, the positioning of adverbs between the auxiliary
Interaction of adverbial and verbal positionsâ•… 123 and the participle suggests a relatively low position for the participle€– at least, lower than the position of the finite verb, which precedes them. Yet the classical account just described leaves clitics completely out of the picture. Recall that in finite sentences these appear quite high before the finite verb; hence their post-participial position would suggest that the participle is higher than the finite verb€– clearly a contradiction. (45)â•… a.â•… b.
Margarita u Q pa pi/ ŋku nEŋ tSama-u he has not any.longer/ yet not called-him ‘He hasn’t called him him yet/ any longer’ Montaldo i pQ ŋku nEŋ dZa fQ-u I have not yet not always done-it ‘I haven’t yet done it already’
(47) c.
Margarita j pa pi nEŋ sEmpre vist-je them I.have not any.longer not always seen-them ‘I haven’t always seen them any longer’
The fact that the post-participial clitics of Piedmontese represent a problem for the general schema of explanation ultimately going back to Pollock (1989) is pointed out by Kayne (1991) in relation to infinitives. If, as already suggested, we take the clitic position as an indication that the participle is in a high position, say the C position, then this forces the conclusion that it defines an independent sentence. The auxiliary in turn defines an independent sentence whose complement is the participial sentence, while the monoclausal interpretation must be the result of incorporation processes applying between the matrix and embedded predicate. This general idea can be fleshed out along the lines of chapter 5 here, but allows for very different implementations, such as that of Kayne (1993). Though the bi-sentential construal of perfect tenses yields a straightforward analysis of clitic placement, a problem now arises with adverbs, specifically those in the sub-hierarchy in (1), which would be expected to appear systematically after the participle in C, given that they appear systematically after the finite verb in the lower I position. Here the theory of semantic selection for adverbs argued for by Svenonius (2002) and Ernst (2002) will help us to resolve the paradox. Participles may define full sentences on the evidence of the inflectional material they are able to associate with, yet traditional monoclausal treatments of perfect tenses embody an important interpretive intuition about participles,
124â•… Sentential negation:€adverbs namely that they lack independent temporal reference (though they may have aspectual/modal properties corresponding to their placement in C). The lack of temporal properties, while not interfering with the inflectional argumental domain (represented by clitics), allows very high syntactic projections like C to preserve a semantic characterization akin to that of lower predicational/eventive projections in tensed sentences. If so, while post-participial adverbs can be connected to the positions they occupy in finite clauses, pre-participial adverbs can be construed as attaching to domains higher than C, which are unavailable to them in ordinary finite clauses. In short, we assume that the participle is inserted in a fixed C position on the basis of equally fixed aspectual/modal properties. The variable order with respect to adverbs, both within a given variety and cross-linguistically, becomes a question about adverbs themselves, some of which will appear in the low eventive domain (after the participle) and others will attach to the higher modal C field (before the participle). One of the arguments that Cinque (1999) deploys in favour of his construal of the evidence is that while the position of the participle (or other verbal head) varies with respect to the adverbs, the latter maintain their order with respect to one another. A good example of this line of argumentation is provided by the relative position of ‘well’ in the hierarchy, following aspectual and quantificational adverbs. Though in Northern Italian varieties negation, aspectual and quantificational adverbs are generally found before the participle, as seen for instance in the data in (45) and (47c) reproduced in this section, ‘well’ follows the participle, as shown in (58). (58) Castellinaldo (Piedmont) a. u piâ•… / aŋkura naŋ maï†−dZ baŋ he has any.longer/ yet not eaten well ‘He hasn’t eaten well yet/ any longer’ a’. u naŋ sampre maï†−dZ baŋ he has not always eaten well ‘He hasn’t always eaten well’ Fontane (Piedmont) b. j aŋ piâ•… / ŋk Jent dymi bEŋ they have any.longer/ yet not slept well ‘They haven’t slept well yet/ any longer’ b’. j aŋ Jent sEmp dymi bEŋ they have not always slept well ‘They haven’t always slept well’
For Cinque (1999) the contrast between (45) and (58) implies that the participle is inserted between ‘always’ and ‘well’ in the fixed hierarchy in (1). For
Interaction of adverbial and verbal positionsâ•… 125 us, the data require a different analysis€– namely that the very same properties that typically associate ‘well’ with the lower position in the adverb rank also associate it with the core eventive domain. In other words, in order to adequately describe the split between the ‘well’ adverb positioned after the participle and other adverbs positioned before the participle, all we need is the notion that manner adverbs like ‘well’ hook on to a nominal N position while other adverbs appear in deictic/quantificational slots. Yet nothing in principle prevents an N adverb from appearing before the participle in the high C domain, which counts as an eventive domain in the absence of temporal specifications. This way of putting things is therefore less restrictive than Cinque’s (1999), for any rearrangement of the hierarchy in (1) is disallowed (at least in principle) by the fact that it is part of the computational system. As far as we can tell, the flexibility of the present theory is precisely what is required by the data. Specifically, ‘well’ is described by Cinque (1999) as occupying a Voice projection. This predicts that ‘well’ does not appear before active participles, while it will appear in front of passive participles that can move to the Voice head. In reality, our intuitions on standard Italian are that both examples in (59) are acceptable, so that pre-participial ‘well’ is independent of voice. In present terms, pre-participial occurrences of ‘well’ need not be restricted to low projections, but may equally appear in higher projections, as long as selection for events is respected. (59) a. b.
La commissione ha bene accolto la nostra proposta The committee has well received the our proposal ‘The committee received our proposal well’ La nostra proposta è stata bene accolta dalla commissione The our proposal is been well received by the committee ‘Our proposal has been well received by the committee’
As for quantificational and deictic (i.e. aspectual) adverbs, Southern Italian varieties of the type in (57) seem to require the fewest stipulations, since all adverbs of the low range in (1) follow the participle exactly as they follow the finite verb, in both cases appearing in the low eventive domain. In Northern Italian varieties of the type in (58), what seems to be happening is that quantificational/deictic adverbs are preferably inserted in the highest possible eventive domain, i.e. the C domain delimited by the participial projection. This parameter in turn seems to be just a manifestation of a very general variation scheme in natural languages. To take another example close at hand, in the perfect tenses of Piedmontese varieties like (45) and (47), the clitic occupies the post-participial position
126â•… Sentential negation:€adverbs corresponding roughly to its lowest possible position (in the inflectional domain of the predicate with which it has a direct selectional relation). In contrast, in standard Italian, as in most Italian and Romance varieties (French, Spanish, etc.), the clitic in the same structures occupies the highest position available to it (in the inflectional domain of the highest verb of the verbal complex formed by ‘restructuring’, i.e. roughly event unification). As far as we can tell, there is a complete parallel in all other respects between the Piedmontese perfects and, say, the standard Italian ones. In general, certain lexical items select certain domains of lexicalization in certain languages€– without any correlation with meaning, or other properties of the grammar. This is precisely the situation with (quantificational and deictic) adverbs, which are restricted to the postverbal lower range of the sentence by the presence of tense specifications in finite clauses, but in untensed clauses can range from lower to higher domains. In itself, our construal of the adverb placement parameter with respect to non-finite (and finite) verbs is neither simpler nor more complex than Cinque’s (1999) account in terms of the placement of the participle, for the lower or higher functional head to which the participle moves does not appear to be determined by general principles. We have seen that for Cinque (1999) the crucial argument in favour of the participle moving€ – and adverbs staying in place€– is that when a participle interrupts a sequence of adverbs, so that some are to its right while some are to its left, the two subsequences are still predicted by the hierarchy in (1). However, we have presented more than one counterexample to this prediction. Thus, the examples relating to Castiglione in (30) or to Quarna Sotto in (56) show that the negation can appear postparticipially, and aspectual adverbs before the participle, though aspectual adverbs are (or can be) preceded by the negation when they all occur in the same domain. Further discussion would require us to increase the already rich repertory of facts in this chapter, at least by adding data concerning the position of adverbs, including the negation, in infinitival sentences. At the same time it is evident that for infinitival sentences we could adopt the same principles of analysis explored here for participles, since the latter are treated here as independent sentences as well. Thus, consider the data in (23)–(24), of which we reproduce below just the Margarita examples. The post-infinitival position of the clitic (shared with many Romance varieties, including standard Italian) leads us to associate the infinitive with a high C projection (Kayne 1991). If so, the ordering of the negation and aspectual adverb before the infinitive represents a shifting of their position with respect to the tensed sentence€– parallel to the one discussed for participial sentences.
Interaction of adverbial and verbal positionsâ•… 127 (23)â•…
c.â•… Margarita j di-te t nEŋ tSamE-uâ•… pi I I.haveâ•… told-youâ•… to not call-him any longer ‘I told you not to call him any longer’
(24) c.
Margarita j di-te t pi nEŋ tSamE-u Iâ•… have told-youâ•… to any longer not call-him ‘I told you not to call him any longer’
Under the present analysis, we expect the relative order of adverbs to reproduce the hierarchy observed in finite sentences when they occur in the same domain, as in (24). But nothing prevents them from occurring in different domains€– and this will create what appear to be reorderings of the (underlying or superficial) hierarchy, as in (23). This type of evidence represents a direct counterexample to Cinque’s (1999) and Zanuttini’s (1997) strongest prediction€ – and is in fact difficult to account for within their analysis, except by stipulation.
4 Sentential negation:€clitics
Many Romance languages have a negative clitic which stands alone (for instance standard Italian) or is doubled by a negative clitic (for instance French). Since Pollock (1989) these have been modelled by a NegP projection in which the negative clitic and the negative adverb occupy the head and Spec position respectively. According to Zanuttini (1997), who extends this model to Italian dialectal variation, languages which require a sentential negation adverb gen erate it in one of the lower Neg positions; if a clitic combines with the adverb, it is generated in the head of the relevant Neg position and moves to the inflec tional domain by cliticization. By contrast, negative clitics in languages like Italian which do not require a sentential negation adverb are hosted by a Neg position generated above I, Neg1. In other words, for Zanuttini (1997) preverbal clitic negations are associ ated with two different structures, according to whether they stand alone or combine with a negative adverb. Moreover, she detects a correlation between these structures and the ordering of the negative clitic with respect to subject clitics in languages that have both of them. In her terms Neg1, hence the stan dalone negative clitic, will precede inflected subject clitics, though it may fol low vocalic uninflected subject clitics. Counterexamples to this generalization are Tuscan varieties in which the standalone negative clitic precedes 1st/2nd person subject clitics but follows 3rd person ones (cf. Càsola in (7) below). Zanuttini (1997:€37) recognizes that these varieties require two different pos itions for subject clitics€– not only one following Neg1 but also one preceding it. Yet this is not sufficient to account for a more direct counterexample like Vagli in (2) below, in which the standalone negation follows all inflected sub ject clitics. In turn, according to Zanuttini (1997) the negative clitic doubled by a negative adverb follows all subject clitics. Yet there are several examples of negative clitics doubled by negative adverbs in which the negative clitic also precedes the 1st/2nd person subject clitic while following 3rd person ones; Oviglio in (20) below is a case in point. Zanuttini (1997:€39) suggests 128
Sentential negation:€cliticsâ•… 129 that this should be explained in terms of two different positions occupied by (inflected) subject clitics, one following and one preceding the negative clitic. But again there are more direct counterexamples in which the negative clitic doubled by the negative adverb precedes the entire series of inflected subject clitics.1 From this evidence we conclude that there is no necessary correlation between the order of the negative clitic with respect to subject clitics and whether it doubles a sentential negation adverb or not. Correspondingly, in the rest of the chapter we discuss negation clitics without taking into account whether they stand alone or combine with an adverb. At the same time, there is a lot of literature which takes the clitic to instan tiate the logical negation operator (Rizzi 1982; Longobardi 1992 on Italian); this makes negative clitics natural exponents for the functional category Neg, on the assumption that its content coincides with the logical negation itself. Yet in chapter 3 we concluded that sentential negation adverbs (which can also stand alone) are nominal constituents and negative polarity items; this led us to recategorize them from exponents of the Neg functional category to quantificational (Q) elements in the eventive (adverbial) domain of the sen tence. In this chapter we argue that clitic negations are also best construed as nominal elements, anchored in the argument structure of the verb, and specific ally the (nominal or eventive) internal argument. Correspondingly they do not instantiate the specialized functional category Neg, but rather they instantiate nominal categories independently necessary for hosting pronominal clitics in the inflectional domain of the sentence. Like negation adverbs, they are sim ply negative polarity items, which need to be read in the scope of the negation (or other polarity operator)€– and therefore imply the introduction of a logical negation operator at the interpretive interface. If we are correct, the latter is never lexicalized in the languages at hand (and possibly in natural languages in general). The evidence to be presented concerns first of all the distribution of negation clitics with respect to subject and object clitics. In section 4.1 we consider data which involve copying of the negative clitic on either side of a subject clitic, as well as the classical distributions studied by Zanuttini (1997) and Poletto (2000). In section 4.2 we consider copying of a negation clitic on either side of an object clitic, and then we move on to simple instances of a negation clitic inside the object string€– both phenomena also observed by Parry (1995, 1997). In both cases the doubling of the negation clitic is sensitive to the socalled person split, roughly between 1st/2nd person and 3rd person. On these
130â•… Sentential negation:€clitics grounds we argue that clitic doubling could not be a result of purely phono logical processes (of the type envisaged by Harris and Halle (2005)), because the notion of person split that it is sensitive to is a syntactic one. Similarly, readjustment rules at Morphological Structure (in the sense of Halle and Marantz (1993)) would have to import this same syntactic notion (and others) into a separate morphological component€– which, even if possible, would amount to a description, rather than an explanation of the phenomena. As for analyses of doubling based on head-Spec configurations and subsequent movement, we note that a base-generated head-Spec configuration is not suf ficient to account for multiple copies of the clitic, since there is a unique head position that the clitic can fill. The surface distribution of the clitic copies, sensitive only to the nature of other elements of the clitic string, also casts doubt on the idea that they are spell-outs of clitic traces, since it remains to be shown that general principles of movement and spell-out could determine such a distribution. Our own analysis is based on the assumption that clitics are base-Â�generated in the positions where they surface, being connected to their copies or other lexical material by the interpretive calculus at the LF interface, as in so-called representational models (Brody 2003). The sensitivity of negation clitics to the person split is explained in accordance with the conclusions of chapter 3 on the similar sensitivity displayed by negation adverbs€ – namely as a result of their nominal character and their connection to the overall argument structure of the sentence. Section 4.2.1 considers potentially problematic cases where one of the apparently negative copies also surfaces in positive contexts. Because we treat the so-called sentential negations as nominal elements, connected with the argument structure of the sentence, in section 4.3 we are led to conclude that doubling of sentential negations is an instance of neg ative concord. We deal with the latter by assuming, as anticipated, that nega tions are negative polarity items, so that negative concord amounts to closure of two or more negative polarity items by the same logical negation (or by the same existential quantifier in the scope of the logical negation). We also briefly consider what lack of negative concord amounts to€– or, equivalently, mutual exclusion of negative polarity elements in the negative concord read ing. We suggest an account modelled on mutual exclusions elsewhere in the grammar, for instance between pronominal clitics. This account correctly predicts that mutual exclusion (lack of negative concord) facts are domain sensitive.
Interactions of negation with subject cliticsâ•… 131 4.1
Interactions of negation clitics and subject clitics
Let us consider a case of doubling of the clitic negation on either side of a sub ject clitic, as seen in the Northern Tuscan variety of Viano in the 2nd person singular in (1ii). In the other persons, reported in (1) under the correspond ing roman numbers, the negative clitic simply follows the subject clitic. The examples in (1ii) show that the doubling of the negative clitic is entirely indif ferent to the composition of the object clitic string. (1)â•… Viano (Tuscany) i. a n drm I not sleep ‘I don’t sleep’ ii. a.â•… a n t n drm ClS not you ClS sleep ‘You don’t sleep’ b. n t (n) m cam not you not me call ‘You don’t call me’ c. n t n l cam not you not him call ‘You don’t call him’ d. n t n t lav not you not you wash ‘You don’t wash yourself’ e. n t ŋ g l dE not you not there it give ‘You don’t give it to him’ iii. i/ la n drm he/she not sleeps ‘S/he doesn’t sleep’ iv. a n dormjaŋ we not sleep ‘We don’t sleep’ v. n durmi not you.sleep ‘You don’t sleep’ vi. i/ la n dɔrmn they.m/they.fâ•… not sleep ‘They don’t sleep’
The pattern in (1) is connected to the fact that cross-linguistically a negative clitic can appear either before or after the subject clitic, as illustrated here in (2)–(3), again with varieties from Northern Tuscany. To be more precise, in a
132â•… Sentential negation:€clitics variety like Vagli in (2), the negative clitic follows the subject clitic, while in a variety like Sillano in (3) it precedes the differentiated subject clitic and it follows invariable e. (2)â•… Vagli di Sopra (Tuscany) i nun dɔrmE tu n dɔrmE i/ Eâ•… nun dɔrmE nun sâ•… dɔrmE nun durmitE i nun dɔrmn I not sleepâ•… etc. ‘I don’t sleep’ etc. (3)
Sillano (Tuscany) (e) n(o) i dɔrma (e) non tu dɔrma (e) no ll dɔrma (e) non dormjaŋ (e) non durmidd (e) no ll dɔrmŋ ClSâ•… not I sleepâ•… etc. ‘I don’t sleep’ etc.
Poletto (2000) accounts for data similar to (2)–(3) by assuming that both subject and negation clitics have more than one position available to them. To begin with, in (3) the invariable subject clitic is higher than the inflected one, since the negation occurs between them. Furthermore, the negation in (3) is higher than the negation in (2) since the former precedes the inflected subject clitic, while the latter follows it. This yields a hierarchy roughly of the type:€SCl (invariable)€– Neg€– SCl (inflected)€– Neg. We adopt the conclusion that the two subject clitics in (3) correspond to two different subject clitic positions. Similarly we accept that there are two different positions for the negations, preceding the inflected subject clitics in (3) and fol lowing them in (2). Following the intuition of Chomsky (1995) as to the nature of the EPP argument, we notate the subject clitic as D. More importantly, we do not identify the position of the negative clitic with a dedicated functional projection Neg, though nothing would prevent us in principle from doing so. Rather, given the existence of an independent hierarchy for pronominal clitics, we simply assume that the negation fits into it€– and specifically corresponds to the R slot (generically suggesting Referentiality), which occurs immediately
Interactions of negation with subject cliticsâ•… 133 after the subject clitic and immediately before object clitics. Furthermore the two different positions of the subject clitic and the two different positions of the negation clitic do not correspond to different functional positions, but sim ply to two different instantiations of the same D and R positions€– internal to the I and C domains respectively. On these grounds we assign the structures in (4) and (5) respectively to, say, the 2nd person singular of the paradigms in (2) and (3). (4)â•… Vagli di Sopra D tu/i/ε R n(un) I dɔ rmε
(5)â•… Sillano D e R no (C)
e
D i/tu/l
I dɔ rmε
The structures in (4)–(5) amount to the proposal that Vagli and Sillano dif fer as to whether the negation is inserted in the domain immediately above C, as in Sillano, or in the domain immediately above I, as in Vagli.2 If the Cand I-domain positions of the negation in (4) and (5) are both instantiated, we derive the doubling of Viano in (1ii), as illustrated in (6).
134╅ Sentential negation:€clitics (6)╅ Viano D a R n (C) R t
R n
e
I dɔ rm
e
e
The question now arises of why the pattern in (6) is restricted to the 2nd person singular. A connection can be established with another pattern which singles out 2nd person clitics in their interaction with the negation. As illus trated in (7) with Càsola, again a Northern Tuscan variety, invariable subject clitics precede the negation, as expected; on the other hand, inflected subject clitics split, in that only 2nd person clitics follow the negation, while 3rd per son clitics precede it. Notice that both Càsola in (7) and Viano in (1) lack inflected clitics for the 1st person. Therefore we may equally well describe the facts by saying that in Càsola in (7) the negation precedes all inflected 1st and 2nd person subjects and follows 3rd person subjects; similarly, in Viano in (1) doubling opposes all inflected 1st and 2nd person subjects to 3rd person sub jects. In other words, both types of language can be described as instantiating a classical person split between 1st /2nd person and 3rd person. (7)â•… Càsola (Tuscany) a n Dɔrm n t Dɔrm i/la nâ•… Dɔrm a n Durmiaŋ a n v Durmi i/la n Dɔrmn ClS Neg ClS sleep etc. ‘I don’t sleep’ etc.
The structures proposed for the negation clitic that follows the inflected subject clitic, as in the Vagli example in (4), or precedes it, as in the Sillano example in (5), suggest an account of the alternating pattern of a language
Interactions of negation with subject cliticsâ•… 135 like Càsola along the lines of (8). In (8a) the negation is inserted within the I domain and therefore follows the 3rd person subject clitic. In (8b), assuming that the position of 1st and 2nd person subject clitics remains constant, the neg ation precedes them in that it is inserted in the C domain. (8)â•… Càsola a. [D i/a [R n b.â•… [R n [C [D t
[I Dɔrm [I Dɔrm
In chapter 3 we analysed the interaction of the person split with the adverbial negation on the basis of the assumption that the so-called adver bial negation is neither adverbial nor, strictly speaking, negative; rather, negative adverbs are nominal elements interpreted as negative polarity items. Furthermore, we suggested a characterization of 1st and 2nd per son arguments as ‘discourse-anchored’ and 3rd person arguments as ‘eventÂ�anchored’. Now, the pattern in (8), far from appearing arbitrary, can be seen as a more abstract version of the Quarna split in chapter 3. In other words, the negation (here the clitic) has two different lexicalizations (here, posi tions) depending on the presence of a 1st/2nd or 3rd person argument (here the EPP argument).3 The Viano example in (6), repeated here in (9b), is a variant of the Càsola example in (8b). Thus, in (9b) a copy of the C-domain negation is inserted when the subject clitic is 1st or 2nd person. By contrast, the I-domain copy is the sole lexicalization of the negation with the 3rd person subject clitic in (9a). From another perspective, Càsola and Viano in (8)–(9) are sensitive to the same parameter between the I-domain and C-domain lexicalizations of the negation as Vagli and Sillano in (4)–(5), except that one of the two negations is generalized to the whole paradigm in (4)–(5), whereas they split according to person in (8)–(9). (9)â•… Viano a. [D i/la [R n [I Dɔrm b. [D aâ•… [R nâ•… [C [D t [R n [I Dɔrm
In the theory proposed by Poletto (2000), where data of the type in (7) are also considered, the relative order of subject clitics and the negation depends on a clitic hierarchy of the type described above, roughly ClS(invariable)€ – Neg€– ClS(3rd)€– Neg€– ClS(2nd)€– Neg. Under this account, the negation of Càsola as in (7) would be the middle one, since it precedes ClS(2nd) but follows ClS(3rd). This hierarchy could also correctly derive the doubling of Viano as in (1), where ClS(2nd) is flanked by two copies of Neg. Poletto’s (2000) model predicts the impossibility of the reverse order to the one found in Càsola, with
136â•… Sentential negation:€clitics the negative clitic preceding the 3rd person subject clitic and following the 1st and 2nd person clitic. This is because her clitic hierarchy, roughly ClS€– Neg€– ClS(3rd)€– Neg€– ClS(2nd)€– Neg, forces any negation that precedes ClS(3rd) to also precede ClS(2nd). However, when it comes to doubling, the same hier archy can yield the unattested string Neg€– ClS(3rd)€– Neg, assuming that the highest and lowest Negs of the hierarchy are instantiated. Here, starting with the two different positions of the negation in (8), we are able to predict the doubling in (9), as opposed to the reverse order. However, we have not yet explained why the negation and the subject clitic pattern as in (8), as opposed to the reverse€– since, contrary to Poletto (2000), the position of the negation with respect to 1st/2nd person clitics and 3rd person ones is not written into a hierarchy. In other words, the question is why the negation cannot be forced into the C domain by an event-anchored (3rd person) EPP argument, while remaining within the I domain in the presence of a discourseanchored (2nd person) argument. If we construe the negative clitic as a nom inal element and a negative polarity item, the generalization is that the presence of a discourse-anchored argument provokes the lexicalization of argumental material (the so-called negation clitic) in the C domain, thus indirectly linking discourse-anchoring and the C domain. On the contrary, event-anchoring is connected to the I domain.4 Independent evidence suggests that the connection just established is cor rect. Thus, in the imperatives of Southern Italian and Albanian varieties, dis course-anchored and event-anchored object clitics are found in the C and I domains respectively€– the former in mesoclisis (Manzini and Savoia (2009a), and references quoted there), and the latter in enclisis. The higher position of discourse-anchored elements with respect to event-anchored ones is also con firmed for non-related languages (for instance Salish, as discussed by Davis (1999)). This casts further doubt on Poletto’s (2000) hierarchy, in which the 3rd person subject clitic is higher than the 1st or 2nd person one. Crucially, the interactions of the negative polarity clitic with the person split require that it also participates in the definition of predicate–argument structure. If it was simply a logical connective, the reason for such an interaction would remain mysterious€– or at the very least could not fall under the generalization just proposed. Other interactions between the so-called negative clitic and subject clitics concern mutual exclusion phenomena which affect the combinations of sub ject and object clitics in Romance varieties. We observe that in some of these varieties the exclusion of the subject clitic (specifically 3rd person) is induced not only by object clitics, but also by the clitic negation. The data in (10a) show that Agliano has subject clitics that are obligatorily lexicalized in the absence
Interactions of negation with subject cliticsâ•… 137 of either negation or object clitics. The negation in (10b) excludes the subject clitic, with the possible exception of discourse-anchored clitics (in practice the 2nd person singular). The examples in (10c–c’) show that object clitics have the same effect of excluding the subject clitic€– though again they optionally combine with a discourse-anchored one. (10)â•… Agliano (Tuscany) a. (i) Drm tu Drmi i/la ddrma/Drma s Drma i Durmit i/la ddrmn/ Drmn ClS sleep ‘I sleep’â•… etc. b. nun drm nunâ•… (tu) drmi nun drma nun s drma nun durmit nun drmn Neg (ClS) sleep ‘I don’t sleep’â•… etc. c. (tu) l/la/ m cami youâ•… him/her/me call ‘You call him/her/me’ c’. l/â•… la/ɟi / m/ tâ•… camn/camat him/her/them/â•… me/us they.call/ you(pl).call ‘They/ you call him/her/them/me/us’
Mutual exclusions between subject and object clitics are similar to a mutual exclusion more often discussed in the literature on Romance, namely that between two object clitics€– for instance that between the dative and the accusative clitic in the Spurious se of Spanish. Current analyses in Distributed Morphology, Optimality Theory, etc. are based on the idea that there is a con straint against the co-occurrence of certain forms, and that one of them is even tually substituted for by a ‘default’. We propose a different model, which does not have recourse to notions of competition or default. Rather, only one of two clitic forms is lexicalized in that this form subsumes the crucial properties of the other (the non-lexicalized one). In cases of generalized exclusion of the subject clitic by the object clitic, as in (10c–c’), we propose that any pronominal clitic, independent of its denotation, is sufficient to lexicalize D(efiniteness) proper ties for the entire clitic string/domain and hence to exclude the subject clitic, a pure instantiation of such properties.5 The interest of data such as the Agliano
138â•… Sentential negation:€clitics examples is that the negative clitic in (10b) behaves exactly like an object clitic in excluding the subject clitic. This is expected if the so-called negation is nothing more than a nominal element, and specifically a negative polarity item. If the negative clitic was the lexicalization of the negative operator, it would be very difficult (or impossible) to see why it should interact with the argumental clitic series in the way it does. Another relevant property of Agliano is the different treatment of eventÂ�anchored and discourse-anchored subject clitics in (10b) and in (10c) vs. (10c’). As just proposed, the pure D(efiniteness) properties of an event-an chored (3rd person) subject clitic are subsumed by the lexicalization of any other element of the clitic string, yielding for instance (10c’); however, this is not necessarily the case for the deictic reference associated with speaker/ hearer, which can therefore be independently lexicalized, as for instance in (10c). Incidentally, since the negation precedes the 2nd person clitic when they co-occur in (10b), we conclude that Agliano has the same structure as Càsola in (8b). 4.2
Interactions of negation clitics with object clitics
In the examples presented so far, the negative clitic systematically precedes object clitics, as is usual in Italian varieties. However, cases in which the nega tive clitic appears inside the object clitic string are noted in the literature, in particular by Parry (1997) for Cairo Montenotte, in which the negative clitic precedes 3rd person objects but follows 1st and 2nd person ones. What is more, Parry (1997) observes that in some varieties of Liguria and Piedmont the nega tive clitic can appear both to the right and to the left of the 1st and 2nd per son clitic. Zanuttini (1997:€18) suggests that only languages with doubling of the negative clitic by an adverbial negation allow this doubling. In reality, the Northern Tuscan varieties that we exemplify below are a counterexample to this generalization. Let us begin by considering the relatively simple case of Bedizzano in (11), in which only one instance of the negation appears, following 1st and 2nd per son object clitics and preceding 3rd person ones. This positioning of the nega tive clitic within the object clitic string is insensitive to the person of the verbal paradigm, as can be seen from the comparison between 2nd and 3rd person in (ii) and (iii) respectively. (11)â•… Bedizzano (Tuscany) ii.â•… a.â•… t m n cam you me not call ‘You don’t call me’
Interactions of negation with object cliticsâ•… 139 b. c. d. iii. a. b. c.
t nâ•… l cam you not him call ‘You don’t call him’ t m n l/n da youâ•… me not it/of.itâ•… give ‘You don’t give me it/any of it’ t n i l da youâ•… notâ•… to.himâ•… it give ‘You don’t give it to him’ i tt/ss/vv n l/n da (pr ent) heâ•… you.sg/us/you.plâ•… notâ•… it/of.itâ•… givesâ•… at all ‘He doesn’t give it/any of it to you/us’ i n i da he not to.him it gives ‘He doesn’t give it to him’ i n i/s da kwEst he not to.him/to.us gives this ‘He doesn’t give this to him/us’
Doubling data are provided in (12) for Colonnata. The negative clitic Â� follows all subject clitics and precedes 3rd person object clitics, both accusa tive and dative (as in (iiib)). At the same time, the negative clitic both pre cedes and follows 1st and 2nd person clitics as well as the se-type clitic (as in (iiia)). (12)â•… Colonnata (Tuscany) i. a. a n t n we I not you not see ‘I don’t see you’ b. a n t n l ‘dag I not you not it give ‘I don’t give it to you’ c. a n l ve I not it see ‘I don’t see it’ ii. a. t n t n lav you not yourself not wash ‘You don’t wash yourself’ b. t n l ve you not it see ‘You don’t see it’ iii.â•… a. i n s n lav he not himself not washes ‘He doesn’t wash himself’ b. i n i l da he not to.him it gives ‘he doesn’t give it to him’
140â•… Sentential negation:€clitics c. i n t n l da he not you not it gives ‘He doesn’t give it to you’
We approach the doubling data of Colonnata in (12) assuming that the nega tive clitic is doubled because it is inserted both in the C and in the I domain; the possibility of a negative clitic being inserted in either domain has been independ ently motivated in section 4.1, and its doubling in both domains allowed us to account for the doubling on either side of a subject clitic. We can account for the Colonnata data if we assume additionally that 1st and 2nd person clitics (P in the present notation, to suggest Person) insert in the C domain, so that they precede the lower copy of the negation and follow the higher copy, as in (13). This ana lysis furthermore requires all subject clitics to be generated within the C domain. Note that we notate 3rd person accusative clitics as N (cf. here chapters 7–8). (13)
Colonnata
D i R n P t
e (C) R n
e N l I da
If we apply the approach in (13) to the non-doubling data of Bedizzano, we obtain structures like (14), which display the person split between 3rd person (N) clitics in the I domain and other clitics in the C domain, so that the former follow and the latter precede the negation in the I domain.
Interactions of negation with object cliticsâ•… 141 (14)
Bedizzano
i
D P t
e (C) R n
e N l/n I da
The question is what explanatory value the structures in (13)–(14) have. It appears that the relevant generalization is the same as that formulated in the conclusions of section 4.1, namely that event-anchored clitics are associated with the I domain, while discourse-anchored ones are associated with the C domain. In (13), the negation splits along the same lines as object clitics, namely one copy occurs in the I domain and one copy in the C domain. Note that in this analysis, all subject clitics are in the C domain€– we may take this to be due to their EPP nature, potentially extraneous to the event- vs. discourse-anchored contrast. Recall that even in the varieties of section 4.1, all inflected subject clit ics occurred in the same domain (the I domain in that case)€– and it was the neg ation that eventually split between two domains. In other words, the behaviour of the negation is closer to that of object clitics than that of subject clitics. In Southern Italian and Albanian varieties in which discourse-anchored and event-anchored object clitics occur in imperatives in the C and I domain respectively, locatives and the si-type clitic pattern with 1st and 2nd person. The same is true of se in the Bedizzano example in (11iiia). Descriptively speaking, this is the 3rd person reflexive, but in present terms it is the free variable of the argumental clitic system (cf. here chapters 5–6). Therefore the split must involve something like the present notion of discourse- and event-anchoring rather than just the opposition of 1st and 2nd person vs. 3rd person. It is important to recall that what we are seeking to analyse is finely struc tured parametric space. What we find is a set of universal categorial distinc tions. One such distinction is that between the C and the I domain. This interacts with the person split either directly, so that event-anchored complement clitics have an I lexicalization and discourse-anchored ones have a C lexicalization, or
142â•… Sentential negation:€clitics indirectly, so that the C negation associates with the discourse-anchoring of the EPP argument and the I negation with its event-anchoring. Despite the unpre dictable way in which the relevant distinctions show up in a given language, what we find is that the patterns they give rise to may not be reversible. Thus it is always discourse-anchored elements that are higher than event-Â�anchored ones€ – more precisely, that trigger the lexicalization of the C domain (for instance by the negative polarity clitic) as opposed to the lexicalization of the I domain in combination with event-anchored elements. The fact that negative clitic doubling interacts with LF interface notions such as that of person split excludes the possibility that it could be a purely morpho(phono)logical phenomenon. For instance, the copying and displace ment processes targeted by Distributed Morphology (Halle and Marantz 1993), or more recently by Harris and Halle’s (2005) theory of metathesis, can affect clitic material entirely comparable to negative n. In other words, since the negative clitic is clearly part of the clitic cluster, there is no reason why its copying (or displacement) should not be handled at the morpheme structure/ phonological form (MS/PF) interface as that of argumental clitics routinely is. However, if the preceding discussion is correct, the notions necessary to account for the present data are syntactic/LF notions, such as those of person split or I vs. C domain. To be more precise, it is possible to import the primi tives necessary to state the correct distribution of the negative clitic (1st/2nd vs. 3rd person, and so on) at the MS/PF interface. The problem, however, is one of explanatory adequacy. On the one hand, a theory in which the relevant notions are available in more than one component must surely count as more complex than a theory in which they are handled in a unified (morphosyntac tic) component. More importantly, if the interaction between the person split and the placement of the negation is to be understood in terms of the lexicali zation of the relevant clitics in I and C domains, then this essential aspect of our explanation cannot be reproduced at MS/PF at all. In our view, this type of consideration argues against any attempt to reduce negative doubling to an MS/PF process, and it also indirectly casts doubt on MS/PF accounts of pro nominal clitics€– whose parallels with negative clitics seem to call for a unified analysis. Next, in (15) we report data similar to those in (11), but taken from a variety (Càrcare) of the type studied by Parry (1997), with the negative clitic doub ling a negative adverb identical to the argument for ‘nothing’. It will be noted that the sentential negation adverb is in complementary distribution with other negative adverbs, as in (15iib–c), and arguments, as in (15iiib), and in fact appears to be altogether optional, as in (15iiic). Regarding this last example, we assume that (true) optionality simply reflects the presence of different
Interactions of negation with object cliticsâ•… 143 grammars in the competence of the same speaker. As for the question of nega tive concord, i.e. the possibility (or impossibility) of combining two or more given negative forms with a single negation meaning, we return to it in section 4.3€– in connection too with the more general problem of how to interpret a doubled clitic, or the doubling of phrasal material by a clitic. What is relevant here is that in (15) the negation is expressed by a triplet consisting of a negative adverb or phrasal argument doubled by a pair of clitics. The position of these is the same as that observed in (12); thus, they precede and follow P clitics and si-type clitics, in all cases preceding accusative/Â�partitive ones. Subject clitics generally precede the higher copy of the negative clitic. However, they can either follow or precede the 2nd person singular subject clitic, as in (15iib) and (15iic) respectively. What is more, these two potential positions of the negative clitic can combine, much as observed in (1); thus we obtain sentences of the type in (15iia), in which three copies of the negative clitic are present alongside the negative adverb. (15)â•… Càrcare (Liguria) ii. a. Eŋ t Eŋ t Eŋ lɔvi nE:nt notâ•… youâ•… notâ•… yourselfâ•… notâ•… washâ•… nothing ‘You don’t wash yourself’ b. Et Eŋ m Eŋ tʃɔmi mɔi you not me not call never ‘You never call me’ c. Eŋ t Em Eŋ lE dɔi mɔi not you me not it give never ‘You never give it to me’ iii. a. u ŋ s Eŋ lɔva nE:nt he not himself not washes nothing ‘He doesn’t wash himself’ b. u ŋ m Eŋ dɔ nE:nt he not me not gives nothing ‘He gives me nothing’ c. u ŋ m Eŋ lE/nuŋ dɔ he not me not it/of.it gives ‘He doesn’t give it/any of it to me’
We can extend to varieties of the Càrcare type the same treatment already proposed for Bedizzano or Colonnata. Specifically, the lower negative clitic appearing after the P object clitic can be lexicalized within the I domain, as in (16) below, while the object P clitic itself and the copy of the negation pre ceding the P clitic are found in the C domain. As we have seen, an interesting property of Càrcare is that the negative clitic can in fact be trebled, with its highest copy appearing in front of the subject clitic. Since the doubling of the negation on either side of the subject clitic is sensitive to a person split, as in
144â•… Sentential negation:€clitics Viano (discussed in section 4.1), it is natural to extend to the cases at hand the analysis proposed in section 4.1. The problem is that in Viano the discourseanchored subject clitic was in the I domain, so the copy of the negation that preceded it could be in the C domain. But since, for the reasons just reviewed, the discourse-anchored subject clitic is in the C domain in (16), the highest copy of the negation must be lexicalized in a still higher domain. More specif ically, we can assign it to the domain immediately above C in a split-C analysis of the type proposed by Rizzi (1997); this is CI in (16) below. (16)â•… Càrcare R εn (C1) D t
R εŋ
P t
(C) R εŋ
I l vi Q nεnt
c
It seems to us that there is no way of accounting for the data (or for their variation) in terms of a base-generated head-Spec configuration to which move ment subsequently applies. Apart from general concerns regarding the status of the head-Spec configuration (Starke 2004; Chomsky 2008), important coun terevidence is represented by the possibility of having more than one negative clitic, hence more than one potential head of the construction. It is true that doubling could be analysed as the result of multiple spell-out of the copies that a single clitic head leaves behind in the course of the derivation. However, the distribution of negative clitics is constrained exclusively by the distribution of other material in the clitic string. In no case is there any evidence that the sur face distribution depends on the presumed base-generated head-Spec config uration. The latter therefore represents an empirically unmotivated enrichment of the theory. Similarly, an explanation of doubling in terms of copying and
Interactions of negation with object cliticsâ•… 145 multiple spell-out presupposes that the placement of the clitic copies can be determined on the basis of general constraints on movement, and their morp holexical properties on the basis of PF rules. It seems to us that the burden of proof falls squarely on the proponents of such a theory€– on both counts. 4.2.1 Non-negative n A potential problem for the analysis of negative clitic doubling comes from the fact that in varieties of Liguria and Piedmont, P clitics can be followed by n morphology even in positive contexts. As it turns out, there are indications that in these contexts as well, the distribution of n is syntactically determined. Thus, in Dego in (17), the 1st person singular alternates between m if an accusative clitic is present and m-ŋ in accusative-less contexts. The alternation cannot be phonologically governed, since both the verb in (17a) and the accusative in (17b) are monosyllabic forms with an initial consonant. A similar contrast is noted by Parry (1997) for Rocca d’Arazzo. (17)â•… Dego (Liguria) a.â•… t m-Eŋ tmi you me call ‘You call me’ b. u m li/ a/ i d he me it-m./it-f./them gives ‘He gives it/them to me’
Furthermore, the n morphology that is in complementary distribution with the accusative in (17) appears after the P clitic in negative contexts, irrespect ive of the presence or absence of an accusative. Thus, in negative contexts, m followed by Eŋ combines with the accusative as well as with the partitive, as in (18). If we take Eŋ in these contexts to instantiate the negation, then this repro duces the distribution studied for the varieties of section 4.2 (without doub ling), for instance Bedizzano. (18)â•… Dego a.â•… u m/t Eŋ li d he me/you not it gives ‘He doesn’t give it to me/you’ b. u m  naŋ d ClS me not of.it gives ‘Nobody gives any of it to me’
nE:nt nothing niSyŋ nobody
Following the conclusions of the preceding section, we assume that the nega tive (polarity) clitic in Dego is inserted in the R position of the pronominal clitic string, as shown in (19a), where it yields no mutual exclusion effects. On
146â•… Sentential negation:€clitics the other hand, the mutual exclusion between the  segment and the accusa tive in the positive contexts in (17) can be accounted for by assigning it to the lowest position in the clitic string, otherwise occupied by the elements with which  is in complementary distribution, i.e. the accusative and partitive. In our notation this position is N, as in (19b). (19)â•… Dego a. D u P m/t (C) R εŋ N li I d�
Q nε:nt
b. D t P m (C) N εŋ I tʃ� mi
If the occurrences of the  morphology in the two contexts in (19) involve two different lexical entries, then not only their homophony but also their simi lar interaction with P clitics remain purely coincidental. Thus, negative  yields person split structures like (19a), where P clitics are inserted in the C domain and N clitics in the I domain. In the same way, positive  appears to be selected by P clitics. Assuming, then, that there is a single lexical entry for , its differing distribution and interpretation in negative and positive contexts need to be accounted for.
Interactions of negation with object cliticsâ•… 147 Before considering this question, we introduce evidence from the variety of Oviglio, in which the complementary distribution between n morphology and the accusative is observed in negative contexts. In Oviglio, what appears to be the higher copy of the negative clitic follows an already familiar pattern, appearing after subject clitics, except for the 2nd person one, which it pre cedes, as in (20). In turn, what appears to be a lower copy of the negative clitic, nu, is inserted after P clitics. However, this differs from the negative clitics considered in section 4.2 in that it is in complementary distribution with the accusative and partitive, as in (21). What is more, the data from, say, Bedizzano show that it is the lower negative clitic (the one inside the object string) that is obligatory in the absence of doubling. By contrast, in the Oviglio variety it is the higher n clitic that is obligatory in non-doubling examples. In all cases the negative clitic or clitic pair is obligatorily doubled by the sentential negation adverb næint(a) ‘nothing’. (20)â•… Oviglio (Piedmont) a n drm a n t drmi u n drm a n drumuma i n drmi i n drmu ClSâ•… not ClS sleep ‘I don’t sleep’â•… etc.
næinta næinta næinta næinta næinta næinta not
(21)â•… Oviglio i.â•… a.â•… a n t nuŋ tSam næinta I not you not call nothing ‘I don’t call you’ b. a n t El dag næint I not you it give nothing ‘I don’t give it to you’ ii. a. a n t um nu tSami næinta ClSâ•… not you me not call nothing ‘You don’t call me’ b. a n t um El dai næint ClS not you me it give nothing ‘You don’t give it to me’ t nu tSama næinta iii a. u n he not you not calls nothing ‘He doesn’t call you’ b. u n El tSama næinta he not him calls nothing ‘He doesn’t call him’
148╅ Sentential negation:€clitics The distribution of the n clitic is sensitive to the familiar person split, whereby it appears before P subject clitics but after 3rd person ones. Under the analysis of section 4.1, the position following 3rd person clitics reflects the lexicaliza tion of the negation clitic within the I domain, as in (22b). The position before the P subject clitic reflects the lexicalization of the negation in the C domain, as in (22a). In both cases the complementary distribution between nu and accusative clitics can be accounted for if nu is inserted in the N position. (22)╅ Oviglio a. D a
R n (C) D t
P um
N nuŋ I t ʃami Q næinta
b. D u
R n (C) P t
N nuŋ
I t ʃama Q næinta
A further twist on this pattern is that the nuŋ negative form has the same form as the partitive, as seen for instance in (23a). To be more precise, partitive nuŋ
Interactions of negation with object cliticsâ•… 149 has an n alternant appearing in front of (auxiliary) verbs beginning with a vowel, as in (23a’). Similarly but not identically, while nuŋ lexicalizes the lower neg ation in (23b–c) before a verb beginning with a consonant, it does not surface in front of (auxiliary) verbs beginning with a vowel, as in (23b’–c’). Incidentally, (23c) establishes that the middle-passive clitic si, despite its association with the 3rd person, behaves like a discourse-anchored clitic, followed by nuŋ. In other words, it is not a pure person split that is relevant, but a more abstract split, such as the one encoded here through the notions of discourse and event anchoring. (23)â•… Oviglio a.â•… u nu da doi a pr e he of.themâ•… gives two to each ‘He gives two of them to each one’ a’. aâ•… n o dahtS du a pr e I of.themâ•… have given two to each ‘I have given two of them to each one’ b. aâ•… m nu o næint lava I myself not am nothing washed ‘I haven’t washed myself’ b’. aâ•… n t o næint tSama I not you have nothing called ‘I haven’t called you’ c. u nu drɔm næinta it M not sleeps nothing ‘One doesn’t sleep’ c’. u E næint lava he not M is nothing washed ‘He hasn’t washed himself’
There are several reasons not to treat the formal identity of the negation and the partitive as a pure case of homophony. One of them is the complementary distribution between the negation and the partitive (more generally the N argu ment, including the accusative). Another reason is the otherwise unexpected restriction to contexts before verbs beginning with a consonant. Suppose then that we provide a unified lexical entry for nu. All of the evidence we have presented suggests that we should characterize it as the partitive. This charac terization is supported by the fact that€– in contrast to the varieties of sections 4.1€– it is the higher copy of the negation that is obligatory in the absence of doubling; thus, nu is not necessary for the negative interpretation of the sen tences in which it occurs. If nu is the partitive, its distribution in negative contexts can be accounted for by assuming that it must be lexicalized in the scope of the negation, if the internal argument is a discourse-anchored element. Indeed, in chapter 3 we
150â•… Sentential negation:€clitics observed that the negative adverb mia of Quarna selects for N objects€– lexi calized as partitives when they are clitics. Similarly, in Oviglio the negation requires the object to be an N€– inserted in the form of the partitive nu when otherwise only discourse-anchored elements are present. In non-negative con texts, nu is inserted under the ordinary circumstances that require a partitive form. On the basis of Oviglio, we can also return to the slightly more complex case of  in Dego. In negative contexts,  has the same form as the nega tive polarity clitic. Now, as discussed in more detail in section 4.3, a negative polarity item is essentially an indefinite, i.e. a free variable, which gets exist entially closed, and assumes its negative value by being interpreted in the scope of a negation (or other modal) operator. Suppose  is such an element. By definition, in positive contexts the negative operator is not present, and  in Dego is read existentially. Following the discussion of Oviglio, we fur ther assume that in positive contexts Dego must lexicalize  in N when the internal argument is discourse-anchored, simply because the lexicalization of the internal argument in this language requires N material€ – essentially as it does in Oviglio in negative contexts. Structurally, the negative reading of  appears to correlate with the R position and the positive reading with the N position. Evidently the negative reading requires  to be in a quantifi cational/left edge position, while N receives the ordinary internal argument interpretation. The various dimensions of language variation implicit in the accounts for Dego or Oviglio can also be shown to be realized independently of one another, which is an important argument in favour of the present analysis. Thus, P clitics ending in n morphology in non-negative contexts are found in varieties which, at least descriptively, do not have any negative clitics, as in S. Bartolomeo in (24). The distribution of n is syntactically determined by the presence vs. absence of an accusative, as shown in (24a) vs. (24b). The sentential negation, involving only a negative adverb, is exemplified in (24c). (24)â•… S. Bartolomeo Pesio (Piedmont) a.â•… u m-╅ d susi he me gives this ‘He gives me this’ b. u m lu d he me it gives ‘He gives it to me’ c. i lu tSam Je I him call nothing ‘I don’t call him’
Interactions of negation with object cliticsâ•… 151 The way we accounted for the complementary distribution of the  morph ology with accusatives in a variety like Dego was to say that  is inserted in N in sentences where the internal argument is otherwise lexicalized by a P element. Data like those of S. Bartolomeo indicate that this type of distri bution is independent of  expressing the negation. In other words, though the two interpretations can coexist in the same lexical item, as in Dego, one is independent of the other. Not only are there very many languages in which only the negative interpretation is attested, but there are also varieties like S. Bartolomeo in which only the non-negative one arises.6 Similarly, the formal identity of what are descriptively the partitive and the negation is not an isolated phenomenon. Thus, in Càrcare the partitive, illustrated in (15iiic), has the same form as the lexicalization of the negation in modal contexts, such as the negative imperative (morphosyntactically an infinitive) in (25). On the basis of the proposals advanced so far, it is natural to assume that the same lexical element, associated with the internal argument, is involved both in (25) and in the partitive. We may further speculate that when it is lexicalized in the I domain, as in (15iiic), its interpretation is partitive. In modal contexts, on the other hand, it must be lexicalized in one of the domains of the articulated C field, since it precedes the verb, which itself is in C or higher. In this position it will be read as a polarity specification, as in (25). (25)â•…
Càrcare nu St-lE a tSamE not stay him toâ•… call ‘Don’t call him’
In short, the presence of elements identical to what we have characterized as copies of the negation in positive contexts might at first suggest that they correspond to uninterpretable material, whose relevance is purely prosodic or computational. In reality, the data reviewed in this section, far from support ing this conclusion, provide evidence in favour of the idea that the negation is nominal (and argumental) in nature. Thus we account for the fact that what appears to be a lower copy of the negation clitic can have the same form as the partitive (Oviglio, Càrcare) or can lexicalize the same N slot as it does (Dego, S. Bartolomeo). If instead the negation copies corresponded to the mul tiple spell-out of the head of a specialized Neg category, their presence in nonnegative environments, and more to the point, their formal identity (lexical or distributional) with the partitive could not be accommodated. Similarly, a prosodic analysis of doubling could not account for the cases in which the socalled negation copy has the same form as a clearly contentful element, such
152╅ Sentential negation:€clitics as the partitive. In either case, all of the lexical coincidences examined in this section would have to be treated as cases of homophony, obscuring the patterns that connect them. 4.3
Negative concord and negative doubling
Our analyses so far motivate the conclusion that a negative clitic is a nom inal element, specifically a negative polarity item; hence it introduces not the negative operator, but simply a variable read within the scope of the negative (or other modal) operator. On this point we differ from much of the literature (Rizzi 1982; Longobardi 1992 on Italian), which takes the clitic to instantiate the negative operator. If the negative clitic is a negative polarity item, it is evi dent that it must itself be read in the scope of a sentential negation operator. The latter is therefore not introduced by any morpholexical constituent, but rather is semantically implied by the presence of the negative polarity clitic (or other negative polarity material). As mentioned in chapter 3, the most imme diate argument in favour of the polarity status of n-words in Romance (Rizzi 1982; Longobardi 1992; Acquaviva 1994 on Italian) comes from the fact that they occur in modal (irrealis) environments without any implication of nega tive meaning. Strikingly, the argument holds not only for n-phrasal units but also for n-clitics whose non-negative occurrences have been studied in the lit erature as instances of ‘spurious’ or ‘expletive’ negation (e.g. Belletti 2000).7 Let us now return to simple examples of the type in (2)–(3) or (7), in which the sentential negation is represented by a single negation clitic. Their LF includes a negative operator, i.e. ⌐. in whose scope the negative polarity item represented by the n-clitic is interpreted, as in (26). If we consider negative polarity items to be indefinites, i.e. free variables (Heim 1982), we may very well assume that the variable introduced by the so-called negative clitic is existentially closed (Acquaviva 1994) in the scope of ⌐. (26)â•… a.â•… b.
Vagli [⌐ [∃x [D i/ E [R nun (x) [I drmE Sillano [⌐ [∃x [R no (x) [D ll [I drm
Consider, then, the case routinely described in terms of the doubling of a clitic negation by an adverbial one, as illustrated here by the paradigm in (20), for instance. Since the so-called negative clitic has exactly the same status as any other polarity argument in the sentence, it is evident that each time it �co-occurs with one of them, the resulting configuration is one of negative concord under which all variables are read in the scope of a single negation
Negative concord and negative doublingâ•… 153 operator. Under the set of assumptions introduced here, negative concord is in fact the expected state of affairs. Quite simply, the variables introduced by the clitic and by the adverb are both interpreted in the scope of the same ⌐ and existential closure operators, as in (27). Thus we predict that there is a single instance of the negation at the interpretive level. (27)
Oviglio [⌐â•… [∃x,yâ•… [D aâ•… [R n (x)â•… [D tâ•… [I drmiâ•… [Q næinta (y)
In turn, the doubling of a negative clitic by another is identical in all relevant respects to its doubling by an adverb. Thus, consider (1) again, with simple negation in the 3rd person (1iii) alternating with the doubling of the clitic in the 2nd person (1ii). The two relevant LFs are provided in (28). (28a) contains a single variable, very much like (26), while (28b) is comparable to (27). In all cases the presence of a single negation operator returns a single negation (negative concord) reading. (28)â•… Viano a.â•… [⌐ [∃x [D i/ la [R n(x) [I drm b. [⌐â•… [∃x,yâ•… [D aâ•… [R n (x) [D t [R n (y) [I drm
Negative concord, though necessary to the interpretation of sentences like (27) or (28b), is not sufficient to derive it. There is another crucial component that enters into the reading for sentences like (27) or (28b), as opposed to the ordinary negative concord reading, which is that the two negations (whether they are both clitics or a clitic and an adverb) are understood as ‘doubling’ one another. In Manzini and Savoia (2007) we deal with doubling in connection with pronominal clitics. In particular, if doubling is not a morphophonological or computational (i.e. multiple spell-out) effect, then the fact that the differ ent instances of a doubled pronominal clitic express the same argument must depend on interpretation at the LF interface. In fact, the doubling interpret ation can be formalized through the notion of chain, which in representational models, in the sense of Brody (2003), is an LF primitive, and not a product of the derivation. Thus the theta-calculus at the LF interface will force all of the different instances of a doubled pronominal clitic (or of a clitic and its doubling full noun phrase) to be in a chain relation€– i.e. to fill the same argument slot. If the so-called negation is a nominal, argumental element, the same solution can be applied to the cases at hand. Specifically, the interactions of the sentential negation adverb with the internal argument of the verb reviewed in chapter 3 (the differing lexicaliza tion of the negation according to the person reference of the internal argu ment, the genitive of negation, and the ambiguity between the adverbial and
154â•… Sentential negation:€clitics argumental readings of ‘nothing’) motivated the assignment of the negation to the internal argument slot. There is a distinct tradition in linguistic studies which identifies the sentential negation with a negative quantification over the Davidsonian event argument of the sentence (Acquaviva 1994). However, we do not conceive of the event argument as an ordinary argument slot in the argument structure of the predicate (Higginbotham 1985), nor do we construe the so-called negation as a visible instantiation of the event argument. Rather, we have proposed that negative adverbs can introduce a variable restricted by the elementary subevent to which the external argument applies or by its internal argument, which is construed as the internal argument of the sentence as a whole. This line of thought is compatible with what we have said so far about the negative clitic, and is applicable to it. Pursuing this, the two negative polarity items in (27), i.e. the clitic and the adverb negation, or in (28b), i.e. the two clitic negations, introduce a variable restricted by the same argument€– i.e. the elementary event of the sentence, or its internal argument. This means that the two variables x and y in (27) and (28b) are effectively identified€– i.e. they enter into a chain(-like) relation. Summing up so far, the (rough) semantics that we have postulated for the so-called sentential negation delivers negative concord as a consequence, without any need for further assumptions. In turn, the latter is a prerequisite for the chain(-like) reading which underlies phenomena usually described in terms of doubling. In the present approach there is no copying (either morpho phonological or syntactic), but rather independent insertion of lexical items€– which are then identified by the argument calculus at the LF interface. The two approaches are equivalent in the interpretive component, as far as we can tell, since the chain relation, whether derived through movement or primitive, holds at the LF interface. The reason why the present analysis is to be preferred is the one we have tried to advance throughout this chapter€– in short, it remains to be shown that the complex interactions of doubling with interpretation-based notions such as that of person split etc. can be captured under alternative mod els. What is more, our argument is that when these interactions are dealt with at the level at which they belong, i.e. the LF interface, morpholexical-level facts can also be handled in a more revealing fashion€– specifically, without recourse to systematic homonymy. It is not any one fact that is compelling in our view, but rather the accumulation of apparently unrelated facts all pointing in the same direction. For the sake of completeness, we also consider briefly how the present approach can account not only for negative concord but also for its impossibility,
Negative concord and negative doublingâ•… 155 i.e. for cases where two n-words cannot combine under a negative concord reading. These cases are potentially problematic to the extent that, as argued in relation to (27)–(28), negative concord is the state of affairs normally pre dicted by the present theory. Specifically, negation clitics generally combine with other negative elements; however, negation adverbs are often in comple mentary distribution with negative arguments and adverbs, as has been noted in passing for Càrcare in (15ii). Very much the same conclusion can be drawn with respect to the 2nd person singular of Oviglio in (20), in which ‘never’ is in complementary distribution with the ‘nothing’-type sentential negation adverb present in all other persons. The first relevant observation is that this pattern is by no means necessary. For ease of exemplification, we stick to the ‘nothing’-type sentential negation and the combination of this negation with argumental ‘nothing’ and ‘nobody’. In (29) we report data from the Mezzenile variety, in which the ‘nothing’-type negation can always combine with a negative argument, whether this is ‘noth ing’ and adjacent to the ‘nothing’-negation, as in (29a), or ‘nobody’ and not adjacent to the ‘nothing’-negation, as in (29b). (29)â•… Mezzenile (Piedmont) a.â•… u fi (Jint) ï†−Ente he does not nothing ‘He doesn’t do anything’ b. u j nt Jint tSama JyN theyâ•… Locâ•… haveâ•… not called nobody ‘The haven’t called anybody’
In other varieties, more restrictive patterns are observed. Thus, there are lan guages in which the combination of the ‘nothing’-type negation with the nega tive argument ‘nobody’ is possible, as in (30a), though the combination with ‘nothing’, as in (30b), is not attested. This in turn holds when the two elements are, or would be, adjacent. Once adjacency is removed, all combinations are possible, as in (30a’–b’). (30)â•… Fontane (Piedmont) a. u i vEN Jent Jy it Loc comes not nobody ‘There doesn’t come anybody’ b. e mdZu Jent they eat nothing ‘They eat nothing’ a’. a Jent vŋgy Jyŋ it Loc has not come nobody ‘There hasn’t come anybody’
156â•… Sentential negation:€clitics b’. j aN Jent madʒa Jent they have not eaten nothing ‘They haven’t eaten anything’
Other data point to a contrast based purely on adjacency. Thus, in several varieties the sentential negation adverb is in complementary distribution with a negative argument in simple tenses, but not in perfect tenses. Incidentally, because this pattern is relatively frequent we can document it easily both in languages which have only a sentential negation adverb, like S. Bartolomeo, and in languages, like Dego, which have a negation clitic as well. The rele vant contrast in S. Bartolomeo is between (31a–b) and (31a’–b’), and in Dego between (32a) and (32b). (32a’) from Dego provides a concrete illustration of the pattern not found in these languages.8 (31)â•… S. Bartolomeo Pesio a. i t:m JyN I call nobody ‘I don’t call anybody’ b. i mï†−dZ Jente I eat nothing ‘I don’t eat anything’ a’. i JeN tSamQ Jy I have not called nobody ‘I haven’t called anybody’ b’. i JeN maï†−dï†ıâ•… Jente I have not eaten nothing ‘I haven’t eaten anything’ (32) Dego a. i J maJduâ•… JEnte theyâ•… not eat nothing ‘They eat nothing’ a’. *i J maJdu nEnt JEnte they not eat not nothing ‘They eat nothing’ b. i n aJ nEnt maï†−d JEnte they not haveâ•… not eaten nothing ‘They have eaten nothing’
Let us consider, then, how the present theory can account for Dego (32a’). Recall that the doubling of a negative clitic by another negative clitic or adverb has been treated as akin to the doubling of a pronominal clitic by another clitic or a noun phrase. If so, the mutual exclusion of the sentential negation adverb and the negative polarity argument in Dego (32a’) is comparable to a mutual
Negative concord and negative doublingâ•… 157 exclusion between pronominal clitics€ – perhaps most famously the dative– accusative mutual exclusion that gives rise to the Spurious se phenomenon in Spanish. Therefore, we propose to account for Dego (32a’) and the like along the lines suggested by Manzini and Savoia (2007) for mutual exclusions between clitics. In particular, we propose that inserting a negative polarity argument associated with the internal argument, such as nte in Dego (32a’), is sufficient to satisfy all of the properties that would otherwise be lexicalized by the sentential negation adverb, effectively excluding its insertion. This pro posal is given theoretical content by our conclusion in chapter 3 that a senten tial negation introduces a variable restricted either by the internal argument of the sentence or by the elementary event that includes it. From this perspective, the negative polarity argument, i.e. nte, subsumes the contribution of the negation adverb to the interpretation of the sentence because it already intro duces a variable corresponding the internal argument of the sentence. As we noted, in languages like (31) and (32), the mutual exclusion holds only when the negation and the negative argument are within the same predica tive/eventive domain, for it is perfectly possible to have the same properties lexicalized twice in two different domains, e.g. by the clitic negation in the inflectional domain and by the negative polarity argument in the predicative domain in Dego (32a). Similarly, according to the conclusions of chapter 3, a negative adverb preceding the participle is inserted in a high C-type domain within the participial sentence, and is thus predicted to combine with a postparticipial negative argument inserted in the predicative/eventive domain, as in Dego (32b). We also expect that the absence of mutual exclusion between elements in different domains of the sentences (as opposed to those in the same domain) has a counterpart with pronominal clitics. In fact, mutual exclusion between subject and object clitics is obviated in interrogatives, where the sub ject clitic remains in the postverbal (enclitic) domain while object clitics are found in the preverbal domain (as proclitics). We further expect that, along with languages of the type in (31), there might be languages like (29) that have no mutual exclusion€– pretty much because mutual exclusion between clitics is a strictly language-specific phenomenon. More to the point, we predict that, given the account we have provided for it, mutual exclusion between sentential negations and negative arguments is constrained by lexical properties. Thus, it is expected that there should be lan guages like Fontane in (30) in which the mutual exclusion is only between the sentential negation adverb and the ‘nothing’ argument, while the ‘nobody’ argument (endowed with a human restriction) does not necessarily subsume the adverb.
158â•… Sentential negation:€clitics Delving further into the patterns of mutual exclusion is beyond the scope of this chapter, since it would require us to introduce a considerable amount of additional data. Before concluding, however, it is worth noting that if the dis tribution of sentential negation adverbs with respect to negative polarity argu ments is sensitive to their domains of insertion, as in (31) and (32), we expect a similar effect to be observable with negative clitics. Indeed, complementary distribution with negative arguments is found in many Italian varieties, includ ing the standard language, in the configuration in which a negative argument would precede the clitic, as in the case of a preverbal subject in (33a). This phenomenon is again parametrized; thus, the co-occurrence of the negation clitic with a negative polarity preverbal subject is attested in Old Italian texts, as noted by Meyer-Lübke (1899:€§695), who quotes the example in (33b). In (33a) we can apply essentially the same analysis proposed for Dego (32a’). We assume that the position of the preverbal subject in null subject and clitic subject languages is in the C field (cf. here chapter 1). Inserting the negative polarity subject in the C domain or higher subsumes all modal properties other wise lexicalized by the negation clitic. Therefore the clitic need not, and must not, be inserted.9 (33)â•… a.â•… b.
Nessunoâ•… (*non)â•… dorme nobody not sleeps ‘Nobody sleeps’ gente neuna non v’ arrivava (Novellino 55) people none not there arrived ‘Nobody arrived there’
It is evident that the line of reasoning deployed so far can in principle be extended to mutual incompatibilities between any two negative polarity items, including two arguments, two adverbs or an argument and an adverb. Thus, we can say that the lexicalization of one subsumes (relevant properties of) the lexicalization of the other (within a given domain) and hence excludes it (i.e. a re-lexicalization of its relevant properties). In any event, the general interpret ive mechanism of negative concord laid out in (26)–(28) remains in place.
5 The middle-passive voice:€evidence from Albanian
Fifty years ago, Chomsky (1957:€§5.4) argued that passive sentences should be excluded from phrase structure grammar and introduced instead by a transformational rule applying to active sentences. This was because introducing passives through rewriting rules would mean doubling the selectional restrictions independently imposed on actives, while a transformational rule would allow them to be stated only once. Chomsky (1965:€103–4) provides what has remained the standard conceptualization of this transformational process by proposing that the Manner Adverbial should have as one of its realizations a ‘dummy element’ signifying that the passive transformation must obligatorily apply. That is, we … may formulate the passive transformation … with an elementary transformation that substitutes the first NP for the dummy element passive and places the second NP in the position of the first NP.
In current practice, the by-phrase is independently generated by Merge; but the analysis whereby passive is defined by ‘substitution’ of an internal argument for the EPP position (second or internal Merge) remains at the core of generative transformational grammar. In this chapter we propose to evaluate this analysis in the light of data from Albanian, which provides two separate and complementary phenomena of interest. On the one hand, the passive (i.e. promotion of the internal argument to the EPP position with the external argument independently interpreted) has the same lexicalization as the reflexive, the anticausative and the impersonal. The question then is whether all of these different interpretations are associated with the same underlying syntax. On the other hand, there is no single lexicalization of this cluster of meanings; rather, it varies according to tense and aspect specifications. Thus, in standard Albanian the middle-passive voice is lexicalized by a specialized (agreement) inflection in the present or imperfective past, by the clitic u combined with the active forms of the verb in the perfective past, and by the periphrasis be–participle in the present perfect and in the pluperfect. The question in this respect is whether (in the passive and in 159
160â•… The middle-passive in Albanian the other interpretations) there is a single underlying syntax corresponding to all of these different morpholexical formats. The discussion in this chapter also provides a background for the discussion of the functional category ‘auxiliary’ in chapter 6, since the notion of voice (together with notions of transitivity and person split) crucially enters into the determination of auxiliary selection patterns. Correspondingly, after introducing the Albanian data in section 5.1, we devote section 5.2 to the clitic lexicalization of middle-passive voice and section 5.3 to the inflectional lexicalization; however, we leave the auxiliary–participle construction till chapter 6. In section 5.3.1 we briefly examine the possibility, also attested in Albanian, of lexicalizing the passive interpretation (in all tenses) with a be–adjectival participle construction. 5.1
Data
5.1.1 Middle-passive morphologies In the present indicative, Albanian has a specialized inflection for the nonactive voice, as exemplified in (1) for Gjirokastër (a Tosk variety, essentially like the standard) and for Shkodër (a Gheg variety). In (1a) we illustrate verbal bases ending in a vowel, while in (1b) we exemplify verbal bases ending in a consonant; the comparison with the active is provided in (1a’) and (1b’) respectively. The most complex instantiation of the middle-passive morphology can be seen in the vocalic bases of Shkodër, which are followed by the h affix, which is in turn followed by inflections for person. In the consonantal bases, as well as in a possible pronunciation of the vocalic bases in Gjirokastër, the affix preceding the person inflections is simply (Trommer 2005). The comparison with the active present allows us to establish that the person inflections are themselves specialized for the middle-passive voice€– very clearly so in the singular, where the active and middle-passive forms bear no relation to one another. (1)â•… Gjirokastër a.â•… la- (h)- m/ S/â•… t/â•… mi/â•… ni/â•… n wash- MP- 1sgâ•… etc. ‘I washed myself’ etc. a’. la- i/ n/â•… n/â•… im@/â•… ni/â•… in@ wash- 1sg etc. ‘I wash (something)’â•… etc. b. viS- - m/ S/ t/â•… mi/â•… ni/â•… n@ dress- MP- 1sg etc. ‘I dress (myself)’ etc.
Dataâ•… 161 b’. vS vS vS vS- im vS- ni vS- in dress- 1sgâ•… etc. ‘I dress (somebody)’ etc. Shkodër a. l- - m/ S/ t/ na/ ni/ n wash- MP- 1sg etc. ‘I wash myself’ etc. a’. l- i/ n/ n/ im/ ni/ in wash- 1sg etc. ‘I wash (something)’ etc. b. veS- - m/ S/ t/ na/ ni/ n dress- MP- 1sg etc. ‘I dress (myself)’ etc. b’. veS- i ve:S ve:S veS- im veS- ni veS- in dress- 1sgâ•… etc. ‘I dress (somebody)’â•… etc.
In the simple past, Albanian resorts to a different morphosyntax for the formation of the middle-passive voice, preposing the clitic u to the verb, as illustrated in (2). The clitic can be taken to correspond roughly to Romance se; it is associated with all the different forms of the paradigm, as is also the case in some Romance varieties (in particular Romansh ones) for se. As for the morphology of the verb, no specialized middle-passive affix is present; furthermore, the person inflections are identical to those of the active, except for the 3rd person singular, whose active form is provided in (2a’) and (2b’). Even there, the middle-passive voice is characterized simply by the omission of the inflection present in the active paradigm, not by a different inflection. We interpret affixes like€–it in the vocalic paradigm of Gjirokastër or€–v in the vocalic paradigm of Shkodër as connected to the expression of the perfective past. (2)â•… Gjirokastër a.â•… u la- it- a/ / -/â•… @m/â•… @t/â•… @n MP wash- past- 1sg etc. ‘I washed myself’ etc.
162â•… The middle-passive in Albanian a’. la- it- i it wash- past- 3sg ‘He washed it’ b. u vS- a/ /â•… -/â•… @m@/â•… @t/â•… @n@ MP dress- 1sg etc. ‘I dressed (myself)’ etc. b’.â•… vS- i him dress- 3sg ‘He dressed him’ Shkodër a. u l- v- a u l- v- u l: u l- m u l:- t u l- nE MPâ•… wash-â•… past-â•… 1sgâ•… etc. ‘I washed myself’ etc. a’. E la- u itâ•… wash-â•… 3sg ‘He washed it’ b. u veS- a u veS- u vS u veS- m u veS- t u veS- n MPâ•… dress- 1sg etc. ‘I dressed (myself)’ etc. b’. veS- i itâ•… dress-â•… 3sg ‘He dressed him’
The perfective past in (2) differs from the present in (1) both in temporal properties and in aspectual ones, under the natural assumption that the present is essentially an imperfective form. Therefore the lexicalization of the Â�middle-passive voice could in principle be sensitive to tense or to aspect. In Tosk varieties, the imperfective past follows the pattern of the present, with specialized middle-passive morphology, as in (3); thus, the split between present and imperfective past on the one hand and perfective past on the other appears to be based on aspect. The morphological analysis of the verb shows that, as in the present, the middle-passive voice is carried by the affix€–, which is followed by a€–S morpheme carrying the past specification; the latter is specialized for the middle-passive, as can be seen by comparing it with the active.
Dataâ•… 163 (3)â•… Gjirokastër a.â•… la- (h)-â•… S- a/ E/ -/â•… im/â•… it/â•… in wash-â•… MP- past- 1sgâ•… etc. ‘I washed myself’ etc. a’. E la- j- a j n- tE n- im n- it n- in itâ•… wash- past- 1sgâ•… etc ‘I washed it’â•… etc. b. viS- - S- a/ E/ -/â•… im/â•… it/â•… in dress- MP- past- 1sg etc. ‘I dressed (myself)’ etc. b’. viS- j- a j- t n- im n- it n- in dress-â•… past-â•… 1sgâ•… etc. ‘I dressed (somebody)’ etc.
On the other hand, in Gheg varieties, the specialized morphology and clitic realizations of the middle-passive voice split according to tense; thus, while the present has specialized middle-passive morphology, not only the perfective past but also the imperfective past in (4) have the u clitic. The data in (4) show that substituting an accusative clitic for the u middle-passive voice clitic yields the active reading with no change in verb morphology. In other words, what the u clitic combines with is the ordinary active morphology of the verb. (4)â•… Shkodër a. u / l- ï†fi-â•… a ï†fi-  t ï†fi- im ï†fi- it ï†fi- in MPâ•… /itâ•… wash-â•… past. impf-â•… 1sgâ•… etc. ‘I.washed myself/ it’â•… etc. b. u/ veS- S- a ï†fi-  t ï†fi- im
164â•… The middle-passive in Albanian ï†fi- it ï†fi- in MPâ•… /himâ•… dress- past.impf-â•… 1sgâ•… etc. ‘I dressed (myself)/ it’â•… etc.
The middle-passive conjugation in Albanian also includes forms consisting of an auxiliary followed by the participle, as illustrated in (5) for the present perfect. In particular, the auxiliary jam ‘I am’ followed by the participle is sufficient to yield the middle-passive voice. The data for comparison in (5a’) and (5b’) show that the active is formed with the same participle but with the kam ‘I have’ auxiliary. Thus, in this case it is the switch from kam ‘I have’ to jam ‘I am’ that yields the switch from active to middle-passive voice. Similarly, the pluperfect is formed with the imperfective past of the two auxiliaries followed by the participle. As for the morphology of the participle, Tosk varieties (including the standard) have a participial ending –r, which is not present in Gheg varieties; it is worth noting that the vocalic bases in the Gjirokastër examples in (5a–a’) also include the perfective€–it morphology. (5)â•… Gjirokastër a.â•… @St la- it- ur he.isâ•… wash-â•… prf-â•… prt ‘He has washed himself’ a’. ka la- it- ur itâ•… he.hasâ•… wash-â•… prf-â•… prt ‘He has washed it’ b. @St vS- ur he.isâ•… dress-â•… prt ‘He has dressed (himself)’ b’. ka vS- ur him he.has dress- prt ‘He has washed him’ Shkodër a. ï†fi l:/ ve:S he.is washed/ dressed ‘He has washed/dressed (himself)’ b. k l:/ ve:S him he.has washed/ dressed ‘He has washed/ dressed him’
5.1.2 The interpretation of the middle-passive morphologies In the preceding discussion, we have illustrated the three basic morphologies for middle-passive voice in Albanian. In each case we have chosen to illustrate the middle-passive voice with verbs which make the reflexive interpretation
Dataâ•… 165 particularly salient, and we have glossed our examples accordingly. In reality, each of the forms that we have exemplified is multiply ambiguous, allowing for a range of meanings that is independently attested, for instance, for the Romance counterpart of the u clitic, e.g. Italian si. In what follows, we review the various meanings, showing that they are associated with all morphological instantiations of the middle-passive voice. The reflexive reading prominent with a verb like ‘to wash’, for instance, implies a single participant in the event, which is both its theme (patient, etc.) and its causer (agent, etc.). This can easily be distinguished from another reading which equally involves a single participant in the event€– which we shall refer to as anticausative. This is the reading where the single participant is the theme (patient, etc.) and no external agency (cause, etc.) is expressed or implied in the event. This is evidently a salient meaning for the middle-passive predicate in (6), which we correspondingly glossed as ‘to wake up’. Of course, although ‘to wake oneself up’ is also a possible predicate, the reading is less salient for pragmatic reasons. What is important to note is that the anticausative reading, like the reflexive reading in the previous section, is associated with all lexicalizations of the middle-passive voice, namely the specialized inflection in the present (6a), the clitic in the perfective past (6c) and the jam–(perfect) participle formation in the present perfect (6d); the imperfective past (6b) has the specialized inflection or the clitic, depending on the variety. (6)â•… Gjirokastër a. zu- (h)-â•… t wake-â•… MP- 3sg ‘He wakes up’ b. zu- (h)- S wake-â•… MP-â•… past ‘He woke up’ c. u z- it MP wake- prf ‘He woke up’ d. ï†fit zu- a he.is wake- prt ‘He has woken up’ Shkodër a. tSo- h- t wake MP- 3sg ‘He wakes up’ b. u tSo- t MP wake- 3sg ‘He woke up’
166â•… The middle-passive in Albanian c. d.
u tSu: MP woke ‘He woke up’ ɐï†fit tSu: he.is woken ‘He has woken up’
The reading of the middle-passive voice that implies two participants in an event, including the theme (patient, etc.) and an external argument (agent, cause, etc.), is what is known as the passive. Again, the passive meaning is available independently of the particular morphology instantiating the middlepassive voice, as illustrated in (7). In these examples, it is really the by-phrase that distinguishes the passive from the other possible readings. Of course, the passive reading implies an agent, a so-called implicit argument, even when no by-phrase is lexicalized. Here we exemplify the present (7a), the perfective past (7c) and the present perfect (7d); the imperfective past reflects the morphology of the present (Tosk) or of the perfective past (Gheg). (7)â•… Gjirokastër a.â•… kt kmiï†fi la- (h)- n ga aj these shirts wash- MP- 3pl by him ‘These shirts are washed by him’ c. ata u z- it- @n ga t@ tir@t they MP wake- prf- 3pl by the others ‘They were woken by some people’ d. kt kmiï†fiâ•… jan la- it- u ga aj these shirts are wash- prf- prt by him ‘These shirts have been washed by him’ Shkodër a. fmia veS- - t pei nns the.childâ•… dress- MP- 3sg by the mother ‘The child is dressed by his mother’ a’. at tï†fi- h- n Q tiErt they wake- MP- 3pl by the others ‘They are woken up by some people’ d. jan tSu Q tiErt they.are woken by the others ‘They have been woken up by some people’ d’. ɐï†fit l: pei nɐns he.is washed by the mother ‘He has been washed by his mother’
It comes as no surprise that middle-passive voice morphology can attach to unergative verbs in Albanian, since these are construed by current theories
Dataâ•… 167 (Hale and Keyser 1993) as concealed transitives, where the verb effectively incorporates an object (cf. also the discussion in chapter 3). Therefore we may expect that the combination of the middle-passive morphology with such a verb will yield an impersonal meaning, essentially as a by-product of passivization, as in (8). Note, however, that in an impersonal passive we would expect it to be possible to lexicalize the external argument independently with a by-phrase. However, this does not seem to be possible in Albanian. (8)â•… Gjirokastër a.â•… ati fl- (h)- t mi there sleep- MP- 3sg well ‘There one sleeps well’ c. ati u fit mi there MPâ•… sleptâ•… well ‘There one slept well’ Shkodër a. atj fl- (h)- t mi there sleep- MP- 3sg well ‘There one sleeps well’ b. atjE u flE- tE mi there MP sleep 3sg well ‘There one slept well’ c. ktu u fje:t mi hereâ•… MPâ•… slept-3sgâ•… well ‘Here one slept well’ d. atjE ɐï†fit fie:t mi (*N ata) there it.isâ•… sleptâ•… well (by them) ‘There one has slept well’
More importantly, the middle-passive voice can also attach to unaccusative predicates€ – i.e. intransitive predicates once again, which cannot reasonably be construed as concealed transitives. Rather, the only argument of such predicates corresponds to their theme, e.g. the element that undergoes the change of location with motion verbs such as ‘to go’, exemplified in (9). With these verbs, therefore, the middle-passive voice cannot be analysed as yielding a sort of passive, albeit an impersonal one. Rather, it yields an impersonal tout court, which must then be entered among the possible interpretations of the middle-passive voice in Albanian; this is confirmed by the impossibility of associating such structures with a by-phrase. Once again, there is a parallelism with Italian si; however, it must be emphasized that the impersonal interpretation in Albanian is not restricted to the clitic morphology, but is equally found with specialized inflection or with jam–participle formations.
168â•… The middle-passive in Albanian (9)â•… Gjirokastër a. ga ati dil- - t from there exit- MP- 3sg ‘One exits from there’ a’. ai del he exits ‘He exits’ b. ga ati u dD from there MP exited ‘One exited from there’ b’. ai dD- i he exited- 3sg ‘He exited’ c. ga ati ɐï†fit dal mi from there it.is exited well ‘One has exited well from there’ c’. ka dal he.hasâ•… gone ‘He has exited’ Shkodër a. pej ktj dl- - t throughâ•… thereâ•… go.out-MP- 3sg ‘One goes out that way’ a’. ai dl heâ•… goes.out ‘He goes out’ b. pej ktj u dl- t through there MP go.out- 3sg ‘One went out that way’ b’. ai dl- t heâ•… go.out-3sg ‘He went out’ c. pej ktj u do:l through there MP went.out ‘One went out that way’ c’. ai dol- i heâ•… went.out- 3sg ‘He went out’ d. pej ktj ɐï†fit d:l mi through there it.is gone.out well ‘One has gone out well that way’ d’. kæ d:l he.has gone.out ‘He has gone out’
(* ata)
Dataâ•… 169 5.1.3 The Arbëresh varieties The Arbëresh varieties of Albanian, spoken in Southern Italy, belong to the Tosk group, and indeed reflect the conditions of the standard (or of Gjirokastër here) in many respects. In particular, with vocalic bases, these varieties lexicalize the middle-passive voice with the specialized verb inflection in the present indicative and in the imperfective past, as illustrated respectively in (10a) and (10b) from Portocannone, in which the middle-passive morphology is€–x; the perfective past has the clitic u, as in (10c). Microvariation is present at various points between the mainland varieties and Arbëresh, as well as within the Arbëresh fold. We note in particular that in Portocannone the perfective past, despite the presence of the u middle-passive clitic, maintains the middleÂ�passive affix€–x– of the present and imperfective past. As for the person inflections, it is not only the 3rd singular that distinguishes active and middle-passive in the perfective past, but also the 1st person singular. (10)â•… Portocannone (Molise) a.â•… la- xE- m/ S/ t/ mi/ ni/ n wash-â•… MP-â•… 1sgâ•… etc. ‘I wash myself’ b. la- x- S- a/ E/ i/ m/ t/ n wash- MP- impf- 1sg/ etc. ‘I washed myself’ c. u la- x- tS/ E/ -/ m/ t/ n MP wash- MP- 1sg etc. ‘I washed myself’ etc. c’. E la- v- a/ E/ i/ m/ t/ n it wash- past- 1sg etc. ‘I washed it’ etc.
A major point of variation between the middle-passive voice of Arbëresh varieties and that of mainland varieties concerns auxiliary–participle formations in the perfect. In Arbëresh varieties these involve the auxiliary kam ‘I have’, exactly as in the active, rather than jam ‘I am’; therefore middle-passive voice is lexicalized by the u clitic. The Portocannone variety displays an interesting further parameter concerning participial morphology. In the active voice in (11b) the participle has recognizably the same form as in the Gjirokastër example in (5), with the verbal base la- followed by the perfective morphology€–it and the participial ending –ur. By contrast, in the non-active voice in (11a), the participle is formed through suffixation of the middlepassive morpheme€–x, followed by the ordinary participial ending€–ur. Thus, in Portocannone and similar varieties the vocalic verb bases bear specialized
170â•… The middle-passive in Albanian morphology throughout the paradigm, even when a u clitic is present, as in the pluperfect in (11), but also in the perfective past in (10c). (11)â•… Portocannone a. at kiï†fin u la- x- they had MP wash- MP- ‘They had washed themselves’ b. at kiï†fin E la- it- they had it wash- prf- ‘They had washed it’
ur prt ur prt
With this morphological background, we are now in a position to consider the readings that are associated with the various forms. In the examples that we provided above, it is of course the reflexive reading that is salient. The possibility of what we have called the anticausative reading is evident in the examples in (12). In the case of auxiliary–participle formations, we provide a comparison of the middle-passive (12d) with the active (12d’)€ – which displays the difference between the two participial morphologies. A further point of variation between Portocannone and other varieties (both mainland and Arbëresh) emerges in the data in (12d)–(12d’), namely that the participle can be introduced by a coordinating/subordinating particle, literally ‘and’. This parameter is essentially irrelevant here (but cf. chapter 6). Another property which singles Portocannone out (and which is largely irrelevant for present purposes) is that the clitic is not positioned before the auxiliary, but immediately before the participle, even in the absence of the particle, as can be seen in (11). (12)â•… Portocannone a.â•… z- - Em wake- MP- 1sg ‘I wake up’ b. z- x- S- a wake- MP- past- 1sg ‘I woke up’ c. u z- tS MPâ•… wake- 1sg ‘I woke up’ d. ai kiS E u tSa- it had and MP break- ‘It had broken’ d’. ai kiS E E tSa- he had and it break- ‘He had broken it’
x- ur MP- prt it- ur prf- prt
Dataâ•… 171 Next, the impersonal meaning is available both with unergative predicates and with unaccusative ones, exemplified here in (13). As usual we provide the contrast between the middle-passive participle formation in (13d) and the active one in (13d’). (13)â•… Portocannone a. ktu vE-â•… xE- t te ha hereâ•… go- MP-â•… 3sgâ•… to the village ‘This way one goes to the village’ c. ktu u va- x te ha here MP go- MP to the village ‘This way one went to the village’ d. ktu kiS u va- x- ur te ha here it.had MP go- MP- prt to the village ‘This way one had gone to the village’ d’.â•… kiS va- t- ur he.hasâ•… go-â•… prf-â•… prt ‘He had gone’
As we fully expect, the range of morphologies that we have considered so far can be associated with a passive reading€ – i.e. a reading characterized, like the transitive one, by the presence of two roles and two event participants, except that of course the theme is found in the EPP position. Relevant examples are provided in (14). It should be noted that, while in mainland Albanian by-phrases normally co-occur with the middle-passive morphology, as illustrated in (7), the MP-passives of Arbëresh are normally impersonal, in the sense that they only allow for an impersonal (i.e. generic) reading of the agent€– that is, they do not normally combine with the by-phrase. (14)â•… Portocannone a. atiE la- xE- n kmiS-t here wash- MP- 3pl shirt-the.pl ‘Here shirts are washed’ c. atiE u la- x- n kmiS-t here MP wash- MP- 3pl shirt-the.pl ‘Here the shirts were washed’ d. atiE kiï†fin u la- x- ur kmiS-t here had MP wash- MP- prt shirt-the.pl ‘Here the shirts had been washed’ d’.â•… kiï†fin i la- it- ur they.hadâ•… themâ•… wash-â•… prf-â•… prt ‘They had washed them’
172â•… The middle-passive in Albanian 5.2
The u clitic
We shall begin our discussion with the structures formed with u, which we have described throughout as comparable to Romance si/se. Just as the distributional properties of Italian si require it to be treated as a pronominal object clitic (as opposed to an affix, a subject clitic, etc.), we can argue that the same holds for Albanian u. Consider, for instance, enclisis–proclisis alternations in Arbëresh varieties. These show that u is sensitive to exactly the same conditions as other object clitics, such as accusative E ‘him/her’ or dative i ‘to him/to her/to them’, down to very fine variation. Thus, we have seen that in a variety like Portocannone, the auxiliary precedes both, as in (11). Another variety where the same holds is S.Benedetto, as in (15), which shows that there is no correlation with the possibility of the participle being introduced by E ‘and’. By contrast, in a variety like Civita, as in (15), the u clitic and the E accusative clitic precede the auxiliary. We refer the reader to Manzini and Savoia (2007) for an analysis of the relevant parameter(s); what is relevant here is the complete parallelism between the two clitics. (15)â•… S.Benedetto Ullano (Calabria) a. kia pa I.had him seen ‘I had seen him’ b. kia u aitu I.had MP washed ‘I had washed myself’ Civita (Calabria) a. u kiï†fi zua MP I.had woken ‘I had woken up’ etc. b.â•… ki pa: him I.had seen ‘I had seen him’
Consider also the imperative 2nd person singular. As exemplified in (16) with the Arbëresh variety of Civita, in the positive forms pronominal clitics are found in enclisis (as a reflex of the high position of the verb); this is true both of accusatives, as in (16a), and of the u clitic, as in (16a’). By contrast, the presence of the negation induces proclisis (as a reflex of the verb staying in its inflectional position), and this affects the accusative and u clitics alike, as in (16b–b’). (16)â•… Civita a. zɟj wake.upâ•… him ‘Wake him up’
The u cliticâ•… 173 a’. b. b’.â•…
zɟj u wake.up MP ‘Wake up’ ms zɟ not him wake.up ‘Don’t wake him up’ ms u zɟ not MP wake.up ‘Don’t wake up’
Given this distribution, it is evident that any adequate theory of Albanian u must take into account the fact that it is an object clitic. Within the object clitic string itself, the u clitic appears to be found in a relatively low position. In particular, as shown in (17), u follows the 3rd person dative, as well as the 1st person clitic. Incidentally, these are examples of what we have called the anticausative interpretation; the dative adds a benefactive/malefactive specification. (17)â•…
Gjirokastër m/ i u i gta to.me/to.him MP broke the.glass ‘The glass broke on me/him’
Shkodër m i/ u y: gota to.me/to.him MP broke the.glass ‘The glass broke on me/him’
Portocannone m/ i u tSa- x bukjer to.me/to.him MP break- MP a glass ‘A glass broke on me/him’
The object clitic nature of u corresponds to a rather natural treatment of at least the reflexive interpretation. Thus, we could say that, exactly like accusative E, reflexive u is a lexicalization of the internal argument of the verb€– with the difference that while E is pronominal, u is anaphorically dependent on the EPP argument. This is the theory proposed by Burzio (1986) for Italian reflexive si. However, by analysing reflexive si as just described, Burzio (1986) is forced to postulate the existence of at least one other homophonous si€– i.e. an impersonal si, which lexicalizes the external argument of the verb as a generic. Its effect is that the EPP position is vacated and the internal argument can and must move into it, yielding the classical movement derivation for middle-passives. The problem, of course, is that if there were two different si’s, one would expect to be able to tell them apart through their syntactic behaviour (and not just their interpretation). In reality, Manzini and Savoia (2005, 2007) show that
174â•… The middle-passive in Albanian Italian si behaves homogeneously under distributional tests, even when object and subject clitics otherwise show a split. What is more, the postulation of two si’s forces other principles to have a disjunctive formulation, as is notably the case for auxiliary selection. As if this were not enough, data like those from Albanian show that the cluster of meanings associated with si forms a natural class, showing up in many diverse languages€– and with many diverse morpholexical instantiations. The main alternative analysis available in the literature involves unifying the various si’s under the movement derivation classically associated with passives. Thus, let us assume that in the passive, si becomes associated with the external theta-role of the predicate, and this forces the internal argument to externalize, yielding the typical promotion of object to subject. Because a reflexive predicate is by definition symmetric, reflexive si could in principle correspond to the internal or to the external argument of the verb. Suppose that, just like passive si, reflexive si is associated with the external argument of the predicate; the derivation that ensues is then identical to that of the passive, with promotion of the object to subject position providing for a unification of the two si’s (Marantz 1984). Needless to say, this derivation not only unifies various interpretations of si, but also does so by extending the classical movement approach to all of them. Despite what may appear to be its theoretical advantages, this analysis fails for empirical reasons. In particular, we note that authors who have espoused the unaccusative theory of middle voice have systematically considered languages like French, which does not have the impersonal (non-passive) reading of the middle voice. For a language like Italian, a theory associating si with movement from the object to the subject position is directly contradicted by the existence of examples like (23a) below in which si co-occurs with an overtly lexicalized accusative. Another important disadvantage of this analysis is that it does not predict that the morphosyntax of si is consistently that of an object clitic. That this is not an idiosyncratic property of si can be seen precisely in the context of cross-linguistic comparison, for instance with Albanian. There is no doubt that the properties of u are in some respects quite different from those of Italian si, in that, for instance, u is associated with all persons as opposed to si, which is only associated with 3rd person (though there is great variability in Romance languages, and in Romansh the si-like form can be associated with all persons). Another difference is that si is associated with all temporal and aspectual specifications of the verb, while, as we have seen, u is restricted to the perfective (standard Albanian) or to the past (Gheg varieties). Precisely because of
The u cliticâ•… 175 this variation, it is all the more striking that what remains constant in the morphosyntax of u and si is that they behave like object clitics; evidently this is a central property of such forms and not merely an accidental one. But although their distribution suggests that elements like si or u are just the middle-passive counterpart of the accusative clitic, this fact cannot be captured by the treatment of middle-passive morphology in terms of movement. Note also that si or u cannot be treated as subject clitics in languages like Italian or Albanian which do not otherwise have such elements€– nor can they be subject clitics in imperatives, which consistently lack such elements even in subject clitic languages. Reinhart and Siloni (2005), basing their work on Chierchia (2004) and Reinhart (1997), introduce as many separate operations on argument structure as there are basic meanings of si€– namely reflexive bundling, which bundles the external theta-role with some other theta-role; saturation/arbitrarization (responsible for passives/impersonals), which saturates the external theta-role through existential closure; and decausativization (responsible for anticausatives), which reduces (i.e. suppresses) an external [+cause] theta-role. The unification of these various rules by a single morphology is imputed to Case theory. The assumption is that the arity reduction operations just mentioned do not affect the Case properties of the verb, leaving an accusative (or a nominative in arbitrarization contexts) potentially unchecked; ‘the clitic (or its equivalent) reduces Case’ (Reinhart and Siloni 2005:€402). As far as we can see, Chierchia (2004), and consequently also Reinhart (1997) and Reinhart and Siloni (2005), are oblivious to the existence of impersonal unaccusatives. The latter exclude the possibility that arbitrarization can simply be construed as reduction of the external theta-role, for there is no external theta-role to be reduced in unaccusatives. Furthermore, Reinhart and Siloni (2005) argue that si is not an ‘object clitic’, but their evidence really shows that si is not an accusative clitic (for instance not triggering faire–à constructions when embedded under a causative). On the contrary, Italian si and its Albanian u counterpart behave like object clitics with respect to their distribution€– contrasting with subject and other clitics. In other words, any theory that does not treat si (or u) as an object clitic is forced to state all of the relevant generalizations (distributional, etc.) twice, once for object clitics and once for si/u. Here, the point is that if si/u is a bona fide object clitic, we expect it to be like any other object clitic (with which it patterns) in having not only Case properties, but also denotational and argumental ones. In this respect, Reinhart and Siloni (2005) represent a step backwards even with respect to Chierchia (1995), who identifies si with the variable existentially bound in passives/impersonals.
176â•… The middle-passive in Albanian A clue that Reinhart and Siloni’s (2005) theory is on the wrong track regarding the status of si is that it yields incorrect predictions as to the connection between si and auxiliary selection. In order to explain the parameter differentiating Italian, in which the middle voice is associated with be, and Dutch, in which it is associated with have, they invoke a distinction between the thematic and structural components of Case. In their account, Dutch zich, ‘though referentially defective, occupies the complement position’ where it can check structural Case, while Italian si cannot, leaving ‘the structural accusative residue to be checked … The auxiliary “be” is used whenever there is such an accusative residue’ (Reinhart and Siloni 2005:€432–3). Leaving aside any other considerations, languages like Soazza, reviewed in chapter 6, show that the clitic or non-clitic status of the middle morphology is completely irrelevant for the auxiliary selection parameter€– thus eliminating a potential argument for treating si as anything but a bona fide clitic counterpart of zich. As stressed by Culicover and Jackendoff (2005), current generative theorizing is strongly biased in favour of what they call Interface Uniformity, i.e. the principle that ‘the syntax–semantics interface is maximally simple, in that meaning maps transparently into syntactic structure; and it is maximally uniform, so that the same meaning always maps onto the same syntactic structure’. From such a perspective, the objections just raised may be considered of little import when weighed against the possibility of maintaining a ‘uniform’ movement analysis for passive. The argument pursued here is that loss of predictive power with respect to the actually observed morpholexical forms is to be taken as seriously as loss of predictive power at the LF interface. Hence the difficulty in predicting the object clitic behaviour of si or u cannot be discounted even in the face of apparent gains in ‘Interface Uniformity’.1 We take the conclusion that si is a bona fide pronominal clitic as our starting point, assuming that as such it fills an argument slot of the predicate (as does its Albanian counterpart u, etc.). Another premise of our analysis is that the denotation of si/u differs from that of other pronominal clitics in that it is an indefinite, i.e. a free variable. This is proposed by Chierchia (1995) for impersonal si and by Manzini (1986) for all si; the various readings of si simply depend on the possible ways of closing the variable. Consider the reflexive interpretation. In the absence of evidence to the contrary, we can assume that the structure of the reflexive sentence matches that of its transitive counterpart; in other words, u in the reflexive sentence occupies an object clitic position, as E does in the active sentence; in (18) it is categorized as Q (as si is) in accordance with its variable interpretation. In turn, the EPP argument, i.e. D in present terms, is lexicalized in (18) by the verb inflection€–in,
The u cliticâ•… 177 as is normally the case in null-subject languages like Albanian. Crucially, the referential properties of u, which by hypothesis are those of a variable, set it apart from other pronominal clitics. In order for the variable to be valued, it must be bound by a referential element. The reflexive interpretation is simply the consequence of the binding of the u variable by the closest available referring element, i.e. the EPP argument. The construal of reflexivization that we have now provided is essentially the traditional one, with the reflexive element (u in this case) associated with an internal argument position and bound by the EPP argument. (18)â•… Shkodër Q u
I I l� ʃ (x,y)
D in
Consider now the passive interpretation, which could equally be associated with the sentence in (18). Our proposal is that the structure of the passive sentence is identical to that of the reflexive sentence, with the u clitic inserted in the internal argument position. Indeed there is no evidence that the reflexive and the passive readings correspond to different underlying structures. Rather, all morphosyntactic evidence points to the conclusion that structures like (18) are genuinely ambiguous, allowing for both readings under consideration. If the same structure underlies both the reflexive and the passive readings, then passives must differ from reflexives only interpretively. Let us assume that in the passive, the dependency between the u variable and the EPP argument corresponds to a chain. If so, the passive is treated exactly as in classical generative grammar, as an instance of chain formation between the internal argument and the EPP argument. The only difference is that instead of a trace (i.e. an empty category or a copy), the analysis we propose has an overtly lexicalized internal argument, whose semantics is that of a variable. There are several respects in which this analysis is at variance with standard generative frameworks, including notably the minimalist programme of Chomsky (1995). While these take a derivational view of movement, the present analysis is representational. Thus, since the two positions related by the chain in (18) under the passive reading are each independently lexicalized, there cannot be a derivational process of movement relating them; rather, movement must reduce to the notion of chain at the LF interface (Brody 2003).
178â•… The middle-passive in Albanian Chomsky (1995) assumes, in a manner consistent with the overall derivational outlook of minimalist theory, that the verb inflection is a cluster of uninterpretable features which drive processes such as movement (agreement, etc.) because of the need for such features to be checked (valued, deleted, etc.). By contrast, our discussion of the structure in (18) presupposes a treatment of the verb inflection as an interpretable element€– specifically as the morphologicallevel lexicalization of the EPP argument of the sentence. Finally, under (minimalist) movement, chains can be motivated by feature-checking requirements or by interpretive requirements. In present terms, they can be motivated only by interpretive requirements, i.e. in the case of the passive chain, by the need to provide a value for the variable internal argument. The representational construal illustrated here for movement affects all transformational processes. Thus, agreement, which is the rule specifically responsible for feature checking in Chomsky’s (1995) framework, is taken to be a relation which must hold if various interpretations (including the chain one) are to hold in turn. For instance, the chain in (18) requires agreement (or, to be more precise, compatibility in referential properties) between the EPP argument and the element lexicalizing the variable, though in this case the requirement is trivially met because of the presence of a variable. In short, we part ways with standard models of generative transformational grammar in adopting a representational model with the properties outlined in the preceding discussion. At the same time, it should be clear that in this representational form, our analysis of Italian si or Albanian u includes standard generative ideas about passive as involving a chain between the internal argument and the EPP argument. Thus, while we share the concern of Culicover and Jackendoff (2005) about Interface Uniformity, we certainly do not subscribe to their view of chain interpretations as mediated by a Grammatical Functions (GF) level of representation. The analysis we propose is no more (and no less) than a representational version of generative transformational models (specifically of minimalist ones). We contend that, at the LF interface, the present theory maintains all of the explanatory power of conventional theories of movement. At the same time, crucial motivation for it comes precisely from the kind of considerations pertaining to morpholexical structure that we have detailed above. In the present theory, it is perfectly possible to maintain that elements such as Italian si or Albanian u are exactly what they appear to be, i.e. normal object clitics. In this respect, the crucial assumption is simply that their content is that of a variable€– in other words, that the variable status is not restricted to traces (i.e. empty categories or copies) created by movement. This latter restriction seems
The u cliticâ•… 179 to us to be an artifact of strictly derivational approaches, while representational approaches can easily handle specialized lexical items with the content of variables. In short, we can maintain what appears to be the transitive structure of sentences like (18), with the clitic instantiating an object, while at the same time incorporating the core generative insight that the passive interpretation involves the chain construal of the internal argument with the EPP argument. Consider now what we have called the anticausative reading, salient in examples like (6). This reading can be obtained on the basis of structures like (19), entirely parallel to (18) above, through the formation of a chain between the variable internal argument, lexicalized by u, and the EPP argument, represented by the verb inflection. This leads to the interpretation under which the EPP argument is interpreted as the internal argument of the verb (roughly the theme undergoing the waking up). Note that the argumental frame of the verb is in itself transitive; quite simply, in the anticausative reading the external argument is not interpreted. Needless to say, the passive interpretation, roughly ‘he was woken up’, is equally predicted to be possible in (19), on the basis of chain formation. Thus, exactly as in standard movement models, passives and anticausatives share the same core syntax. The difference is that in the passive reading, the implication is preserved that the event takes place through an external agency or cause, corresponding to the external argument of the transitive argument frame of the verb. The latter can receive independent lexicalization through a by-phrase, or it can be interpreted through generic binding of the argument variable€– yielding a so-called ‘implicit argument’, while in the anticausative reading the implication is that the theme is not acted on by another agent/cause. (19)â•… Shkodër Q u
I I tʃo (x,y)
D tε
Going back now to the reflexive interpretation, the maximally simple assumption about the nature of the dependency between the u variable and the EPP argument is, again, that it is a chain. This assimilation of the reflexive to the passive/anticausative does not prevent their respective meanings from being clearly differentiated. As in the anticausative interpretation, in the reflexive there is no implication of an external agency or cause, differentiating both from
180â•… The middle-passive in Albanian the passive. At the same time, the reflexive and the anticausative are differentiated in that, in the reflexive, some degree of intentionality is associated with the argument of the si sentence; thus, reflexive readings are available only with EPP arguments capable of a mental state. In other words, pragmatic knowledge about the event of ‘waking up’ excludes the reflexive reading in (19). The preceding discussion implies that we agree with Chierchia (2004) and Reinhart and Siloni (2005) that the external argument configurations in the middle are responsible for differences between anticausative, passive and reflexive readings. Yet it seems to us that they correspond to all and only the logically possible such configurations€– and as such they do not need to be stated, but follow as a matter of logical necessity. Thus, in the anticausative the external theta-role is not interpreted. In the passive, it is interpreted through quantificational closure (passive); alternatively, it is interpreted through assignment to an adjunct, i.e. the so-called by-phrase. As for the reflexive, it seems to us that the arguments put forth by Reinhart and Siloni (2005) are not decisive as to whether the external argument is to be imputed to the si chain (reflexive). For instance, ne cliticization from postverbal subjects is not a reliable test of the internal or external argument status of the latter, depending instead on the presentational (focus) properties of the sentence (Belletti and Rizzi 1981; Belletti 1988; Saccon 1992, pace Burzio 1986). Thus, under the right conditions we accept ne-extraction from the inverted subjects of reflexives (varying pragmatic contextualizations are in fact a better explanation for the variability in grammaticality judgements noted by Reinhart and Siloni (2005) among Italian speakers). Therefore we side instead with Chierchia (2004) in assuming that reflexives are a subclass of anticausatives, in which agency is imputed to the sole lexicalized argument€– by a meaning postulate in Chierchia (2004). Note now that there are also verbs which display the middle-passive conjugation without having a transitive counterpart; an example is ulm ‘I sit down’ in (20a–c). One may legitimately wonder how these differ from active unaccusatives, i.e. predicates whose only argument is a theme (an internal argument), and which are not formed through middle-passive morphology. An example is provided by a motion verb like dal ‘I go out’, exemplified in (9) and here in (20a’–c’), which combines with the active person ending and in the present perfect with the auxiliary kam ‘I have’. (20)â•… Shkodër a.â•… ul- E- t sit-â•… MP- 3sg ‘He sits down’
The u cliticâ•… 181 b. c. a’. b’. c’.
ai u ul he MP sit-3sg ‘He sat down’ ɐï†fit u:l he.is sat ‘He has sat down’ ai dl he goes.out ‘He goes out’ ai dol- i he went.out- 3sg ‘He went out’ kQ dQ:l he.hasâ•… gone.out ‘He has gone out’
For verbs like dal, we simply assume that their single argument slot (a theme), as in (21a), is assigned to the obligatory argument of the sentence, i.e. the EPP argument. This yields an unaccusative reading comparable to the anticausative reading in (19)€– but does not imply the presence of non-active morphology. At the same time, the grammar must provide a way to distinguish between verbs like dal ‘I go out’ and verbs like ulEm ‘I sit down’. A way to formalize this distinction is simply to associate the latter with an argument frame of the same type found on tSo- in (19), as shown in (21b). The fact that, in (21b), the potentially transitive frame is restricted to the anticausative reading will have to be learned as a lexical property.2 (21) a. dEl (x) b. ulEt (x,y)
The final reading of Albanian u sentences that remains to be considered is the impersonal one, most clearly present in sentences involving unaccusative predicates such as (9). Under the line of explanation pursued throughout this section, one may be led to conclude that, in the absence of distributional or morphological evidence to the contrary, the structure underlying (9) is the same as that already indicated for the other interpretations of u in (18)–(19), as in (22). The crucial difference is that (18)–(19) contain not only the variable u clitic, but also some independently referring EPP argument€– even if only represented by the inflection of the verb. By contrast, (22) contains no independently referring EPP argument. Indeed, the obvious construal of the generic (or ‘impersonal’) interpretation associated with the EPP argument in (22) is that the u variable itself supplies it, through closure by a generic operator. This interpretation, in turn, can correspond to a syntax in which, just as in the
182â•… The middle-passive in Albanian other cases considered before, the u clitic forms a chain with the EPP argument represented by the D inflection of the verb. We assume that a generic interpretation cannot simply be associated with the 3rd singular inflection of the verb; this necessitates the introduction of the variable, i.e. u, which can be bound by the generic operator, as detailed above. (22)â•… Shkodër Q u
I I dεl (x)
D tε
Recall, as we noted above, that the arbitrarization rule of Chierchia (2004) and Reinhart and Siloni (2005) (i.e. existential closure of an external argument) could not yield impersonals with unaccusative verbs. If our discussion is on the right track, this problem can be avoided simply by taking arbitrarization to correspond not to an operation on argument structure, which needs to be stated in terms of such primitives as external argument, but to an interpretation at the LF interface, which can apply to any variable not independently closed. We also noted above that the parallel between Italian si and Albanian u is all the more interesting because the two elements also display important points of variation. A relevant observation in this respect concerns the fact that impersonal si in Italian can combine with accusative objects, as in (23a). This possibility is not available in Albanian€– and in fact it is excluded in many Italian varieties as well, where the only possible combination between si and a transitive predicate is the counterpart of standard Italian (23b), i.e. a passive. (23) a. b.
Li si chiamerebbe volentieri them MP would.call gladly ‘One would gladly call them’ Si chiamerebbero volentieri MP would.call gladly ‘They would gladly be called’
In present terms, the crucial difference between the impersonal in (23a) and the passive in (23b) is that in (23b), the si clitic is construed as an instantiation of the internal argument slot, bound by the independently lexicalized EPP argument. On the other hand, in (23a), si itself satisfies the EPP position, while the internal argument is independently lexicalized by the accusative
The u cliticâ•… 183 clitic. The impossibility of the Albanian counterpart of (23a) can then simply be described in terms of a necessary association of Albanian u with the internal argument slot. On the assumption that the only argument slot of unaccusatives is an internal argument slot (albeit assigned to the EPP position), we still derive the impersonal reading of u with unaccusatives. In conclusion, we have now provided an answer to one of the key questions raised at the beginning€– namely whether a single syntax underlies all of the different meanings of the middle-passive voice. At least in the case of si/u sentences, what gives rise to all of the different readings is the property of si/u that it is a variable. As a result of the presence of this variable, one of the argumental positions of the predicate remains unassociated (in anticausatives/reflexives) or is interpreted only in that it is existentially closed (passives, impersonal) and/or associated with an argument which is external to the structure of the predicate (the by-phrase). This is not the characterization associated with (middle-)passive by standard generative theory€– according to which (middle-)passive is reduced to movement from object to subject position. However, the construal of the notion that we suggest is in a way equally traditional€– having to do with the non-closure (or generic closure) of the argument structure of the verb. Importantly, the grounds for this switch are entirely empirical, in particular that the standard generative construal in terms of object to subject promotion stands no chance of unifying the passive/ reflexive reading with the impersonal one.3 Needless to say, (the representational counterpart to) object to subject movement still characterizes the bound readings of the variable when construed as an internal argument (in passives, anticausatives and reflexives). This latter fact also provides us with a key for understanding languages, like those routinely considered in the literature, in which the impersonal reading illustrated here in Italian and Albanian is not available (except perhaps in the shape of impersonal passives, with unergatives). A language like French, in which the impersonal has the specialized subject clitic lexicalization on (‘one’), provides a good minimal contrast with Italian.4 A good minimal contrast with Albanian is provided by Greek (Roussou 2008; Manzini, Roussou and Savoia forthcoming), in which there are specialized middle-passive inflections which split according to aspect as in Albanian, but these do not yield the impersonal reading. In terms of the present approach, the difference between the se of French and the si of Italian can be straightforwardly described by assuming that French se is a bound variable which does not admit of simple quantificational closure. We shall return to Greek and Albanian in the next section.
184â•… The middle-passive in Albanian 5.3
Specialized inflections
On the basis of the conclusions reached in section 5.2, we are in a position to consider the second key question raised at the beginning, namely whether there is a single underlying syntax corresponding to all of the different morpholexical instantiations of the middle-passive voice€– or whether instead what we are dealing with is merely the interpretive equivalence of (slightly) different syntaxes. Consider, then, the lexicalization of middle-passive voice in Albanian with specialized verb inflections, as in (1) and (3). In Gjirokastër and in the vocalic bases of Shkodёr, this non-active morphology can be seen to include an invariable affix –hE. With the consonantal bases of Shkodёr, we can take the€ –E extension of the base to represent the middle-passive morphology (Trommer 2005); the same will hold for the reduced form with€–E in Gjirokastër. As for the person inflections, these differentiate the active and the middle-passive in the present, and more specifically in the singular; in the plural it is only the 1st person that appears to be sensitive to voice. In the imperfective past of Gjirokastër in (3), there is substantial identity of the person endings in the active and middle-passive voices, the only difference being that the 3rd person singular is not lexicalized in the middle-passive. An insight into the nature of the middle-passive€ –h infix is provided by the fact that in an Arbëresh variety like Portocannone, illustrated in section 5.1.3 above, –hE (or –x in Portocannone) systematically co-occurs with the u clitic in all forms of the verb in which the latter is present, hence not only in the perfective past (10c), but also in the perfect, which in Arbëresh varieties is formed with kam ‘I have’ and the u clitic, as in (11a). One possible conclusion suggested by this pattern is that it represents a case of doubling, whereby the u clitic and the€–hE infix lexicalize essentially the same properties, at the syntactic level and at the morphological level respectively. If so, given the characterization provided for the u clitic in section 5.2, we are led to suppose that the –hE morphology represents a Q-type variable attaching to the verbal base I, as illustrated in (24) for the present indicative. The specialized€–t person ending in turn lexicalizes the D argument of this verbal constituent. (24)â•… Shkodёr D t I l�(x,y)
Q hε
Specialized inflectionsâ•… 185 Given the structure in (24), the computation of the different meanings associated with it proceeds in the same way as detailed in the previous section for u clitic structures. If the€–(h) variable, filling the internal argument slot of the predicate, is bound by the EPP argument, represented in (24), by the€–t inflection, the passive, reflexive and anticausative readings are obtained, depending on the interpretation of the external argument. If this is generically closed, we obtain the passive interpretation, while the anticausative/reflexive interpretation depends on the external argument remaining unlinked. In the case of a predicate like ‘to wash’ in (24), the salient meaning is reflexive rather than anticausative because agency/intentionality is normally attributed to the only lexicalized argument. Generic closure of the u variable leads in turn to the impersonal interpretation, seen in connection with unaccusative predicates like dl in (25). Overall, the effect of the€–(h) variable, just as with the u variable, is to associate the verb with an unsaturated or generically closed argument position€– namely the external argument with transitives and the only argument with unaccusatives. (25)â•… Shkodёr D t I dε l (x)
Q ε
As already noted, in the Arbёresh variety of Portocannone the€–x morphology combines with the u clitic to yield an instance of doubling, illustrated in (26) for the past perfective. The distribution of the u clitic in Portocannone, as in Arbёresh varieties in general, is determined by the perfect/imperfect split, so that u co-occurs with the perfective forms of the verb, including auxiliary–Â�participle constructions, formed with the kam ‘I have’ auxiliary. By contrast, the€–x morphology enters into the formation of the entire middle-passive paradigm, including the existence, noted in section 5.1.3, of a specialized middle-reflexive participial form. These separate distributions for€–x and u, when combined, yield the correct distribution of the doubling configuration (as their intersection). (26)
Portocannone Q u D tʃ I la(x,y)
Q x
186â•… The middle-passive in Albanian Assuming that the discussion of the€–h morphology is on the right track, it still remains for us to clarify the nature of the specialized agreement endings that characterize middle-passive inflectional formations. In particular, the singular –m/–ʃ/–t inflections of the present bear no relation to their active counterparts. In her discussion of Greek middle-passive voice, Roussou (2008) notices that in imperfective tenses, the middle-passive voice of Modern Greek is entirely realized by a series of specialized agreement inflections, which can be analysed into a thematic vowel –e/–o and specialized person endings –me, –se, –te, etc. for the present in (27a) or –mun, –sun, etc. for the past in (27b) (Philippaki-Warburton 1973; Ralli 1988). (27)â•… a. a’. b. b’.â•…
plen- ome/â•… ese/â•… ete/ omaste/ osaste/ onde Wash-â•… MP.1sgâ•… etc. ‘I wash (myself)’ etc. plen- o / is/ i/ ume/ ete/ un wash-â•… 1sgâ•… etc. ‘I wash (something)’â•… etc. plen- omun/â•… osun/ otan/ omastan/ osastan/ ondan wash-â•… MP.1sgâ•… etc. ‘I was washing (myself)’ etc. eplen- a/ es/â•… e/â•… (plen-) ame/â•… (plen-) ate/â•… (eplen-) an wash-â•… 1sgâ•… etc. ‘I was washing (something)’ etc.
The middle-passive perfective tenses are formed instead by the agreement inflections of the active voice combined with the affix –th for the present tense,5 as in (28a) and the affix –th–ik for the past tense, as in (28b); in this second case –th can be taken to realize middle-passive voice again, and –ik can be taken to realize past tense (Philippaki-Warburton 1973). (28)â•… a.â•… a’. b. b’.
pli -th o/ is/ i/ ume/ ite/ un wash-â•… MP-â•… 1sgâ•… etc. ‘I wash (myself)’ etc. plin o/ is/ i/ ume/ ete/ un wash- 1sg etc. ‘I wash (something)’ etc. pli- th- ik- a/ es/ e/ ame/ ate/ an wash MP past 1sg etc. ‘I washed (myself)’ etc. eplin- a/ es/ e/ (plin-) ame/ (plin)-ate/ (eplin-) an wash 1sg etc. ‘I washed (something)’ etc.
Roussou (2008) notices that the split in lexicalization between the imperfective and the perfective has some of the same properties in Greek and in
Specialized inflectionsâ•… 187 Albanian. In particular, in both Greek and Albanian, the imperfective has specialized middle-passive agreement inflections, while the perfective has the ordinary active agreement inflections, which combine with the –th affix in Greek and with the u clitic in Albanian. On the basis of this parallelism, Roussou (2008) suggests that Greek perfective tenses involve a structure along the lines of the Albanian u formations, where the –th affix plays the same role at the morphological level as the clitic plays at the syntactic level€– as a variable lexicalization of the internal argument. Crucially, Roussou (2008) goes on to propose that while the active agreement inflections associate with the EPP argument, the middle-passive agreement inflections associate with the internal argument. In other words, while active agreement is nominative, middle-passive agreement is absolutive. Extending this analysis to Albanian, we would conclude that the difference between the active agreement inflection (in the perfective tenses) and the middle-passive agreement inflections (in the imperfective tenses) is that while the former corresponds to a nominative case configuration, the latter correspond to an absolutive one, picking up the internal argument. At this point a potential problem arises. In Greek, specialized middlepassive agreement inflections are incompatible with –th morphology:€ thus, *pli–th–ome or *pli–th(ik)–omun are ruled out. Yet in Albanian, specialized middle-passive agreement co-occurs with the€–hE affix, which we construe as a morphological-level counterpart of the middle-passive clitic u, exactly as Roussou (2008) suggests for the –th affix in Greek. Now, note that Roussou (2008) does not provide an analysis of the thematic vowel specialized for the middle-passive agreement endings; in fact, she adopts a segmentation that does not separate the thematic vowel from the person ending with which it combines (following Ralli 2005). However, Manzini and Savoia (2005, 2007) treat Â�so-called thematic vowels as variables closing off the internal argument of the verb. From this perspective, the specialized thematic vowel of Greek middlepassive agreement inflections could be the real counterpart of Albanian –(h)E. This could also mean that the mutual exclusion between the middle-passive agreement inflections and the€ –th affix of Greek should be understood as a mutual exclusion between the€–th affix and the thematic vowel of these inflections, rather than with their (absolutive) person endings.6 We can now return to the distribution of the different lexicalizations of the middle-passive voice according to aspect (Tosk Albanian and Greek) or tense (Gheg Albanian). In Tosk Albanian the imperfective (including the present and the imperfective past) lexicalizes the middle-passive voice with specialized morphology, while the perfective lexicalizes it by syntactic means, i.e. through the u clitic. In Gheg varieties of Albanian, a split is also found, except that it
188â•… The middle-passive in Albanian is temporally based, distinguishing the present (with specialized morphology) from the past (with u). The importance of the aspectual/temporal split is underscored by the comparison with Greek, which also has a middle-passive voice whose lexicalization is differentiated according to aspect. Roussou (2008) suggests that this distribution involves an ergativity split, with a set of nominative inflections (the active ones) and a set of absolutive inflections (the middle-passive ones) apportioned to the imperfective and perfective tenses respectively. Nevertheless, in other attested ergativity splits, the nominative pattern tends to associate with imperfective tenses, and the ergative pattern with perfective ones. We therefore leave this matter open, having noted its interest. We started the discussion in this section by remarking that comparison between the lexicalization of the middle-passive with an u clitic, as analysed in section 5.2, and its lexicalization with specialized inflections could shed some light on the question of whether a single syntax underlies all of the forms allowing for the cluster of middle-passive interpretations. Since we have assumed that the€–(h)E represents a morphological-level counterpart to the u clitic, the answer is in a sense positive:€in both cases a crucial role is played by the variable instantiation of the internal argument. Yet while one of the structures has a syntactic-level variable (in the shape of the u clitic), the other does not; crucially, no abstract structure, specifically no movement structure, intervenes to make these two morpholexical instantiations of middle-passive isomorphic. 5.3.1 Be–participle The question of whether the same syntactic structure underlies the same (cluster of) interpretations, i.e. the uniformity question in the sense of Culicover and Jackendoff (2005), can equally be asked with respect to the third possible lexicalization of Albanian middle-passives, namely the combination of the jam ‘I am’ auxiliary with a participle. We delay our account of these structures till chapter 6, where they are considered in relation to other instantiations of perfect tenses. In the final part of this chapter, we consider a different lexicalization available in Albanian for the passive reading, again involving the jam ‘I am’ auxiliary. This auxiliary, however, embeds a participial form which is different from those reviewed in section 5.1, yielding passive for all of the various tenses of the paradigm. Both in being restricted to the passive reading, and in not being restricted temporally/aspectually, the relevant construction in Albanian is close to the English be–en passive and its Romance counterparts (essere V–to in Italian). It is useful to begin by reviewing what is perhaps the most basic occurrence of jam, i.e. as a copula. In (29), we provide various examples of this configuration in the variety of Shkodёr. The paradigm of adjectival embedding in (29a–b) is the same as in the standard. This requires the full adjectival
Specialized inflectionsâ•… 189 inflection, including a preposed article and a postposed inflection, both agreeing for number, gender, definiteness and Case. (29)â•… Shkodёr a. t i kutS/ kutS- it.is m.sg red/ f.sg red-f. ‘It is red’ b. jan t kutS/ kutS- they.are pl red/ red-f ‘They are red’
The examples in (30) illustrate the embedding of participles (consonantal base) in the copular construction, yielding a structure identical to that seen for adjectives in (29). It is interesting that in the variety of Shkodёr, the adjectival inflection is not simply added to the bare bases with which participles have been seen to coincide in (5) above. Rather, the base shows a participial inflection€–m or€–un depending on verbal class (vocalic and consonantal respectively). Verbal adjectives are regularly formed from transitive predicates (‘to dress’) as in (30a–a’), from intrinsically middle-passive ones (‘to sit down’) as in (30b), and from active unaccusative ones (‘to come’), as in (30c). The adjective in (30d), from unergative ‘to sleep’, does not have the meaning of ‘slept’, but rather of ‘asleep’. The example in (30e) establishes that the participial form with adjectival inflection is found not only in copular constructions, but also in other predicative constructions, here in a small clause embedded under kam ‘I have’; incidentally, the same example also allows us to illustrate participles formed from vocalic bases. As shown in (30a–a’), copula–participle formations based on transitive verbs admit of the passive meaning, disambiguated here by the presence of a by-phrase. A comparable example is provided in (31) for Gjirokastër. (30)â•… Shkodёr a. t i veS-un/ veS-un s/he.is m.sgâ•… dress-ed/â•… f.sgâ•… dress-ed ‘S/he is dressed up (by his/her mother)’ a’.â•… jena t veS-un/ t veS-un-a we.are pl dress-ed/ pl dress-ed-pl ‘We are dressed up (by our mother)’ b. guja ɐï†fitâ•… ulun the.womanâ•… is f.sg seated ‘The woman is seated’ c. ï†fit i arDun he.is m.sg arrived ‘He has arrived’ d.â•… ai ɐï†fit i fjetun he is m.sg asleep ‘He is asleep’
(pi s Qms) (by gen mother) (pi s Qms) (by gen mother)
190â•… The middle-passive in Albanian e. i kam kmiSa-t /t lQ-m- themâ•… I.haveâ•… shirt-s Art wash-ed-f ‘I have the shirts washed’ (31)â•… Gjirokastër jan t@ vu/ vua they.are pl dressed.up/dressed.up-f. ‘They are dressed up’
The data in (29)–(31) can be usefully integrated with those of Arbёresh varieties, which also have the construction in which the copula is followed by the inflected participle. This is illustrated for Portocannone in (32a) with a transitive predicate and in (32b–c) with unaccusative predicates. In this construction, transitive predicates associate with by-phrases, as in (32a); by contrast, in the discussion of (14) we saw that the middle-passive voice is normally restricted to a generic agent; i.e. it excludes the by-phrase. The participle that is involved in the copular construction is the ordinary active form, despite the existence in the Portocannone variety of a specialized middle-passive participle, as discussed in section 5.1. As shown in (32a’–b’), the same variety also admits of copular (non-perfect) jam–participle constructions in which the participle is uninflected. (32)â•… Portocannone a. kt kmiS jan/ kjetn t@ la- it- these shirts are/ were pl wash- prf- ‘These shirts are/were washed by them’ a’.â•… kt kmiS jan/ kjetn la- it- ur theseâ•… shirts are/ were wash- prf- prt ‘These shirts are/were washed by them’ b. iSt i uj- ur he.is m.sg seat- prt ‘He is seated’ b’. iSt uj- ur he.is seat- prt ‘He is seated’ c. iSt i vdEk- ur he.is m.sg die- prt ‘He is dead’
ur- a (tE a’ta) prt fpl (by them) (tE a’ta) (by them)
For bases ending in a vowel, as in (33a), the adjectival participles of Shkodër shows an –m suffix which can be analysed as the bearer of the aspectual, perfective properties of the participle. Thus, the –m inflection is an I head which takes the verbal base as its complement. The verbal base, including the thematic vowel, appears independently of further inflectional material as the invariable participle in (5), i.e. lQ. Verbal bases ending in a consonant, in turn, form the
Specialized inflectionsâ•… 191 participle with a suffix –(u)n, as in (33b). We assume an analysis parallel to that for bases ending in a vowel, so that the –n inflection is an I aspectual, perfective head, which takes as its complement the verbal consonantal base combined with a thematic vowel€–u, selected in this case by the perfective aspect itself. Entirely parallel structures can be provided for Tosk varieties (including both the standard and Arbëresh dialects, like Portocannone), with the only difference that –r is the participle-forming I inflection in these languages with both vocalic and consonantal bases. (33)â•… Shkodër a. e�
I m
veʃu
I n
b.
The agreement inflections following the verbal subconstituents in (33) are the same as those found in adjectives and nouns. Following the analysis of nominal inflections in Albanian in chapter 7, we assume that at least vocalic endings correspond to N inflections, whether they lexicalize so-called accusative, as in (34a) or so-called nominative, as in (34b). (34)â•… Shkodër a. N ε e�
I m
b. N a veʃu
I n
192â•… The middle-passive in Albanian The article preceding the participle can be treated, as in standard analyses of the noun phrase, as a D element at the phrasal level, giving rise to structures like (35). In present terms, as detailed once again in chapter 7, the agreement between the determiner and the Case inflection of the participle is a reflex of the fact that they form a chain€– in other words, together they satisfy the same argument slot(s). (35)â•… Shkodër a. D ε
I l� mε
D t
I veʃuna
b.
From the point of view of the present discussion, the crucial question concerns the passive interpretation of sentences like (30a–a’). Given the complete formal similarity to (29), we may once again begin by asking how an ordinary copula–adjective construction is interpreted. In the case at hand, it seems evident that the embedded predicate kutʃ ‘red’ has a single argument, lexicalized by its N inflection and by the determiner that it agrees/forms a chain with. Agreement and chain formation in turn hold with the D (i.e. EPP) argument of jam ‘I am’, i.e. its inflection. The overall result is, correctly, a raising interpretation, whereby the EPP argument of the matrix sentence is assigned the argument slot of the embedded (adjectival) predicate. The same raising interpretation characterizes examples like (30a–a’), the only difference being that they involve transitive, i.e. two-place, participial predicates. In this case, the determiner and the agreeing inflection of the participle pick up its internal argument slot; this means that participial structures are absolutive, in the sense that the D argument which closes them (the determiner) is associated with their internal argument slot. The latter ends up being associated with the EPP argument of jam ‘I am’ by the same mechanism just detailed for the copular sentences in (29)€– so that, from this perspective, the passive interpretation of sentences like (30a–a’) (with the internal argument of the participle assigned to the EPP argument of the copula) is simply an
Specialized inflectionsâ•… 193 ordinary adjectival interpretation. The external argument slot of the participial predicate is in turn closed by an adjunct by-phrase or by generic quantification, in this latter case yielding the implicit agent reading. We noted in passing that in Portocannone, the copular passive can either have the agreeing form in (32a) or a non-agreeing form as in (32a’). In other words, the structure of (32a’) appears to be entirely analogous to that found in, say, the Gjirokastër example in (5) for the perfect middle-passive. Nevertheless, we maintain the conclusion that all copular passives in (30)–(32) result from a nominal embedding of the predicate; the lack of agreement in the Portocannone example in (32a’) represents an independent parameter. By contrast, in chapter 6 we pursue the conclusion that the structure of embedding in the perfect tenses is not nominal, but sentential€– so that examples like (5) have a bi-clausal, restructuring syntax. Once again, Greek presents a constellation of facts strictly comparable to those of Albanian. The middle-passive voice of Greek includes a perfect formed by an auxiliary followed by a participle, as in (36a); (36b) provides the comparison with the active. In this case, the middle-passive configurations of Greek are strictly comparable to those of Arbёresh varieties, since the auxiliary is exo ‘have’ both in the active and in the middle-passive, and the difference between the two voices is found in the participle. In (36a) it bears the middle-passive morphology€–th, coinciding in particular with the 3rd person singular of the middle-passive present perfective, while in (36b) it coincides with the 3rd person singular of the active present perfective. What is immediately relevant for present purposes is that Greek has another participial form which is created by attaching to the verb base the affix€–men followed by the adjectival (nominal) agreement for gender, number and Case. The Greek copular (‘periphrastic’) construction formed by this participle and the copula has a passive interpretation, as in (37), under which, as discussed at length by Anagnostopoulou (2003), it systematically co-occurs with by-phrases. (36)â•… a. b.
exo, exis, exi, etc. pli-th-i have.1sg, 2sg, 3sg, etc. washed ‘I have been washed’ (‘you have/he has been/ etc. washed’) exo, exis, exi, etc. plin-i have.1sg, 2sg, 3sg, etc. washed ‘I have washed (something)’ (‘you have /he has etc. washed’)
(37) Afta ta pedhia ine dolofoni-mena these the children are murder-ed ‘These children are murdered’
194â•… The middle-passive in Albanian A question that we have so far left implicit is how the copular structure relates to the generative notion of adjectival passive. All the evidence at our disposal suggests that the adjectival, i.e. stative, reading of passives can be associated with the Albanian examples in (30)–(32) involving transitive predicates, both with and without a by-phrase. On the other hand, the verbal, i.e. eventive, passive reading seems to be responsible for the fact that Arbёresh speakers routinely offer these structures€ – rather than structures like (14) above€– when passive is elicited. An important line of thought in generative grammar, dating back at least to Wasow (1977), associates verbal passives with a syntactic derivation, and adjectival passives with a lexical one. The latter analysis is adopted by Terzi and Wexler (2002) for Greek. We take it that the evidence from Albanian runs counter to the traditional account of verbal vs. adjectival passives in terms of syntactic vs. lexical derivation, as also argued for Greek by Anagnostopoulou (2003). Instead, we conclude that the so-called adjectival and verbal passive are just interpretations attached to the same copular structures€– which, everything else being equal, are systematically ambiguous between the two. Anagnostopoulou (2003), basing her work on Kratzer (2000), introduces in the structure of adjectival (i.e. stative) passive a category ‘Stativizer’ differentiating it from the verbal, i.e. eventive passive. But this at best succeeds in translating the interpretation into the syntax, without any explanatory gain, while at worst it obscures the fact that the stative/eventive opposition (i.e. in traditional generative terms the adjectival/verbal one) is precisely a matter of ambiguity at the LF interface, not corresponding to any independently observable morphosyntactic difference. Finally, the fact that in Arbёresh, by-phrases, which normally occur in the copular construction (32), are incompatible with middle-passive voice forms, as in (14), is reminiscent of a restriction found in Italian (and generally in Romance) against by-phrases in si-passives. In both instances, the external argument is interpreted, but the only possible reading is through binding by a generic operator; in other words it cannot be linked to a referential noun phrase within an adjunct noun phrase. In Romance, one may be tempted to relate this restriction to the presence of si, which is sometimes construed in the literature as an absorber of the external theta-role. In Arbёresh, however, the impersonal reading of the passive characterizes not only the perfective forms with the u clitic, but also the imperfective forms with the specialized verb morphology. What is more, mainland Albanian varieties seem to allow by-phrases in contexts that are morphosyntactically
Specialized inflectionsâ•… 195 identical to that of Arbёresh. Therefore we conclude that there is an independent parameter defining the possibility of middle-passive forms co-occurring with a by-phrase€ – i.e. whether the external argument of a middle-passive allows quantificational closure only, or can also be closed by an adjunct (essentially a form of predication, external to the core event of the sentence).
6 The auxiliary:€have/be alternations in the perfect
The discussion in this chapter is based on a set of data which reflect microvariation in a closely related set of languages (Romance varieties) as well as variation between more distant languages (Romance and Albanian). In section 6.1 we present the basic evidence and review the notions of person split, transitivity and voice, in terms of which we analyse the data in sections 6.2 and 6.3. We discuss person-split systems in detail in section 6.2, and we concentrate on auxiliary selection according to transitivity/voice in section 6.3, arguing in particular that the notion of transitivity should be split from that of voice, and a characterization not based on movement should be adopted for the latter (cf. chapter 5). In section 6.4, we introduce some Italian varieties in which selection according to transitivity/voice and the person split cross-cut, as well as varieties in which the have/be distinction is neutralized in parts of the paradigm. As discussed in section 6.1.1, the account we provide is crucially based on the assumption that the embedded participle does not select the auxiliary, in the sense in which a lexical category could be said to select its functional projections. Rather, the auxiliary and the participle define two independent sentences (Kayne 1993), and the selectional relation is the ordinary one from matrix predicate (the so-called auxiliary) to embedded sentence. Adopting the bi-clausal analysis just outlined for perfects amounts to eliminating auxiliaries, i.e. functional verbs, from the present grammar. 6.1
Evidence
We begin with the well-known evidence concerning standard Italian. As argued by Burzio (1986), auxiliary selection in Italian is sensitive to the distinction between transitive/unergative and unaccusative verbs; the former take have as in (1a), and the latter take be, as in (1b). (1)â•… a.â•… Ho lavato (la camicia) I.haveâ•… washed the shirt ‘I have washed (the shirt)’ 196
Evidenceâ•… 197 b. Sono arrivato/arrivata I.am arrived.m/f. ‘I have arrived’
The most notable complication concerning Italian is that the presence of the si element correlates with the selection of be, as in (2), independently of the many readings available for si:€namely reflexive, as in (2a), anticausative, as in (2b), passive, as in (2c) or impersonal, as in (2d). Most importantly, in the impersonal in (2d), the auxiliary be induced by si co-occurs with an overt accusative clitic€– i.e. with a transitive frame. The fact that si cannot be treated simply as a valency-reducing morpheme (an ‘intransitivizer’) is underscored by examples like (2d’), where it can be seen that si (impersonal again) combines with an unaccusative predicate. (2)â•… a.â•… b. c. d. d’.
Gianniâ•… si è lavato G. MPâ•… is washed ‘Gianni has washed himself’ Gianniâ•… si è svegliato G. MP is woken ‘Gianni has woken up’ Le camiceâ•… si sonoâ•… lavate (a secco) The shirts MPâ•… are washed dry ‘The shirts have been dry-washed’ Li si è mangiati themâ•… MPâ•… isâ•… eaten ‘One has eaten them’ Si è arrivati One is arrived ‘One has arrived’
In (2) we have glossed si as MP to suggest ‘middle-passive’. In fact, the closest morphology to si we know of is the so-called middle-passive morphology of languages like Greek or Albanian analysed in chapter 5. In (3) we reproduce the basic data showing that the middle-passive morphology of Albanian has the same range of interpretations as the si morphology of Italian, including reflexive (3a), anticausative (3b), passive (3c) and impersonal (3d). Note that, crucially, in (3d) the middle-passive morphology attaches to an unaccusative verb, just as si does in Italian. (3)â•… Gjirokastër a.â•… la- (h)E- m wash- MP- 1sg ‘I wash myself’ b. zu- (h)E- t wake- MP- 3sg ‘He wakes up’
198╅ The auxiliary:€have/be alternation c. d.
kt kmi la- (h)E- n ŋga aj these shirts wash- MP- 3pl by him ‘These shirts are washed by him’ ŋga ati dil- E- t from there exit- MP- 3sg ‘One exits from there’
The comparison between the middle-passive voice of Albanian and Italian si is made particularly direct by the fact that, whereas in the present, the middlepassive voice of Albanian is lexicalized by the specialized inflection in (3), in the perfective past it is lexicalized by a clitic€– namely u€– which combines with normal active inflections, as shown in chapter 5 and below in (4). (4)â•… Gjirokastër a.â•… u la- it- a MP wash- Prf- 1sg ‘I washed myself’ b. u z- it MPâ•… wake- Prf ‘He woke up’ c. ata u z- it- n ŋga t tiErt they MP wake- prf- 3pl by the others ‘They were woken by some people’ d. ŋga ati u d from thereâ•… MP exited ‘One exited from there’
There is a third way of instantiating middle-passive voice in Albanian through auxiliary selection€– this is the one that concerns us directly, and we will come to it shortly. Returning to the issue at hand, namely the distribution of have and be in the Italian perfect, the descriptive conclusion is that be is found with unaccusatives and with middle-passives. Have is restricted to the complementary set, i.e. transitive/unergative actives. Looking back to the data in (2), it appears that the notions of transitivity and voice in Italian cannot be collapsed; surely, example (2d) is not intransitive, even though it is middle. Indeed it is precisely because of si that Burzio (1986) could not provide a unified characterization of be-selection, but ended up with a disjunctive statement. As indicated at the beginning, here we shall consider the problem of auxiliary selection in a parametric context. Hence, before evaluating whether selection frames for have and be can be unified in Italian, we shall turn to other parametric settings. One of them is especially well-known from Germanic languages like Dutch or German, but is also robustly attested in Romance varieties such as Soazza in (5)–(6). Like Dutch, and like Italian, Soazza has be with
Evidenceâ•… 199 unaccusatives as in (5a) and have with unergatives/transitives as in (5b). The interesting point is that exactly the same distribution holds in the presence of si morphology, as in (6). Thus, the reflexive in (6a), the anticausative in (6b) and the impersonal/passive in (6c), which are formed with transitive/unergative verbs, have auxiliary have. Only the impersonal in (6d), formed with an unaccusative verb, has auxiliary be. (5)â•… Soazza (Grisons) a. som rivo/ rivada I am arrived.m/f ‘I have arrived’ b. o dormi:t I.have slept ‘I have slept’ (6)â•… Soazza a. El/la s a lavo/lavada he/she MP has washed.m/f ‘He/she has washed himself/herself’ b. alâ•… s a Zmortso al tSar it MP has gone.off the light ‘The light has gone off’ c. s a sempro dormit beŋ MP has always slept well ‘It has always been slept well’ d. s e sempro rivo tart MP is always arrived late ‘One has always arrived late’
Reference to a Romance variety like Soazza, as opposed to better known languages like Dutch, allows us to establish a few important points. First, in comparing Italian with Dutch, one may be tempted to conclude that the different auxiliary selection with si and zich respectively is due to their different clitic vs. non-clitic status; the account of Reinhart (1997) and Reinhart and Siloni (2005) is built on this premise. However, comparison of sufficiently close languages like Italian and Soazza shows that different auxiliary selection properties can correlate with essentially stable properties of the middle-passive morphology. Thus, there is no doubt that the s morpheme in (6) has exactly the same clitic prosody, distribution etc. as its Italian counterpart. Second, Dutch zich and Italian si also differ in that zich does not combine with unaccusative verbs to yield the impersonal reading (again we take this to be an independent parameter). But s in Soazza is like Italian si in this respect€– which allows us to show that s simply does not influence auxiliary selection. Rather, selection
200â•… The auxiliary:€have/be alternation is determined by verb class (transitive/unergative vs. unaccusative) in (6) as well. Albanian is in a way the mirror image of Soazza, in that it displays no sensitivity to transitivity alternations, but selects have and be according to voice. Recall from chapter 5 that the present perfect of Albanian, like that of Romance and Germanic languages, is formed by a combination of auxiliary and perfect participle. In the active voice, the auxiliary is have, both with transitives and with unaccusatives, as in (7a) vs. (7b). However, in the middle-passive€– i.e. as part of the conjugation that includes the present in (3) and the simple past in (4)€– the perfect is formed with be. In fact, be followed by the very same perfect participle found in the active (i.e. without the support of either suffixal or clitic material) yields the set of middle-passive interpretations, namely the reflexive (8a), the anticausative (8b), the passive (8c) and the impersonal (8d). (7)â•… Gjirokastër a.â•… E ka la- it- ur it he.has wash- prf- prt ‘He has washed it’ b. ka dal he.has gone.out ‘He has gone out’ (8)â•… Gjirokastër a. t la- it- ur he.is wash- prf- prt ‘He has washed himself’ b. t zu- a he.is wake- prt ‘He has woken up’ c. kt kmi jan la- it- u ŋga aj these shirts are wash- prf- prt by him ‘These shirts have been washed by him’ d. ŋga ati ɐt dal mi from thereâ•… it.is exited well ‘One has exited well from there’
Incidentally, it may be noted that the perfect participle in Albanian never agrees with any of its arguments€– though a full agreement paradigm is available for it and emerges in adjectival contexts (cf. chapter 5). By contrast, Soazza has the same perfect participle agreement system as standard Italian, even if the auxiliary is have, for instance in the reflexive in (6a). In other words, perfect participle agreement is entirely independent of the selection of have or€be.
Evidenceâ•… 201 Another parametric choice attested by well-known Germanic and Romance languages (such as English or Spanish) is formation of the present perfect with the auxiliary have independently of transitivity and voice. In (9) we illustrate this parametric option with a Southern Italian variety which (like Spanish) shows the insensitivity of auxiliary selection to si; and (unlike Spanish) provides an example of the independence of participle agreement from auxiliary selection, since unaccusative participles as in (9b) agree with their argument. (9)â•… Verbicaro (Calabria) a.â•… aj laßa:t (a makn) I.have washed the car ‘I have washed (the car)’ b. a mmurt / mmɔrt s/he.hasâ•… died.m/f ‘S/he has died’ c. s a llaßa:t MPâ•… he.hasâ•… washed ‘He has washed himself’
Another logical possibility, i.e. selection of the be auxiliary in all of the present perfect, appears to be more rarely instantiated€– but can also be found in Romance varieties such as Pescolanciano in (10). While the selection of be with the unaccusative in (10a) or si in (10c) may be familiar from Italian, this is a language that also has be with transitives, as in (10b). Incidentally, it will be seen that the agreement pattern is the one familiar from Italian whereby the feminine morphology in (10b) singles out the internal argument, lexicalized by the accusative clitic. (10)â•… Pescolanciano (Abruzzi) a. sŋg mnu:t I.am come ‘I have come’ b. la her they.are called ‘They have called her’ c. ts E lava:t@ MP he.is washed ‘He has washed himself’
Finally, there is another major pattern of the have vs. be split instantiated by Romance languages, since in some varieties auxiliary selection is sensitive to the reference of the EPP argument. The most widely known split opposes the 1st and 2nd person with be to the 3rd person with have (Kayne 1993; Cocchi 1995; D’Alessandro and Roberts 2010), as in the paradigm in (11) (in
202╅ The auxiliary:€have/be alternation the examples A indicates the selection of have auxiliary and E the selection of be auxiliary). This classical person split is oblivious to verbal class, as shown in (11), as well as to si.1 (11)╅ S.Benedetto del Tronto (Marche)
s vnu:t/ vist E Si E a A mâ•… E t E a A ‘I have come/seen’ etc.
Note that there is no intrinsic association of be with 1st and 2nd person and of have with 3rd person. Thus, varieties are found in which it is the 1st and 2nd person that are associated with have, while the 3rd person is associated with be, as in (12). This parametric choice is less robustly attested, so that in Morcone in (12) it only characterizes the singular, while the plural has have in all persons. However, the limitation of the person split to the singular is an independent parameter, since it can also be seen in a variety like (13) which in the singular has be in the 1st and 2nd person and have in the 3rd. Incidentally, the morphophonology of Morcone (with non-neutralized final vowels) allows us to see that the participle agreement once again follows exactly the same lines as standard Italian€– namely, agreement with unaccusatives (‘come’) and lack of agreement with unergatives (‘slept’). (12)â•…
Morcone (Campania) addo menuto/ durmuto a E emo menuti/durmuto ete ao ‘I have come/slept’ etc.
A A E A A A
(13)â•…
Bisceglie (Apulia) s drmmi:t/ vni:t si a mm av:tâ•… onn ‘I have slept/come’ etc.
E E A A A A
Evidenceâ•… 203 The varieties in (11)–(13) prompt a few general remarks. First, we have seen that although the pattern in (11) may be more robustly attested, the reverse pattern in (12) is equally possible. However, not all parametric choices are reversible. Thus, we are aware of no language in which auxiliary selection is the reverse of, say, Soazza, so that be is associated with transitive and have with unaccusatives. Also, we do not know of any language which patterns in the opposite way to Albanian, selecting have with middle-passives and be with actives. This lack of reversibility will have to be accounted for as part of the overall parametric picture. Second, the picture of auxiliary selection according to person provided in (11)–(13) is greatly simplified with respect to the actually observed variation. A fuller picture will be presented in section 6.2. Importantly, it is not only auxiliary selection according to person that is subject to this fine variation, but also transitivity and voice alternations, as discussed in section 6.4. 6.1.1 Theoretical background Our study of parametrization in auxiliary selection presupposes a number of assumptions concerning the structure and the interpretation of perfects. Perhaps the single most important assumption is that perfects are not monoclausal, consisting of a verb associated with an auxiliary functional projection€ – as in the conception of English auxiliaries of Chomsky (1957, 1981, 1995). Rather, the embedded participle and the matrix auxiliary each define a separate sentential unit (Kayne 1993). The importance of this point cannot be overstated. A decade after Kayne’s (1993) discussion, Bentley and Eythórsson (2003:€460) maintain that perfective auxiliaries are morphological exponents of tense/aspect features on a par with past tense markers in English … On the basis of the above analysis, the selection of perfective auxiliaries in Italian, Dutch and other languages (‘auxiliary selection’) can be considered to be an instance of allomorphy. Â�Accordingly, ‘have’ and ‘be’ are allomorphs of a tense/aspect morpheme which in combination with the past participle forms the analytic perfect. This type of selection might be taken to be comparable to the formation of the past with weak and strong verbs in English. This would imply that ‘have’ is added by a rule and that the smaller class of verbs selecting ‘be’ would have to be memorized … we propose, in the first instance, that auxiliary selection in languages like Italian is due to a morphological rule which is sensitive to the lexical semantics of verbs.
Their evidence includes the fact that ‘agreement markers (qua clitics) attach to the tense/aspect markers (qua auxiliaries)’. They argue that on a ‘clitic climbing/argument composition’€– i.e. restructuring€– analysis, the ‘obligatoriness’ of the positioning of clitics must be stipulated.
204â•… The auxiliary:€have/be alternation Now, in some Romance languages, enclisis on the participle is actually observed. Belletti (1990) notices that standard Italian has ‘absolute’ (i.e. adjunct) participle constructions, in which the participle is associated with sentence-level inflectional properties such as the ability to support an enclitic. If we adopt Kayne’s (1991) theory of enclisis, in which enclisis depends on the high position of the verb in the sentence, we conclude that the participial clause has a C layer as well, hosting the participle itself in enclitic contexts. What is more, the Piedmontese varieties reviewed in chapter 3 show enclisis on the participle in the perfect as well, as in (14). We take this to provide evidence that the participle is a full clause, with the verb in C. In particular, Forno in (14) displays not only enclisis on the participle, but also proclisis on the auxiliary, as in (14a’–b’). This optionality of clitic positioning in Forno parallels that of clitic climbing in Italian ‘restructuring’ constructions, in the sense of Rizzi (1982)€– suggesting that they admit of the same explanation.2 (14)â•… Forno Valle Strona (Piedmont) a.â•… l a viStu -n he has seen us ‘He has seen us’ a’. a n a viSt he us has seen ‘He has seen us’ b. l E lava -s heâ•… isâ•… washed -MP ‘He has washed himself’ b’. a s E lava he MP is washed ‘He has washed himself’
Albanian varieties provide even stronger evidence that the embedded participle syntactically represents an independent sentence (Savoia and Manzini 2007). Thus, in the Arbëresh variety of Portocannone the participle can be introduced by the sentential connective ‘and’, and pronominal clitics are associated with the participial clause rather than with the matrix clause, as illustrated for the active in (15a) and for the middle-passive in (15b). Note that, while in Romance varieties like (14) a clitic associated with the participle appears in enclisis (showing that the participle is in a relatively high position, i.e. C), in Albanian the clitic precedes the participle, as with any other finite verb. This leads us to conclude that participles and finite verbs occupy exactly the same position in the sentence, i.e. I.
Evidenceâ•… 205 (15)â•… Portocannone a.â•… ai kiS E E he had and it ‘He had broken it’ b. ai kiS E u it had and MP ‘It had broken’
tSa- it- ur break- prf- prt tSa- x- ur break- MP- Prt
We can associate with the active in (15b) a structure of the type in (16). Both the kam auxiliary and the participle head their own sentential projection, each of which is associated with a full argument structure (witness the clitics in the participial clause). We assimilate the coordinating particle to a subordinating one, as suggested by Kayne (1994, and references quoted there). In accordance with the theory of complementizers in chapters 1–2, we then construe as an autonomous nominal head, which is selected by the matrix predicate and takes the participial clause as its complement. (16)â•… Portocannone D ai I kiʃ I ε Q u I tʃaxur
Besides the conclusion that perfect participles are bi-sentential structures, we share with Kayne (1993) the conclusion that the have and be auxiliaries are exactly the same verbs that appear in possessive or copular constructions respectively, descriptively as main verbs. The main argument in favour of this conclusion is that the coincidence of the different contexts in which the same lexical forms appear is systematic€– and cannot therefore be imputed to simple homophony. Bentley and Eythórsson’s (2003) approach is incapable of capturing this obvious and important fact. On the other hand, Kayne (1993) also argues that have and be are transformationally related. His idea, which is based on the possessive construction, is that have is essentially an applicative of be, derived through the incorporation
206â•… The auxiliary:€have/be alternation of a (dative) preposition; in other words, possessive have is be–to, as proposed by Benveniste (1966). Similarly, auxiliary have is derived from auxiliary be through the incorporation of a prepositional complementizer. The problem is that, as far as we can see, there is no Romance variety, of the many examined, that provides any morphophonological clue in favour of the proposed derivation (nor any other language that we know of).3 Therefore we simply assume that have and be are two independent lexical entries. In other words, have and be are just what they appear to be€– namely the independently known possession and copular verbs, eventually embedding a participial clause. There is one major piece missing in the structural sketch of participles provided above concerning the EPP argument of the participle. As in all non-finite clauses (including, notably, infinitivals) the EPP position of the participial clause is not overtly realized. According to the theory of empty categories in Chomsky (1981) maintained by Chomsky (1995), the EPP position of nonfinite sentences corresponds either to a trace€– i.e. a copy€– in raising constructions or to a base-generated empty category, conventionally PRO, in control constructions (whether obligatory or arbitrary control). On the other hand, some authors have sought a unified characterization of these environments in terms of the minimalist notion of trace, i.e. copy. Thus, for Hornstein (1999), (obligatory) control as well as raising is derived through movement; raising and control are simply the names for different interpretations of the same syntactic object, i.e. a chain. Because this theory seeks to unify only the bound readings of the embedded subject of non-finite sentences, so-called arbitrary control is excluded from it€– being represented presumably by an empty category again (say pro). Another possibility that has been explored in the literature is that in fact there is no embedded EPP position in non-finite sentences. This stance is explicitly taken by Epstein and Seely (2006), but is also implicit in Manzini and Roussou (2000). It is interesting, therefore, that both Roussou (2009) and Manzini (2009) reject this possibility on the basis of control and raising from finite complements in Greek and Albanian respectively. Since finite complements involve a nominative position, it is evident that control and raising interpretations cannot be linked to the absence of an EPP position. We therefore adopt the standard assumption that all sentences have an EPP argument. At this point, it also becomes relevant that we adopt a representational model of grammar, as defined by Brody (2003), in which chains are an interpretive construct at the LF interface€– and do not depend on a computationally driven process of movement. At the LF interface, the reading of a trace (i.e. a copy) is that of a variable (bound by its antecedent, i.e. the wh–operator
Evidenceâ•… 207 in wh–chains and so on). Similarly, the base-generated empty category (i.e. PRO/pro) is generically closed in the arbitrary reading, so is once again a variable (Lebeaux 1984). Therefore we simply assume that in non-finite clauses, in which the EPP argument is not provided by the morphosyntactic (agreement) structure, it enters the LF interface computation in the form of a variable. In raising and obligatory control, the EPP variable has an antecedent-bound reading; in arbitrary control, it is interpreted by generic closure (in the absence of an antecedent). Finally, we need to say a few words on the notions of voice and transitivity. Descriptively, in standard Albanian have forms the perfect in the active voice, as in (7), and be forms the perfect in the middle-passive voice, as in (8); in present terms, this means that have selects for an active embedded sentence and be for a middle-passive one. Similarly, in Italian (1)–(2) the middle-passive voice, as instantiated by si, is associated with be independently of the transitivity of the verb, while in the active have and be alternate according to transitivity. The latter is the only alternation present in Soazza in (5)–(6) independently of voice. Evidently, explaining these patterns requires a preliminary account of what exactly is meant by the descriptive labels of active and middle-passive or transitive and non-transitive. Furthermore, the two splits, according to voice and transitivity, cannot be unified, since Albanian or Romance varieties like Soazza are sensitive to one but not the other. What appears to be the more complex and therefore problematic of the two notions, i.e. voice, has been studied in chapter 5 in relation to Albanian. Our conclusion regarding the Albanian clitic u and its Italian counterpart si was that what ultimately gives rise to all of the different readings of the middle-passive voice is the property of si/u being a variable. As a result of the presence of this variable, one of the argumental positions of the predicate remains unassociated (in anticausatives/reflexives) or is interpreted only by quantificational closure (passives, impersonal) or is finally associated with an argument which is external to the core predicate (the by-phrase). Therefore middle-passive voice is defined by the lack of closure (or the generic closure) of the argument structure of the verb. By contrast, the theoretical characterization for the notion of transitivity would appear to be a settled question on the basis of Burzio (1986) and the relational grammar work on which it is based. According to Burzio (1986), transitive verbs have both an internal and an external argument€– linked to the sister of V position and to the sister of I (EPP) position respectively. Unergatives only have an external argument, which is linked to the EPP position. Finally, unaccusatives have only an internal argument, which is promoted to the EPP
208â•… The auxiliary:€have/be alternation position. Regarding the latter, it should not be forgotten that the movement derivation of unaccusative verbs is meant to unify them with middle-passives for the purposes of be-selection. Yet we have just seen that what is classically taken to be the strongest argument in favour of a uniform movement structure for middle-passives and unaccusatives in reality works against it. For in languages like Albanian, it is only middle-passives that combine with be, to the exclusion of unaccusatives€– and the reverse is true in Soazza. We have now adopted a characterization of middle-passives as predicates with an open variable in their argument structure, either not interpreted or interpreted through quantificational closure. But the standard construal of unaccusatives as predicates with a single argument slot, saturated by the EPP argument, leaves no room for non-interpreted or quantificationally closed argument slots. Therefore this simple characterization is sufficient to distinguish unaccusatives from middle-passives for present purposes€ – as already suggested in chapter 5 in relation to Albanian. The same characterization is also sufficient to distinguish them from transitives and unergatives€– for transitives have two arguments, and, following Hale and Keyser (1993) and Chomsky (1995), unergatives can be treated essentially as concealed transitives in which the internal argument has incorporated into the verb. 6.2
Auxiliary selection independent of transitivity/voice
Consider Pescolanciano in which the auxiliary of the perfect is be, as in (10). From the present perspective, Pescolanciano can simply be described as a language in which be embeds any participial clause without any restrictions. By contrast, in Verbicaro, the perfect is systematically formed with the auxiliary have, as in (9). However, we cannot simply say that have in Verbicaro embeds just any participial clause, with no restrictions, since passives in this language, as in standard Italian or English, are formed by the participle embedded under be. Recall that in the middle-passive perfect of Albanian varieties, jam ‘be’ is followed by an invariable (non-agreeing) participle, as in (7)–(8). By contrast, in the passive it is followed by a participle featuring both an agreement inflection and a preposed article, as discussed in chapter 5; a relevant example is reproduced in (17). (17)â•…
Gjirokastër jan t@ vESu/ vESua they.are Art dressed/dressed-f. ‘They are dressed’
Auxiliary selection independent of transitivity/voiceâ•… 209 We take this contrast to imply that the present perfect in (7)–(8) and the passive in (17) involve two different embedding structures. While the evidence reviewed in section 6.1 leads to the conclusion that the embedding in (7)–(8) is sentential, a nominal embedding is involved in (17), as discussed more in detail in chapters 5 and 7. We tentatively propose that the same distinction between sentential and nominal embedding of the participle holds (in some form or other) in Romance as well. Indeed, a contrast between lack of agreement in the perfect and agreement in the copular (passive) construction can be found in some Romance varieties. Thus, in (18), the free alternation between have and be in the perfect does not have any effect on the participle, which maintains its invariable, non-agreeing format. As shown in (18b–b’), on the other hand, in the same languages be requires an agreeing adjective in the copular construction. (18)â•… Montebello Ionico/ Saline Ioniche (Calabria) a. Era/ Eri/ Era/ Erumu/ Eruvu/ Erunuâ•… durmutu/ vinutu I.was etc. slept/ come a’. aiva/ aivi/ aiva/ aivumu/ aiuvu/ aivunuâ•… durmutu/ vinutu I.hadâ•… etc. slept/come ‘I had come/ slept’ etc. b. Era staŋku I.was tired b’. Erumu staŋki we.were tired(pl)
We thus factor out passive as involving a different structure of participial embedding with respect to perfects. If so, the comparison between Pescolanciano in (10) and Verbicaro in (9) yields a simple parameter whereby participial clauses are selected by either have or be to the exclusion of the other. 6.2.1 Auxiliary selection according to person The patterns of auxiliary selection according to person in (11)–(13) can be obtained at this point by letting the parameter just formulated€– i.e. the selection of participial clauses by have or be€– interact with the person split between 1st/2nd person and 3rd person. The latter is independently motivated in the typological literature. Thus, DeLancey (1981) argues that languages with so-called ergativity splits, i.e. alternations between the ergative (/absolutive) case system and the nominative (/accusative) system, most commonly oppose 1st and 2nd person to 3rd. Person splits are also found in Romance languages. Restricting ourselves to clitics, 1st and 2nd person object clitics differ
210â•… The auxiliary:€have/be alternation from 3rd person ones with respect to their distribution, their morphological make-up (gender and Case distinctions), their agreement properties, and more (for instance, the drop phenomenon studied by Savoia and Manzini (2010)). According to Manzini and Savoia (2007), the person split, in its various manifestations, depends on the fact that the speaker and the hearer (and the sets including them) are anchored directly in the universe of discourse, independently of their role within the event; on the other hand, non-participants in the discourse depend directly for their characterization on the position assigned to them within the structure of the event.4 The same split can be defined in more traditional frameworks by a feature [±discourse participant] (e.g. Legendre (2010) has [±local]). The reason we do not adopt such a characterization is that in our morphosyntax, there are only positively specified properties. In other words, the 3rd person cannot simply be characterized in negative terms, as excluding speaker and hearer reference (Benveniste 1966). Rather, 3rd person reference is characterized in positive terms, by its necessary anchoring in the event. In this sense, we also differ from Harley and Ritter (2002), who assume monovalent (privative) features, but at the same time adopt the default view of 3rd person, as characterized by the absence of the Participant feature node. Let us now cross the have vs. be parameter defined with respect to Pescolanciano and Verbicaro with the person split€– so that have and be can be sensitive to the reference of the EPP argument. The crossing should in principle yield two systems. In one system, be is associated with a 1st and 2nd person EPP argument while the 3rd person requires have€– while in the mirrorimage system, it is the 1st and 2nd person that require have while the 3rd person is associated with be. In other words, the crossing of the two parameters yields the systems observed, at least in the singular, in S.Benedetto in (11) and Morcone in (12) respectively. It is important to stress that we take the person split to interact with the selectional properties of auxiliaries at the core syntactic level. By contrast, for Bentley and Eythórsson (2001:€ 71), ‘alternation according to person is part of a grammatical person marking system on verbs’. Similarly Loporcaro (2007:€ 186) argues that ‘where precisely (in which verb persons) the morphemes ‘have’ and ‘be’ occur … is however a matter of morphology, not syntax’. What seems crucially to motivate their conclusions is the fact that there exist many more possible distributions than the 1st/2nd vs. 3rd exemplified so far. Thus, for Bentley and Eythórsson (2001:€71), ‘the mixed paradigm does not appear to be attributable to syntactic or semantic principles associating each auxiliary with a particular grammatical person. This possibility is ruled
Auxiliary selection independent of transitivity/voiceâ•… 211 out by the variety of existing patterns … Thus we would not subscribe to the view that there is a rule of auxiliary selection according to person’. Similarly, for Loporcaro (2007:€186), ‘what is inconceivable … is for this empirical variability to be directly encoded into structural categories in a one-to-one correspondence’. Here we will argue that, on the contrary, a syntactic account of this considerable variation is exactly what is needed. Quite intricate patterns can be gleaned from the tables in Manzini and Savoia (2007) and Legendre (2010). The number of languages described by Manzini and Savoia (2005, 2007) is obscured by the fact that several varieties display one or more alternations, both free and not free (such as alternations according to transitivity/voice and according to whether the following verb starts with a consonant or a vowel). In order to make the possible patterns emerge more clearly, in Table 6.1 we have reformatted the data in Manzini and Savoia (2007, Appendix to chapter 6). Wherever an alternation was present, we have written down two separate paradigms. Free alternations have been treated so as to minimize the resulting pattern€– i.e. where several of them were present, we chose be or have throughout. Even a cursory look at Table 6.1 reveals that the singular and the plural do not necessarily pattern alike€– in fact they most often display separate patterns. This has already been illustrated above by the contrast between S.Benedetto in (11) and Bisceglie in (13)€– where the same person split occurs throughout the paradigm in S.Benedetto and only in the singular in Bisceglie. Let us now consider the singular. The richest possible grammar is one in which not only the 1st/2nd vs. 3rd person split is allowed, but also 1st person vs. 2nd/3rd and 2nd person vs. 1st/3rd. Once these three possibilities are crossed with the two auxiliaries, six patterns emerge. All of these are attested in Table 6.1. Now the splitting of 1st from 2nd person€– i.e. of the two members of the discourse-anchored set€– is also a long-standing issue in the typological and formal literature. In typological approaches, where the discourse- vs. non-discourse-anchored split is taken to be a cut along an animacy/agenthood hierarchy (Dixon 1994), the 1st singular is taken to precede the 2nd singular in the hierarchy in that ‘the speaker is for him- or herself, the quintessential agent’ (Dixon 1994:€90). Here, of course, we have discounted the idea that animacy and/or agenthood enter into the determination of person-split phenomena, since our system does not contemplate the possibility of hierarchies, but only of discrete splits. Similarly, the formal feature hierarchy of Harley and Ritter (2002) yields ‘speaker’ as the unmarked member of the ‘discourse participant’ set. As mentioned above, we cannot have recourse to feature geometries which, even if they only make use of privative features, crucially imply a notion of underspecification (thus 3rd person is lack of specification for ‘discourse participant’). At
212╅ The auxiliary:€have/be alternation Table 6.1 Distribution of be (E) and have (A) according to person in the present perfect (in Central and Southern Italian varieties)
(ii) (vi) (xi) (i) (iv) (v) (vii) (viii) (ix) (iii) (xii) (x) (xiii) (xiv) (xv) (xvi) (xvii) (D)
Roccasicura Vastogirardi Secinaro cf. (x) Poggio Imp. Gallo Matese Monteroduni = (i), (iv) Colledimacine S.Benedetto T. =(vii) Viticuso cf. (vii) Capracotta Agnone = (xiv) Sassinoro Bisceglie Ruvo cf. (x), (xiii), (xv) Popoli cf. (xiii) Padula = (x), (xiii), (xiv), (xv) Molfetta = (xiii) Gravina
1sg
2sg
3sg
1pl
2pl
3pl
A A E E A
E E E E E
E A A E E
E E E E E
E E E E E
E E E A A
E
E
A
E
E
A
A A E A E A
E E E E E E
A E A A A E
E A A A A A
E A A A A A
A E E A A A
E
E
E
A
A
A
A E E
A A A
E E A
A A A
A A A
A A A
the same time, the data reviewed here show that these approaches are problematic, in so far as the pattern isolating the 2nd person from the 1st person and the 3rd is also possible. Manzini and Savoia (2007) state that ‘the property that sets apart the 1st singular is one of pragmatic salience’. This universal, in itself obvious, yields one of two logically possible results. Either 1st singular is isolated while 2nd and 3rd are treated alike, or its pragmatic salience allows for it to be treated together with event-anchored 3rd person, allowing 2nd person to be isolated. In other words, if a schematic representation is desired, our system amounts to (19). (19a) and (19b) generate the split between discourse-anchored and eventanchored referents as well as the internal splitting of the discourse-anchored referents that the auxiliary selection data require. (19) a. split between discourse-anchored referents and event-anchored referents b. pragmatic salience of speaker reference
Auxiliary selection independent of transitivity/voiceâ•… 213 Let us now consider the plural. The plural can pattern independently of the singular, but it appears to be more severely restricted than the singular. In our data, the plural of person split varieties only has two options, namely no split or the canonical split opposing 1st/ 2nd person to 3rd person. It is true that the varieties tabulated by Legendre (2010) on the basis of the existing literature show the possibility of other person split patterns in the plural as well. Interestingly, however, in these cases the plural is consistent with the singular. Therefore we may tentatively add to our picture the general possibility of the plural being consistent with the singular, although we will mostly disregard this aspect of the problem. In Manzini and Savoia (2007), we concentrate on non-split plurals of person-split varieties. In order to explain the lack of person split in the plural, ‘considerations concerning the actual denotation of the so-called 1st and 2nd person plural’ are invoked. ‘Thus the 1st person plural does not necessarily denote a plurality of speakers (though it may), or the speaker and hearer only (though again it may); rather its denotation routinely involves one speaker and a certain number of other individuals that are being referred to together with the speaker. The same is true for the 2nd person singular, which does not necessarily (or normally) denote a plurality of hearers but simply refers to the hearer taken together with a certain number of other individuals’. In more traditional terms (cf. Bobaljik 2008), ‘we’ is 1st (+ 2nd) (+ 3rd) while ‘you’ is 2nd (+ 3rd). Because of this referential structure of the so-called 1st and 2nd plural, it is reasonable to propose that even varieties that activate the person split (19a) in the singular may not do so in the plural. Let us now consider those varieties in which the plural displays a split Â�pattern and the latter is not simply consistent with the singular. If the generalization suggested by Table 6.1 is correct, namely that plural splits are restricted to the canonical discourse- vs. event-anchored sets, this can again be derived on the basis of (19). What we have just suggested is that the referential structure of the plural may block the expression of (19a). If the same reasoning is applied to (19b), we obtain the result that distributions isolating the 1st person from the 2nd are blocked. In other words, (20) yields the asymmetry between singular and plural person splits. (20)â•… (19a), (19b) are not defined in the plural
There is one further question left, namely how the singular and plural patterns generated so far can or cannot combine. A glance at Table 6.1 shows that when the plural is associated with have, all eight logically possible combinations for the singular are attested, including the seven in Table 6.1 and those
214â•… The auxiliary:€have/be alternation that show have throughout the paradigm. Matters change when we consider be or split patterns in the plural. In this case, the patterns found in the singular are severely cut down to about half of those which are possible in principle. The best predictor of the actually observed distribution, at least in Italian varieties, is the 2nd person singular. If this is associated with be, then the plural can have be; if the 2nd person singular is associated with have, then no combination with be in the plural is possible. As ever, our problem is not stating the restriction, which emerges fairly clearly from our data, but rather understanding why it would hold. Two sets of considerations appear to be relevant. First, as discussed above in connection with (20), the referential structure of the plural is more complex than that of the singular; in particular, the elementary discourse-anchored ‘speaker’ reference is simpler than the plural referents. Second, though we are concentrating here on the embedding of nominal predicates (participles) under have and be in perfects, we have mentioned in section 6.1 that the languages at hand also have copular embeddings (including passives) which involve be. Therefore a language in which have embeds nominal predicates (participles) in the perfect has a categorial split with be when nominal embeddings as a whole are considered. We surmise that if the 2nd singular is associated with the relatively complex grammar characterized by have in the perfect and be in copular embeddings, then the more complex plural referents must also be associated with have in the perfect (vs. be as the copula), along the lines of (21).5 Since the 2nd singular is crucial in defining this implication, we must conclude that the ‘hearer’ is singled out by the lack of any referential property, including pragmatic salience, beyond mere anchoring in the discourse. (21)â•…Categorial splits (e.g. copula be vs. auxiliary have), if supported by simple referents, are supported by complex referents as well.
In purely extensional terms, once we take into account the simpler structure of the plural with respect to the singular in (20) and the predictor role of the 2nd singular in (21), we are left with only two predicted and unattested languages, namely *AEA AAE and *EEE AAE. If we allow for all plural splits identical to those in the singular, two further languages are predicted and not attested, namely *AEE AEE and *EAE EAE. By comparison, Legendre’s system allows for too much variation in the plural; if we exclude plurals which do not split along the canonical discourse- vs. event-anchored axis, we cut down the overgeneration from thirteen to three. Second, the system undergenerates, by incorrectly excluding the possibility of be in the plural combining with
Auxiliary selection independent of transitivity/voiceâ•… 215 person splits in the singular. Counterexamples noted by Legendre (2010:€198) include the patterns exemplified by Vastogirardi in Table 6.1 (vi) and Secinaro in Table 6.1 (xi), but there is also Poggio Imperiale in Table 6.1 (i) (cf. Manzini and Savoia 2007, Appendix to chapter 6). She also undergenerates three other languages, namely Sassinoro (Table 6.1, x) (associated with Arielli by Legendre (2010:€198)), as well as Gallo Matese (Table 6.1, iv) and Agnone (Table 6.1, xii). As far as we can tell, the present grammar does not undergenerate. But this is not the main aspect of the problem that we are concerned with. For us the issue is not simply devising a system that describes the observed facts€ – but rather devising a sufficiently restrictive system. A crucial role in our attempt at restricting the variation space is played by the person split in (19a). Now, Loporcaro (2001, 2007) criticizes our previous work (Manzini and Savoia 1998) for proposing that what he calls ‘mixed auxiliation systems’ are driven by a person split. He objects to this proposal on both empirical and theoretical grounds. His empirical objection is that ‘the distribution of aux E/H in mixed systems is not predictable, overall, on the basis of an alleged person ergativity split’ (Lopocaro 2007:€196). The point of much of the previous discussion is precisely that the fundamental person split, i.e. (19a), underlies the observed variation, although it is masked by other parameters (in particular (19b)). Loporcaro’s second objection is more theoretical in nature, namely that ‘since … we are dealing with closely related varieties, the null assumption is that they should not diverge on such an essential property as dominant alignment’ (2007:€196). The relevant notion of dominant alignment is defined by Nichols (1992:€ 92) as ‘the pattern found in the majority of parts of speech … or the nominal rather than pronominal pattern, or … the most semantic of the patterns’. In such terms, an ergative alignment found in some persons in the present perfect would simply not misalign Romance varieties from what is presumably considered to be their dominant accusative pattern, making Loporcaro’s point irrelevant. Even so, Loporcaro is correct in perceiving a fundamental contrast between his position, reducing person split to a ‘morphological’ property (as discussed above), and the present construal of this split as no less a core syntactic phenomenon than ergativity splits. Even deeper than this, the contrast is between externalist views of parametrization, correlating the structural depth of parameters with the genetic/typological distance between the languages involved, and the present internalist view, under which no correlation is expected between structural depth and genetic/typological divergence. Our position, as discussed in more detail in the introductory chapter, is that macrophenomena can be
216â•… The auxiliary:€have/be alternation decomposed into the same elementary conceptual components that determine local lexical variation€– and in fact the latter is the true matrix of perceived macroparameters. 6.3
Splits according to transitivity/voice
6.3.1 Auxiliary selection according to voice Let us turn to Albanian. Going back to our discussion in section 6.1, we should be able to derive the auxiliary selection patterns from the simple statement that in Albanian, be selects the properties that characterize middle-passive voice in the embedded participle, i.e. in present terms the non-closure (or generic closure) of the argument structure of the verb. Have selects the complementary properties. We begin with the passive interpretation of a sentence like (22), cf. (8a). The internal argument slot of the participial predicate is associated with the embedded EPP variable (cf. section 6.1) ultimately controlled by the matrix EPP argument. The external argument slot is interpreted through binding by a generic operator (i.e. as an implicit agent). What jam ‘be’ selects for is this latter property, i.e. the generic closure of the argument structure of the participial predicate. To be more precise, in passives the external argument can also be assigned to the object of a by-phrase. The latter, however, is an adjunct€– so that we may consider the external argument to remain unbound as far as the core embedded predicate/event is concerned. (22)â•… Gjirokastër I
Əʃt
I laitur (x,y)
Consider next the reflexive and anticausative readings. The relevant structures are entirely parallel to that of the passive, as sketched in (23) for (8b), which has a salient anticausative reading. The internal argument slot of the participle is once again associated with the embedded EPP variable, which in turn is controlled by the EPP argument of the matrix clause (represented in (23) by the verb inflection). The same is true in the reflexive reading, which is harder to obtain in (23) merely for pragmatic reasons (cf. the English ‘I woke myself up’). The difference between the anticausative and the reflexive
Splits according to transitivity/voiceâ•… 217 reading is interpretive, having to do with the degree of intentionality (agency, etc.) attributed to the single participant in the event. Under both readings, the external argument of the embedded predicate is an open variable in the LF structure, neither being associated with any argument nor being quantificationally closed. Thus jam ‘be’ selects for reflexives and anticausatives in that their external argument remains unassigned/open. (23)â•… Gjirokastër I ǝʃt I z� ua� (x,y)
Finally, the impersonal construal of jam–participle structures, exemplified in (8d), corresponds to structures like (24). In (24), the unaccusative predicate dal ‘I go out’ is associated with a single argument slot, assigned to the EPP variable of the participial clause. The latter, in accordance with section 6.1.1, can be generically bound (and determine the generic closure of the matrix EPP argument as well). This means that again jam ‘be’ selects for an embedded structure containing a generically closed variable€– not assigned to argumental material. (24)â•… Gjirokastër I ǝʃt I dalǝ (x)
Consider, by contrast, the selectional properties of kam ‘have’. In (25a), corresponding to the transitive active sentence in (7a), both argument slots of la- ‘wash’ are filled by argumental material:€the internal slot by the accusative clitic ‘him/her’, and the external slot by the variable EPP argument of the participle, ultimately linked to the matrix EPP argument of kam. Similarly, in (25b) the single argument slot of dal- is satisfied by argumental material, ultimately corresponding to the matrix EPP position. Therefore kam selects predicates with a closed argument structure, in the sense that no free or generically closed
218â•… The auxiliary:€have/be alternation variables are instantiated within it. This restriction on kam, or equivalently the restriction on jam ‘be’, define the meaning of descriptive terms such as active and middle-passive as applied to Albanian auxiliary structures. (25)â•… Gjirokastër a.â•… [NE [I ka [I laitur (x, y) b. [I ka [I dal@(x)
As discussed in more detail in chapter 5, in Albanian the jam–participle construction6 alternates with other lexicalizations of the middle-passive voice, namely through a specialized –(h) inflection and through the u clitic. Nevertheless, although the morphosyntax involved in u/–h structures and in jam–participle structures differs, they yield interpretively equivalent results. In a theory upholding Interface Uniformity in the sense of Culicover and Jackendoff (2005), one would be led to assume that their fixed meaning corresponds to a fixed underlying structure, embedded by the varying surface realizations. For instance, the underlying structure could involve movement from the object to the EPP position. The objection that we raised in chapter 5 is that such an approach does not capture important properties of the structures involved, for instance the fact that u is a bona fide object clitic. Thus, under the present approach, different surface realizations can correspond to different underlying structures€– whose interpretive equivalence is established directly at the interface. 6.3.2 Auxiliary selection according to transitivity Let us now turn to the Soazza pattern. If the discussion in section 6.1.1 is correct, this can be accounted for simply by saying that be selects mono-argumental predicates€– which is the construal suggested here (and in chapter 5) for the notion of intransitivity. This notion excludes transitives/unergatives, both in the active and middle-passive voices, which will instead combine with have. Recall that in section 6.1.1, we construed unergatives as two-place predicates (i.e. transitives) with an incorporated internal argument. As an illustration, (26a–b) provide structures for the active of an unergative verb (5b) and for its impersonal (6c); (26c) provides the structure for the reflexive of a transitive (6a). The transitive/ergative properties of the verbs embedded in (26) are sufficient to determine selection of have in the Soazza language, though in other respects their structures differ. Thus, in the active (26a), the external argument slot is ultimately saturated by the finite verb inflection. In the impersonal in (26b), the external argument is quantificationally closed, and in the reflexive (26c) it remains unassigned€– while the finite verb inflection binds the internal argument slot.
Splits according to transitivity/voiceâ•… 219 (26)â•…
Soazza a. [I [I dormit (x, y) b.â•… [D Elâ•… [Q tsâ•… [I a [I lavo (x, y) c. [Q s [I a [I dormit (x, y)
The structures in (27a–b) concern the active (5a) and the impersonal (6d) of an unaccusative, i.e. by hypothesis, a verb with a single argument slot. Again, this property determines the selection of be independently of the other properties of the structure. Thus, in the active (27a) the single argument slot is closed by the verb inflection, while in the middle-passive (27b) it is closed only by generic quantification (yielding the impersonal reading). (27)â•… Soazza a.â•… b. [Q s
[I som [I rivo (x) [I e [I rivo (x)
At this point, on the basis of the discussion of Soazza and Gjirokastër, we can also define auxiliary selection in Italian. Simply put, in Italian the set of contexts that involve be with the perfect participle results from the union of the Albanian contexts (middle-passives) with the Soazza ones (unaccusatives). In (28) we provide logical forms for the active transitive (1a) and for its reflexive and passive counterparts (2a) and (2c). The two argument slots of the transitive predicate in (28a) are satisfied by le camicie and the 1st person singular verb inflection. By contrast, in the middle-passives (28b–c) the internal argument slot of the verb is ultimately assigned to the EPP argument Gianni or the finite verb inflection. The result is that the external argument is not assigned in (28b), yielding the reflexive reading, and is only generically closed (i.e. interpreted as an implicit agent) in the passive in (28c). Therefore, as in Albanian, the auxiliary will be have in (28a), but be in (28b–c). (28)â•… a. [I ho [I lavato (x, y)â•… le camicie b.â•… [D Gianni [Q si [I è [I lavato (x, y) c. [D le camicieâ•… [Q si [I sonoâ•… [I lavate (x, y)
In (29a–b), on the other hand, we provide the structures for an unaccusative (monoargumental) predicate, both in the active (1b) and in the impersonal (2d’). In this case, the auxiliary is be, as in Soazza, independently of how the unique argument slot is satisfied, namely by an argument (i.e. the verb inflection), as in the active (29a), or by generic closure of the si variable, as in the impersonal (29b). (29)â•… a. [I sono [I arrivato(x) b.â•… [Q si [I è [I arrivati(x)
220â•… The auxiliary:€have/be alternation Now, the fact that the contexts that involve be in Italian are the union of those that involve be in Soazza and Albanian could mean that a disjunctive statement is present in the lexicon of Italian€– or that unaccusatives and middle-passives form a natural (super)class. The latter is of course the line pursued by classical generative theorizing, which construes movement as the common characteristics of the two subgroups. Here we propose a distinction between ‘defective’ argument structures, including those with an unbound or generically bound variable (middle-passive), or those with one argument only (unaccusatives) and ‘non-defective’ argument structures (defined as the complement of the above). From this perspective, Albanian is sensitive to voice, Soazza is sensitive to transitivity, and Italian is sensitive to ‘defectiveness’ in defining the selection frames for auxiliaries. Going back now to more descriptive matters, we note that Albanian, as shown in (22)–(25), differs from Romance languages like Soazza, as in (26)– (27), or Italian, as in (28)–(29), not only with respect to auxiliary selection, but also because in Albanian the be auxiliary does not co-occur with other middle-passive morphology€ – such as the si clitic that co-occurs with be in Italian. Moreover, there is no necessary connection between these two facts. Thus, consider the Romansh variety of Trun in (30). Trun has the same auxiliary selection pattern as Italian, since be combines with unaccusatives in (30c) and with middle-passives (reflexives) in (30b), and have with transitive/unergative actives in (30d). As in Italian, furthermore, the reflexive is formed with the se clitic in the simple tenses in (30a). Yet in the perfect, the reflexive can be formed without the se clitic, so the contrast between the middle-passive (30b) and the active (30d) rests solely on the fact that the active has the auxiliary have, while the middle-passive has the auxiliary be€– reproducing in this respect the pattern of Albanian. (30)â•… Trun (Grisons) a.â•… Elts se lavan theyâ•… MPâ•… wash ‘They wash themselves’ b. Elts ain (se) lavai they are MP washed.pl ‘They have washed themselves’ c. jau sUn veJia I am come-f. ‘I have come’ d. jau ai durmiu I haveâ•… slept ‘I have slept’
Splits according to transitivity/voiceâ•… 221 One parameter that we have disregarded in the preceding discussion concerns whether the middle-passive morphology can be associated with a dative argument slot or not; for instance, it can in Italian, as in (31), but not in Albanian, as in (32). (31)â•… Gianni si è lavato i piedi G MP is washed the feet ‘John washed his feet’ (32)â•…
Gjirokastër ai lan fitirn he washes face-the ‘He washes (his) face’
Whatever the explanation turns out to be,7 what interests us here is that in a language like Italian, dative si is treated like the other instances of si considered so far for the purposes of auxiliary selection, so that in (31) the auxiliary is be. In other languages which otherwise follow the auxiliary selection pattern of Italian, however, dative si co-occurs with auxiliary have, as illustrated in (33a) for Làconi; (33b) provides a comparison with si sentences with be. (33)â•… Làconi (Sardinia) a.â•… s a ssamunau i mmanuzu MP has washed the hands ‘S/he washed his/her hands’ b. Es samuna-u/a MP is washed-m./f. ‘S/he washed him/herself’
Since we theorized that in Italian-type languages, be selects a ‘defective’ argument structure in the embedded participle (i.e. with an open or generically closed variable), we must conclude that in Italian itself, dative si structures count as defective in the relevant sense; however, in languages like Làconi they do not. Therefore, for Italian (31) we maintain an analysis parallel to that of other si sentences, where the argument slot filled by the si variable is ultimately assigned to the EPP argument. In contrast, we must conclude that in Làconi, the same structures are treated as active, with the dative argument slot assigned to si and the external argument slot assigned to the EPP argument. This parameter in turn seems to belong to a family of parameters having to do with how the dative is computed for the purposes of argument structure, i.e. roughly as a second internal argument, or rather as an oblique/adjunct of sorts. The latter construal still leaves open the possibility of having a bound reading of si€– but without triggering the properties of middle-passive voice, including be-selection.
222â•… The auxiliary:€have/be alternation 6.3.3 Irreversibility One major fact not yet accounted for is that auxiliary selection according to voice and/or transitivity is non-reversible, in the sense of section 6.1. In other words, we know of no language in which the middle-passive is selected by have and the active by be (i.e. the reverse of Albanian) or in which unaccusatives are selected by have and transitives by be (i.e. the reverse of Soazza). Loporcaro (2007:€189–92) addresses a similar issue in rather different terms. For him, ‘E/H selection varies stepwise, across Romance, along [a] scale’, roughly unaccusative€ – reflexive€ – transitive/unergative. Crucially ‘varieties do not seem to occur that show any one of the scattered distributions of the two auxiliaries’, i.e. have in unaccusatives, be in reflexives and have in transitives/unergatives. On the basis of the evidence discussed here, this approach is descriptively inadequate, since Albanian is a language which has have with unaccusatives and transitives while be is associated with ‘reflexives’€– in fact middle-passive voice. From a theoretical point of view, the available evidence supports the conclusion that there are no parametric scales, but, rather, fine parametrization results from the crossing of discrete properties associated with lexical items (see also the discussion in section 6.4). Returning to the question of (ir)reversibility, a relevant observation is that in all Romance and Albanian varieties, the be auxiliary of the perfect is also the copula (cf. section 1).8 Consider, then, a simple Italian copular sentence like (34). This has a single argument slot, that of the embedded predicate ‘happy’, which becomes associated with the matrix EPP argument (represented by the finite inflection of the copula). This raising analysis (Moro 1997) implies, of course, that be is ‘defective’ in the sense defined at the end of the last section, since at the very least it lacks an external argument. (34)â•… Sono contento I.am happy ‘I am happy’
As for have, in both Romance and Albanian varieties it is independently attested as a possession verb and as a necessity modal, as in Italian (35). Have as a possession verb in (35a) is fairly uncontroversially a transitive (i.e. twoplace) predicate. In turn, the modal reading of have9 in (35b) can be construed by analogy with that of the necessity modals in other Romance languages, e.g. dovere ‘must’ in Italian. Independently of the present theory of auxiliaries as main verbs, the relevant modals in Italian are widely assumed to be main verbs subject to ‘restructuring’ (Rizzi 1982, pace Cinque 2006). In particular, the deontic reading is taken to depend on the fact that the necessity modal
Finer parametrizationâ•… 223 behaves as a control verb. In other words it selects a sentential complement and an external argument, the latter controlling the EPP argument of the embedded sentence. Thus, on both readings, have can be characterized as a ‘nonÂ�defective’ predicate. (35)â•… a.â•… b.
Ho tre figli I.have three children ‘I have three children’ Ho da fare questo I.have to do this ‘I have to do it’
Given this background, the non-reversibility of the auxiliary selection patterns according to transitivity and/or voice can be described by saying that the ‘defective’ be is restricted to selecting ‘defective’€– i.e. intransitive and/or middle-passive€– predicates; conversely, the transitive active have is restricted to selecting transitive and/or active predicates. In other words, the nonÂ�reversibility of the selection pattern corresponds to a fairly obvious uniformity requirement on the argumental structures of the auxiliary and the embedded predicate. 6.4
Finer parametrization
The observed parametrization in auxiliary selection is even more fine-grained than we have illustrated so far. In particular, in section 6.4.1 we shall consider varieties in which auxiliary selection according to transitivity/voice and according to person interact in various ways. Before we do so, however, we should briefly mention the literature concerning variation in have/be alternations according to transitivity. Sorace (2000, 2004) argues that only a core class of unaccusative verbs takes auxiliary be in languages sensitive to transitivity, notably change-of-location verbs like come. With these verbs, furthermore, the choice of auxiliary be is stable within each given language. With other verbs, auxiliary selection varies across languages and correspondingly there is a greater or lesser amount of variation within each given language. For instance, Hoekstra and Mulder (1990) argue that in Dutch, the telic construal of manner-of-motion verbs results in be-selection, while atelic construals result in have-selection. Yet Sorace (2000) points out that Italian follows this pattern much less systematically than Dutch, and that in French the auxiliary is have independently of telicity. Sorace (2000, 2004) concludes that intransitive verbs are organized along an auxiliary selection hierarchy (ASH)
224â•… The auxiliary:€have/be alternation defined primarily by the degree of telicity of the verb and secondarily by its degree of agentivity ranging from verbs of ‘telic dynamic change’ that categorically select BE to verbs of atelic ‘non-motional activity’ that categorically select HAVE in languages with auxiliary selection. Between the two extremes are verbs that display variable behaviour.â•…â•… (Cennamo and Sorace 2007:€67)
The ASH is an ‘empirical generalization’ and ‘needs to be accounted for within a formal model of the syntax-lexicon interface’ (2007:€67–8).10 As discussed in section 6.3.3 in connection with Loporcaro’s (2001, 2007) work, we have reason to reject the idea that parametrization is organized along scales. More specifically, consider Cennamo’s (2001) evidence from Campanian varieties. These data are somewhat at variance with the ASH€ – since they seem to indicate that definite change-of-state verbs (‘to be born’, ‘to die’) are core unaccusative verbs. Verbs denoting inherently telic change of location (‘to come’, ‘to go’) seem to occupy a peripheral place in the unaccusativity hierarchy. For, in some varieties they are syntactically encoded as unergatives (Cennamo 2001:€433). Since, in terms of the ASH, change of location > change of state, Cennamo’s evidence could be taken to show that the be/have parameter does not cut a point on a scale, further undermining the idea that the variation space could be generally organized through scales.11 Therefore, for the purposes of the present discussion, we simply maintain that where the same predicate takes have or be depending on the language, it is construed as monadic (unaccusative) with be and as dyadic (i.e. a concealed transitive) with have, although admittedly we have no insight to offer on this matter.12 6.4.1
Interactions between auxiliary selection according to transitivity/ voice and according to person In what follows, we consider the fine variation associated with interactions between auxiliary selection according to person and according to transitivity and voice. The data have attracted some attention in the literature, though not necessarily for the reasons deemed relevant here. For instance, Cennamo (2001, 2008) notices that in Sorrento, be occurs in the 3rd person singular with (change of) state verbs; in all other cases the auxiliary is have. However, for Cennamo, ‘what is interesting about the contemporary Campanian data on auxiliary selection is the fact that the variation appears to follow the Unaccusativity gradient’ (2008:€135), since she dismisses the person split as ‘part of a system for marking persons on the verb’ (2001:€447, our translation). Thus, what we
Finer parametrizationâ•… 225 take to be complex variation involving the crossing of two parameters (transitivity and person split) is considered in Cennamo’s work only for its interest with respect to one of these. Let us begin by illustrating the languages in (36)–(37), in which some persons in the paradigm are associated with a single auxiliary, while the others alternate between have and be essentially according to the pattern of standard Italian. Specifically, in Colledimacine, as shown in (36), the 3rd person is sensitive to transitivity and voice, as shown in (the relevant forms of) (36a) and (36c), while the 1st and 2nd persons are associated with auxiliary be, as in (36a–b). In Aliano, as shown in (37), the 3rd person singular alternates between have and be according to transitivity/voice, as in (37b) and (37c), while the other persons take have, as in (37a). (36)â•… Colledimacine (Abruzzi) a. so/ Si/ e/ semm@/ se:t@/ I.am/you.are/he.is/ we.are/you.are/ ‘I have come’ etc. b. so/ Si/ semm@/ se:t him I.am/you.are/ we.are/ you.are ‘I/you/we have called him’ c.â•… a cama:t@ him he.has/they.have called ‘He has/ they have called him’
e m@nu:t@ they.are come cama:t@ called
(37)â•… Aliano (Lucania) a.â•… Eddï†ı/ (G)ei/â•… Emm@/ aves@/ En@ v@nut@/durmut@ I.have/you.have/we.have/you.have/they.have come/slept ‘I have come/slept’ etc. b. E vv@nut@ he.is come ‘He has come’ c. a durmut@ he.hasâ•… slept ‘He has slept’
In both of the examples in (36)–(37), it is the 3rd person (singular) that displays auxiliary selection according to transitivity and voice€– while the 1st and 2nd persons do not display such sensitivity and are associated with a single auxiliary; incidentally, this latter auxiliary varies between have in Aliano and be in Colledimacine. Loporcaro (1988, 2007) and La Fauci and Loporcaro (1989:€167) discuss similar data from Altamura, in which the 1st and 2nd persons freely alternate between have and be, while 3rd person shows a residual sensitivity to verbal class; thus, 3rd singular has only be with unaccusatives, while 3rd plural has only have with transitives.
226â•… The auxiliary:€have/be alternation There are also varieties in which auxiliary selection according to transitivity and voice is observed in other persons in addition to the 3rd. For instance, the split paradigm of Vastogirardi in Table 6.1 (vi) holds for transitives/unergatives, while unaccusatives involve be throughout the paradigm. This means that both the 3rd and 1st singular alternate between have in transitives and be in unaccusatives, as illustrated in (38a–b), while the other persons do not alternate, as in (38a–b’). (38)â•… Vastogirardi (Molise) a.â•… sig@/â•… si/â•… E/â•… sem@/â•… set@/â•… soâ•… I. amâ•… etc. ‘I have come’â•… etc. b. r aj@/ a camat@ himâ•… I.have/he.hasâ•… called ‘I have/hehas called him’ b’. r@ si/ sem  himâ•… you.are/we.are/you.are/they.areâ•… ‘You/we/they have called him’
mnut come
camat@ called
These can be described as person split languages where have/be alternate according to transitivity/voice in some persons, typically the 3rd person, while the other persons have a single auxiliary, namely have in Colledimacine and Vastogirardi and be in Aliano. An equivalent way of stating this is what to Loporcaro (2007:€ 198) ‘makes little sense’, namely that ‘the distribution of auxiliary E/H … suggests a person ergativity split’. The languages in (39)–(40) illustrate another interaction of the person split with auxiliary selection according to transitivity/voice. In Molina di Ledro, the reflexive in (39a) has be in the 1st and 2nd persons and have in the 3rd person. However, unaccusatives are associated with be throughout the paradigm, as shown for the 3rd person in (39b), and transitives/unergatives are associated with have, as shown for the 1st person in (39c). (39)â•… Molina di Ledro (Trentino) a. me so lava E te te se E el/ laâ•… s a A neâ•… sumeâ•… laveâ•… E ve se E i/ le s a A ClS MP Aux washed ‘I have washed myself’ etc. b. l/ i E viï†−u ClS Aux come-m./f. ‘I have come’ etc
Finer parametrizationâ•… 227 c. o miGa durmi I.have not slept ‘I have not slept’
Example (39) is an instance of what Loporcaro (2007) calls ‘triple auxiliation’. Since there are two auxiliaries, it is evident that ‘triple auxiliation’ will involve the alternation of be with have and with the have/be split according to person, as in Molina’s unaccusatives (be) vs. reflexives (have/be according to person) vs. transitives (have). In terms of the parametrization picture laid out in this chapter, therefore, Molina di Ledro can be described as a language in which the 3rd person displays a have/be alternation according to transitivity only (as in Soazza), while the 1st and 2nd persons involve auxiliary selection according to transitivity and voice (as in Italian). The result is that the 1st and 2nd persons of the middle (reflexive) are associated with be (as in Italian), and the 3rd person is associated with have (as in Soazza). As expected on the basis of the general reversibility of person split patterns, the reflexive in Buonabitacolo, as shown in (40a), has the reverse split from Molina€– namely, be with the 3rd person singular and have with the other persons (the 1st and 2nd singular and the plural). Like Molina, this language systematically associates be with unaccusatives as in (40b), and have with transitives/unergatives, as exemplified in (40c) with the 3rd person. In the present terms, Buonabitacolo, therefore, is a language in which auxiliary selection in the 3rd person singular is sensitive to transitivity and voice (like standard Italian); hence we find be in the reflexive (40a) and in the unaccusative (40b). By contrast, the 1st and 2nd persons (and all persons in the plural) are sensitive only to transitivity; hence be combines with unaccusatives, as in (40b), but otherwise we find have. (40)â•… Buonabitacolo (Campania) a.â•… m addZaâ•… lavato A t a A s E E JtSâ•… amma lavati A v aita A s anna A MP Aux washed ‘I have washed myself’ etc. b. so/â•… si/â•… E/â•… simmo/â•… siti/â•… soâ•… vv@nuto/ v@nuti Aux come/ come-pl ‘I have come’ etc. c. iddu a camatu/ a patitu he has called to your father ‘He called your father’
228â•… The auxiliary:€have/be alternation In a nutshell, the data in this section illustrate the important point, concerning parameters in general, that complex surface patterns are obtained through freely crossing more elementary patterns, attested in isolation in other systems.13 Specifically, no scales are necessary (or even sufficient) to account for the observed variation. 6.4.2 The third auxiliary Finally, we examine a number of Romance varieties which alternate between be, have and what can be described as syncretic or neutralized forms of have and be. Consider, for instance, the perfect of Briga in (41). This displays a transitivity split between be with unaccusatives, as (41a–a’) and have with transitives/unergatives, as in (41b). In the 2nd singular, however, both classes of verbs combine with the same auxiliary form, namely . The latter also occurs in copular contexts as in (42a) and in possession contexts, as in (42b). (41)â•… Briga Novarese (Piedmont) a.â•… sum/ l E Jy I.am/â•… he isâ•… come ‘I have/ he has come’ a’. sumŋ/ si/ iŋ Jyi we.are/ you.are/ they.are come ‘We/ you/ they have come’ b. / l a/ um/ i/ (i) aŋ drumetS I.have/ he has/ we.have/ you.have/ they have slept ‘I/ he/we/you/ they have slept’ c. t E Jy/drumetS youâ•… Aux come/ slept ‘You have come/ slept’ (42)â•… a.â•… b.
t E kunte:nt youâ•… areâ•… happy ‘You are happy’ ti g E fa:m you Loc have hunger ‘You are hungry’
In the reflexive€ – i.e. what in the present terms is an instance of middleÂ�passive voice€– Briga alternates between be in the 1st person and in the 2nd plural, as in (43a), and have in the 3rd, as in (43b), reproducing a pattern similar to Molina, as in (39). Interestingly, although have in (43b) combines with si, be in (43a) appears without si. In other words, the reflexive with be lacks any clitic middle-passive morphology, just like Trun, as in (30), or in fact Albanian. The form of the 2nd singular appears without si, as in (43c).
Finer parametrizationâ•… 229 (43)â•… Briga a. sum lava I.am washed ‘I have washed’ a’. sumn/ si lavai we.are/ you.are washed.pl ‘We/ you have washed’ b. s a lava MP he.has washed ‘He has washed’ b’. s aŋ lavai MP they.have washed.pl ‘He has washed / they have washed’ c. t E lava you Aux washed ‘You have washed’
The data in (41)–(42) may suggest that a language like Briga has the two ordinary auxiliaries have and be€– while the rule in the morphological component takes care of the fact that there is a single lexical exponent for both of them in the 2nd person singular. The relevant lexical item could be associated just with [2nd singular, Aux] specifications, and in a Late Insertion framework like Distributed Morphology this could simply be inserted under richer syntactic specifications in the absence of better (more highly specified) candidates. The absence of a se form in the reflexive in (43c) suggests that the neutralized form of the auxiliary can be functionally equivalent to be. In the pluperfect, we also find unique auxiliary forms, as illustrated for Briga once again in (44). In the 2nd person and in the 1st person plural, be is lexicalized in combination with the unaccusative in (44a–a’) and have with the unergative in (43b). In the 3rd person and the 1st person singular, there is a generalized form of the auxiliary, as in (44c–c’). The data in (45) show that this unique form for have/be also occurs in main verb contexts, i.e. in possessive contexts in (45a) and in copular contexts in (45b). (44)â•… Briga a.â•… sevi Jy I.wasâ•… come ‘I had come’ a’. sevŋ/ sevi Jyi we.were/ you.were come ‘We/ you had come’ b. evi/ evŋ/ evi drumetS I.had/ we.had/ you.had slept ‘I/ we/ you had slept
230â•… The auxiliary:€have/be alternation c. c’.
t evi/ l eva Jy/ drumetS you Aux/he Aux come/ slept ‘You/ he had come’ evŋ Jyi/ drumetS they.Auxâ•… come.pl/ slept ‘They had come/ slept’
(45)â•… Briga a.â•… ti g evi/ l g ev/ i g evŋ fa:m you Loc had/ he Loc had/ they Loc had hunger ‘You were/ he was/ they were hungry’ b. t evi/ l ev/ (i) evŋ kunte:nt you were/ he was/ they were happy ‘You were/ he was/ they were happy’
This type of pattern in the pluperfect is not infrequent in Italian varieties. Table 6.2 presents a summary of the relevant Northern Italian varieties exemplified by Manzini and Savoia (2005); the * symbol refers to the descriptively neutralized auxiliary, A-E refers to have/be alternations according to transitivity and/or voice, while a parenthesized (A) indicates the optionality of have with transitives/unergatives. It can be seen that the distribution of unique auxiliary forms gives rise to a person split of sorts. For instance, the more robustly attested pattern, that of Grumello and other varieties, has have/be alternations in the 1st and 2nd persons, both singular and plural, and the unique form of the auxiliary in the 3rd person€– which is the classical discourse-anchored vs. event-anchored split discussed in section 6.2. Comparable data are found in Southern Italian varieties (Manzini and Savoia 2005; Cennamo 2010). Just as we have done for the unique auxiliary of the 2nd person singular, one may envisage a morphological-level solution for the unique auxiliary of the pluperfect. In these terms, the syntax would have distinct categorizations for have and be throughout the pluperfect paradigm€ – while underspecified elements would be inserted under some of the terminal nodes, depending on morphological rules. However, this approach to syncretism, requiring Late Insertion, Impoverishment and default lexical entries, is rejected here for theoretical (as well as empirical) reasons (cf. in particular chapters 7 and 8). The discussion of the person category in section 6.2 is directly relevant in this connection. The three traditional persons are not captured in present terms by simply letting the 3rd be unspecified (or negatively specified) for the hearer and speaker features. Similarly, the three-way split found in (44) between differentiated be, differentiated have and the unique eva forms cannot be described by simply letting eva be underspecified for the have and be properties. Rather, a categorization system must be devised so that each of the three auxiliaries in (44)€– or in (42) above€– is positively specified.
Finer parametrizationâ•… 231 Table 6.2 Distribution of be, have and syncretic forms in the present perfect in Piedmontese and Lombard varieties Trecate Briga, Montebruno Cerano Masserano Grumello, Passirano, Civate, Casorezzo, Inveruno, Arconate, S.Fedele, S.Bartolomeo C. Olgiate Quarna Sotto
* A-E
* *
* *
* A-E
* A-E
* *
(A)-E A A-E
* * A-E
* * *
(A)-E A A-E
(A)-E A A-E
* * *
A-E A-E
A-E A-E
* A-E
A-E A-E
A-E A-E
A-E *
This takes us back to a question that we only partially dealt with in section 6.3.3, namely the lexical content of have and be. Let us begin with be, which in section 6.3.3 we simply characterized as a verb lacking an external argument. In copular contexts, we can take the lexical content of be to be the logical relation of set membership. In other words, the logical form of John is intelligent is roughly ‘John ∈ {intelligent}’ or ‘John is a member of the set of intelligent individuals’. Following Moro (1997), we conclude that there is no separate identificational reading of be. Let us now consider have, which we can take to be fundamentally the verb denoting possession. In section 6.1.1 we rejected Kayne’s (1993) analysis of have as derived from be through the incorporation of a prepositional complementizer, mostly because this applicative derivation doesn’t seem to be reflected in the morphologies which are actually observed.14 At the same time, it is evident that have, i.e. possession, is closely connected to be, i.e. set membership. We propose that this connection should be captured directly at the interpretive level, in the sense that have can be lexically characterized as the reverse relation to be. Thus, have fundamentally denotes set inclusion. In fact, in certain cases have is equivalent to ‘include’ as in Italian has two auxiliaries (or more abstractly This set has two members). The inclusion relation also yields inalienable possession in a particularly natural way (John has nice legs)€– and we can take all possession to fall under a reasonable extension of the same relation. Of course, the ‘include’ relation is transitive, whence the transitive properties of have discussed in section 6.3.3.
232â•… The auxiliary:€have/be alternation This proposal is close to that advanced by Belvin and den Dikken (1997:170), according to whom the ‘meaning’ of have … denotes a special kind of inclusion relation … dubbed ‘zonal inclusion’ … Entities have various zones associated with them, such that an object or eventuality may be included in a zone associated with an entity without being physically contained in that entity … The type of zones which may be associated with an entity will vary with the entity.
In short, be and have are not only characterized by a transitivity opposition, but more to the point they express relations which are in some sense the mirror image of each other. Against this background, we propose that the unique eva auxiliary in (44) denotes a generalized ‘set relation’, which according to context can be read as ‘member-of’ or as ‘includes’. If this is on the right track, we expect other lexicalizations of this interpretive category to be independently attested in natural languages. We surmise that the Romance preposition di ‘of’ is a good candidate for this role. Thus, di quite obviously corresponds to the inclusion relation in partitive contexts, as for instance in tre di queste mele ‘three of these apples’ (roughly ‘three individuals included in the set of apples here’)€– and we extend the same construal to inalienable possession, as in i capelli di Gianni ‘lit:€the hair of John’), and possession in general. At the same time, di is compatible with a reversal of the inclusion reading in such expressions as quell’idiota di Gianni ‘that idiot of John’ (cf. Kayne 1994; den Dikken 1998), where John is attributed the property of being idiot, i.e. ‘John ∈ {idiot}’. The analysis just proposed extricates us from the more traditional way of approaching the descriptive problem posed by (42)–(43) or (44)–(45), namely whether the unique auxiliary forms should be identified with have or with be. Cennamo (2010), in a historical perspective, suggests that eva is derived from have, criticizing the analysis in Manzini and Savoia (2005), whereby eva is taken to instantiate be, on the basis of the fact that it appears as the copula. Despite the obvious difference between the present proposal and our previous analysis, the same underlying intuition remains, namely that the eva type represents a particularly elementary semantics. Systems like Briga could be seen as a true instance of the ‘triple auxiliation’ of Loporcaro (2007). Thus, in Briga and similar varieties have and be alternate with what we may call the ‘third’ auxiliary (eva). In the Briga perfect, in particular, have and be split according to transitivity/voice, while the ‘third’ auxiliary alternates with both of them on the basis of the person split. In other varieties, such as Trecate, as shown in (46), have, be and the ‘third’
Some conclusionsâ•… 233 auxiliary can be seen to lexicalize the 1st, 3rd and 2nd persons respectively, as in (46a). (46)â•… Trecate (Piedmont) a.â•… / t E/ l E ï†−y/ drumy/ tSama-r I.have/ you Aux/ he is come/slept/ called-him ‘I have/ you have/ he has come/ slept/ called him’ b. suk/ t E/ l E kunte:nt I.am/ you Aux/ he is happy ‘I am/ you are/ he is happy’ c. i / t E/ l a tri fiØ Loc I.have/ you Aux/ he has three children ‘I have/ you have/ he has three children’
At the same time, we do not find what Loporcaro (2007) or other supporters of transitivity scales predict to exist€– i.e. a language in which, for instance, be lexicalizes one extreme in the scale, have lexicalizes the other extreme and there is a middle zone associated with the ‘third’ auxiliary. To us this is further evidence that while the whole approach based on hierarchies may have descriptive value, it certainly has no theoretical substance. 6.5
Some conclusions
The review of auxiliary selection phenomena in what precedes is consistent with the conclusion that there is a limited range of descriptive categories that are involved in the definition of superficially very articulated parametric systems. This includes the split between event-anchored and discourseanchored referents, the split between active and middle-passive voice and the split between transitive and unaccusative predicates. The interaction of these primitives results in fine parametrization. As for the transitivity split, we proposed that what is involved is simply the n-adicity of the predicate€– the basic split being between monadic and polyadic. As for voice, we concluded that the unification of reflexives and passives with impersonals requires reference to a notion of open/generic variable in the argument structure. The person split has been characterized in terms of event- vs. discourseanchoring. The kinds of facts that we are thinking of when we use labels like these are precisely person splits of the type presented here. What is at stake is not so much the referential properties of 1st/2nd person vs. 3rd person per se as the way in which the anaphoric properties of the 3rd person (which can function as a variable) vs. the purely deictic properties of the 1st and 2nd person interact with their anchoring in argument structure. For instance, if we think of
234â•… The auxiliary:€have/be alternation the argument slots associated with a given predicate as variables (Adger and Ramchand 2005), we may suggest that 3rd person referents provide (quantificational) binders for these variables€– while the mechanism for linking 1st and 2nd persons to the argument structure must necessarily follow some other route. Here and in chapter 5 we have illustrated the reasons why we conclude that transitivity and voice alternations cannot be characterized in terms of movement from object to subject position€– either separately or together. Faced with a set of empirical issues, such as those encountered in applying the classical movement analysis to the middle-passive voice of Italian or Albanian, one way out is simply to add provisos to the theory, so as to fit the new evidence. However, in the case of middle-passive voice, it seems to us that the extensive literature on this topic has shown that no simple manipulation of the movement theory suffices to capture the facts. Therefore there are reasons why the alternative we are offering should be least be considered. Importantly, the primitives that we employ, namely the notions of selection, LF variable, generic binding, predicate–argument structure and so on, all independently enter into minimalist theories, so no questions of explanatory adequacy should arise in this respect. We also went into some detail on the empirical evidence in order to draw general conclusions from it about the nature of parametrization. A core pursuit of linguistic theory is to define the categorial distinctions of universal grammar (UG). Languages vary as to whether these categories are or are not instantiated (and how) in their lexicon. These categorial distinctions are the true ‘atoms’ of language (Baker 2001); on the one hand they are part of the universal competence, and on the other they enter into the differentiation of the various language-particular grammars. If what precedes is correct, what appear to be complex syntactic patterns of auxiliary selection reduce to the lexical, selectional properties of have and be. For the sublexicon consisting of have and be, the relevant categories are ‘defectiveness’ for standard Italian, transitivity for Soazza, voice for Albanian and the person split for the varieties in Table 6.1; both ‘defectiveness’ and the person split are relevant for the varieties in section 6.4. The distinction between microparametric and macroparametric approaches to variation has been so often discussed that the contours of the debate have become somewhat blurred. It is evident that, to the extent that the primitives manipulated by variation are macrocategories like transitivity or voice, we could describe our approach as macroparametric€ – though the fact that the unit of variation can be as small as a single lexical item qualifies it as
Some conclusionsâ•… 235 microparametric. What is clear is that the empirical evidence at our disposal appears to be incompatible with macroparameters in the sense of Baker (2001)€– i.e. structural parameters. None of our findings implies that the distinction between lexical and functional categories has any import for variation. Thus, all syntactic structures can be projected from lexical terminals, and there is neither a specialized morphological component nor specialized lexicalization principles applying to abstract functional nodes. The mechanisms that determine variation in so-called functional categories (such as those relevant for the selection of perfect participles by have and be) are the same ones responsible for variation in the substantive lexicon (including natural kinds, spatial relations, and similar universals connected with general cognition). The argument developed here in favour of this hypothesis is essentially based on economy; since the lexical/functional divide is not necessary, it can be dispensed with.
7 The noun (phrase): agreement, case and definiteness in an Albanian variety This chapter considers nominal inflections in Albanian, a language with case morphology. We argue that the deepest layer of inflection in Albanian nouns (including the predicative lexical base and the nominal class vowel) is more or less comparable to the inflection of Romance nouns. Above this layer, Albanian nouns add what is descriptively a case ending; we analyse it as an inflection specialized for the satisfaction of properties of the superordinate structure (nominative for the EPP argument, accusative for the internal argument etc.), as discussed in section 7.2. In the minimalist work of Chomsky (1995) and in its morphological implementation in Distributed Morphology (Halle and Marantz 1993), case and phifeatures are functional properties, interpretable at best on a subset of lexical heads, and as such triggering feature-checking operations€– as well as undergoing morphological rules and providing some of the key arguments in favour of Late Insertion. Here, on the contrary, we reduce case and phi-inflections to elements legible at the LF interface, and more precisely to arguments. Agreement between such inflections means that the arguments they represent form chains. Our crucial motivation is that abandoning schematic ideas about the division of labour between syntax and morphology and between functional and lexical categories (features) makes it easier to account for the finer articulation of the data and their variation. A particular point of interest as regards Albanian is that nominal inflections display not only agreement and case properties, but also definiteness properties. In Balkan linguistics (cf. also the discussion of Romanian in chapter 8), definite inflections are often (though not always) treated as postposed articles. In present terms, however, there is no advantage (descriptive or theoretical) to adopting a similar treatment; note that not even complementary distribution with the preposed article holds, since in Albanian definite inflections can combine with prenominal definite determiners, at least with kinship terms. Preposed articles also combine with adjectives (section 7.3). The pre-adjectival article, in turn, is not restricted to NP-internal positions, but is 236
Theoretical and empirical backgroundâ•… 237 also found in predicative contexts (copular and others). This property, which sets Albanian apart from other Balkan languages, indicates that the article is part of the adjective phrase€– not of the noun phrase that may or may not embed it. 7.1
Theoretical and empirical background
Under present assumptions, predicative elements such as the verb in the sentence and the noun in the noun phrase project a set of referential contents (D, Q, P, Loc, N) which build a structure of the type in (1). In this, I(nflection) corresponds to the core position of the lexical item (the verb or the noun) and D, Q, P, Loc, N form its (projection) domain. D corresponds to the (in)definiteness properties which characterize the EPP argument of the sentence and its equivalent within the noun phrase; that is, the determiner. P(erson)€– i.e. 1st and 2nd person€– Q(uantifier) and Loc(ative) lexicalize referential properties of deixis (P, Loc) and quantification (Q), satisfying (or participating in the satisfaction of) the argument slots of the predicate. In particular, N(oun) is the dedicated category for the internal argument of the predicate. We can think of the hierarchical order in (1) as a way of representing the relative scope of these elements. There are other positions available to the lexical head in the sentence/ noun phrase other than the I position mentioned in (1). These Â�correspond to different interpretive domains associated with predicative content (√, i.e. root, replacing N, V etc.), quantification over events, quantifications over situations, hence in particular modality (C), etc. Each of these positions projects the entire nominal string. (1) D Q P Loc N I
We extend the categories in (1), and hence the scope relations that order them, to what is conventionally thought of as morphological structure. Thus, the so-called agreement inflection of the verb is but an exponent of the D (EPP)
238â•… The noun:€agreement, case and definiteness argument within the verb structure. Similarly, consider the Italian noun phrase in (2a). The noun macchina ‘car’ can be analysed as a structure in which the lexical base, indicated as √, expressing predicative content, combines with a nominal class (gender) inflection –a. As illustrated in (2b), this inflection is identified with the N category, associated with the internal argument of the predicative base (also avoiding the need for a functional n category à la Marantz (1997)). The article in turn lexicalizes the D category at the phrasal level, as illustrated again in (2b). (2)â•… a.
la the
macchina car
b. D la
N macchina a
Current generative approaches conceive of agreement between, say, the different elements within the noun phrase as the result of the checking of features (of gender, number, etc). In contrast, we construe agreement between, say, the article, D, and the inflection, N, in (2) as the sharing of referentially relevant properties, which are interpreted by reference to the same individual, and jointly satisfy the internal, and only argument of the noun. In other words, since N and D jointly satisfy the same argument slot€– i.e. they form a chain€– their referential properties must be compatible, which is what agreement means in present terms. To say that macchin- is just a predicative base, as in (2), is to say that it cannot satisfy any argument slot. Rather, the argument in (2) is –a, whose referential properties are further specified by the Determiner with which it agrees€– i.e. in present terms, the Determiner it forms a chain with. Therefore it is€ –a, or the (la, –a) chain, that satisfies, say, the internal argument of the matrix predicate when (2) is further embedded, as in (3). Note that the inflection of the verb is construed as the verb-internal realization of the EPP argument of the sentence (Manzini and Savoia 2005, 2007, 2008b), notated as D, in keeping with Chomsky’s (1995) suggestion. The D/EPP argument saturates one of the arguments of the predicative base, i.e. here its external argument.
Theoretical and empirical backgroundâ•… 239 (3)â•…
a.
Lavo la macchina I. wash the car ‘I am washing the car’
b.
lav (x,y)
D ox
D l
N N ay
macchin
N ay
At this point a difficulty seems to arise, in that€–a in (3) appears to satisfy two argument slots (theta-roles) at once, namely that of the noun and that of the verb. Now, there are independent proposals in the minimalist literature, beginning with Hornstein (1999) and Manzini and Roussou (2000), that such configurations are independently found in the syntax. (Obligatory) control is precisely such a configuration:€ one in which the same argument satisfies an argument slot of a matrix and an embedded predicate. In the same way, we can say that the argument –a (equivalently its chain) in (3) satisfies the internal theta-role of the predicate base macchin- and controls the internal argument slot of the verb. Incidentally, while control configurations in languages like Italian or English have the controller higher than the controlled position, reverse configurations are possible, as in so-called backward control (Polinsky and Potsdam 2006). For a discussion of control along the lines assumed here, the reader is referred to Manzini (2009). Now, since Albanian nominal inflections register properties such as case and definiteness, it is evident that an extension of the model in (2)–(3) to Albanian requires further discussion of case categories in particular. In section 7.1.1 we will present the basic data, and in section 7.1.2 an overview of the theoretical literature on case. 7.1.1 Nominal inflections in Albanian In Albanian varieties, including the standard (Solano 1972; Camaj 1984; Giusti and Turano 2007) nouns have an indefinite and a definite declension. The definite inflection characterizes nouns typically occurring without determiners, to which it confers definite reference. The indefinite declension characterizes
240â•… The noun:€agreement, case and definiteness nouns that occur together with indefinite determiners. The indefinite paradigm is characterized by poorer case inflections, in particular in the sense that nominative and accusative are differentiated in the definite but not in the indefinite. These general properties are illustrated here for the Arbëresh (Italo-Albanian) variety of Vena, whose conditions are quite close to those of the standard, though not identical. For instance, demonstratives in Vena combine with the indefinite declension on nouns, while in the standard they combine with definites.1 In (4) we illustrate the definite singular paradigm, in (5) the indefinite singular paradigm and in (6)–(7) the definite and indefinite plural paradigms. In (4)–(7), the examples in (a) illustrate the nominative (for instance the subject of a finite verb, i.e. its EPP argument), the examples in (b) the accusative (for instance the internal argument of a transitive active verb) and the examples in (c) the dative (i.e. the second internal argument of a ditransitive like ‘give’). The latter is syncretic with the genitive, to be examined in detail in section 7.3; because of this syncretism we shall refer to the case in (c) simply as oblique. The other major case recognized in traditional paradigms is the ablative, which is, however, not preserved in Vena, allowing us to simplify the presentation somewhat.2 Consider first the definite paradigm in (4). The nominative is lexicalized by the –i/–u inflections for the masculine and by –a for the feminine, as in (4a); –u is selected by a lexical base ending in a velar or a vowel. The accusative in (4b) is lexicalized by the –n() inflection preceded by –i/u for the masculine and -@/ for the feminine. It is natural to assume that –i/u and –@/ are nominal class endings, while only –n@ marks accusative case. In turn, the oblique displays –i/u–t(@) in the masculine and –@/E–s@ in the feminine, which we analyse again as complex inflections formed by the nominal class morphology and the case morphology proper; here, therefore, case endings select for nominal classes (–t@ in the masculine and€–s@ in the feminine). (4)â•… a.â•… kriatur-i/ ‘the boy’/ b. kriatur-i-n/ ‘the boy’/ c. kriatur-i-t/ ‘to the boy’/
ri-u/ ‘the man’/ ri-u-n@ / ‘the man’/ ri-u-t@ / ‘to the man’/
vazd-a/ ‘the girl’/ vazd-@-n@/ ‘the girl’/ vazd-@-s@/ ‘to the girl’/
matS-a ‘the cat’ matS--n@ ‘the cat’ matS--s@ ‘to the cat’
In the indefinite paradigm in (5), the masculine nominative and accusative coincide with the uninflected nominal base; in the feminine, an – nominal inflection crops up with a subset of lexical bases, as in matS–. The oblique has –i/u–ç@ in the masculine and ––j in the feminine, again formed by the nominal class morphology followed by the oblique morphology. As for determiners,
Theoretical and empirical backgroundâ•… 241 the indefinite article @ in the nominative/accusative alternates with iç in the oblique. The paradigm of the demonstrative is ai, at atiç@ for the masculine vs. aj, at@ asaç for the feminine. Note that in the oblique feminine it combines with the noun inflected simply by nominal class. (5)
a. a’. b. b’. c. c’.
@ kriatur/ ‘a boy’/ ai kriatur/ ‘that boy’/ @ kriatur/ ‘a boy’/ at kriatur/ ‘that boy’/ iç@ kriatur-i-ç@/ ‘to a boy’/ atiç@ kriatur-i-ç@/ ‘to that boy’/
@ ri/ ‘a man’/ ai ri/ ‘that man’/ @ ri/ ‘a man’/ at@ ri/ ‘that man’/ iç@ ri-u-ç@/ ‘to a man’/ atiç@ ri-u-ç@/ ‘to that man’ /
@ vazd@/ ‘a girl’/ aj vazd@/ ‘that girl’/ @ vazd@/ ‘a girl’/ at vazd@/ ‘that girl’/ iç@ vazd--j/ ‘to a girl’/ asaç@ vazd@/ ‘to that girl’/
 matS- ‘a cat’ aj matS- ‘that cat’  matS- ‘a cat’ at matS- ‘that cat’ Jiç@ matS--j ‘to a cat’ asaç@ matS- ‘to that cat’
The plural is illustrated in (6) for the definite conjugation and in (7) for the indefinite. In the indefinite in (7), a specialized plural –a inflection emerges in several nominal classes in the nominative and accusative, as for instance in vazd–a. Other nominal classes show the same nominal class inflection as in the singular, notably matï†fi or have allomorphs specialized for the plural, as in r@s@ ‘men’ vs. ri ‘man’. In the definite conjugation in (6), nominative and accusative are lexicalized by the same formative€ –t added once again to the nominal base inclusive of the plural/nominal class vocalic inflection –a/. The oblique, both in the definite and in the indefinite, is lexicalized by the specialized (case) inflection –v, added to the nominal base inclusive of the nominal class vowel. As for the determiner system, the demonstrative has a generalized form at for nominative and accusative, masculine and feminine, while in the oblique it takes the form atir the indefinite quantifier tï†fi@d is inflected as tï†fi@d–v in the oblique, itself following the indefinite declension. (6) a.â•… b. c. (7) a. a’.
kriatur-a-t@ / ‘the boys’/ kriatur-a-t@ / ‘the boys’/ kriatur-a-v/ ‘to the boys’/
r@s@-t@/ ‘the men’/ r@s@-t@/ ‘the men’/ r@s@-v/ ‘to the men’/
tï†fid kriatur-a/ ‘some boys’/ at kriatur-a/ ‘those boys’/
vazd-a-t@/ ‘the girls’/ vazd-a-t@/ ‘the girls’/ vazd-a-v/ ‘to the girls’/
tï†fid r@s@/ ‘some men’/ at r@s@/ ‘those men’/
matï†fi-t@ ‘the cats’ matï†fi-t@ ‘the cats’ matï†fi-v ‘to the cats’
tï†fid vazd-a/ ‘some girls/ at vazd-a / ‘those girls’/
di matï†fi ‘two cats’ at matï†fi ‘those cats’
242â•… The noun:€agreement, case and definiteness b.â•… tï†fid kriatur-a/ tï†fid r@s@/ tï†fid vazd-a/ di matï†fi ‘some boys’/ ‘some men’/ ‘some girls’/ ‘two cats’ b’. at kriatur-a/ at r@s@/ at vazd-a/ at matï†fi ‘those boys’/ ‘those men’/ ‘those girls’/ ‘those cats’ c. tï†fid-v kriatur-a(-v)/ tï†fid-vr@s@(-v)/ tï†fid-v vazd-a(-v)/ div matï†fi ‘to some boys’/ ‘to some men’/ ‘to some girls’/ ‘to two cats’ c’ atir kriatur-a(-v)/â•… atir r@s@(-v)/â•… atir vazd-a(-v)/ atir matï†fi(-v) ‘to those boys’/ ‘to those men’/ ‘to those girls’/ ‘to those cats’
Kinship terms (as often happens in natural languages) show special morphosyntactic behaviours, beginning with the fact that they combine the definite inflection of the noun with a preposed article. That the kinship terms in (8) have the definite inflection in the nominative (a), accusative (b) and dative (c) can be seen by comparing them with the forms in (4). As for the preposed article, in nominative contexts it corresponds to the nominal class morphology i for the masculine and for the feminine. In the accusative and oblique the preposed article is t@. (8) a. b. c.
r i kuSirir-i/ kuSirir-a cameâ•… theâ•… cousin.m/ the cousin.f ‘The/his/her/their cousin came’ p t@ kuSirir-i-n@/ t@ kuSirir-@-n@ I.saw the cousin.m/ the cousin.f ‘I saw the/his/her/their cousin’ j-a D t@ kuSirir--t@/ t@â•… kuSirir-@-s@ her-it I gave to.the cousin.m / to.the cousin.f ‘I gave it to the/his/her/their cousin’
In (9) we illustrate contexts in which kinship terms are preceded by an indefinite quantifier or a demonstrative; correspondingly they bear an indefinite inflection, reproducing the conditions already detailed in (5). (9) a. b. c.
r  kuSiri/  kuSirir came a cousin.m/ a cousin.f ‘There came a cousin (of his/ hers/ theirs)’ p  kuSiri/  kuSirir I.sawâ•… a cousin.m/ a cousin.f ‘I saw a cousin (of his/ hers/ theirs)’ j-a D iç@ kuSir--j / asaç@ kuSirir her-it I.gave to.a cousin.f/ to.that cousin.f ‘I gave it to a/that cousin’
Finally, in (10) we illustrate the definite paradigm of plural kinship terms, from which it can be seen that the prenominal article is t for the three cases
Theoretical and empirical backgroundâ•… 243 we are considering; (10a’) also provides an example of the indefinite paradigm, which (as already shown for the singular) simply reproduces the conditions observed with non-kinship terms. (10) a. a’. b. c.
r t@ kuSiriç@-t@ came the cousins.nom ‘The/his/her/ their cousins came’ r tï†fid kuSiriç@ came some cousins ‘Some cousins (of his/hers/ theirs) came’ p t@ kuSiriç@-t@ I.saw the cousins.acc ‘I saw the/his/her/ their cousins’ ja t@ kuSiriç@-v them-it I.gave to.the cousins.dat ‘I gave it to the/his/her/ their cousins’
The data in (4)–(7) illustrate the existence in Vena of several inflectional endings which are associated with two or more interpretations, yielding instances of so-called syncretism. In our examples we find two types of syncretism:€(i) some inflections correspond to two (or more) cases; (ii) some inflections correspond to both a case interpretation and a nominal class interpretation (the traditional gender and number). For instance, the –a inflection lexicalizes the nominative definite (for the feminine singular class) in (4a) and the nominative/accusative indefinite (for the plural class) in (7a) and (7b). Thus, vazd–a is ambiguous between ‘the girl(Nom)’ and ‘girls (Nom/Acc)’. At the same time, the –a morphology also appears as a thematic vowel in plural formations involving specialized consonantal/syllabic case endings, for instance in the oblique (definite and indefinite) in (6c), (7c) and in the nominative/accusative definite in (6a), (6b). Similarly, the –i inflection, corresponding to the thematic vowel for the masculine singular, lexicalizes the nominative definite in (4a). In turn, the –t(@) inflection is associated with the oblique (singular masculine) in (4c) and with the nominative/accusative (plural) in (6a) and (6b). The –v inflection is uniquely associated with the oblique plural, yet it includes both the definite and the indefinite reading, as in (6c) and (7c). In Table 7.1 we list the morphological endings associated with case in the Vena nominal system in (4)–(7). For each of the forms we indicate the traditional case, definiteness, and number features they are associated with. The fact that most entries are associated with more than one row of values implies that they are syncretic. We abstract away from the thematic vowels, i.e. the vocalic formatives that appear between the nominal root and consonantal/
244╅ The noun:€agreement, case and definiteness Table 7.1 Distribution of nominal inflections in Albanian Vena Nom a i, u
* * * * *
@
*
t(@)
*
Acc
Def *
*
*
* * * *
v n(@) s@ j ç@
Obl
*
* * * * * *
* * * * * *
Indef * * * * * * * *
Sg * * * * * * *
* * *
* * * *
Pl * * * *
* * * *
syllabic endings like –t(@),€–v etc; rather, we have tabulated the vocalic formatives only as they occur word finally. We have also left out the traditional gender (nominal class) from the properties being tabulated. 7.1.2 Generative approaches to case In the minimalist approach of Chomsky (1995), properties such as gender (nominal class), number and person, which are intrinsically associated with nominal constituents, are bona fide lexical features. However, relations, such as thetaroles, are not features at all, but correspond to syntactic configurations. From this perspective, it is potentially problematic to find that case is a feature. The fact that case is the only feature in Chomsky (1995) which is radically uninterpretable (i.e. which does not have an interpretable counterpart) is a reflex of the deeper difficulty of reconciling its relational core with its feature status. The solution at which Chomsky (2008) arrives is effectively to deny that case has a primitive relational content. In technical terms (and contra Chomsky 1995), case does not enter into any feature checking. Rather, the real underlying relation between case assigner and case assignee is an agreement relation, involving phi-features; case is but a reflex of this relation which appears on nominal constituents.
Theoretical and empirical backgroundâ•… 245 Similarly, Pesetsky and Torrego (2007) treat case as a temporal property. The presence of interpretable but unvalued features T[x] on the functional head T acts as a probe for the subject DP associated with the same T[x] feature, both uninterpretable and unvalued, yielding agreement between these two elements. The value of the feature is assigned through a further agreement operation with the corresponding T feature associated with the verb, which is uninterpretable, but valued. We agree with Chomsky, Pesetsky and Torrego that case cannot be a primitive of grammar. However, it seems to us that Chomsky (2008) falls short of implementing the reduction of case to agreement. For, simply put, if case is reduced to other primitives, why do we need to keep the case label at all? In other words:€what is the difference between a language which has just agreement (say, Italian) and a language like Latin which has the ‘case’ reflex of agreement? Similarly, saying, as Pesetsky and Torrego (2007) do, that (nominative) case is merely the name that Tense takes when lexicalized on a noun leaves us without a clue as to why we still need to refer to this Tense on nouns as case. Otherwise stated:€ where is the evidence, either morphological or interpretive, that independently connects the Tense of verbs and the supposed ‘Tense’ of nouns? The works reviewed so far consider so-called ‘abstract’ case, i.e. a case property independent of morphological realization, and as such found (by hypothesis) in all languages. In turn, case inflections have been the target of considerable morphological discussion. Within Distributed Morphology, Halle and Marantz (1993) propose a treatment for the case inflection of Potawatomi. In this language, the form /–mun/ denotes the 1st person plural as a subject and, in the context preceding the preterite affix, as an object. Halle and Marantz (1993:€157) conclude that /–mun/ is specified in the lexicon just for the features [+1], [+pl], and that it is inserted as, say, an accusative because of a rule that deletes [ACC] in front of the preterite, as in (11). (11) [ACC] → φ / ╇ ____╇ [+preterite] â•… [+1]
To generalize, syncretism corresponds to the lack of isomorphism between interpretive categories€ – e.g. in (11) the cluster [accusative, 1 pl]€ – and Â�morphological categories, e. g. /–mun/, which is just [1 pl]. In Distributed Morphology, this lack of isomorphism is circumvented by assuming that at the syntactic level, all semantic properties relevant for interpretation are abstractly represented€– while some categories to which syntactic computation applies do not have any morphological expression. This result is achieved through a model crucially involving the assumption that lexical insertion applies after
246â•… The noun:€agreement, case and definiteness morphological rules, such as Impoverishment in (11), have operated on the abstract terminal nodes (Late Insertion). What interests us here directly is that under the Distributed Morphology approach, a morpheme traditionally associated with case, e.g. accusative, turns out not to have any such property, but only a 1st person plural (‘we’) denotation. In general, in case-inflected languages, the presence of morphological entries associated with several case contexts (i.e. syncretic in traditional terms) leads to the conclusion that these case morphologies have a purely denotational content, devoid of case properties, and associated only with nominal class, number, and possibly definiteness, etc. We agree with Halle and Marantz on the content of actual case terminals€– i.e. the fact that their intrinsic properties may be just nominal class and the like. However, we differ from them in that we assume a unified morphosyntactic component, where Late Insertion is replaced by projection of syntactic structure from lexical terminals (Manzini and Savoia 2005, 2007, 2008a). We take this to be the theoretical position implied by the minimalist programme of Chomsky (1995). In such a framework, syncretism cannot be the result of the fact that morphological rules allow for radically underspecified lexical items to be inserted under richly detailed syntactic nodes. On the contrary, the denotational content that even conventional morphological models impute to terminals is all that is projected to the syntax and handled by the computational component. If, therefore, case is not a property of syntactic representations at all, we will have to show that syntactic and semantic composition can be successfully effected on the basis of this more restrictive approach. 7.2
Analysis of Albanian nominal inflections
Kinship terms provide a useful starting point in the analysis of the Albanian noun (phrase) because they present a more complex structure, where determiner and nominal inflection combine. This co-occurrence seems to result in a doubling of sorts, since, for instance, in the nominative singular definite, the nominal inflection and the preposed article converge on the nominal class morphology i for the masculine. The article that appears elsewhere in the paradigm, i.e. t, is also identical to the definite inflection for the nominative/ accusative plural, as well as for the definite oblique masculine singular. These parallelisms may invite a treatment of at least the definite inflections of Albanian as postnominal articles, i.e. as a prenominal D element to which the N head is left-adjoined through movement. This analysis is proposed, among others, by Dobrovie-Sorin (1987) for Romanian, Taraldsen (1990) for
Analysis of Albanian nominal inflectionsâ•… 247 Norwegian and Turano (2003) for Albanian. The N-to-D analysis is also proposed by Longobardi (1996, 2001) for a different type of structure, in which N is in complementary distribution with the article, as in Italian kinship terms or the casa ‘home’ type. Nevertheless, the N-to-D analysis of postnominal definiteness morphology has been questioned more than once in the literature. Thus, Longobardi (2000) points out that in Scandinavian languages, the postnominal definiteness morphology co-occurs with prenominal adjectives as in Icelandic frabæra bokinn hans ‘beautiful book-def his’, which Longobardi (2000) interprets as conclusive evidence in favour of a relatively low position of N and against movement of N to the highest D position. Similarly, for Romanian, Dimitrova-Vulchanova and Giusti (1998) argue that the so-called postnominal article is in fact generated directly on the noun; movement of N is involved in the derivation of definite noun phrases, but it is to a Focus position where the N-D head controls the scopal D position. In presenting the data, we have already referred to the postnominal definiteness morphology as an inflection. In fact, the presence of both a postnominal definiteness element and a preposed article in kinship terms undermines the main empirical basis for the postposed article analysis€ – namely, that such an analysis predicts complementary distribution between postnominal definiteness and definite articles. This complementary distribution is simply not observed in Albanian. Besides this, and the other empirical problems noted in the literature cited, the apparent simplicity of the movement analysis conceals a certain amount of stipulation. In particular, from a minimalist perspective, it is not made clear why a noun would move to D€– or for that matter to Focus, as proposed by Dimitrova-Vulchanova and Giusti (1998). We started from the observation that in examples like (8a) and (10a–b), the i and t articles coincide with the –i and –t(@) inflections. Despite this coincidence, and their shared definiteness properties, the lack of complementary distribution between articles and inflections leaves little room to analyses deriving one from the other. We assume that in sentences like (8a) and (10a–b), the preposed article i, t lexicalizes the D position within the noun phrase, cf. Italian (2b). The postnominal inflections –i, –t(@) cannot be the syntactic-level determiners at any stage of the derivations for the simple reason that the syntacticlevel D position is already filled by the article. Let us focus on the simpler vocalic inflection, e.g. –i. This element combines with€–n(@) for the accusative singular definite, –t(@) for the oblique singular definite and –ç@ for the oblique singular indefinite. Such a distribution suggests that the lexical entry for –i is associated with what in traditional terms is
248â•… The noun:€agreement, case and definiteness an agreement property, namely nominal class, like the vocalic endings of languages like Italian, as in (2). In keeping with the discussion in section 7.1, we then assign to –i an N categorization. The combination of a nominal root with an N morphology and a D determiner gives rise to structures like (12)€– entirely parallel to those of Romance, cf. (2). (12)â•… D i
N i
kuʃirir
The same treatment can be extended to the vocalic inflections€–a,€–,€–@ and –u. In the plural, the€–a and€– forms (depending on nominal class) combine with –t()) in the definite non-oblique and with€–v in the oblique. In turn, - combines with€–n(@)@) in the definite singular accusative, –s@ in the definite singular oblique and –j in the indefinite singular oblique. This distribution leads us to conclude that the lexical entries for€ –a and€ – are again associated with nominal class N properties. As for –u, it fairly obviously has the same distribution as –i, though with different lexical bases. The combination of nominal roots with an N morphology therefore gives rise to structures like (13). Nominal class morphologies are restricted to subclasses of roots; this can be expressed as a selectional relation whereby –a,€–,€–i etc. select the relevant lexical bases. (13) a. kriatur ɲeri
N i u
matʃ vazd
N ε ǝ
kriatur/ vazd
N a
b.
c.
Analysis of Albanian nominal inflectionsâ•… 249 Still, a couple of essential pieces are missing from the analysis of Albanian vocalic inflections. One is an account of how shift in nominal class can have a number interpretation; the other is an account of case. As already mentioned,€–a combined with vazd- or matï†fi- (conventionally the feminine) corresponds in traditional terms to the definite nominative singular; combined with vazd- again or with kriatur-, as in (13c), it corresponds to the indefinite nominative/accusative plural. Similarly,€ – in (13b) combined with matʃ- yields the nominative/accusative indefinite singular and plural. In turn, the masculine singular –i/–u in (13a) is the nominative (definite). We will return to vocalic inflections in section 7.2.2. Before addressing these various questions, however, we consider consonantal inflections, beginning with –t(@), which we have seen to coincide with the pronominal article in examples like (10a–b). In traditional terms,€–t(@) forms the plural definite nominative/accusative and the definite singular oblique, depending on nominal class in the oblique. It is evident even from this list that every occurrence of –t(@) in the nominal paradigm contributes definiteness properties to the base to which it attaches. This suggests that€–t(@) is (or includes) a definiteness operator taking in its scope structures like (13a) or (13c). Correspondingly, we may assign the categorial signature Q to the position projected by€–t(@) as a nominal inflection€– at least if we reserve the D projection for the prenominal article, where present. This yields structures of the type in (14) for the kinship terms in (10a–b). (14)â•… D tǝ kuʃiriçǝ
Q tǝ
The same analysis of€–t applies to plurals of non-kinship terms, as in (15a); and we can provisionally take it to hold for the oblique singular (masculine), as in (15b).
250╅ The noun:€agreement, case and definiteness (15)╅ a.
vazd/kriatur
N a
kriatur
N i
Q t(ǝ)
b. Q t(ǝ)
Once more, what is missing from the account in (14)–(15) is a match with the traditional notion of case. Using the traditional case terminology, the€–t(@) morphology in the singular (15b) is restricted to the oblique, while in the plural (15a) it is restricted to the non-oblique. Avoiding the case terminology altogether, the problem is how to limit the€–t(@) morphology to certain syntactically defined environments, whatever their exact definition turns out to involve. A similar problem arises for vocalic inflections. Thus, why is the definite reading of –a and€–i morphologies in isolation restricted to what is conventionally known as the nominative context? At this point it is unclear whether these restrictions can even be stated if case is not an available primitive. We examine this question item by item, beginning with –t(@) in the next section. 7.2.1 Consonantal inflections Let us consider the masculine singular oblique in (15b). The question is:€how do the quantificational properties of€–t(@) relate to the context of appearance? Abstracting away from case terminology such as ‘oblique’, which is what we are trying to explain, the contexts in which –t(@) masculine singular appears correspond to the second argument of ditransitives (the so-called dative) and to the genitive, which we will examine in more detail in section 7.3.1. The dative–genitive syncretism is widely attested, characterizing, for instance, Modern Greek, Romanian, and the pronominal clitic system of some Romance varieties in which genitive and dative are syncretically lexicalized by ne. Manzini and Savoia (2005, 2007, 2008a) conclude that ne denotes a superset in relation to which some other argument is interpreted. This superset-of denotation is fairly obvious in partitives (e.g. three of the boys), where the boys specifies a larger set to which the three singled out belong; inalienable
Analysis of Albanian nominal inflectionsâ•… 251 possession and attribution of mental states are equally clear instances, since in John’s nose or John’s fears, the nose or fears are part of the collection of properties that we call ‘John’. Similarly, in chapter 6 we argue that the predicate have fundamentally denotes set inclusion (or ‘zonal inclusion’ in the sense of Belvin and den Dikken (1997)). In fact, have is sometimes equivalent to ‘include’, as in Italian has two auxiliaries (or more abstractly This set has two members); the same relation can be expressed by ‘genitives’, e.g. the two auxiliaries of Italian or the two members of this set. Although the inclusion relation yields inalienable and psych-state possession in a particularly natural way€– we can take all possession to fall under a reasonable extension of the same relation.3 Possession, hence in present terms inclusion, is a natural characterization for the dative as well. In particular, the second internal argument of ditransitives has been argued to be connected to possessives at least since Kayne (1984). English He gave a fright/a book to everybody corresponds to the attribution of a mental state or a material possession to the ‘dative’ argument€– and Romance languages also have inherent possession datives, as in Ho lavato i capelli a Maria lit:€‘I have washed the hair to Mary’, i.e. ‘I washed Mary’s hair’. In terms of the preceding discussion, the dative–genitive syncretism seen in the Albanian oblique points to a superset-of characterization for the relevant morphology, including –t(@) in (15b). This, in turn, appears to be compatible with the Q quantificational characterization that we have assumed for –t(@), given the set-theoretic calculus involved. From this perspective, we propose that there is no oblique case involved in (15b). All there is is a quantificational element –t(@) capable of satisfying the superset-of (roughly possessive) specification required for the satisfaction of the second arguments of ditransitives Â�(so-called dative) and more (i.e. the genitive to be considered in section 7.3.1). The other major context in which –t(@) is found to occur is the nominative/ accusative plural. The syncretism of oblique (i.e. dative) singular with nominative/accusative plural is again independently attested in the Romance clitic system. Thus, the standard Italian dative singular gli is an allomorph of li for accusative plural; other varieties display exactly the same (l)i form. In Manzini and Savoia (2005, 2007, 2008a), we explain this syncretism by proposing that the –i morphology of Italian varieties has a quantificational content. This has two instantiations:€superset-of (i.e. dative) and plurality. We take it that in the plural interpretation, the Q specifications of€ –i take only the lexical base to which they apply in their scope. In turn, the superset-of reading depends on the quantificational specifications of –i taking sentential scope.
252â•… The noun:€agreement, case and definiteness We apply this analysis to the fact that Albanian has€–t(@) for both oblique singular and non-oblique plural readings. Thus, when€–t(@) is read as plural, as in (15a), it takes in its scope the nominal class€–a (or€–) specifications. When it is read as superset-of, its scope is sentential. Interestingly, it follows that the two readings are in complementary distribution:€–t(@) can be oblique, but it will not be plural; or it can be plural, but it will not be oblique. Similar syncretisms are found in other languages. As discussed in chapter 8, Romanian –i is the oblique (dative/genitive) singular and the (masculine) nominative/accusative plural. Latin –i in turn shows up as genitive and/or dative singular and nominative plural in both the I and II classes; Latin –s is (among other things) genitive singular and nominative/accusative plural in the III, IV and V classes. What we characterized as the sentential scope of Q must be construed more precisely as scope over the internal arguments of the verb. Indeed, the –i dative of Romance or the€–t(@) oblique of Albanian are dyadic operators establishing a superset-of relation between the argument to which they attach (the descriptive dative) and the internal argument of the verb, excluding the external argument. The simplest way of achieving the correct scope within the present framework is to assume that the Q operator, which we will henceforth also notate as Q(⊆), has scope over the elementary event (the verb plus its internal argument(s)) (on Q elements within the predicative domain, see also the discussion of adverbs of quantification in chapter 3). We will occasionally use ‘sentential’ scope as a shorthand. Let us now consider the so-called accusative singular definite, which is associated with the dedicated morphology€–n(@). Observe that for the plural we already have structural schemas for the so-called accusative definite, as in (15a). In particular, we have argued that in (15a) the Q specification added to the nominal class morphology has a number interpretation. Thus, it would appear that it is N morphology that satisfies the so-called accusative context. Now, the syntactic properties of N that we have discussed so far relate it to the internal structure of the noun (phrase). We have proposed that the N nominal class morphology lexicalizes the internal argument of the predicative base of the noun. Therefore we are not surprised to find that the same morphology can satisfy the internal argument of the verb. Applying this reasoning to the€ –n(@) morphology of the singular definite forms, we can conclude that the –n(@) ending simply has N properties. In other words, it is nothing but an N inflection, further specialized for definiteness, as illustrated in (16). In these terms,€–n(@) therefore introduces reference to a specialized nominal class€– specialized both in that it is definite and in that it is contextually restricted to the internal argument-of relation.
Analysis of Albanian nominal inflectionsâ•… 253 (16)â•… a.
kriatur
N i
vazd/matʃ
N ε
N n(ǝ)
b. N n(ǝ)
In short, we argue that accusative is nothing but the name given to the satisfaction of an internal argument slot by nominal N morphology4. A potential problem for our proposal arises in connection with the fact that accusative is standardly taken not to be linked to any particular theta-configuration, and so is a structural case in the sense of Chomsky (1986). The classical motivation for this in English is so-called Exceptional Case Marking (ECM), whereby the EPP argument of some infinitival complements turns up in the accusative (as can be overtly seen with pronouns). Needless to say, an EPP argument can correspond to an internal argument slot (as in unaccusatives) but also to an external argument (as in transitives). Interestingly, there is a tradition in the generative literature (Chomsky 1975 [1955]) which treats English ECM as an instance of restructuring of the propositional attitude verb with the embedded verb, making the embedded subject into the thematic object of this complex predicate. An independent tradition proposes, furthermore, that the position of the accusative is not that of an embedded subject but that of a matrix object (Postal 1974; Johnson 1991). Assuming the (basic) validity of our results concerning –t(@) and –n(@) inflections, the entire set of consonantal inflections in Table 7.1 more or less falls into place. Thus –t(@) has the oblique definite singular reading only in the masculine. In the feminine, the oblique definite singular is lexicalized by –s@. In the plural, all nominal classes and both definites and indefinites are associated with –v. In the indefinite singular, again€–ç@ and –j represent specialized inflections for the masculine and feminine respectively. It stands to reason that if€–t(@) is able to lexicalize the second argument of ditransitives etc. in virtue of its superset-of denotation, then€–s@,€–v and –ç@ are characterized by essentially the same denotation, projecting the Q category as well, as in (17). We come back to the somewhat special properties of –j in section 7.2.2.
254╅ The noun:€agreement, case and definiteness (17)╅ a.
kriatur/vazd
N a
Q vε
b. Q çǝ kriatur
N i
matʃ
N ε
c. Q s(ǝ)
While, as discussed in the previous section, –t(@) has either sentential scope (oblique) or scope over the noun (plural),€–ç@,€–s(@) and –v will have only the sentential scope that in our terms corresponds to the oblique. Since –s@ only appears in the definite feminine, we will say that its Q properties include definiteness besides the superset-of denotation, and it also has a selectional property, namely for some particular nominal class(es). The same selectional property characterizes –ç which lacks definiteness properties. In turn,€ –v only appears in the oblique plural. If we take it that its Q(⊆) properties lead to the oblique interpretation because of their sentential scope, then by the present reasoning they should not be sufficient to determine plurality as well. In other words, plurality should be a property of the bases selected by –v. Indeed, the oblique plural can always be obtained by adding –v to the indefinite plural (kriatura–v, matï†fi–v, etc). This takes us back to a problem left open so far and to which we return directly below in section 7.2.2, namely how nominal class inflections can determine number interpretation€– and of course case. Summarizing so far, the traditional idea that there is a relation of case assignment between, say, a verb and an embedded complement€– and that this relation is lexicalized through case morphemes€– translates here into an altogether different picture. This is that the intrinsic denotational properties of nominal inflections enable them to satisfy the argument slots of predicates; nominal inflections that are sensitive to the particular argument slot they satisfy yield
Analysis of Albanian nominal inflectionsâ•… 255 what are conventionally called case inflections. In our model of case, there is no case checking or evaluation, for there truly is no case. Rather, case is a descriptive label for relations that are much more primitive (essentially satisfaction of predicate–argument frames), as are the categories that enter into them (nominal class, quantification).5 7.2.2 Vocalic case inflections and lack of inflections We will now consider the structures in (12)–(13), where the inflection consists only of a nominal class morpheme. Let us begin with –i/–u and€–a as the singular definite nominative endings in the masculine and feminine respectively, as in (13a) and (13c). Following Chomsky (2001) or Pesetsky and Torrego (2007), nominative case is the reflex on a noun of a feature-checking relation it bears to a finite inflection. In present terms, however, agreement is not defined in terms of feature checking, for the finite inflection of a verb, as for instance in Italian (3), does not represent a bunch of non-interpretable features, but is a verb-internal EPP argument associated with the categorial signature D. As argued at the outset, different argumental constituents which jointly fill the same argument slot enter into a chain relation, which presupposes the compatibility of their denotational properties and hence the surface effect of agreement. Within this framework, saying that inflections such as –a, –i and –u agree with the finite verb means saying that they combine with the finite verb inflection in fixing the reference of the EPP argument of the sentence. Saying that they are nominative translates into saying that they provide a link in this chain. In short, there is no nominative case. There is on the one hand nominal class morphology, and on the other hand the chain it forms with the D verb inflection. Nominal class morphology is sufficient to satisfy this context in the singular. Masculine€–i/–u appears as the nominal class vowel throughout the singular declension, including the nominative definite. In the so-called feminine, however, while –a is the definite nominative, the nominal class vowel appearing elsewhere (eventually followed by consonantal endings) is€–/–@. This asymmetry is not due to the incompatibility of the –/–@ nominal classes with the nominative environment (i.e. agreement, or chain formation, with the D inflection of the verb), for –/–@ show up in the nominative context at least in the indefinite of the matʃ- and vazd- classes respectively. Rather, we conclude that while –a is compatible with definiteness, –/–@ are inherently indefinite. This is confirmed by the fact that, when appearing alone, –/–@ only have an indefinite interpretation (in the nominative/accusative of the matʃ- class and in the nominative/accusative singular of the vazd- class). We propose that in the relevant configuration, nominal class inflections can be read as definite precisely
256â•… The noun:€agreement, case and definiteness because they are in the scope of the D finite inflection. Masculine singular –i/–u requires such D closure when lexicalizing the inflection alone, since it does not appear in the indefinite. Consider now non-definite inflections, including –/–@ in the feminine and –a in the plural, appearing as the nominative/accusative. These nominal class inflections are either in the scope of quantifiers, for instance the indefinite article, with which they agree in indefiniteness, or of deictic specifiers (demonstratives) which have definiteness properties of their own. Both of these elements close (quantificationally or deictically) the reference of the N inflection.6 Within the structure of the noun phrase, a Q position can be assigned to the indefinite article/quantifier€– and a Loc position to the demonstrative element, essentially a deictic element, as in (18). (18)â•…
a. Q/Loc ɲǝ/aiɔ matʃ
N ε
kriatur/vazd
N a
b. Q/Loc tʃǝdɔ/ atɔ
In this connection, we should also consider another fact, namely that in the singular indefinite, at least the masculine bases kriatur and ri can satisfy the nominative/accusative environment in the absence of any inflectional specifications€– and this is true of the accusative as well. In the present framework, zero morphology is excluded€– fundamentally for reasons of restrictiveness. Therefore, what appear to be inflectionless terminals really are treated as such within the present analysis. We should then ensure that their inflectionless status does not create problems. Recall that we formalize theta-roles/argument slots as variables introduced by a predicative base and bound in all instances considered so far by referential material (nominal class specifications or quantificational/definiteness specifications). Lack of inflectional specifications does not yield ungrammaticality (characterizing, for instance, a language like English) as long as the variable is closed by syntactic-level material. In particular, in examples like (5a–b), it is closed by the indefinite article or by the demonstrative (a deictic referent) again, as in (19).
Analysis of Albanian nominal inflectionsâ•… 257 (19)â•…
Q/ Loc ɲǝ/ai
kriatur
Interestingly, in the variety we are considering here, Vena, the oblique demonstrative is followed in the feminine singular by a form of the noun identical to the nominative/accusative, hence inflected only for nominal class. By contrast, the masculine singular is regularly inflected for the oblique, by the –ç@ morphology. In present terms, –ç@ projects a Q category, which satisfies the dative argument slot of a selecting verb, as already schematized in (17b). Feminine –j, i.e. its counterpart for feminine bases, seems to have different intrinsic properties, since it appears to include the feminine nominal class morphology. In fact, we can take –j to be an allomorph of€ – specialized for phonological contexts where it is preceded by vowels. Therefore –j also projects an N position, as schematized in (20). If so, the indefinite oblique feminine singular, the€–j ending in the scope of indefinite determiners and the – ending in the scope of demonstratives are essentially the same element, i.e. the N nominal class morphology. In the scope of indefinite determiners this morphology is doubled, while it is not in the scope of demonstratives. This latter distribution can be treated by assuming that the demonstrative, because of its deictic reference, does not require the lexicalization of the€–j inflection doubling, i.e. it subsumes it in terms of the theory of mutual exclusions discussed in chapters 3–4 in connection with negative doubling. (20)â•…
ɲ
N i
Q çǝ vazd
N ε
N jε
In the indefinite plural oblique, the€–v inflection is optional in the scope of both indefinites and demonstratives€– though it obligatorily occurs in the definite paradigm. We can express this distribution in terms similar to those suggested for the distribution of the feminine singular€–j, i.e. as a by-product of the fact that the indefinite quantifier or the demonstrative can subsume the lexicalization of the oblique properties by the noun. Incidentally, the structure in (20) shows that the same morphological analysis that we have so far applied to nouns can be extended to determiners. Thus, the indefinite in (20)
258â•… The noun:€agreement, case and definiteness can be analysed as consisting of a lexical n- base to which properties of indefinite (existential) quantification can be imputed, followed by the nominal class inflection€ –i and by the oblique inflection€ –ç. Interestingly, demonstratives also follow the indefinite declension. This is compatible with our characterization of their referential properties not in terms of definiteness but rather of deixis (here Loc). When we consider the overall distribution of nominal class inflections, an interesting pattern seems to emerge. We note that they can appear as definite, but this will exclude plural and oblique, as in the nominative singular –a, –i, –u; or they can appear as oblique, but this will exclude definiteness and plurality, as in the oblique indefinite singular –(j); or they can appear as plural, but this will exclude definiteness and oblique interpretations, as in the indefinite non-oblique plural –a,€ –. In short, definiteness, oblique case and plurality appear to be compatible with (all or some) nominal case inflections, but only as long as no two of them co-occur. This complementary distribution is reminiscent of the facts reviewed in section 7.2.1 for the definite morphology –t(@), which can either have superset-of properties (oblique) or plural properties, but not both (i.e. cannot be both oblique and plural). For –t(@) we proposed that both so-called oblique and plural correspond to quantificational properties, inherently associated with –t(@). But since plurality depended on –t(@) taking noun phrase scope and so-called oblique on sentential scope, the two readings were predicted to be in complementary distribution. Definiteness, by hypothesis, is inherently associated with –t(@). There is an obvious difficulty in extending this treatment to nominal class inflections, namely that no quantificational properties have been imputed to them€– nor can they be, since we find such elements in contexts that do not warrant a quantificational treatment (e.g. singular, indefinite, non-oblique, as in matʃ-). This difficulty can, however, be circumvented if we assume that the definite, plural and oblique properties accruing to nominal class inflections depend on their closure by quantificational operators. Let us begin with definiteness, which is associated with vocalic inflections in the so-called nominative singular. On the basis of the discussion at the beginning of this section, in the so-called nominative configuration the nominal class inflection satisfies the EPP argument of the verb, forming a chain with the finite inflection of the verb, i.e. in present terms a D specification. In this configuration we propose that it can be read as definite€– because it is in the scope of D. No such independent closure is available for the oblique interpretation of nominal class inflections, in the indefinite singular, e.g. (20). We speculate, therefore, that it derives from a quantificational closure at the sentential level,
Analysis of Albanian nominal inflectionsâ•… 259 licensing the superset-of (possessor) interpretation. In other words, the Q(⊆) quantificational property that we associate with terminals such as –t(@),€ –s@ and –v is also available in the form of an abstract closure at the LF interface. It is the presence of this abstract quantifier, with sentential scope, that licenses the oblique (superset-of) interpretation in (20). In relation to –t(@), we have further proposed that the same Q(⊆) quantificational properties can be read as plurality when they take the nominal base in their scope. If Q(⊆) is available as a closure at the LF interface, then we also predict that nominal bases inflected only with a nominal class vowel could have a plural interpretation. In Albanian in particular, it is – and –a that allow for this closure. At this point, the objection may be raised that, although we do not allow abstract morphosyntactic material into our grammar (in the form of impoverished features, zero morphemes, silent categories à la Kayne, the string lexicalization of nanosyntax, or uninterpretable/unvalued properties), we do allow for abstract quantificational closures, which may do part of the work (for instance, the indefinite plural interpretation). It seems to us that this cannot represent an objection to our analysis. This is because at least existential closure for indefinites, and generic closure for PROs, are independently needed and generally postulated. In other words, even if we abstracted away entirely from syncretisms, we would need quantificational closures in natural languages. But as far as we can tell, the existence of impoverished or non-lexicalized features is entirely motivated by the syncretism phenomena they are devised to explain. 7.2.3 Prepositional contexts Prepositions in Albanian assign all the cases that are independently found in sentential contexts. Thus, they can select accusative, as in (21), and oblique, as in (22); finally, as shown in (23), nominative can also be embedded under prepositional elements. (21)
Preposition€– Accusative bɐri â•… p/ m â•… â•… ri-u-n it he.made â•… for/ with â•… man-Acc.def ‘He made it for/with the man’
(22)
Preposition€– Oblique vura prpara ╇ ri-u-t it I.put before â•… man-ms-Obl.def/ ‘I put it in front of the man’
(23) Preposition€– Nominative a. ai r t gruaj-a/ ri-u he goes to girl-Nom man-Nom ‘He went (close) to the woman/ the man’
260â•… The noun:€agreement, case and definiteness b. kij ki ban ga ri-u thisâ•… wereâ•… Artâ•… madeâ•… by man-Nom ‘This has been made by the man’
The fact that prepositional phrases license all cases that sentences do, would appear to be incompatible with the idea that prepositions assign a specialized oblique case in the sense of Chomsky (1995). One could object that the incompatibility of data such as (21)–(23) with the oblique case proposal of Chomsky (1995) disappears if the morphological component is taken into account. In the framework of Distributed Morphology, it could be assumed, for instance, that the insertion of at least some of the cases that prepositions select is due to impoverishment rules. Thus, we could assume that an impoverishment rule deletes oblique case from the prepositional contexts. However, prepositions in Albanian can select not one, but two different non-oblique cases, i.e. not just accusative, but also nominative. Therefore, the system has two different non-oblique entries (i.e. the nominative and the accusative) whose distribution in prepositional contexts cannot be described simply in terms of the underspecification of oblique€– but rather requires some positive characterization. Incidentally, the fact that the nominative is selected by prepositions is equally problematic for Chomsky’s (2001, 2008) construal of nominative as a reflex of agreement with the finite verb. As for Pesetsky and Torrego (2007), they identify prepositions with an aspectual type category:€‘a species of T merged below D and above NP’. They suggest that the selection of specific cases by certain subsets of prepositions must be connected with particular features associated with the varying properties of the event. However, when it comes to prepositions selecting the nominative, it is far from clear that Tprep can instantiate properties parallel to those of sentential T, precisely because sentential T is related to time reference while Tprep is related to the event. Let us now consider what can be said about cases selected by prepositions within the present approach. Prepositions are two-place predicates whose internal argument is independently lexicalized, while the external argument is controlled by some argument of the matrix predicate. For instance, in (22), the external argument of ‘in front of’ is controlled by the matrix accusative clitic ‘it’. Indeed, what ‘before’ denotes is a spatial relation between ‘it’ and ‘the man’. Consider now prepositional contexts like (21) which require the so-called accusative. In present terms, the prepositions that determine these contexts behave like transitive verbs in that their internal argument is satisfied, if definite, by the specialized nominal class morphology€–n(@) in the singular and by the nominal class + quantificational inflection –(V)t(@) in the plural. In
Analysis of Albanian nominal inflectionsâ•… 261 the indefinite, it is sufficient to have nominal class morphology or a bare nominal base, quantificationally closed. As for prepositional contexts selecting the oblique, as in (22), in present terms they require the satisfaction of their internal argument by argument morphology with Q specifications. These Q specifications correspond to a superset-of interpretation when taking a scope wider than the word; superset-of will then be the property selected by the preposition. Consider finally prepositional contexts requiring the so-called nominative. The gist of the present proposal is that there is no case, but only denotational properties capable of fixing argument reference in certain syntactic contexts. In particular, so-called nominative inflections are associated with contexts where they are in the scope of the D argument of the sentence (the finite verb inflection). Prepositions like t/ga in (23) then select inflections with the denotational properties found in the scope of D; these are the properties relevant for fixing the reference of their internal argument. 7.2.4 Summary The starting point of the present discussion was the data in Table 7.1, which laid out the nominal inflection system of Vena, classified in terms of the traditional case, definiteness and number categories. The aim was to show that case categories could be abandoned in favour of denotational primitives. Our perspective rests on a strict adherence to the (minimalist) postulates of projection of the syntax from actual terminals (no impoverishment, etc.) and of the ‘perfection’ of the computational module (no uninterpretability etc.). The overall picture that emerges from Albanian nominal inflections can be summarized in a fairly compact table (Table 7.2) which can now be substituted for Table 7.1. What we are perhaps most interested in is the fact that Table 7.2 is a genuinely different way of cutting the data, so that we expect direct empirical evidence to be able to discriminate between Table 7.1 (or its rendering by conventional morphosyntactic theories) and Table 7.2. Roughly speaking, there are three types of properties relevant for the nominal inflections of Vena:€ N(ominal class), Q(quantification), D(efiniteness). Vocalic N elements can have definite, or plural, or superset-of (oblique) interpretation, but we do not seem to see two or more of these combining. If the discussion at the end of section 7.2.2 is correct, this is due to the fact that such properties are not intrinsic to vocalic inflections, but contributed by the context of insertion in the shape of quantificational closures€– whence the parentheses in Table 7.2. It will be noted that in some instances the properties of two inflections overlap, such as those in the first line of the€–a and€–i entries
262â•… The noun:€agreement, case and definiteness Table 7.2 Denotational properties of Albanian (Vena) nominal inflections N -a -i,€–u -(j) -@ -t(@) -v -n(@) -ç@ -s@
+ + + + + +
Q=pl
Q=obl
(+)
(+) (+) +
+
Def
(+) + + + +
(+)
+ + + +
respectively; this, of course, corresponds to the fact that entries with the same content select for different sets of nominal bases. Among the other consonantal/syllabic endings,€–t(@) and –s@ are definiteness elements, while –ç@, –v are not; all are quantificational and will be plural (–t@) and/or oblique (–s@, –t(@),€–v,€–ç@). 7.3
The Albanian noun phrase
7.3.1 The genitive In presenting the major case configurations in (4)–(10), we omitted the genitive on purpose. The reason is twofold. On the one hand, genitive morphology overlaps with the morphology we have exemplified for dative. In other words, there is a single oblique case, which covers both dative environments and genitive ones. On the other hand, genitive contexts are distinguished from other oblique contexts in that the genitive is introduced by an article agreeing in number, gender and case with the noun that it is a complement of. These properties are illustrated by the data in (24)–(25), where the genitive is the complement of a definite noun, specifically a nominative in (24). In all the examples, the genitive can be seen to bear the same inflection as the dative; thus the definite genitive has€–s in the feminine singular,€–v in the plural, –t in the masculine singular; the indefinite genitive (introduced by an indefinite article or by a demonstrative) has€–ç in the masculine singular, –j in the feminine singular and again –v in the plural. The article that introduces
The Albanian noun phraseâ•… 263 the genitive agrees with the head noun. With a nominative head noun, i can be the pre-genitival article when the head noun is masculine singular, as in (24a), while can correspond to a feminine singular head noun, as in (24b). At the same time a certain degree of free variation is observed, whereby the t form, which lexicalizes the plural, can alternate with nominal class ones in the singular. (24) a. a’. a’’. b. c.
kutuâ•… biSt-i i/ t matS--s@ here is tail.nom.m the cat.gen.f ‘Here there is the tail of the cat’ biSt-i t asa-ç@ matS- tail.nom the of. that cat.f ‘the tail of that cat’ kj St t i-ç@/ ati-ç@ ri-u-ç@ this is the of.one/ of.that man ‘This is of a/ that man’s’ kjâ•… kmb-a matS--s@ this is leg.nom.f the cat.gen.f ‘This is the leg of the cat’ k@t jan biSt-@t matï†fi-v these are tails.nom.pl the cats.gen.f ‘These are the tails of the cats’
When the head noun is accusative, as in (25), the genitive can be introduced by , which again appears to be in free variation with t. The examples in (25b) and (25c) show that if a kinship term is the genitive complement of a head noun, it keeps its preposed article in addition to being embedded under the article agreeing with the head noun. (25) a. a’. b. b’. c.
p biSt-i-nâ•… matS--s@ I.sawâ•… tail.m the cat.gen.f ‘I saw the tail of the cat’ p biSt-i-n / t@ asaç@ matï†fi I saw tail.m the of.that cat.f ‘I saw the tail of that cat’ mra dr-@-n@ t@ mtr-@-s@ I.took hand.f the of.the sister.gen ‘I took the hand of the/his/her/their sister’ p kmb--n / t@ Ji-ç@ matS--j I.saw leg.f the of.one cat.f ‘I saw the leg of a cat’ mra kuputs--t t t nip-i-t I.took shoes.acc the of.the grandchild.gen-m ‘I took the shoes of the/his/her/their grandchild’
264╅ The noun:€agreement, case and definiteness Finally, when the head noun is indefinite, the article that introduces the genitive is always t, as seen in (26) for articles agreeing with a nominative singular, masculine in (b) and feminine in (a). (26) a. b.
/ aj kmb t@ matS--s@ a/that.nom leg the cat.gen.f ‘a/ that leg of the cat’ kj St t@ ri-u-t@ this is the man.gen ‘This is of the man’s’
The most notable structural property of genitives is that they are introduced by a determiner agreeing with the noun of which they are a complement€– or of which they are predicated, as in the copular structures in (24a”) and in (26b). As we will see in section 7.3.2, adjectives are characterized by a similar configuration, in which the adjective is preceded by a determiner agreeing with the noun it is predicated of, either within the noun phrase or in a copular structure. We conclude that in Albanian, all genitives (like all adjectives) are introduced as predications, not only in copular structures but in noun phrases as well. The structural representation of a DP like (24a) poses no particular problem. Quite simply, the head noun biʃti selects a noun phrase complement headed by a determiner, which agrees with the noun itself. In turn, this determiner takes the genitive noun (phrase) as its complement, as shown in (27a). The determiner agreeing with the matrix noun combines with the determiner introduced by kinship terms, as in (25b), showing that two different structural positions are involved. In (27b) it is particularly evident that, while the embedded determiner takes the form t for the oblique, the higher determiner takes the form agreeing in accusative properties with the head noun. (27)â•… a. N i
biʃt
D i matʃ
N ǝ
b. D ε
D tǝ mɔtr
N ǝ
Q sǝ
Q sǝ
The Albanian noun phraseâ•… 265 A similar structure will of course characterize indefinite genitives, preceded by an indefinite determiner or a demonstrative, and embedded under a determiner agreeing with the head noun, as in (28), corresponding to example (25b’). (28)â•… D tǝ
D ɲiçǝ matʃ
N ε
N jε
What first interests us here is what in traditional terms would be described as the syncretism of the genitive inflection with the dative. In section 7.2.1. we argued that this syncretism is based on the fact that the relevant Q morphology specifies a superset-of denotation. Thus, the second internal argument of ‘give’, i.e. the traditional dative, in present terms participates in fixing the reference of the first internal argument, i.e. the accusative, by denoting a superset including it. Similarly, the traditional genitive specifies a superset in terms of which the reference of the head noun is fixed. In Albanian, there is of course a formal difference between datives and genitives, namely that genitives are embedded under a determiner agreeing with the head noun; this is not true of datives. This difference correlates simply with the different points of merger of the genitive and the dative. Merger within a noun phrase requires the predicative layer provided by the determiner agreeing with the head noun (also found in copular structures); in dative environments, the oblique merges directly as complement of a verb.7 Let us consider the articles that appear in front of genitives. These belong to the same set that we have independently seen for kinship terms, including i, and t. We have already characterized t as a pure definiteness morpheme capable of carrying definite denotation independently of nominal class, number and case specifications. This is precisely what it does when introducing genitive complements of indefinite head nouns, and optionally of definite head nouns as well. Recall that if the head noun is indefinite, the pre-genitival article is t in both the nominative and in the accusative, irrespective of gender and nominal class. What the data suggest is that with indefinite head nouns, the predicative structure introducing the genitive must overtly lexicalize definiteness. If the head noun is definite, we find i in the nominative masculine singular, exactly as for kinship terms. On the other hand, can be found not only in the
266â•… The noun:€agreement, case and definiteness nominative singular (feminine, as with kinship terms), but also in the accusative singular (irrespective of nominal class). We have characterized the i and morphemes as nominal class elements:€there is no reason why this characterization cannot be maintained for their occurrence as articles. Thus, there is a single lexical entry for all of the occurrences of i and , though their points of merger vary. In particular, the article is lexicalized by in all accusative contexts; this is consistent with the conclusion that accusative is satisfied by nominal class properties N. In the nominative, article again occurs with plural and feminine head nouns, but the masculine requires i, in other words the morphology specialized for the masculine nominal class. Evidently, the fact that the article ultimately participates in lexicalizing the EPP (D) argument of the sentence requires the emergence of specialized lexical class morphology (i for the masculine, and for the feminine). The oblique and the plural have t throughout. 7.3.2 The adjective The normally attested adjectival construction in Albanian, including Arbëresh varieties, has the inflected adjective preceded by the article. Within the noun phrase, the adjective appears postnominally. Although in the standard variety a subclass of adjectives can appear prenominally, this possibility appears to be missing in Vena. In turn, the article–adjective sequence is not restricted to noun-phrase-internal contexts, but appears in predicative contexts as well, including the copular construction. This observation leads us to conclude that the pre-adjectival article is part of the structure of the adjective phrase, and not of the noun phrase in which the adjective is eventually inserted. Consider first the nominative, illustrated in (29) with the copular construction and in (30) with adjectives embedded in noun phrases. As shown in (29b), the plural bears the€ –a inflection, while in the singular the masculine is not inflected, having the same form as the bare adjectival base, and the feminine bears a vocalic - inflection in some subclasses, as in (29a’). Nominal class (gender) and number may equally be lexicalized by stem allomorphies, as for vk@ç@ ‘small.m’ vs. vg@λ@ ‘small.f’. In turn, pre-adjectival articles take the i, , t@ form that we have already seen for kinship terms and pre-genitival articles; i appears in the masculine singular, in the feminine singular and t@ in the plural. The adjective agrees with the noun (phrase) it is predicated of; the pre-adjectival article agrees with both the noun and the adjective in turn. (29) a. St i traS/ traS s/he.isâ•… the.m fat/ the.f fat ‘S/he is fat’
The Albanian noun phraseâ•… 267 a’. b.
St i ma/ ma s/he.is the.m big/ the.f big ‘S/he is big’ jan t@ traSa/ â•… t mbi they.are the fat/ the big ‘They are fat/ big’
(30) a.â•… a’. b. b’ c’. c.
r dia-i i ma/ vk@ç@ came boy the big/ small ‘The big/ small boy came’ r  diaç@ i vk@ç@ came a boy the small ‘A small boy came’ r vazd-a ma/ vg came girl the big/ small ‘The big/ small girl came’ r  vazd@ vg came a girl the small ‘A small girl came’ r aj vazd@ vg/ ma came that girl the small/ big ‘That small/ big girl came’ r kriatura-t@â•… t@ mbia/ vga cameâ•… boys theâ•… big/ small ‘The big/ small boys came’
Vena displays no sensitivity to case in the adjectival embedding. Thus, in the accusative in (31) and in the dative in (32), both the adjectival ending and the preposed article have exactly the same form as in the nominative in (30)€– displaying sensitivity only to nominal class and number. (31) a. a’. b. b’. c.
p dia-i-n i vkiç@ I.saw boy-acc the small ‘I saw the small boy’ p at@ diaâ•… i vkiç@/ i ma@ I.sawâ•… thatâ•… boy the small/the big ‘I saw that small/ big boy’ p vazd-@n@ vgi I.saw girl-acc the small ‘I saw the small girl’ mra @ mbisal maar- I. took a tablecloth the biggish ‘I took a biggish tablecloth’ p kriatur-a-t t@ vga I.saw boys-acc the small ‘I saw the small boys’
268â•… The noun:€agreement, case and definiteness (32) a. b. b’. c.
ja dia-i-t i vkiç@ him-it I.gave boy-dat the small ‘I gave it to the small boy’ ja vazd-s vogi@ her-it I.gave girl-dat the small ‘I gave it to the small girl’ ja â•›iç@ vazd-j vgi@/ asaç@ vazd@ ma her-it I.gave to.a girl the small/ to.that girl the big ‘I gave it to a small girl/ to the big girl’ ja kriatur-a-v t@ vga them-it I.gave boys-dat the small ‘I gave it to the small boys’
In short, the adjectives in (29)–(32) are formed like nouns by a predicative base followed by an inflection sensitive to nominal class and number, but not to case. We take it that these adjectival inflections are argumental elements capable of satisfying the obligatory (internal) argument of the predicative base. The presence of the article in front of the adjective can be imputed to a requirement that adjectival structures be closed by a D (EPP) element, very much as noun phrases are in Italian or English. This article picks up the same argument as the adjectival inflection with which it agrees, as shown for a feminine singular in (33). (33)â•… D ε maD
N ε
Consider the embedding of a structure like (33) in a copular context. Agreement between the adjective and the EPP argument of the copula corresponds to the fact that the latter forms a chain with the adjectival inflection and article. In virtue of this chain relation, the internal argument slot of the adjective is ultimately assigned to the EPP argument of the copula. Consider, then, the embedding of structures like (33) inside a noun phrase, as schematized in (34) for the nominative feminine singular. We know that both of the predicative bases present in (34), i.e. the adjectival and the nominal one, have (at least) one argumental slot, filled by the inflection€– and for adjectives also by the article. In present terms, these two (sets of) arguments slots are satisfied by the same individual(s). On the one hand, this provides a basis for the
The Albanian noun phraseâ•… 269 intersective interpretation typical of adjectival modification, whereby a ‘big girl’ is an individual which is both ‘a girl’ and ‘big (for a girl)’. On the other hand, it means that agreement holds of the noun, the pre-adjectival article and the adjective. (34)
vazd
N a
D ε
maD
N ε
We have noted that in Vena, the adjective does not agree with the head noun in case (nor is it sensitive to its definiteness properties). In other words, the Vena variety has a fully developed case system in the nominal domain, but not in the adjectival domain. This is not simply a morphological property of adjectival bases, for when the same bases are the head of a noun phrase (rather than predicated of the head noun), they are inflected just like nouns are (section 7.3.3). In standard Albanian (as described by Solano 1972; Turano 2002), by contrast, the pre-adjectival article is sensitive to case (and to the definiteness of the head noun). In classical frameworks in which gender, number and case form a bundle of features of the noun, or in a framework like Chomsky’s (2008) in which case is merely a consequence of agreement rules applying to number and gender, the fact that case is lexicalized in certain contexts but not in others must be the result of morphological rules. In the present framework, no such rule suppressing the overt realization of an abstractly present case category is necessary€– or possible. Rather, syntactic contexts differ as to whether they are or are not associated with case. In section 7.2, we have argued that case is merely the name of a specialized argument; if so, the question as to its absence becomes a question as to the absence of this specialized argument. It is natural to assume that the absence of the case argument on the adjective in (34) is connected to the fact that the argument slot of the matrix predicate is independently satisfied by the case argument of the noun. Therefore, the identification of the argument of the adjective with that of the noun does not require a case argument to be lexicalized on the adjective as well. Importantly, the basis for distinguishing the descriptive category ‘noun’ from the descriptive category ‘adjective’ is not that the predicative bases occurring in adjectival and nominal contexts are marked for one or the other category. Rather, it is only the contexts of occurrence that can be labelled in that way. In the variety under consideration, therefore, case
270â•… The noun:€agreement, case and definiteness discriminates between adjectival embeddings (case-less) and nominal embeddings (case-marked)€– not between nouns and adjectives. Let us now turn to the pre-adjectival article. The literature concludes€– much as we do here€– that the article is part of the adjectival constituent (DimitrovaVulchanova and Giusti 1998; Turano 2002, 2003; Giusti and Turano 2007). However, according to Dimitrova-Vulchanova and Giusti (1998), the pre-adjectival article is just an agreement marker (and a ‘redundant’ one). For them, the order Noun–Adjective in (34) is generated by movement of the noun to a Focus position, licensing the D position of the noun phrase. We have already argued against this derivation in discussing the internal structure of kinship terms in section 7.2. In turn, Turano (2002, 2003), following Cinque (1995), argues that the adjective is generated in the Spec of a functional projection dominating the noun, so that the Noun–Adjective order in (34) is derived by movement of the Noun to D. If our discussion in section 7.2 is on the right track, kinship terms exclude such a derivation, since they show that postnominal definiteness inflections and prenominal articles can co-occur; therefore the former cannot be derivationally related to the latter. At the same time, there are some differences between the distribution of preadjectival articles and that of articles in front of kinship terms. In particular, the data concerning kinship terms in (8) vs. (9) or in (10) show that demonstratives and indefinite quantifiers, including ted ‘some’ in (10) and ‘a’ in (9), are in complementary distribution with the prenominal article. Elements quantifying over the adjective are not in complementary distribution with the pre-adjectival article, but rather precede it, as in (35). (35)â•… St m@/ Sum i ma/ ma s/he.is more/ very the.m big/ the.f big ‘S/he is bigger/ very big’
A minimal contrast with the quantifier ï†fium in (35) is provided by the occurrence of the same element as a quantifier of the noun in (36a)€– where it is in complementary distribution with the prenominal article. Furthermore, quantificational (and other) material can appear between the article and the kinship noun, as illustrated in (36b) with a numeral quantifier. (36)â•… a.â•… b.
∫um ku∫iriç many cousins ‘many cousins (of his/her/theirs)’ t katra ku∫iriçt the four cousins ‘the/his/her/their four cousins’
The Albanian noun phraseâ•… 271 The contrast between (35) and (36b) suggests that the determiner is inserted in a lower position within the adjective phrase than it is within the noun phrases. The high position of the determiner within the noun phrase has led the literature (Szabolcsi 1994) to the conclusion that the determiner has a strict affinity with the C position of the sentence. By contrast, we can analyse the pre-adjectival determiner as an inflectional-level element, filling the D position of the I domain, as in (37), where it is preceded by the indefinite quantifier in Q of the C domain. In this way, we also capture Dimitrova-Vulchanova and Giusti’s (1998) intuition that the pre-adjectival determiner is an agreement element of sorts.8 (37) Q mǝ/ʃum
(C) D ε
N ε
maD
On the basis of the contrast between (36) and (35), we conclude that the prenominal article of kinship terms is a determiner of the English/Romance type, inserted in the D position of the C domain, as in (38), where it can be followed by quantificational material in the same domain or in the lower inflectional domain, such as the numeral. The complementary distribution between the definite determiner and the indefinite quantifier seen in (8) vs. (9), or in (35), is not necessarily due to competition for the same position, but rather to the fact that their interpretations are mutually exclusive. (38) D tǝ
Q katra kuʃiri
...
It is also worth noting that possessives precede kinship nouns, and are in turn preceded by prenominal determiners, as in (39). This further confirms the availability of inflectional domains (for the positioning of the possessive clitic) between the determiner and the noun. The (1st person) possessive itself has an inflection sensitive to nominal class and number, alternating in particular
272â•… The noun:€agreement, case and definiteness between the bare stem im, with a masculine singular head noun, and the€ – inflection with a feminine head noun. This agreement of possessives with the head noun is observed in Romance as well, and is analysed by Manzini and Savoia (2005) for those languages.9 (39)â•… a. b.
p t im kuSiri/ t im kuSirir/ t im@ kuSiriç@ I.sawâ•… the my cousin.m/ the my cousin.f/ the my cousins ‘I saw my cousin/ cousins’ ja t im kuSirir-i-ç@/ s ╇ im kuSirir/ t im kuSiriç-v him-it I.gave the my cousin.dat.m/ the my cousin.dat.f/ the my cousins.dat ‘I gave it to my cousin/ cousins’
7.3.3 Adjectives as heads of the noun phrase The insensitivity of Albanian adjectival inflections to case (and in Vena also to the definiteness of the head noun) is not a lexical property of the bases involved. Rather, the same bases that have been illustrated above as heads of adjective phrases embedded in a noun phrase can also occur as heads of noun phrases. If so, they display full nominal inflection, and like kinship terms they are also preceded by the definite determiner, as illustrated in (40). (40)â•… a. a’. a’’. b. b’’.â•… c.
r i vgi/â•… vga came theâ•… little.m/ the little.f ‘The little one came’ r t@ vga-t@ came the little-pl ‘The little ones came’ r m i maD-i came more the big.m ‘The bigger one came’ p t@ vgi-n@/â•… t@ vgn@/ t@ I.saw theâ•… small.m/ theâ•… small.f/ the ‘I saw the small one(s)’ p m t mbiDç--t I.saw more the big.pl ‘I saw the bigger ones’ t@ vgi-t@/ t@â•… vgs@/â•… ja him.itâ•… I.gaveâ•… theâ•… small.m/ the small.f/ ‘I gave it to the small one(s)’
vga-t@ small-pl
t@ vga-v the small.pl
These adjectival heads differ from kinship terms in some important respects, in which they parallel the structure of adjectival phrases. In particular, as can be seen in (40a’’), the adjective and its article can be preceded by a quantifier,
The Albanian noun phraseâ•… 273 such as the degree element m ‘more’. Similarly, unlike what happens in kinship terms, where the determiner excludes indefinite quantifiers and demonstratives, the latter can precede adjectival bases and their articles, as shown in (41). On the other hand, the demonstrative and the indefinite article combine with the indefinite inflection on the adjectival/nominal head. The latter is sensitive only to nominal class and number, except for the oblique plural, which optionally displays the case ending, as in (41c’). (41) a. a’. b. b’. c. c’. c’’.
r ai i vkiç@/ aj vgi came that the little.m/ that the little.f ‘That little one came’ r  i vkiç@ / vgi came a the â•… little.m/ the little.f ‘A little one came’ p at╅ i vkiç@/ vgi/ at t@ vgia I.sawâ•… that the little.m/ the little.f/ those the little.pl ‘I saw that little one/ those little ones’ p ╅ i vkiç@/ vgi I.sawâ•… a theâ•… little.m/ the little.f ‘I saw a little one’ ja atiç@ i vkiç@/ asaç@ vgi him.itâ•… I.gave to.that the little.m/ to.that the little.f ‘I gave it to that little one’ ja atir/ atir@v t@ vogia(v) them.it I.gave to.those the little.pl ‘I gave it to those little ones’ ja Jiç@ i vkiç@/ vgi him.it I.gave to.a the little.m/ the little.f ‘I gave it to a little one’
The examples in (40)–(41) cannot be accounted for through ellipsis of the head noun. In (40), in particular, such an analysis would force us to say that in the context of a deleted definite noun, the adjective takes on the definite inflection. This in turn requires a framework which uses morphological spellout of syntactic features€– so that definiteness features can be passed from the deleted/silent head noun to the adjective before they are actually instantiated in the morphological component. By contrast, we are working with a unified morphosyntactic component. Given present assumptions, a sentence like (40a) can take the structure in (42). The nominative inflection N and the D article satisfy the internal Â�argument slot of the adjectival base vg@ and at the same time the argument slot of the matrix predicate. So-called nominative case in (42) corresponds to
274â•… The noun:€agreement, case and definiteness the D closure of the N inflection, as detailed in section 7.2.2. As we have just seen, in a noun ellipsis treatment, the distinction between nouns and adjectives, built on the presence of empty structures, determines the surface distribution of cases via morphological manipulations. By contrast, in the present framework, the projection of a case category, as determined by the requirements of predicate–argument structure, effectively distinguishes what we call a noun from what we call an adjective. (42)
I εrθ
D i y
vɔgƏ
N i
The nominal constituent in (42) and the adjective phrase in (37) are alike in that the article can combine with higher quantifiers that precede it, specifically degree quantifiers, as illustrated in (43). In (43), the article is therefore in the inflectional D position, as we have assumed for the adjective phrase in (37), rather than in the high C-domain position that we have assumed for kinship nouns in (38). (43)
Q mǝ
(C) D tǝ mbiD εçǝ
N tǝ
On these grounds, we equally expect that, while the determiner of kinship nouns is in complementary distribution with demonstratives and indefinites, the article in front of adjectival bases combines with them, as shown in (44).
Concluding remarksâ•… 275 (44) Q ŋǝ
(C) D ε
7.4
y
vɔgi
N ǝ
Concluding remarks
The proposals in this chapter merely represent a particular instance of a more general perspective we take on the minimalist framework. This perspective restricts current models by rejecting abstract terminals in favour of projection from the lexicon. The analysis of Romance and Albanian phenomena in this and other chapters of the book aims to demonstrate that the sharp divide between the functional and substantive lexicons that current generative literature often takes for granted does not have any real empirical motivation. Rather, the syntactic computation is built on the properties of the actual lexical terminals€– but it does not necessarily register every component of what we call the meaning of a sentence€– which the syntax merely restricts (cf. Culicover and Jackendoff 2005). The specific aim of the present chapter was to account for traditional case inflection within this general framework. The latter consists entirely of referential properties, including nominal class, definiteness and quantification, and it is these properties that enter into the projection of the syntactic tree. The traditional notion of case corresponds simply to the fact that different sets of denotational properties satisfy different syntactic environments, defined by agreement, theta-assignment and in general by the primitive relations of minimalist theory. We would argue further that by cutting away a lot of abstractness, our approach ends up being simpler than other possible solutions to what is by and large a commonly perceived problem. Most of the discussion was devoted to the crucial issue of empirical adequacy. In this respect, we have illustrated what is required for our approach to account for a fairly complex nominal inflection system, such as the one found in the Vena variety of Albanian.
8 (Definite) denotation and case in Romance:€history and variation In historical accounts of the transition from Latin to the Romance languages, the loss of case morphology is related to the evolution of Latin demonstratives into articles and to the typological change from SOV word order to SVO word order. Starting from the observation that change from Latin to Romance languages involved a shift from head-final order to head-initial order, Renzi (1987) proposes that in a string of the type N–k, the case suffix k, identified with a functional head, can no longer stay to the right of N, but must be reordered and move into first position.1 In terms of this explanation, the article is the lexical support for the case affix; the resulting sequence Art+k N realizes the desired head-initial order. In other words, the development of the Latin demonstrative into an article (cf. Vincent 1997) makes it possible for case to be lost as a nominal suffix. This account is not without problems. For instance, it predicts that we should find some Romance language (or language stage) in which case is realized exclusively on the determiner€– but in Old French both determiners and nouns have case (though the determiners may be argued to have a fuller set of case distinctions); and in Italian neither does. A potentially better argument for determiners bearing case to the exclusion of nouns is Romanian; in Romanian, however, the article is postnominal, and can be argued to be an inflection (section 8.2). More to the point, nouns of the feminine class have a residual inflection for dative/genitive as opposed to nominative/accusative. Giusti (1995, 2001) develops Renzi’s (1987) idea further. She separates the article from the other determiners of the noun and identifies it with a functional head also subsuming case; in her terms, F ‘is a nominal functional category which subsumes D[eterminer] and K[ase]’ (1995:€79). The article, therefore, is only ‘a syntactic means of expressing case’. She notices that her approach is ‘in contrast with current semantic theories which take the article as a kind of quantifier’. She questions, however, the ability of these theories to explain ‘how languages with no article can implement the mechanisms of nominal 276
The Latin case systemâ•… 277 interpretation currently attributed to the article in English’ (1995:€89). We shall return to this question in the concluding section. From a diachronic perspective, Giusti (2001:€168) suggests that in a language like Latin, ‘the rich morphology on the noun makes the N-[to-D]chain visible … even if N has not moved’; in Romance languages like Italian, ‘although the case morphology is not strong enough to make the N-chain visible … the presence of the newly formed article complies with the same function’. For both Renzi (1987) and Giusti (2001), therefore, case is a primitive category of grammar, and the difference between Latin and Romance is that, whereas Latin realizes this category as a nominal inflection, Romance lexicalizes the same category as part of the determiner (Renzi) or as the determiner (Giusti). Giusti dissociates the determiner from definiteness and other denotational properties, strengthening the link between determiner and case to a formal equivalence. In essence, therefore, what survives is only case, and its expression either by morphological case or by the determiner. Now, the theory that emerges from our discussion of Albanian case in Â�chapter 7 is that, if case and definiteness/quantification are merely two names for the same fundamental categorial specifications, the characterization of this category is D/Q rather than K€– in other words, Giusti’s stance is reversed. In this chapter, we seek to confirm the conclusions arrived at in the previous chapter on the basis of an analysis of the Latin case system (section 8.1) and its developments in Romance, including in particular the case systems of Romanian (section 8.2) and Old French (section 8.3), as well as survivals of nominative –s in Romansh varieties (section 8.4). 8.1
The Latin case system
The gist of the present proposal regarding case, as laid out in chapter 7, is that the nominal class inflection of, say, the modern Romance varieties is not substantially different from case systems such as that of Albanian, in the sense that ‘case’ inflections, like ‘agreement’ inflections, lexicalize denotational properties of nominal class (gender), quantification, definiteness, etc. In virtue of these denotational properties, they satisfy the argument specifications of the predicate base (the ‘noun’) to which they attach. ‘Case’ inflections differ from ‘agreement’ inflections in that they are restricted to certain syntactico-semantic configurations of embedding. In other words, they are specialized for attachment of the noun (phrase) as the complement of a superordinate verb (‘accusative’), as an EPP argument (‘nominative’), as a complement of a superordinate noun (‘genitive’) etc.
278â•… Denotation and case in Romance:€history & variation These conclusions, if correct, should easily extend from Albanian, considered in chapter 7, to other case languages. From this perspective, we briefly consider the case system of Latin, with a view to studying its development in the Romance languages. A morphemic analysis of the Latin case system is provided by Halle and Vaux (1997). In (1), we match the case endings resulting from this analysis, as schematized in Halle and Vaux’s Table (14), to the stems of the five standard conjugations, with classes II–IV also displaying separate forms for neuter. For Halle and Vaux (1997), these stems are formed by the root followed by the thematic vowels –a for the I class, –o for the II class, –i for the III class, –u for the IV class and –e for the V class. Phonological readjustment rules which delete, lengthen, shorten and change the quality of thematic vowels, as detailed in Halle and Vaux’s (19), would be responsible for the rather more complex thematic vowel schema emerging from (1). (1)
Singular
Plural
(I) Nom.â•… Gen. Dat. Acc. Abl.
ros ros ros ros ros
ros ros ros ros ros
(II) Nom. Gen. Dat. Acc. Abl.
-a -a -i: -a -i: -a -m -a:
-a -a: -i: -a: -i:
-i: -r-um -s -s -s
lup -u -s lup -i: lup -o: lup -u -m lup -o:
lup lup -o: lup -i: lup -o: lup -i:
-i: -r-um -s -s -s
(IIn.) Nom. Gen. Dat. Acc. Abl.
bell -u -m bell -i: bell -o: bell -u -m bell -o:
bell bell bell bell bell
(III) Nom. Gen. Dat. Acc. Abl.
can can can can can
(IIIn.) Nom. Gen. Dat. Acc.
corpus corpor -i -s corpor -i: corpus
-i -s -i -s -i: -e -m -e
-a -o: -r-um -i: -s -a -i: -s
can -e: can can -i can -e: can -i
-s -um -bu-s -s -bu-s
corpor corpor corpor -i corpor
-a -um -bu-s -a
The Latin case systemâ•… 279
Abl. (IV) Nom. Gen. Dat. Acc. Abl. (IVn.) Nom. Gen. Dat. Acc. Abl.
(V) Nom. Gen. Dat. Acc. Abl.
corpor -e
corpor -i
-bu-s
fruct fruct fruct fruct fruct
-u -s -u: -s -u -i: -u -m -u:
fruct fruct fruct fruct fruct
-u: -u -i -u: -i
-s -um -bu-s -s -bu-s
corn corn corn corn corn
-u: -u: -s -u: -u: -u:
corn corn corn corn corn
-u -u -i -u -i
-a -um -bu-s -a -bu-s
di di di di di
-e: -s -e: -i: -e: -i: -e -m -e:
di di di di di
-e: -e: -e: -e: -e:
-s -r-um -bu-s -s -bu-s
We adopt the inventory of case endings from Halle and Vaux (1997) (cf. also Calabrese 1998, 2008). We differ from them only in excluding zero endings; therefore, in instances where they would have a zero morpheme, we have just left a blank space in the relevant column of the paradigm in (1). As we have already done for Albanian, we assume that forms lacking a case ending are simply closed off by a thematic vowel (e.g. the I class abl. singular ros–a:€‘with the rose’) or correspond to bare roots (e.g. the III class neuter nom./acc. singular corpus ‘the body’). For Halle and Vaux (1997), the Latin case dictionary involves eight morphemes, including zero, for which they provide the lexical entries in (2). Cases are characterized by the features [±structural], referring to whether the case is structural or not in Chomsky’s (1986) sense, [±superior], referring to whether the case is assigned under government (-superior) or not, and [±oblique]. Halle and Vaux develop the view that the –r–um genitive plural of classes I, II and V depends on the combination of –s with –um morphology followed by rhotacism of€–s (i.e. conversion to€–r in intervocalic position). But even taking –um and –rum to be allomorphs which are not further analysable would not change the table of lexical entries in any essential respect. Similarly, for Halle and Vaux the –bu–s dative/ablative plural of classes II, IV and V is derived by combining –bu with –s; again, assigning a lexical entry to –bus rather than to –bu would not change the picture in (2) in any crucial way. We will therefore disregard this internal analysis in the discussion to follow.
280╅ Denotation and case in Romance:€history & variation (2)╅
umâ•… bu i: i: a m Ø s
[+obl +struct -sup [+obl +sup [+obl +struct [-obl +sup [-obl [-obl [ [
+pl] +pl] / [III, IV, V]+ -pl] +pl] / [-neut, I, II]+ +pl] / [+neut] + -pl] -pl] ]
Various general assumptions underlying the lexicon in (2) have been rejected as part of the discussion in chapter 7€– notably that the grammar includes relational features such as [±superior] (i.e. [±governed]), etc. and that lexical items may be specified for negative properties or radically underspecified. There are less general aspects of the Latin case lexicon in (2) that interest us here. For instance, the different occurrences of –i cannot be unified, but require two separate homophonous entries, one for the nominative plural occurrences, and one for the genitive/dative occurrences. Interestingly, the distribution of –i is very close to that of Albanian –t. In chapter 7, we saw that –t occurs in the nominative/accusative plural and in the oblique singular; thus, it can be plural or it can be oblique, but it cannot be plural and oblique. Analogously, Latin –i occurs as the dative and/or genitive singular (either genitive, or dative, or both according to inflectional class) and as the nominative plural; but it does not occur as the oblique plural. The analysis of Albanian –t in chapter 7 allows us to provide a unified entry for Latin –i. As for Albanian –t, we propose that Latin –i has quantificational properties. When they take scope over the word, they yield the plural reading (classes I and II, nonneuter). When they are construed as having wider (phrasal/ sentential) scope, they are read roughly as superset-of specifications€– i.e. as genitive/dative (in classes I and V), as genitive (noun phrase scope, in class II), or as dative (sentential scope in classes III and IV).2 Thus, a class II form like lup–i in (3) is ambiguous between the reading ‘the wolves’ (plural) and the reading ‘of the wolf’ (genitive singular). There is only one real difference between Albanian –t and Latin –i, namely that€–t characterizes all non-oblique contexts of insertion in the plural. By contrast, Latin –i in the plural is restricted to the so-called nominative context, which in terms of the discussion of chapter 7 is characterized by agreement (i.e. chain formation) with the D inflection of the finite verb. (3)
lup
Q i
The Latin case systemâ•… 281 It is evident that the present approach is not a notational variant of the more traditional approach represented by such works as Halle and Vaux (1997), since their respective dictionaries are clearly different. In particular, under the present approach, we are able to unify what for Halle and Vaux are two separate€–i morphemes into a single lexical entry. What is more, under the present approach the underlying distribution of –i is seen to be the same as that of Albanian€–t. The fact that the same pattern can be found with genetically unrelated inflections strengthens the argument against it corresponding simply to accidental homophonies or to default mechanisms. Johnston (1997:€102–7) shows that the same syncretism between genitive singular and nominative plural is also found in Russian. His discussion is interesting because he recognizes the systematicity of the pattern, though the ‘homonymies’ cannot be modelled within the ‘geometrical’ model he advocates, ‘because the elements involved have no element in common along any inflectional dimension’ (102). In fact he even predicts the possibility of a ‘geometric constraint that the relevant paradigm cells not be continuous’. This is precisely what our model (unlike Distributed Morphology) can provide. Another major syncretic morphology is€–s, which is treated as the general default of the system by Halle and Vaux. The notion of default is extraneous to our grammar€– where lexical terminals project syntactic structure on the basis of their positively specified properties. In fact, by and large the distribution of –s does not differ substantially from that of€–i. Consider, for instance, class III, where€–s occurs as the conventional nominative singular in (4a), the genitive singular in (4b) and the plural, both nominative and accusative, in (4c). Nominative singular is also attested for class II, genitive singular for classes II, IV and V, nominative/accusative plural for classes IV and V and accusative plural for all classes. In classes I and II, –s also occurs as the dative/ablative plural, which we will treat last. (4) â•… a. b. c.
Canis currit dog.sg.nom runs ‘The dog is running’ canis cauda dog’s tail ‘the dog’s tail’ Canes currunt/ video dogs run/â•… I.see ‘(The) dogs are running’/ ‘I see (the) dogs’
The lack of any feature characterization for –s in the lexicon of Halle and Vaux (1997) means that€–s can automatically be inserted in any of the contexts in (4). However, the Subset Principle that crucially governs Lexical Insertion
282â•… Denotation and case in Romance:€history & variation requires that ‘the most highly specified Vocabulary Item whose features are a subset of the features of the terminal node is inserted’. Consider, then, the nominative singular, which is characterized as [–oblique, +superior, +structural,€–plural] because it is an argument of the verb ([–oblique]), it is not governed ([–superior]) and [+structural]. For independent reasons one must assume that there is a lexical entry in Latin which is specialized for [–oblique,€–plural], namely –m, which occurs in the accusative singular of all nominal classes, as well as in the nominative of the neuter 2nd class. Therefore, –m should take precedence over –s in the nominative singular. In order to avoid this unwelcome result, an Impoverishment rule is postulated deleting the feature [–plural] in the nominative singular, yielding a node which is compatible only with the –s default terminal. An extension of this rule is further envisaged to allow for the insertion of –s in the genitive singular. However, Impoverishment processes such as the one just described are evidently ad hoc. In turn, paradoxically, a default item like –s, whose insertion is permitted by such processes, finds itself filling a multiplicity of case slots not in virtue of its rich case properties, but in virtue of their total absence. What we propose is that –s, far from being an empty element as in Halle and Vaux (1997), has a denotational content, which we identify again with Q, as in (5), corresponding to the structure for nominative singular and plural. The plural interpretation depends on the quantificational specifications of –s taking scope over the noun, as already discussed for –i in (3). To account for its occurrences in the singular, we assume, exactly as for –i in (3), that Q elements in morphology have scope properties€– a notion that ultimately goes back to Pesetsky (1985). The singular readings of the –s morphology, in turn, correspond to Q taking a scope wider than the noun (phrasal or sentential). Specifically, in the singular nominative configuration, we interpret the scope of€–s as sentential. We assume that the EPP argument provides a D closure of the sentence, and correspondingly requires quantificational Q properties, supplied by –s, to satisfy the syntactic context of insertion that it defines. (5)
can
N i/e
Q sx
currx
D itx /untx
In the so-called genitive, we can take the scope of –s to be the entire noun phrase; the genitive argument is then interpreted as ‘including’ the head
The Latin case systemâ•… 283 referent. The genitive reading of€s naturally leads us to the other oblique reading, as the syncretic dative/ablative plural. Ablative has not been discussed so far; since in the variety of Albanian considered here there is no systematic attestation of ablative. The range of readings associated with ablative in Latin is roughly comparable to that of Romance clitics like Italian ci, French y etc.€– i.e. basically locative and instrumental. Similarly, the syncretism of the ablative with the dative in the plural (in all classes), as well as in the singular of class II, is strongly reminiscent of Romance clitic systems in which 3rd person dative is lexicalized by the ci-type (locative) form (Manzini and Savoia 2005, 2007, 2008a; Kayne 2008a). Now, we have already seen that in Latin dative is also syncretic with genitive (cf. the –i inflection), giving rise to a threeway connection between dative, genitive and ablative. This same connection is independently known from possessive constructions (Freeze 1992) in which the possessor can be a descriptive genitive (the book is mine), a descriptive dative (Latin liber mihi est ‘lit:€the book is to me’), or a descriptive locative (Russian u menja est’ kniga ‘lit:€ the book is at me’). In present terms, this conceptual closeness underlying the observed syncretisms can be captured by the notion of superset-of or ‘zonal inclusion’ in the sense of Belvin and den Dikken (1997), as discussed in chapters 6–7. The latter can be construed as the possessive proper (genitive), or it can be defined spatially, resulting in the locative. The dative can correspond to either of these. On the basis of these various observations, we tentatively construe the so-called ablative as a quantificational Q element in turn.3 Coming back to the syncretic dative/ablative plural in –s, it is especially interesting in the present context in that it seems to attest a combination of plural and oblique values that is excluded both for Albanian –t and for Latin –i. This exclusion led us to theorize that the ‘crossed’ syncretism (either plural, or oblique, but no oblique plural) was due to the fact that Q specification of Albanian –t and Latin –i could take scope over the word (plural) or over sentential constituents (oblique), but not both. Interestingly, the oblique (dative/ ablative) construal of Latin –s appears to allow for this possibility. In reality, we think another analysis more likely. Note that in the dative/ ablative plural, the vowel preceding –s is always –i. The discussion of Halle and Vaux (1997) seems to imply that this is one of the many phonological readjustments involving thematic vowels in the Latin case declension. However, it seems unlikely that there is a phonological process leading from –a, –o thematic vowels in I, II class to –i. Therefore we assume that the –i–s ending of classes I and II oblique (dative/ablative) is really made up of the two quantificational elements –i and –s. The first plausibly takes the same word scope that
284â•… Denotation and case in Romance:€history & variation yields the plural reading in the (3) and in its class I counterpart; –s can then introduce the oblique (sentential scope) reading. In short, there are intrinsic lexical properties of –s and there are syntactic environments that it can satisfy. One traditionally labels these environments as case configurations and the terminals such as€–s as case terminals. Yet case has no reality, for the real properties of argumental terminals are denotational, such as Q, and the real nature of the configurations they enter into is that of argument–predicate (thematic) structures, agreement structures, etc. The same holds for€–i, reviewed above, and for environments where the nominal base is simply closed by the nominal class vowel. Regarding the latter, note first that Halle and Vaux (1997) have two separate entries for the same vocalic morpheme€–a, which is both the thematic vowel for class I and the nominative/accusative case ending for all neuter classes (II–IV). In the present approach, the same nominal class morphology N (specifically€–a) can either be followed by other nominal inflections (e.g. the specialized case inflections –m, –rum, etc. in class I) or follow them (e.g. the thematic vowels€ –i- of class III,€ –u- of class IV in the neuter plural). This means that –a can be given a unified lexical entry. In general, the distribution of nominal class morphology in the case paradigm of Latin is highly reminiscent of that found in the case paradigm of Albanian in chapter 7. In the Albanian indefinite conjugation, we found it as the nominative/accusative singular or plural as well as the oblique, but only in the Â�singular; similarly, we found it as the nominative singular, but not the plural, in the definite conjugation. We interpreted this distribution as a reflex of the need for nominal class morphology to be in the scope of quantificational closures. The latter have the same range of interpretations as overt Q/D morphology. Thus, nominal class morphology in the range of such a closure can be interpreted as plural or oblique, but not as plural oblique; it can also be interpreted as definite, but only in the nominative, i.e. in the context defined by the presence of the D inflection of the finite verb. As we commented in chapter 7, closures by abstract operators are not a notational variant of abstract (zero) morphology, given that Q/D closures are restricted in ways in which zero morphologies are not. Going back to Latin, nominal class morphology appears as the sole inflection in the dative singular (class II), in the nominative singular (class I) and in the nominative/accusative neuter, both singular (class IV) and plural (all neuter classes). In all classes, the ablative singular corresponds to the bare thematic vowel. In the plural reading, we take the nominal class morphology to be closed by a Q specification with word-internal scope, yielding set formation, i.e. plural; in the oblique (dative), the same Q specifications take sentential
The Latin case systemâ•… 285 scope, yielding the superset-of (possessor) reading. The nominative singular depends on the nominal class inflection being in the scope of the D (EPP) inflection of the finite verb. On the basis of the discussion of ablative –s (or at least of its locative value, cf. fn. 2) we take it that the dative/ablative syncretism of class II corresponds to the Q closure of the thematic vowel –o at the sentential level. The ablative (locative) reading of the thematic vowel in all classes will correspond to the same structure. We also follow the treatment provided for Albanian bare lexical bases in chapter 7 in dealing with Latin bare lexical bases such as the class III neuter corpus ‘the body’. We take it that in the absence of nominal class vowels and/ or of specialized ‘case’ inflections, the internal argument of the predicative base corresponds to a variable, interpreted in the scope either of some overt operator (for instance an indefinite quantifier within the noun phrase) or in the scope of the usual Q closure. Finally, in (1) we have chosen to illustrate what traditional grammars describe as class III bases in –e. According to traditional grammars, there are class II bases in –i, and Halle and Vaux (1997) consider –i to be the thematic vowel for class III in general, on the basis in particular of plurals such as animal–i–a ‘the animals’ or animal–i–um ‘of the animals’. If, therefore,€–i is also a nominal class vowel, we must assume that the Q properties that we have intrinsically associated with it in the discussion and in Table 8.1 are the result of quantificational closures€– no less than for the other nominal class vowels. On the basis of the preceding discussion, we are able to produce (a first version of) a case lexicon of Latin, on the model of that discussed in considerably more detail for Albanian in chapter 7. The lexicon in (1) appears to have more entries than that in (2)€– but this is only because Halle and Vaux (1997) list the five thematic vowels separately. Therefore they have thirteen entries, to which our nine entries in Table 8.1 compare favourably. In Table 8.1, we have followed the same notational conventions as in Table 7.2, parenthesizing those properties that do not appear to be intrinsic to the single morpheme, but are rather to be provided by quantificational closure available at the interpretive interface. For nominal class vowels, the tabulated values correspond to the instances in which they are not followed by other inflections. Three inflections (or inflection sets) have not been discussed, namely –m,€ –bu–s and –r–um. However, their specialized nature means that they can be slotted in fairly uncontroversially. The descriptive labels in the rightmost column are merely given to facilitate comparison with the data in (2). Let us summarize so far. One of our purposes in analysing the so-called case system of Latin (at least of the particular variety of Latin described by grammarians) was to verify whether the pattern uncovered by our study of Albanian
286╅ Denotation and case in Romance:€history & variation Table 8.1 Denotational properties of Latin nominal inflections
-a -o -e -u -i -s -bu-s -m (-r)-um
N
Q.pl
Q.obl
D
+ + + + + + + + + + + -
(+) (+) + + +
(+) (+) (+) (+) (+) + + +
(+) (+) (+) + + -
nom sg nom/acc pl abl sg dat/abl sg abl sg nom sg acc sg dat/abl sg nom pl gen/dat sg. gen/dat/abl nom/acc pl nom sg dat/abl pl acc sg nom sg gen pl
could account for other systems of roughly equal complexity. We found that the same basic set of N, Q and D specifications, together with scopal and selectional mechanisms, could adequately describe both systems. Importantly, although Albanian and Latin are genetically related languages, their case lexicons only marginally overlap (for instance in the –m/–n of the accusative singular)€– so that what we have uncovered is not a historical relation. Our other purpose in studying Latin was to set the stage for the analysis of the relation connecting it to the modern Romance languages, which mostly lack case, but are endowed with a richer determiner system, including the definite article. In what follows, we will argue that the account of the Latin system summarized in Table 8.1 provides for a natural account of the development of Romance case systems, including Romanian and Old French€– as well as for the relation between loss of case and introduction of determiners. We argue that the view of case that emerges from Halle and Vaux (1997) provides for a rather more opaque account of the historical development (cf. Calabrese 1998, 2008). 8.2
Romance case systems:€Romanian
Modern Romance languages generally preserve a case system in the pronominal domain; however, nouns (and adjectives) display no case inflections. The clitic systems of Romance varieties are analysed by Manzini and Savoia (2005,
Romance case systems:€Romanianâ•… 287 2007, 2008a), in terms of the same syntactic and interpretive categorizations of Table 8.1 (cf. section 8.5). Here we concentrate on nominal inflections€– beginning with the only modern Romance language which does have so-called case, namely Romanian. Romanian, like Albanian, is described as a language with postnominal articles (Dobrovie-Sorin 1987). On the model of Albanian, however, we will assume that the postnominal article is generated as a nominal inflection within the noun€– and in this sense we will simply refer to it as the definite declension of Romanian nouns. Romanian has an indefinite declension in the feminine, since the singular direct case (nominative/accusative) is –ă while the oblique (dative/genitive) is –e. There are clearer case distinctions in the definite declension and on quantifiers preceding indefinite nouns. In (6) we exemplify the nominative/accusative singular, in (7) the oblique singular and in (8)–(9) their plural counterparts. The (a) examples involve indefinites and the (b) example definites. Note that the dative can also be expressed through the preposition la; we exemplify this in the indefinite plural in (9a). (6)â•… Nominative/ accusative singular a.â•… a venit/ am văzut un băiat/ o fat-ă has come/ I.have seen a boy/ a girl-fsg ‘There came/ I saw a boy/ girl’ b. a venit/ am văzut băiat-ul/ fat-a has come/ I.have seen boy-def.msg/ girl-def.fsg ‘There came/ I saw the boy/ girl’ (7) Dative singular a. (i)-l am dat un-u-i băiat/ un-e-i fet-e him.itâ•… I.haveâ•… givenâ•… a-msg-obl boy/ a-fsg-obl girl-fsg.obl ‘I gave it to a boy/ girl’ b. (i)-l am dat băiat-ul-u-i/ fet-e-i him.it I.have given boy-def-msg-obl/ girl-fsg-obl ‘I gave it to the boy/ girl’ (8)â•… Nominative/ accusative plural a. au venit /am văzut do-i băieţ-i/ dou-ă fet-e have come/ I.have seen two boy-mpl/ two girl-fpl ‘There came/ I saw two boys/ girls’ b. au venit/ am văzut băietţ-i/ fet-e-l-e have come/ I.have seen boy-mpl-mpl/ girls-fpl-def-fpl ‘There came/ I saw the boys/ girls’ (9) Dative plural a. (i)-l am dat la doi băieţ-i / două fet-e him.itâ•… I.haveâ•… givenâ•… toâ•… two boy-mpl/â•… two girl-fpl ‘I gave it to two boys/ girls’
288â•… Denotation and case in Romance:€history & variation b. (i)-l am dat băieţ-i-l-or / fet-e-l-or him.it I.have given boy-mpl-def-obl/ girl-fpl-def-obl ‘I gave it to two boys/ girls’
A close point of contact between Albanian and Romanian is represented by genitive structures. So-called genitive and dative inflections are identical; however, genitives, i.e. obliques depending on a noun, are introduced by a (prenominal) determiner agreeing with the head noun, al (m.sg.), a (f.sg.), ai (m.pl.) or ale (f.pl.). The pre-genitival article can also take an invariant form a (which has the same form as the feminine singular) instead of the agreeing form. We exemplify this in predicative structures in (10). We progressively simplify morphological segmentation and glosses in the interest of readability. (10)â•… Genitive a.â•… ăsta e al/a om-ul-ui/ oamini-l-or / un-ei fet-e this.m is the.msg/the man-the-obl/ men-the-obl/ a-obl girl ‘This is (that) of the man/ the men/ a girl’ b. asta e a băiat-ul-ui/ un-ui om this.f is the(fsg) boy-the-obl/ a-obl man ‘This is (that) of the boy/ a man’ c. ăştia sîntâ•… ai/ a copil-ul-ui/ fet-ei / fete-l-or these.mâ•… are the.mpl/theâ•… child-the-obl/â•… girl-obl/â•… girls-the-obl ‘These are (those) of the child/ the girl/ the girls’ d. astea sîntâ•… ale/ a băiat-ul-ui/ un-ui băiat these.fâ•… are the.fpl/the boy-the-obl/ a-obl boy ‘These belong to the/a boy’
When it comes to embedding inside a noun phrase, a pre-genitival article agreeing with the head noun is necessary if the head noun is indefinite, as in (11), although a definite head noun licenses the straight embedding of a genitive, as in (12). (11)â•… Genitive a.â•… un pahar al/ a băiat-ul-ui/ un-ui băiat a glass the boy-the-obl/ a-obl boy ‘a glass of the/ a boy’ b. o carte a un-ui băiat/ băiat-ul-ui / un-ei fet-e a book the a-obl boy/ boy-the-obl/ a-obl girl-obl ‘a book of the boy/of a woman’ c. doi pantof-i ai/ a băiat-ul-ui / fet-ei/ fete-l-or two shoes-mpl the boy-the-obl/ girl-obl/ girls-the-obl ‘two shoes of the boy/ girl/ girls’ d. două kămăş-i ale băiat-ul-ui / un-ui băiat two shirts-fpl the boy-the-obl/ a-obl boy ‘two shirts of the/ a boy’
Romance case systems:€Romanian╅ 289 (12)╅ a.╅ b. c. d.
pahar-ul băiat-ul-ui / fete-i / une-i fet-e glass-theâ•… boy-the-obl/ girl-obl/ a-obl girl-obl ‘the glass of the boy/ the girl/ a girl’ carte-a om-ul-ui book-fsg man-the-obl ‘the man’s book’ pantofi-i băiat-ul-ui/ fete-i/ fete-lor shoes-mpl boy-the-obl/â•… girl-obl/ girls-the-obl ‘the boy’s/ girl’s/ girls’ shoes’ kămăşi-le băiat-ul-ui / fete-i / une-i fet-e shirts-the.fpl boy-the-obl/ girl-obl/ girls-the-obl ‘the boy’s/ girl’s/ girls’ shoes’
It is a crucial property of the present framework that lexical terminals have the properties necessary and sufficient to project morphosyntactic structures. In such a framework, the variation between lexicons is expected to Â�(exhaustively) define the variation between grammars€– providing a straightforward implementation of Chomsky’s (1995) minimalist programme in this respect. What holds for variation holds for change, from one grammar to another; from this perspective, the change from the Latin ‘case’ system in Table 8.1 to the Romanian one should be definable in terms of their respective lexicons. Three of the case endings of Table 8.1 survive in Romanian, namely –a of the nominative singular, -(o)r of the oblique plural and –i of both the oblique singular and the nominative plural (masculine). In present terms, the distribution of –i preserves that of Latin –i in fundamental respects. Suppose, then, we assume that Romanian –i is to be characterized essentially like Latin –i, as a Q element. As such, we predict that it will have the plural reading when taking scope over the words€– or the possessive (dative/ genitive) reading when taking sentential scope. This is precisely what we find, leading us to conclude that Latin –i has essentially been preserved in the lexicon of Romanian. Recall further that Latin –r–um in turn represented specialized oblique plural morphology€– whose value is again preserved in Romanian -(o)r. The –i and –or endings exhaust the case morphology of the masculine nominal class; in the nominative/accusative singular, the –l morphology, to which we return immediately below, is either bare, as in băiat–ul ‘the boy’, or followed by a nominal class inflection –e, as in câine–le ‘the dog’. The masculine and feminine in turn are alike in the oblique. In the nominative/accusative, both singular and plural, they again display the pure nominal class endings –a and –e. We have already commented in chapter 7 (cf. here section 8.1 concerning
290â•… Denotation and case in Romance:€history & variation in particular the –a neuter plural of Latin) on the fact that plurality can simply correspond to a change in nominal class. Next we take the –l morphology into account. In connection with this, we come back to the classical issue in Romance historical linguistics that introduced our discussion€– i.e. the relation between the partial or complete loss of case in Romance languages and the development of a determiner system. In Romanian, the question is set in slightly different terms, since this language has developed a set of specialized definite inflections with€–l morphology. In discussing Latin, we effectively assumed that so-called case inflections are sufficient to provide quantificational closures for the noun; when nouns are closed by thematic vowels, abstract closure at the interpretive interface is available. This means, among other things, that Latin does not need a system of determiners, which provide a quantificational/ definite closure in syntax. By contrast, in Romanian, the introduction of –l morphology corresponds to the need for a specialized definiteness closure of the noun (phrase). In the singular, on the other hand, this morphology is present in the masculine, but not in the feminine; in the plural nominative/ accusative, the relation between the two nominal classes is reversed in the sense that the –e feminine inflection is supported by –l, while€–i suffices to determine a definiteness closure. Let us illustrate the structural analysis of the definite nouns (noun phrases) of Romanian that results from the lexicon just reviewed. The most complex organization corresponds to the oblique singular (masculine) in (13a), the oblique plural in (13b) and the nominative/accusative feminine plural in (13c). These have three separate layers of inflection, including the quantificational specifications –or for the oblique plural in (13b) and –i for the oblique singular in (13a), as well as for the plural (masculine) in (13b). Note that when it represents the oblique, –i is attached above the –l definiteness ending, as in (13a), while when it represents the plural, it is attached below the l- definiteness morphology, as in (13b). The same –i entry is involved in both instances; only its position changes. Similarly, the same nominal class morphology –e is involved twice in (13c). The other type of inflection found in (13) is the nominal class vowel –u in (13a); this raises the question of whether –ul in (13a) should not be further decomposed. Here we chose to treat –ul as an allomorph of –l because there is no –u nominal class ending independently instantiated as the sister of a root.
Romance case systems:€Romanian╅ 291 (13) a.
N u
D ul
bǎiat
Q i
b.
oamen fet
Q or
D l
Q i e
c.
fet
N e
D l
N e
The genitive structures of Romanian can in turn be accounted for along the lines of our treatment of Albanian in chapter 7. The most notable structural property of genitives emerging from (10)–(11) is that they are introduced by a determiner agreeing with the noun of which they are a complement. Cornilescu (1995:€ 49) already concludes that ‘the Gen assigning functional head AL is not a preposition, but a functional nominal, a D° in the same family of Genassigning morphemes with English –s’. We also treat al/a/ai/ale as D elements€– but not as genitive case assigners, but rather as heads of a predicative structure. Following the discussion of the parallel Albanian structures in chapter 7, we associate phrases like (11a) with structures of the type in (14), where the head noun pahar selects a determiner which in turn takes the genitive noun (phrase) băjatului as its complement. The fact that the embedded determiner does not agree with the genitive noun shows that it cannot be the determiner
292╅ Denotation and case in Romance:€history & variation of the noun itself. Rather, the genitive, instead of being inserted directly as the complement of the head noun, is construed as the complement of a determiner predicated of the head noun. (14)
Q un
D al
pahar
bǎiat
Q i
N u
D ul
Romanian differs from Albanian in that there is at least one context in which the genitive is directly embedded under the head noun, namely when the head noun is definite and the genitive is adjacent to it, as illustrated in structure (15) for example (12a). (15)
pahar
D ul bǎiat
D ul
N u
Q i
We propose that the lack of a pre-genitival determiner in structures like (15) is due to a mutual exclusion of the type analysed in the Romance literature for sequences of two l- clitics (in particular in Spurious se environments, cf. Halle and Marantz 1993). In the terms suggested by Manzini and Savoia (2005, 2007, 2008a), the lexicalization of one l- form succeeds in lexicalizing definiteness properties for the entire clitic domain, essentially by taking scope over it. In (15), therefore, we propose that the lexicalization of the pre-genitival article agreeing with the head noun is excluded by the fact that it would be in the immediate scope of definite morphology. The fact that the pre-genitival article is excluded only under adjacency is again very similar to what we have observed elsewhere for l- clitics; thus, mutual exclusion is not observed if one of the two clitics is in enclisis and the other is in proclisis on the verb (cf. here also chapter 4 on the mutual exclusion of two negative n- forms).
Romance case systems:€Romanianâ•… 293 As for the syncretism of the genitive inflection with the dative, in chapter 7 we argued that it is based on the fact that the relevant Q morphology has a superset-of/inclusion denotation. Thus, the second internal argument of ‘give’, i.e. the traditional dative, in present terms participates in fixing the reference of the first internal argument, i.e. the accusative, by denoting a superset including it. Similarly, the traditional genitive specifies a superset in terms of which the reference of the head noun is fixed. As we have just seen, in Romanian, as in Albanian, genitives are (typically) embedded under a determiner agreeing with the head noun€– which is not true of datives. This difference straightforwardly correlates with the different points of merger of the genitive and dative. Merger within a noun phrase requires the predicative layer provided by the determiner agreeing with the head noun; in dative environments, the oblique merges directly as the complement of a verb. In recent work, particularly Cuervo (2003) for Spanish, based on Pylkkänen (2002) (cf. Torres Morais and Salles (to appear) on European vs. Brazilian Portuguese), the dative interpretation depends on the presence of an Applicative head; Diaconescu and Rivero (2007) embrace this model for Romanian. In their conception, an Applicative head is a functional category which takes the internal argument of the predicate as its complement, and expresses a ‘possession’ or other (source, goal) relation between this argument and the Â�argument in its Spec, i.e. the descriptive dative. Under these analyses, the Romance dative clitic is a lexicalization of the Applicative head. In the present proposal, we capture the relevant empirical generalizations without having recourse to the postulation of an Applicative head. Thus, oblique morphology is itself characterized as a dyadic (superset-of/inclusion) operator with scope over the elementary event taking the descriptive dative and the internal argument of the verb as its arguments. To the extent that the sentential scope position of the Q(⊆) operator is identified with that of the dative clitic, we also capture the connection between the latter and oblique case. In short, Applicative heads are a primitive of the theory which assumes them, no less than ‘dative’ is in other theories, and in this sense it is just a restatement of the problem. Here we have proposed a genuine reduction of the notion ‘dative’ to more elementary primitives, i.e. the Q(⊆) operator. Under the Applicative approach, datives and genitives are distinguished not only by their different structural attachment€– to the sentence and to the noun phrase respectively€– but also by the presence of the Applicative functional head for datives. So one may wonder why the genitive–dative syncretism could arise, except as an accident of the PF interface. Furthermore, under the Applicative approach, it is the dative that has (or can have) the more complex embedding
294â•… Denotation and case in Romance:€history & variation structure; the overt evidence tells us that the more complex embedding structure characterizes the genitive. Summarizing so far, the general approach to nominal inflections, and specifically to case, elaborated in the discussions of Albanian (chapter 7) and Latin (section 8.1) easily accommodates the system of Romanian and the variation between Latin and Romanian€– and hence the historical change between them. As in the minimalist programme of Chomsky (1995), the variation between two grammars is (exhaustively) determined by the variation between their two lexicons€– and so is the change from one to another. As for the classical problem in Romance historical linguistics that we started with, namely the relation between (partial or complete) loss of case and the introduction of the l determiner/clitic/inflection system, what we proposed in the discussion surrounding (13) is that a nominal inflection system like the Latin one was sufficient to determine a (definite) closure of the noun phrase in the absence of syntacticlevel quantificational specifications. This is no longer true in Romanian, which introduces –l morphology to produce such a closure. Recall the comparison in chapter 7 between the nominal inflections of Albanian, a ‘case-marked’ language, and of Italian, a ‘case-less’ language. In both languages (as in Latin, Romanian, etc.), contentive terms correspond to properties, relations, etc.€– i.e. to predicates. Thus, in Italian macchin- is the property of being a ‘machine/car’; in Albanian kriatur- is the property of being a ‘child’, etc. These properties cannot denote individuals and sets of individuals, i.e. elements that can provide arguments for other predicates, unless some semantic type-shifting occurs. In the classical semantic view (Higginbotham 1985; Chierchia 1998), this type-shifting is provided by the determiner system. Under the present conception, nominal morphology is sufficient to shift the predicative base to an argument (in our terms itself providing an argument capable of saturating both the nominal base and other superordinate predicates). The inflectional specifications of a language may in fact suffice for a referential closure. This is what happens in Albanian and Latin. In other languages, the argument closure of the predicative base (its shift from predicate to individual) requires the additional presence of determiners, specifically definite determiners. This is what happens in Italian€– and it is also the core of the innovation introduced in Romanian by the –l morphology. Within this model, the relation between (partial or complete) loss of the nominal inflections of Latin and the introduction of determiners (or determinerÂ�like morphology) which characterizes the passage from Latin to Romance can simply be described as the change from a certain system of ‘type-shifter’ (an entirely morphological one) to another one (broken down in inflections and
Other Romance case systems and alternative accountsâ•… 295 determiners). Recall that Giusti (2001) proposed to capture the same relation by taking the ‘functional’ F properties of case to be primitives and attributing the definite determiners of Romance to this F category. It seems to us that Giusti’s (2001) proposal has the same problems as any theory that takes case (or a no better-defined functional specification including it) to be a primitive of grammar. If the present analysis is on the right track, the historical connection between Latin case and Romance determiners€– and more generally between nominal inflection and determiner systems€– can be captured, while at the same time eliminating the spurious category of ‘case’. 8.3
Other Romance case systems€– and alternative accounts
The development of Latin into the medieval Gallo-Romance languages is characterized by a rather different set of changes from those observed in Romanian. Old French preserves a reduced case system in the masculine, based on the distinction of nominative and objective case. In the singular, –s, taken to be derived from the nominative singular of Latin, characterizes the nominative, while an alternant devoid of –s appears in non-nominative contexts, continuing an original accusative/oblique form. In the plural, the alternant with€–s, corresponding to the plural accusative of Latin, characterizes objective contexts, while the alternant without –s, taken to continue non-sigmatic plurals, appears in the nominative (Brunot and Bruneau 1969:€133 ff.). Relevant examples from Old French are provided in (16). (16)â•… a.â•… b.
Dur sunt li colp e li caples est grefs Hard are the blows and the scuffleâ•… is hard (Chanson de Roland, 1678) Guardez le champ … e le-s munz (munt-s) Watch over the field … and the mountains (Chanson de Roland, 2434)
Given the preceding discussion, our take on the change from Latin in Table 8.1 to Old French in (16) is fairly straightforward. Recall that in the discussion surrounding (4), we have associated Latin –s with a quantificational content, which takes on different interpretations according to different scope instantiations. Thus, its plural interpretation corresponds to scope over the word, its genitive interpretation to scope over the noun phrase and its dative interpretation to sentential scope. Consider Old French, then. The facts in (16) are consistent with it preserving both the€–s morphology and a quantificational content for it. The only change that intervenes is that the distribution of Old French –s is slightly simplified with respect to that of Latin. Either –s takes scope over
296â•… Denotation and case in Romance:€history & variation the word, yielding the plural reading, as in (17b), or it takes sentential scope, yielding the nominative reading, as in (17a). Other case terminals of Latin (in particular, consonantal specialized endings such as€–bus for oblique plural,€–m for accusative and neuter nominative, etc.) are simply not continued in Old French. (17) a. [ [√ caple] [Q s]] b. [ [√ munt] [Q s]]
The other development observed in (16), besides the loss of nominal inflections other than –s, is the development of the definite determiner. This is explained along the same lines as the development of the –l inflectional morphology of Romanian in section 8.2, i.e. as a result of the fact that nominal inflections no longer suffice to provide a referential closure for the noun (phrase). This is then independently provided by the determiner system, along the lines discussed in section 8.2 in connection with the development of Romanian. Leaving aside the determiner issue for now, an account of the development of Latin nominal inflections up to both Old French and Romanian is provided by Calabrese (1998) within the framework of Distributed Morphology. The gist of his argument is that one should distinguish contextual from absolute syncretism, where the latter corresponds to lack of attestation in a given language for a particular feature opposition. Contextual syncretism is treated by the mechanisms of underspecification and default described above in relation to Halle and Vaux’s (1997) analysis of Latin –s. However, absolute syncretism requires the postulation of a set of universal constraints disallowing certain feature combinations, ordered in a markedness hierarchy. In a language which has all possible case oppositions, none of the case constraints applies; languages that disallow certain cases activate one or more restrictions, in the order defined by the hierarchy. The activation of these constraints means that the relevant feature combinations are removed from the terminal nodes generated by the syntax; to be more precise, repair rules apply to reduce disallowed feature combinations into allowed ones. What interests us here directly is the account that Calabrese (1998) proposes for the change from Latin to Old French and Romanian in terms of his system. In both languages, the reduction in case oppositions is due to the activation of feature constraints that were inactive in Latin, with repair rules responsible for the consequent reduction of certain cases to others. In particular, according to Calabrese, the activation of constraints disallowing dative and ablative and the subsequent application of repair rules reduced the oblique to the [+possessor,€ –location] form, i.e. the genitive. This yielded the Proto-Romance
Other Romance case systems and alternative accountsâ•… 297 Â� three-case nominative–accusative–genitive system reconstructed by the historical literature (De Dardel and Gaeng 1992; Zamboni 1990, 1998). From this system, according to Calabrese, both Old French and Romanian were derived by the activation of further constraints. In Old French, the *[+possessor,€–location] constraint is activated, disallowing genitive. This triggers the repair of the offending terminal nodes [–subject,€–direct, +possessor,€–location …] into [–subject, +direct,€–possessor,€–location …], so that genitive (the surviving oblique of the system) ends up coinciding with accusative. Romanian activates the *[–subject, +direct] constraint, disallowing accusative. This triggers the repair of the offending terminal nodes [–subject, +direct …] into [+subject, +direct …], so that nominative exponents take the place of accusative ones, giving a system of nominative (direct) vs. genitive (oblique). Calabrese (1998:€111–12) explicitly rejects the idea that changes are due to the simple loss of lexical items, in this instance case endings, with the concomitant extension of the use of other lexical items … One could propose that more marked case endings are lost and replaced by less marked ones. Such an analysis would simply not work. For example … the ending /–s/ has a special status being the Elsewhere case … If we assume that syncretic changes are due to the loss of lexical items, we should expect … the ending /–s/ to be extended to uses that it did not have before.
In short, for Calabrese: the evolution of these case systems seems to operate only through operations on cases regardless of the lexical items composing the case systems. The best way of representing these changes is therefore by modifications in the morphosyntactic component, i.e. through the activation of case restrictions.
Let us consider first the problems that Calabrese envisages for lexically based accounts of historical development. His criticisms presuppose the system of Distributed Morphology, in which syntactic structure is projected from abstract feature bundles and actual terminals typically have a radically underspecified content, as in (2). In such a system, it is not possible to describe the change from Latin to Old French by saying that Old French simply keeps the –s entry of Latin. Since the –s entry of Latin is empty, its distribution in French is not predictable. However, Calabrese’s objections do not apply to lexicons consisting of positively specified entries such as ours, as summarized in Table 8.1. In our lexicon, the –s morphology of Latin has Q properties; it is these, rather than an Elsewhere characterization, that determine its distribution. Therefore it becomes possible for us to assume that the lexicon, and hence the grammar, of Old French are characterized by the preservation of –s, whose distribution is predicted as discussed above.
298â•… Denotation and case in Romance:€history & variation It is true that other –s inflectional systems could equally be derived on the basis of the preservation of this morphology from Latin into Romance, for instance one in which –s covered the nominative/accusative plural. But this is a general property of lexical change. Thus, contentive items can shift in interpretation in constrained ways€– it is not predictable which particular value in the permitted range any given language will pick. Essentially the same will be true of pieces of the functional, morphological lexicon. Conversely, not any system could be derived, for instance, not one where –s corresponded to the accusative singular€– which is lexicalized in present terms by N elements and hence excludes a Q lexicalization. By contrast, Calabrese’s account of historical change maintains the same opaque relation between the underlying abstract morphosyntax and the terminal nodes forming the lexicon, which characterizes the analysis of single linguistic systems, such as Latin in (2). Thus, the development of the abstract morphosyntax, involving case constraints and readjustment rules, and the development of the lexical terminals run on parallel tracks. For instance, the preservation of –s in Old French as the default exponent runs on a parallel track with respect to the survival of the sole nominative–accusative (non-Â�nominative) opposition. Therefore the fact that the two developments match up is just a matter of chance; other matches are equally possible, even assuming the same underlying development and the same dictionary with –s as default. In other words, there is no reason inherent to the system or the historical process why –s couldn’t show up, say, in the objective singular. Furthermore, since, on the account of Calabrese (1998, 2008), what happens on the way from Latin to Old French is the activation of several constraints and the application of the corresponding repair rules, a language with few case contrasts like Old French turns out to have a grammar as rich in case specifications as a language much richer in case contrasts like Latin. In terms of the classical discussion by Kiparsky (1982), the paradox is that the superficially simpler language, here Old French, is more opaque than a superficially more complex one, here Latin. For Kiparsky, a rule A → B / C___D is opaque to the extent that there are surface representations of the form (i) A in environment C___D or (ii) B in environment other than C___D.
He goes on to speculate that opaque rules are a cognitive burden, and that historical development could or should operate in the direction of a simplification of such contexts. Yet in the development from the Latin to the Romance nominal inflectional systems, the progressively more pervasive syncretisms
Other Romance case systems and alternative accountsâ•… 299 hypothesized by Calabrese (1998, 2008) would take us to a progressively more opaque system, hence a progressively more complex one on the basis of what we indeed consider to be a reasonable definition of complexity. Thus, from the perspective of the child who is learning the language, an opaque rule (as defined above) is obviously difficult to learn from the available evidence, since there is no surface cue for it. In the present model, morphosyntactic structures are projected from lexical terminals. In turn, lexical items are associated just with the denotational properties (nominal class, quantification, definiteness for arguments) that characterize them independently of the position of insertion; case is merely the name given to lexical items which in virtue of these properties are specialized for the satisfaction of certain syntactic contexts of embedding. In this model, there is no abstract case system and no constraints and rules mapping it to the PF interface. Therefore the change from Latin to Old French cannot be a change in these constraints and rules. All that is affected is the lexicon of so-called case inflections€– which in Old French is reduced to –s, maintaining its core characterization as a quantificational element. It should be noted at this point that we are not isolated in the current literature in taking issue with the Distributed Morphology model of the morpholexical interface. Others have noted that restrictiveness issues arise in connection with underspecified lexical items and the morphological rules (impoverishment etc.) justifying their insertion. Correspondingly, alternative models have been proposed. In particular, case systems, and specifically case syncretisms, are the core evidence considered by Caha (2009) within the Nanosyntax model (Starke 2009). There are points of contact between Caha’s model and the present one, namely the idea that categories cannot be negatively specified or underspecified for certain properties, and correspondingly that there are no morphological rules (Impoverishment) manipulating abstract terminal nodes to allow for the insertion of underspecified lexical items. Despite the importance of this assumption (cf. Kayne (2006, 2008a)), Nanosyntax remains conceptually close to Distributed Morphology in other respects€– representing, so to speak, its obverse. In particular, both Nanosyntax and Distributed Morphology assume that syntax is projected from abstract categories which are lexicalized only at the PF interface. For Distributed Morphology, the abstract constituents of syntax are bundles of features, while for Nanosyntax they are structured trees; it is not terminal nodes that receive a lexical realization, but entire strings of terminals. Correspondingly, where Distributed Morphology has the Subset Principle to determine lexical insertion, Nanosyntax has a Superset Principle, according to
300â•… Denotation and case in Romance:€history & variation which ‘a phonological exponent is inserted in a node if its lexical entry has a sub-constituent that is identical to the node’. Where two or more lexical items satisfy either the Subset or the Superset Principle, the one that is more similar to the abstract terminal wins the competition by an Elsewhere Condition. Under the Subset Principle, it is the item with more specifications that wins; in the Superset model, it is the less rich item. The underlying similarities between Distributed Morphology and Nanosyntax (as well as their differences) emerge fairly clearly from the comparison between Calabrese’s (1998, 2008) and Caha’s (2009) theories of case. Calabrese assumes a markedness hierarchy of case constraints NOM< ACC< GEN< DAT< ABL such that the realization of any case along the hierarchy implies that of all cases to its left. Caha (2009) transfers the same hierarchical order into a functional tree [Instr [Dat [Gen [Acc [Nom N]. He obtains two results from this. First, he argues that the head Noun moving leftwards along the hierarchy determines the cut-off point between what will appear as a case suffix (i.e. the nodes below N) and what will be realized by prepositions (i.e. the nodes above N). He then argues that the functional sequence of cases, together with the Elsewhere Condition, yield a Law of Adjacency to the effect that ‘only adjacent cases show non-accidental syncretism’. For instance, dative and genitive, or genitive and accusative, can be syncretic, but not dative and accusative, skipping over the genitive. By definition, nonaccidental syncretism must involve ‘various different exponents’ and ‘show up across paradigms’. Recall that lexical items cover terminal strings, and that among various competing items the smallest is chosen. A syncretism between, say, accusative and dative skipping the genitive would mean that the same lexical item lexicalizes both the Dat and the Acc node. But if there is a smaller lexical item for Gen, then this would always win the competition to realize the Acc node. In other words, the Nanosyntax model, like ours, cuts away the intermediate morphological level, so that obscuring rules are abandoned and only syntactic structures and lexical entries have theoretical status. At the same time, Caha maintains some of the postulates of Distributed Morphology that we have argued against here€– in particular, a Late Insertion model in which abstract syntactic structures are realized by lexical entries rather than being projected from them. Since, in the present model, projection of syntax from lexical terminals implies a non-trivial redefinition of syntactic categorization and structure, its empirical predictions quickly diverge from those of Nanosyntax no less than from those of Distributed Morphology. More specifically, our picture of syncretic phenomena does not distinguish between absolute/non-accidental
Other Romance case systems and alternative accountsâ•… 301 and other syncretisms, implying the abandonment of extrinsic criteria such as those that define non-accidental syncretisms for Caha. Indeed, syncretisms like that of oblique singular with non-oblique plural (Albanian –t, Latin –s, Latin –i, Romanian –I, etc.) do not fall under Caha’s Law of Adjacency. Thus, Latin –i as class I/II nominative plural and genitive singular is a syncretism which skips the accusative, contravening Caha’s hierarchy. This pattern therefore constitutes a problem for the Nanosyntax model. Nor does the definition of non-accidental syncretism help in excluding the undesired pattern. The discussion in Johnston (1997) is especially useful in this respect, in so far as it provides a completely independent observation of the facts. Concerning the ‘homonymy’ of genitive singular and nominative plural in Latin, he notes that ‘it occurs in the first, second and fourth declension’, and that it ‘involves at least two pairs of two homonymous affixes … if a less abstract analysis is adopted there are three different pairs of affixes involved:€–ae, –i and –u:s for I, II and IV class respectively’. In other words, the criteria for ‘non-accidental’ syncretism are satisfied. More generally, we consider it unjustified to suspend all of the criteria that we normally employ to establish significant generalizations (notably the relatively regular cross-linguistic appearance of certain subpatterns of lexical identity) in favour of an extrinsic criterion such as the definition of non-accidental syncretism€– which is ad hoc in so far as it applies uniquely to the domain of data that are targeted by the explanation. Furthermore, Caha (2009) does not tell us what happens with accidental syncretisms€– for which some conventional morphological component is presumably presupposed. How does this system work, since there is no longer underspecification and a Subset Principle? Is homophony going to take the lion’s share of the work to be done? In short, syncretism brings into sharp relief the different predictions of the various models reviewed. For Distributed Morphology, a syncretic case entry is not specified for any of the syncretic categories, leading to underspecification. For us, it must be positively specified for some category able to project all of the syncretic environments. For Nanosyntax, the syncretic case entry must be specified for all cases that enter into the syncretism. From the perspective of the present work, the content of lexical entries that results from Nanosyntax is just as undesirable as that resulting from Distributed Morphology. When it comes to historical variation, i.e. the process of change from one grammar to another and from one lexicon to another, a model, like the present one, in which the lexicon is the grammar (up to UG), as in the minimalist programme of Chomsky (1995), simplifies the terms of the problem€– and gives us a better handle on it. Indeed, we are able to capture change in the so-called case
302â•… Denotation and case in Romance:€history & variation system in terms of a straightforward (dis)continuity of certain lexical entries. By contrast, in Calabrese’s (1998, 2008) account, the point that remains fundamentally unclear to us is how the underlying change in rules and constraints relates to change (or not) in the actual lexicon. As far as we can see, the same issue arises for Nanosyntax, since its separate syntax and lexicon cannot but change separately. Similarly, if we are correct, the connection between determiner and case systems depends on the denotational categorization of case entries; as such, it seems to be beyond the reach of the superset case entries of Caha (2009) no less than the default ones of Distributed Morphology. 8.4
Loss of case in Romance:€Romansh –s
In most modern Romance languages, the nominal inflection system only reflects nominal class and number€– as already analysed for Italian in chapter 7. Consider, for instance, modern French, in which the€ –s inflection, whose development we have followed all the way from Latin, only functions as the plural ending. In present terms, the quantificational element –s is limited to taking scope over the noun, and is therefore interpreted as a plural. All wider scope interpretations are absent€– which in traditional terms means that it is no longer a case ending. Thus, the restructuring from Old French to Modern French is entirely straightforward. Systems like Old French are not attested among the modern Romance languages; nevertheless, a more restricted distribution of what appears to be a nominative –s ending for the masculine singular survives in Sursilvan Romansh varieties (Schmid 1951/52; Haiman 1988). As illustrated here with the variety from Vella (Lumnezia Valley, Grisons), this nominative inflection characterizes masculine singular adjectives, as in (18), and participles, as in (19), in predicative contexts. Note that in (18e), the predication takes as its subject the accusative el ‘him’, so that even in traditional terms the characterization of –s as nominative does not appear to be adequate. As generally happens in Romance, possessives pattern with adjectives, as in (18c) (cf. Manzini and Savoia (2005) on Italian varieties). (18)â•… Vella a.â•… el/ kwai om ai kwrt-s/ grnd-s/ bu-s he/ that man is short/ tall/ good ‘He/ that man is short/tall/good’ b. kwai rakwnt ai ver-s that story is true ‘That story is true’
Loss of case in Romance:€Romansh -s╅ 303 c. d. e.
kwai kodi∫ ai me:-s that book is mine ‘That book is mine’ el ai meer-s ke ju heâ•… isâ•… better than me ‘He is better than me’ juâ•… vai viu el kuntent-s I have seen him happy ‘I have seen him happy’
(19)â•… a.â•… b.
i afn ai iu-s the boy is come ‘The boy has come’ ∫tru∫ arivau-s va el a durmi once arrived goes he to sleep ‘Once arrived, he goes to sleep’
The data in (20) illustrate adjectives and participles occurring in attributive contexts, inside the noun phrase, where the –s inflection is excluded. As we can see in (21), the€–s inflection is also excluded in nominalized adjective contexts, even if they are predicates in a copular structure.4 (20)â•… Vella a. in om kwrt a man short ‘a short man’ b. in bien om a good man ‘a good man’ c. miu kodi∫ my book ‘my book’ d. il tSp lavau the jacket washed ‘the washed jacket’ (21)â•… a.â•… b.
kwel ai il grnd that one is the big ‘He is the big one’ kwai ai il miu this one is the mine ‘It is the one of mine’
Otherwise the –s inflection realizes the plural of adjectives in the masculine, as in (22a–a’), and in the feminine, as in (22b–b’), independently of the predicative or attributive context€– as well as the plural of nouns and determiners.
304â•… Denotation and case in Romance:€history & variation Participles have –s in the feminine plural and with a subset of masculine bases, as in (23a’–b)€– roughly, strong ones; weak participles have an –i plural, as in (23a). (22)â•… Vella a.â•… kwe-s omn-sâ•… kwrt-sâ•… those men short ‘those short men’ a’. kwe-s omn-s ain kwrt-s those men are short ‘Those men are short’ b. kw:la-s dona-s kw«rta-s those women short ‘those short women’ b’. la-sâ•… ain grsa-s they.f are big ‘They are big’ (23)â•… a.â•… a’. b.
il-s t∫p-s ain lava-i the jackets are washed ‘The jackets are washed’ il-s e∫-s ain aviart-s the doors are open ‘The doors are open’ la-s kamia-s ain lavada-s the shirts are washed ‘The shirts are washed’
In the framework that we have been defining here, we can construe Sursilvan€–s as a quantificational element attaching to adjectival bases, as in (24a), and to nominal bases, as in (24b)€– including feminine bases, i.e. those which include the nominal class morphology –a, as in (24a’–b’). In all contexts, its scope can remain word-bound, resulting in the plural reading. This does not exclude the possibility that specialized subclasses may form the plural by different means, as with weak perfect participles, which display€–i. (24)â•…
a.â•… b. a’. b’.
[√ kwrt] [Q s] [√ tp] [Q s] [[√ grs] [N a]] [Q s] [[√ don] [N a]] [Q s]
With masculine bases like (24a), i.e. bases without –a nominal class morphology, –s need not have the plural reading, but can attach to a singular. We can explain this distribution as an effect of –s taking sentential scope, so that it contributes a Q closure at the sentential level, rather than plurality. This non-plural –s occurs in predicative contexts, including copular sentences, as in (18a–d),
Loss of case in Romance:€Romansh -sâ•… 305 and small clauses, as in (18e). We conclude that in predicative contexts, adjectives (participles, etc.) require the lexicalization of a Q closure, which is provided by –s. Embedding in a noun phrase puts predicative bases in the scope of the D (definite) or Q (quantificational) closure provided by the determiners and quantifiers of the noun phrase. In predicative contexts, however, such a closure is not provided at the phrasal level; therefore the –s supplies it at the inflectional level. This also accounts for the contexts in (21) where the adjective is in the scope of a D or Q element, although it itself heads the noun phrase.5 Further empirical evidence shows that the notion of predication may be necessary, but is not sufficient to account for the distribution of€–s, which interacts with complex notions of referential content and quantification. To begin with, even in copular sentences, the –s inflection is not found on adjectival predicates in contexts like (25a). The appropriate distinction between (25a) and (18a) cannot be that between neuter and masculine, since there is no evidence for the two genders besides the need to somehow describe the contrast at hand. We propose that the relevant distinction is between the individual reference of the subject of predication in (18a) and the propositional reference of the subject in (25a). We can exclude the possibility that animacy plays a role on the basis of examples like (18b) above. We also note that the –s inflection, while possible, is not necessary with quantificational subjects, including wh–phrases, as in (25b), negative phrases, as in (25c), and existentials, as in (25d). (25)â•… Vella a. kwai ai ver that is true ‘That is true’ b. t∫i ai iu(-s) who is come ‘Who has come?’ c. nidzin ai iu(-s) nobody is come ‘Nobody has come’ d. tsit∫i ai iu(-s) someone is come ‘Someone has come’
As may be expected on the basis of (25a), the€–s inflection is absent from expletive contexts with a sentential subject, as in (26a). Note that we can exclude the possibility that certain types of adjective simply do not take –s on the basis of a comparison with (18d). More interestingly, the –s inflection is also missing in expletive sentences with a nominal correlate, as in (26b); these are characterized by the well-known pattern of French or Northern Italian
306â•… Denotation and case in Romance:€history & variation dialects whereby the verb does not agree with the postverbal subject, but shows up in the 3rd singular masculine. (26) Vella a. i() aiâ•… meer da klama tai it is better to call you ‘It is better to call you’ b. i() ai iu in afn/ afns it is come a boy/ boys ‘There came a boy/ boys’ b’.â•… ils afns ain ii the boys are come ‘The boys have come’
The evidence in (25)–(26) fits in fairly easily with our conclusions. Thus, when we say that –s provides a quantificational closure for the adjective, we should say more precisely that it provides such a closure for the argument slot associated with the adjectival base. It is natural to propose that the latter is not only unnecessary, but also impossible when this argument corresponds not to an individual variable but to a propositional variable€– thus explaining the contrast between (18) and (25a). Similarly, the optionality of the –s morphology in (25b–d) corresponds to whether the negative, wh– or existential quantifier receives a specific reading or an indefinite reading. Under the latter reading, it is incompatible with the –s closure, assuming that –s has definiteness/specificity properties. Turning now to expletive contexts, we assume that in non-agreeing expletive sentences, the expletive introduces a variable whose value is fixed by the correlate via predicational identification (as in Manzini and Savoia 2005, 2007, 2008a).6 This structure corresponds to the focus reading that characterizes postverbal-subject sentences in Romance. Under these assumptions, the lack of€–s inflection in (26) is expected on the same grounds as it is in (25). Quite simply (independently of the definiteness properties of the correlate), the presence of a variable as the EPP argument of the sentence is incompatible with the definiteness/specificity closure provided by the –s inflection. It is interesting to consider how Romansh –s could be accounted for in a more conventional framework€– such as the one discussed by Calabrese (1998, 2008). The continuity with systems such as Old French could be captured by assuming that –s represents a direct case inflection; but still its exclusion from noun phrase embeddings would have to be stipulated, as would the subtler contrasts in (25)–(26). Furthermore, it is hard to see how even the connection between case –s and plural –s could be captured at all. The strong points of the
Loss of case in Romance:€Romansh -sâ•… 307 present account are that it makes possible a unified lexical entry for –s and at the same time it correctly captures its complementary distribution with syntactic determiners/quantifiers. It does not achieve this through additional stipulations€– but through the very same means by which it accounts for Latin and Old French –s, conventionally two pure case inflections. A prediction of the analysis of Romansh€–s proposed here is that we could see an analogous distribution with historically unrelated morphology, as long as this had Q/D properties. This is what happens in Romance varieties of the Montefeltro (Marche, Italy), in which adjectives (and participles) in predicative contexts have a feminine plural inflection€ –le including the Romance definiteness morphology –l, and in fact having the same form as the definite determiner, as illustrated in (27a), as well as the clitic. Crucially, the€–le adjectival inflection occurs in predicative contexts like (27b), but not in attributive DP-internal contexts like (27a’), recalling the distribution investigated for –s in Vella. In predicative contexts, it also appears on the participle, as in (27c). (27)â•… Urbino a. le dn the.fpl women ‘the women’ a’. k-le brv dn those-fpl good women ‘those good women’ b. en brv-le they.are good-fpl ‘They are good’ c. n dvntt-le alt-le they.are become-fpl tall-fpl ‘They have become tall’
We can apply the same analysis adopted for Vella to the Urbino data. When adjectival bases occur inside noun phrases, they find themselves within the scope of the quantificational/definite closures of the DP. However, in predicative contexts, where such a closure is lacking at the phrasal level, it is supplied at the inflectional level, at least in the (feminine) plural, by the€–le element, to which we assign the structure in (28). (28)â•… [√ brv [Q le]]
The Urbino data are particularly interesting in that they show that the Vella pattern can be found independently of any continuity with Latin and medieval Romance case; therefore it need not involve a continuation of Latin case declensions such as€–s, but it may involve a new Romance formation like le.
308â•… Denotation and case in Romance:€history & variation The account that we have provided for –s from Latin through Old French to Romansh explains why a definiteness l form should be involved in the Urbino data. It seems to us that all of this remains coincidental in more conventional approaches. 8.5
Pronouns€– and some conclusions
In the preceding discussion we have considered nominal inflection systems to the exclusion of pronouns. As is well known, most modern Romance languages do in fact maintain a case system for pronouns. Italian is a good example of this, though 3rd person full pronouns vary according to gender and number but not case, as illustrated in (29). Note that the –ui,€–ei,€–oro inflection attached to the l- base is the same as that found in the Romanian oblique. (29)â•… a.â•… a’. b. c.
Lui/lei viene he/she comes ‘S/he is coming’ Loro vengono they come ‘They are coming’ Vedo te non lui/lei/loro I.see you not him/her/them ‘I see you, but not him/ her/ them’ Lo do a lui/lei/ loro it I.give to him/her/them ‘I give it to him/ her/ them’
On the other hand, a two-case system characterizes full (strong) pronouns in the 1st and 2nd person singular, distinguishing nominative (30a) from objective (accusative and oblique) in (30b). (30) a. b.
io, tu ‘I’ ‘you’ me, te ‘me’ ‘you’
Interestingly, in Romanian, the two-case system analysed in section 8.2 holds for nouns and for 3rd person pronouns, where nominative/accusative el/ea/ei/ele forms alternate with dative lui/lei/lor (the genitive corresponds to a possessive form, with adjectival agreement, of the type familiar from other Romance languages). However, 1st and 2nd person (singular) pronouns display a three-case system, distinguishing nominative, as in (31a), from accusative, as in (31b), and dative, as in (31c). In Albanian, the 1st and 2nd person
Pronouns€– and some conclusionsâ•… 309 system has three cases like the nominal system€– but differently distributed; in traditional terms, there is a nominative form, a specialized ablative form, and a single form for accusative and dative (cf. Manzini and Savoia (to appear) on the Gheg variety of Shkodër). (31) a. b. c.
eu dorm/ tu dormi I sleep/ you sleep ‘I/you sleep’ el m-/t- a văzutâ•… pe mine/ tine heâ•… me/you has seen me/ you ‘He saw me/you’ mi-/ţi- l au dat mie/â•… ţie me/youâ•… it they.have given to.me/ to.you ‘They gave it to me/you’
In short, data such as (30)–(31) illustrate a split between 1st/2nd singular bases and other (pro)nouns. We interpret this as indicating that the discourse-linked content of 1st/2nd singular requires the lexicalization of properties related to the different argument contexts (subject vs. dative vs. arguments of prepositions)€ – which are not required by other referential (event-linked) bases (3rd person and 1st/2nd plural, involving reference to ‘others’ besides the ‘speaker’ and ‘hearer’). Therefore the particular shape that the split takes has richer case specifications associated with the lexical bases characterized by the maximally specific denotation (‘hearer’ and ‘speaker’). Some of the richest case systems in Romance languages are to be found in clitic systems, specifically in the 3rd person involving the l- definiteness base, as well as its allomorphs -, -, -, etc. In (32) we illustrate a Tuscan system studied in more detail by Manzini and Savoia (2005, 2007, 2008a), in which the nominative series in (32a) differs from the accusative series in (32b), though there is a syncretism between the nominative masculine (singular and plural) and the accusative masculine plural. The dative in (32c) is in turn syncretic with these two forms, though its reference extends to masculine and feminine and to singular and plural, and is distinct in this sense from both nominative and accusative.7 (32)â•… Vagli Sopra (Tuscany) a.â•… iɟ/ e a/ ann@ durmite he-they.m/ she-they.f has/ have slept ‘S/he has / They have slept’ b. l/ la/ ɟi/ l@ v@jaŋ him/ her/ them.m/ them.f we.see ‘We see him/her/them’
310â•… Denotation and case in Romance:€history & variation c. da ɟɟi le/ ɟi give to.him/her/them it.m/ them.m ‘Give it/ them to him/her/them’
A general point of interest about Vagli is that it has a three-case distinction for l- bases€ – i.e. essentially the richest possible system in a Romance language, and a system as rich as that of Vena, which was discussed in chapter 7. Like standard Italian (to which Tuscan varieties are closely related), Vagli of course has no case declension on nouns. This distribution, and those concerning, in particular, 1st and 2nd person full pronouns above, lead us to wonder what constitutes an ‘absolute’ syncretism in Calabrese’s (1998, 2008) terms€– or what it means to activate a case constraint. Calabrese seems to imply that an entire grammar is involved; but it must really be lexical classes which are involved€– a fact which his theory is not best suited to capture. Another point of interest for the present discussion concerns the syncretism observed with (i) in (32) between the nominative/accusative plural masculine, the nominative singular masculine and the dative. This is essentially the same observed with –s of class III in Latin:€nominative singular, nominative/accusative plural and oblique (genitive singular, dative/ablative plural). At the same time, the (i) form in Vagli continues Latin ill–i based on the –i morphology, which in Latin did not characterize the nominative singular or the accusative plural. Now, the Latin –i morphology, which displays the oblique singular/nonoblique plural syncretism, has been associated in section 8.1 with a Q denotation (cf. also the discussion of Romanian in section 8.2); it is this denotation that is evidently continued by the clitic system in Vagli, at least for l- bases. However, the range of contexts that it satisfies is slightly extended, covering not only oblique singular (dative) and plural, but also nominative singular. When examining the outcome of Latin –s in medieval Romance (section 8.3), we proposed that, while its Q denotation was preserved, the range of contexts in which it appeared became restricted to the nominative singular and the plural. The extension of the contexts relevant for the lexicalization of –i in Vagli with respect to Latin –i represents the reverse phenomenon, as expected. In this connection, we should also return to the Old French example in (16), repeated below in (33). While in section 8.3 we discussed the nominal declension, the definite determiner also displays case inflections which are slightly at variance with the noun. Thus, in the masculine, the nominative singular and plural is li, while in the accusative, le and les alternate in the singular and plural respectively (as in Modern French). Evidently, the determiner system of Old French has an important point of contact with Vagli (and other modern
Pronouns€– and some conclusionsâ•… 311 Italian varieties) in the adoption of –i as nominative morphology. In the present conception, there is on the one hand continuity between the quantificational properties of –i in Latin and in Romance, and on the other hand a change in the context(s) that it satisfies. (33)â•… a.â•… b.
Dur sunt li Hard are the Guardez Watch over
colp e li caples est grefs blows and the scuffle is hard le champ … e le-s munz (munt-s) the field … and the mountains
The discussion of the syncretism between dative –i and masculine plural –i, as also seen in li ‘them’ and gli ‘to him’ in standard Italian, was one of the starting points for the common categorization (‘Q’) of obliques and plurals that we proposed in Manzini and Savoia (2005, 2007, 2008a). The conclusion reached in those works by inspecting the Romance clitic systems are now confirmed by the evidence from the case systems of Albanian, Latin and Romance itself€– as well as by the historical development we are able to trace from Latin to Romance in terms of straightforward lexical (dis)continuity. Indeed, a key property in terms of which our approach differs from the others reviewed is its uncompromisingly lexicalist stance, whereby structures are projected from actual lexical terminals. It is not abstract structures independent of the lexicon that determine the basic categorizations of natural languages and therefore design the basic ontology underlying them. It is the lexicon that does that.
Notes
1
The structure and interpretation of (Romance) complementizers
1. The first published version of this analysis is found in Manzini and Savoia (2003), where ‘we propose that the complement sentence in (1a), the question in (1e) and the wh-phrase in (1g) are all preceded by the same lexical item in that they have a similar LF with che introducing a variable. The different interpretations stem from the fact that in (1e) che binds the internal argument of a verbal predicate, as illustrated in (10a); in (1a), che binds a variable with sentential content, as in (10b); and in (1g) che binds the internal argument of a nominal predicate, as in (10c). (10)â•… a.â•… [che xâ•… [fai (x)]] b. [che x [x:€vieni domani]] c. [che x [camicia (x)]]’
(p. 95).
2. The obvious comparison is with N-words in Romance, discussed in chapters 3–4. These are not intrinsically negative, but rather introduce a variable which can receive a negative interpretation in the scope of a negative operator, though it can equally well be licensed in non-negative contexts. 3. It is important not to confuse I in (10) with some specialized position for the complementizer, since the ordinary noun in (6) is also in I. Note also that we assume that no empty heads are present in the structure; in (10) they are represented essentially as diacritics to help the reader process the tree, as they are in the bracketed representations in (5)–(7). One may wonder what the C head corresponds to under the new conceptualization of the complementizer in (10). Recall that one of the reasons for our original puzzlement at the currently standard analysis of complementizers was the fact that such clearly nominal elements would occur in positions otherwise taken by verbal (modal) heads. The present view, while reforming the conception of complementizers, does not affect the current analysis of verb placement. Thus, interrogative inversion is still construed as the occurrence of the verb in a higher position than I€– i.e. C etc. In this sense, if there are several positions of the verb higher than I, there will be a recursive C field, as in Rizzi (1997). These points are developed in section 1.3. 4. This is the C position that Rizzi (1997) characterizes as Force, i.e. illocutionary force. McCloskey (2004), who supports this characterization, also clearly notes that in semantics ‘there is a well-established and widely-held view that there can be no 312
Notes to pages 21–32â•… 313 such thing as an embedded illocutionary force indicator’. We share these misgivings, and therefore avoid referring to illocutionary force in the present account. ╇ 5. This is the C position that Rizzi (1997) characterizes as Finiteness. ╇ 6. For the reasons mentioned in fn. 4, we resist characterizing the properties instantiated by embedded V2 as illocutionary force. It is worth recalling that the incompatibility of embedded V2 with complementizers in, say, German was what provided the initial support for the identification of the complementizer with a verbal position€– namely the V2 position (den Besten 1984). In this respect two observations are in order. First, the argument only has any force in theories without multiple CPs, hence in practice none of the theories addressed by the present discussion. Second, when it comes to stating the relevant generalization, the simplest thing to say within the present theory is that in complementizer deletion no complementizer (here nominal) layer is present at all; a discussion of complementizer deletion in Romance is provided in chapter 2. This suggests that languages like German allow for the selection of the properties triggering V2 only directly on an embedded sentence€– requiring the ‘deletion’ (i.e. absence) of the complementizer. ╇ 7. We assume that the embedded subject doubled by the subject clitic is a topic. There is an ongoing discussion on this point, on which Rizzi (1986) holds a contrary position to ours:€ namely, that the clitic-doubled subject is in IP. This is entirely independent of the issue at hand; for instance, Poletto (2000) also takes the lexical subject of clitic subject languages to be a topic. ╇ 8. This doubling is independent of the phenomena being discussed here, being found in many other Northern Italian varieties:€ for instance, Viguzzolo, as exemplified immediately below in (24), or Castellazzo as in (25). ╇ 9. In the interest of not overloading the presentation with accessory information, we omitted the structure of the object clitic cluster in (29). 10. As for her Turinese and Ligurian varieties, Paoli (2007) proposes that the higher che lexicalizes Force in the schema of Rizzi (1997), while the lower che lexicalizes Fin. The connection with subjunctive is established by the fact that che is actually merged in the head of a Mood projection from where it is moved to Fin. In present terms, this would translate into a grammar in which a complementizer head selected by another one is necessarily an irrealis one. We believe, however, that the discontinuity not only with respect to our own Piedmontese data but also to the Medieval Romance data of Paoli herself warrants further empirical research. 11. It does not escape us that making such a claim amounts to entering the complex debate concerning s-selection and c-selection (Grimshaw 1979), taking a position in favour of s-selection to the exclusion of c-selection. This position has been Â�argued for independently in the literature, in particular by Pesetsky (1991), who only supplements s-selection by what he calls l(exical)-selection€– a notion also available under present assumptions. 12. Suppose that some notion of phase holds; then, is C (now an exclusively verbal head) the phase head or is that role taken over by the Nominal complementizer head? Suppose we take the most reductive construal of phases, as the locality domains for movement. A structuring of the sentence along the lines of (10) yields
314â•… Notes to pages 32–42 a straightforward reason why the bounding nodes for Subjacency (Chomsky 1973) were originally taken to be the IP (S) and the CP (S’):€namely, because the CP is in reality an independent noun phrase. This suggests that (at least) the so-called complementizer is a phase head in current terminology€– assuming that phases are indeed relevant for locality. In a separate development of the theory, phases have become responsible for the distribution of features in phasal domains (Chomsky 2008). As for C–T agreement generally, expressing it in terms of phases and operations on features is certainly possible€– but so is expressing it in terms of properties selected by a head on another head; we are not aware of any piece of evidence that would make this choice anything but a theory-internal matter. Complementizer agreement in Germanic languages cannot be construed as such a piece of evidence. It is certainly compatible with agreement analysis, but it can also be analysed as depending on an enclitic subject appearing on the complementizer (Shlonsky 1994). Under this latter analysis there is no C–T agreement, but simply an agreement between different copies of the EPP argument in different structural domains. Multiple copies of subject clitics are independently attested in Romance, where the borders between what are traditionally called agreement inflections and clitics are also quite labile. 13. Matrix lexicalizations of the che-type complementizer are possible when its left Â�periphery is lexicalized not only by a wh-phrase, but also by a focus, both in Northern varieties, as in (i), and in Southern ones, as in (ii). The same possibility is present for instance in Brazilian Portuguese, as noted by Mioto (2003), quoted by Belletti (2008). (i) (ii)â•…
Casaccia (Grisons) naZy k al ve nobody that he comes ‘Nobody comes’ Morano (Calabria) smbri ki ti wardnu always that you they.look.at ‘They always look at you’
14. Subsequent work on the left periphery further refines the general schema provided by Rizzi (1997). Thus, Benincà and Poletto (2004) and Benincà (2006) recognize several articulations both in the Focus field (Contrastive€– Informational) and in the Topic field (Hanging Topic€– Scene Setting€– Left Dislocation€– List Interpretation), with the Topic field ordered above the Focus field in accordance with the conclusions of Benincà (2001). In turn, Frascarelli and Hinterhoelzl (2007) articulate the Topic/Focus field into Shifting Topic€– Contrastive Topic€– Focus€– Familiar Topic. These further proposals do not change the substance of the more limited discussion in the text concerning Rizzi’s (2001) articulation of the wh-field into an Int€– (Topic)€– Focus hierarchy. Assuming the empirical correctness of (a subset of) their partitions of interpretive space and of their correlation to syntactic orders, the general question remains the one in the text, namely whether this correlation is encoded by means of functional categories and hierarchies. The alternative is, as in the text, that syntactic order restricts the range of available semantic Â�interpretations.
Notes to pages 42–46â•… 315 Thus, as discussed in the text, a Focus (a wh–phrase) taking a Topic (a lexical subject) in its scope can only have a certain range of interpretations; this reasoning can in principle be replicated for sequences of Topics and so on. More recent work has addressed the left periphery of the dialects of Italy. Cruschina (2010) for Sicilian, Mensching and Remberger (2010) for Sardinian and Paoli (2010) for Triestino all agree on the possibility of a left-peripheral Informational Focus as an answer to questions. Cruschina characterizes this as an ‘Emphatic’ IF, while Paoli suggests that ‘it clearly expresses new information, but simultaneously conveys an implicit sense of contrast’. All of these authors contrast the situation they describe for their respective varieties with that of standard Italian, in which, following Belletti (2004), they assume that answering a question with a left-peripheral Focus is ungrammatical. On the other hand, left-peripheral answers to questions are perfectly possible in our own competence of (standard) Italian, as in (i), from Cruschina (2010:€249), possibly with the pragmatics that the authors quoted associate with their varieties. (i) a. b.
Che cosa What Le chiavi, The keys,â•…
hai dimenticato? have.you forgotten ho dimenticato I.have forgotten
Similar judgements are found in Brunetti (2004), who also stresses that the truly ‘unmarked’ answer in (standard) Italian, as in English, is the elliptical one. 15. In Manzini and Savoia (2005) we propose that the different interpretations of whphrases determined by their ordering with respect to lexical subjects may be codified in terms of D-linking, or even more generally definiteness/specificity. Correspondingly, we introduce a specialized domain C (Definiteness) that supports both specific wh-phrases and topics, allowing them to intermingle. This would be consistent with the conclusion of Benincà (2006) and Benincà and Poletto (2004) that the topic field is superordinate to the focus field. Note, however, that verb positioning provides independent evidence for at most two head positions in the C field, namely the interrogative position considered here and a higher position for imperatives/infinitives, which must be lower than any wh-position. What is more, the behaviour of ‘why’ does not easily follow from Manzini and Savoia’s (2005) proposal. Thus, in that paper we introduce an altogether different analysis for the ‘why’ facts€– as sketched here. For these reasons we consider it more likely that it is this latter line of analysis that will ultimately provide a unified account of the facts. 16. Languages like Trun in (50) have a generalized V2 property. However, the possibility of the subject position in (50) does not correlate with it. Several Romansh and Ladin varieties with generalized V2 do not allow for the lexical subject to appear in the inverted subject position where subject clitics are found, as illustrated in (i) with a question and in (ii) with examples of generalized V2. (i)â•… Zernez (Grisons) ilts ufants an -i dOrmi the children have they slept ‘Have the children slept?’
316â•… Notes to pages 46–59 (ii)â•… La Pli de Mareo (Alto Adige/ South-Tyrol) a.â•… le liber a -l be lit ZaN the book has he well readâ•… John ‘John has indeed read the book’ b. ZaN le liber a -l lit John the book has he read ‘John has read the book’
2
Variation in Romance k-complementizer systems
1. In order to avoid tedious references to Manzini and Savoia (2005), let us state once and for all here that all the empirical generalizations drawn in the text are to be understood as holding for the set of data published by Manzini and Savoia (2005, I:€404–79). 2. In the judgement of one of the authors (Savoia), who is a native speaker of the Florence variety, not only is it possible to delete the che complementizer in (7a), as shown in (i) and as also indicated by Cocchi and Poletto (2002)€– it is also possible to delete relative che (as it is in Medieval Italian texts) as in (ii), and even se ‘if’, as in (iii). It remains impossible to delete wh- che, as in (iv). (i) (ii) (iii) (iv)
Hannoâ•… dettoâ•… viene domani They. have said he.comes tomorrow ‘They said that he will come tomorrow’ Ho contato tutti quelli sono venuti I.have counted all those are come ‘I counted those that came’ Mi chiedevano viene o no me they.asked he.comes or not ‘They asked me whether or not he was coming’ Non sanno *(che) fare not they.know what to.do ‘They don’t know what to do’
3. Làconi will be discussed in section 2.2, because of its special lexicalization of ‘if’; as will be seen, it is comparable to Paulilàtino in all respects considered here. 4. This is important in view of the recent discussion of Eastern Abruzzese by D’Alessandro and Ledgeway (2010), where it is proposed that ‘the Fin° head may be spelled out at PF as ca (realis specification) or chi (irrealis specification)’€ – while Force, ‘independently of the underlying modal specifications of Fin° is invariably spelt out as ca’. We comment on the reversal of the positions of ka and k/ ki in connection with the Cosenza variety in section 2.1.2 (in the text preceding fn. 6). D’Alessandro and Ledgeway (2010) also propose an empirical generalization whereby chi is restricted to contexts not ‘directly selected by a matrix predicate’. Again we do not find any reflex of this in our data; furthermore, D’Alessandro and Ledgeway do not consider potentially relevant contexts such as those selected by ‘before’ and ‘after’, to be discussed in section 2.1.2.
Notes to pages 62–64â•… 317 5. The analysis of temporal clauses as involving a wh-like temporal operator goes back at least to Larson (1987), who in turn quotes Geis (1970). Apart from the fact that temporal subordinators can overtly correspond to wh-phrases (e.g. when in English, quando ‘when’ in Italian etc.), the classical argument in favour of this conclusion is that temporal sentences can display ambiguities corresponding to what the basegenerated position (or the reconstruction position) of the temporal operator is taken to be. Thus John left after Mary told him that he should leave can be interpreted as John leaving after he should have left (according to Mary) or as John leaving after Mary told him something (namely that he should leave). Recent work by Haegeman (2007) takes up this evidence again, arguing that operator movement accounts for an altogether different set of data, namely for the impossibility of embedding left-peripheral material under temporal adjuncts. However, as noted by Haegeman, Romance clitic left dislocation is possible under temporal adjuncts. In turn, embedding of foci in Italian is taken to be ungrammatical by Haegeman (2007:€316), but it can be quite acceptable in our judgement, as in (i); note the lack of resumptive clitic, which excludes clitic left dislocation. (i)â•… Quando un dieci prenderà, sarò contento whenâ•… a tenâ•… he.will.get,â•… I.will.be happy ‘I will be happy when he gets an A’ We return to this matter in connection with conditionals, where it is more directly relevant for our discussion, cf. fn. 12. 6. The varieties close to Cosenza exemplified in Manzini and Savoia (2005), namely S. Fili and Montalto Uffugo, display a single ka complementizer, evidently belonging to less conservative varieties in Ledgeway’s description. 7. As part of the differences between CHE in Early Southern Italian texts and chi in the Cosenza variety, Ledgeway (2009) also notes a reversal of the data concerning the embedding of topics and foci. These are acceptable under ca but not under chi. We have already commented in section 2.1.1 (fn. 4 and text preceding it) on the analogous reversal proposed for Eastern Abruzzese in D’Alessandro and Ledgeway (2010). For Ledgeway (2009), the the Cosenza variety distribution involves yet another parameter€– namely that with chi the verb would move very high in the sentence (depriving even adverbs of preverbal space), while it would remain lower with ca. We refer the reader to chapter 3 for the reasons why we reject this line of argumentation, inaugurated by Cinque (1999). Of particular relevance is the literature (Svenonius 2002; Ernst 2002; Nilsen 2003) showing that adverb hierarchies cannot be used to argue for verbal positions without leading to contradiction. 8. Damonte (2006) provides similar data for Baunei, in which chi introduces finite sentences introduced independently of modality. 9. A few varieties of the Montenerodomo/Arena type tabulated in (A) display free alternation between their two complementizers as relative clause introducers€– namely Guardiaregia and Gizzeria. We interpret free alternation as the alternation between two different grammars. Therefore one of the grammars of these varieties is parallel to that of Montenerodomo/Arena in (14)–(15), with relative clauses introduced by the generalized wh-binder ki. The other grammar with ka as the relative clause introducer simply shows that relative clauses can be treated as an ordinary case of declarative embedding.
318â•… Notes to pages 69–77 10. This account of the se complementizer is already found in Manzini and Savoia (2003), essentially in the same terms as here. Thus ‘in the tradition of Lewis (1975) the sentence introduced by se is interpreted as the restriction of a propositional variable, bound in particular by adverbs of quantification … In the absence of an overt quantification, the operator that binds the quantificational variable can be the generic/universal one … Given these interpretive properties of hypotheticals, it is natural to assume that se “if” introduces an indefinite propositional variable. As such we expect its syntax to parallel that of the propositional che’ (p. 105). 11. Again the account of interrogative se goes back to Manzini and Savoia (2003). ‘For embedded yes–no questions we can again assume that se introduces an indefinite propositional variable and this in fact establishes a wh-question ranging over the proposition’ (p. 106). 12. Haegeman (2009) adopts the Larson/Bhatt-Pancheva syntax for conditionals with a view to accounting for restrictions on the left-peripheral material they embed, similar to those discussed in fn. 5 for temporal clauses€ – despite the fact that the classical argument in favour of treating temporal subordinators as wh-operators, namely the possibility of matrix or embedded construals, does not apply to conditionals. The issue is largely orthogonal to those debated here. In any event, our own judgements on conditionals are that left-peripheral material does not create intervention effects€– providing an unsurprising parallel to temporal clauses. (i) Se un sei prendi, sono contento If a six you.get I.am happy ‘I am happy if you get even a D’ (Context:€‘Will you be happy if I get a C?’) 13. Bhatt and Pancheva (2006) note another problem for Kayne’s (1991) proposal that if is excluded from finite sentences because it governs PRO, namely the fact that conditional if can select participial predicates like (i), whose only argument is controlled by the matrix argument. Notice further that there is no interrogative analogue to this pattern, as in (ii).
(i) If elected, he has said that … (ii) *By now he will be wondering if elected Sentences like (i) are problematic for the present account as well, since they appear to include a variable EPP argument controlled by a matrix argument. But maybe they don’t. Note that any nominal predicate can be embedded under if, as in (iii), making it far from obvious that (i) involves a sentential embedding. On the other hand, (ii) could be automatically excluded if interrogatives require a sentential embedding. (iii)â•… If beautiful, a woman has many chances What kind of embedding (i) represents, if not a sentential embedding, remains an open question; but notice that temporal operators equally allow for it, as in (iv). (iv) When still a child, he met his future wife
Notes to pages 82–90â•… 319 3
Sentential negation:€adverbs
1. Our first explicit statement of this conclusion goes back to Manzini and Savoia (2002):€‘In what follows, we shall concentrate on negative adverbs, arguing that they do not belong to a specialized Neg category; nor are they specialized Adv(erbs), but rather nominal elements. We shall briefly indicate how these conclusions may extend to Adv(erbs) in general and argue that clitic negations are in turn nominal in nature’, and later ‘The treatment suggested … for bare-N adverbs categorizes them as straightforward nominal heads, i.e. N. In turn the negative polarity item/negative quantifier status of ‘nothing’-type adverbs suggests that they are to be assigned to the category Q’. Manzini and Savoia (2005) bring further empirical evidence to bear on this analysis. Only recently, Garzonio and Poletto (2009) conclude that ‘the position of the negative marker nen in Piedmontese … is a dedicated position to quantifiers’, and that ‘in the case of m[inimizer] negation a noun becomes a quantifier and then raises to a position dedicated to quantifiers’ so that ‘the positions where the negative markers occur are not to be labelled as NegP’. Despite the fact that the authors take a ‘grammaticalization’ perspective, their results reproduce ours€– a fact of which they are ostensibly unaware. 2. We agree on the ungrammaticality of certain cases, though for instance (i), which was suggested to us, involves an idiomatic expression: (i) Non ci si capisce *(un) accidente Not there one understands a thing ‘One doesn’t understand a thing in there’ 3. We are aware that in Germanic or Slavic languages there is dative vs. accusative marking in 1st and 2nd person pronouns as well (cf. Fici, Manzini and Savoia 2001). A uniform behaviour of all persons, say with respect to Case, is not a problem for the present theory€– to the extent that the same treatment can be uniformly given to all nominal elements. All that is relevant here is that there are some languages in which the person split can be observed€– and those languages pattern in the way noted. 4. Data like those we are considering (and other data, such as auxiliary selection according to person, as discussed in chapter 6) require not so much a characterization of the intrinsic denotational properties of 1st/2nd person vs. 3rd person, but rather a linking of their referential properties to the properties of the various persons in linking to predicate–argument structures. Suppose, for instance, that we conceive of argument slots in predicate–argument structures as variables (Adger and Ramchand 2005). Everything we have said is compatible with the idea that only 3rd person arguments really bind these slots€– and this is no doubt connected to their ability to bind other arguments in anaphoric relations. As for 1st and 2nd person pronouns, we may consider that they satisfy the relevant argument slot through some different formal means€– an identity relation, a predicative relation or some other relation. We stress again that connecting the referential to the argumental properties is necessary to an understanding of how the person split operates in morphosyntax. Therefore we retain, as a first approximation, the ‘discourse-anchored’ and ‘event-anchored’
320â•… Notes to pages 90–135 characterizations, as opposed to other characterizations hinting at purely referential properties (‘deictic’ vs. ‘anaphoric’, etc.). The reflexive paradigms in (15)–(16) are also relevant, to the extent that the 3rd person, anaphoric si clitic patterns with 1st and 2nd person clitics. 5. Bayer (2006) notices construals of nothing in English of the type in She looks nothing like that, concluding that nothing can be either ‘an argumental category which occupies an argument position in the sentence’ or ‘an adjunct’, as in the sentence quoted. From the present point of view, this is just more evidence in favour of the nominal, argumental status of adverbial negations. For Bayer (2006), however, the implications of these data lie again in the domain of ‘grammaticalization’. 6. Here and throughout, generalizations not supported by the examples in the text presuppose the data published in Manzini and Savoia (2005). 7. Table (38) also provides counterexamples to Garzonio and Poletto’s (2009) recent claim that ‘Neg2 markers derive from grammaticalized minimizers … Neg3 markers derive from the bare inanimate quantifier corresponding to English ‘nothing’’. In fact, even leaving aside the separate question of Lombard nɔ negations, it is evident that the higher position of the negation (Zanuttini’s Neg2) characterizes not only bare-N (minimizer) adverbs, but also a bona fide Piedmontese ‘nothing’-type adverb like nitQ in Castellazzo Manzini and Savoia (2005) tabulate it together with nn in Dego and nnt in Garessio, as well as with nia in Colfosco and La Pli de Mareo. By contrast, leaving aside Lombard n again, the lower position of the negation (Zanuttini’s Neg3) characterizes the minimizer bo in Mustér. In Manzini and Savoia (2005), bit in Donat and buk in Trun are tabulated together with it. 4
Sentential negation:€clitics
1. A case in point is Bagolino in (i). (i)â•…
Bagolino (Lombardy) no drmo no te drme no l/l drma no dormo no dormif no i/le drm not ClS sleep ‘I don’t sleep’ etc.
mi mi mi mi mi mi not
2. An interesting question is whether the positional difference between I and C negations corresponds to an interpretive difference. All the evidence points to a different conclusion, namely that the interpretive value of the negation remains the same, though its lexicalization varies according to a number of ultimately interpretive differences (including in particular the shape of the event, as defined by the nature of the arguments satisfying the predicate). 3. The two negation positions postulated here do not share any of the properties of Zanuttini’s (1997), since they relate neither to a split between presence or absence of
Notes to pages 135–152â•… 321 negative adverbs nor to a split between base-generated and derived positions of the negation. 4. In the present approach we do not adopt what Culicover and Jackendoff (2005) call Interface Uniformity, hence two logically equivalent sentences need not have an identical syntax at some (abstract) level of representation. Rather, it is perfectly possible for two different syntaxes to yield the same semantics (and, vice versa, different semantics can correspond to the same syntax). Therefore we maintain that in languages in which the negation clitic follows the 1st and 2nd person clitic as well as 3rd person clitics, it is simply insensitive to the person split€– there is no LF movement scoping the negation out of the inflectional domain, or any other mechanism making its (abstract) syntax parallel to that of the person split cases. 5. Mutual exclusion between subject and other clitics has hardly been dealt with in the formal literature (as opposed to mutual exclusions, say, between datives and accusatives as in the Spurious se of Spanish). A notable exception is Roberts (1993), who considers only a highly specialized set of facts from Franco-Provençal dialects, in which the relevant alternations between subject and object clitics characterize the context before auxiliaries. According to Roberts (1993:€ 330), ‘some clitic always appears when an auxiliary is present. If an O[bject]CL is available this element may precede the auxiliary, or an S[ubject]CL precedes the auxiliary’. The explanation is that there is an Agr (i.e. clitic) position in front of the auxiliary which must be filled; furthermore, ‘clitics cannot adjoin to other clitics’ (Roberts 1993:€332), so that at most one clitic can fill Agr. The theory we propose in the references quoted is not the same as Roberts’ (1993). In the present model, clitics have dedicated (argumental, not Agr) positions in the sentential structure; this means that the failure of subjects and object clitics to combine cannot be due to competition for the same position. Rather, under the explanation pursued in the text, the Definiteness properties of the object clitic (as lexicalized, say, by the l lexical base of the 3rd person) scope over the entire string€– and it is this that prevents their lexicalization by the subject clitic. Furthermore, the same model is extended by Manzini and Savoia (2005, 2007) to mutual exclusions between object clitics (hence Spurious se and the like), while as far as we can tell this connection escapes Roberts (1993) altogether. 6. Manzini and Savoia (2005) argue that the difference between negative and nonÂ�negative n segments resides in the fact that they are inserted as sentential constituents in negative contexts, but as morphological constituents of the P clitic in positive contexts. As far as we can tell, the best evidence that n segments may turn up as word-internal constituents is provided by data from Felizzano, not reviewed here. 7. It may be worth providing some examples of the latter€ – also stressing that the arguments in the typological literature that ‘in questions, negation is neutralized […]:€Can you hear nothing? and Can you hear anything? have identical truth conditions’ (Haspelmath 1997:€121) are orthogonal to the arguments being advanced in the text; whether they stand further scrutiny or not (Weiss 2002) is very interesting in itself but not relevant for present purposes. Consider Italian (i). It seems fairly clear that there are two truth-conditionally different readings associated with the absence or presence of the negative clitic non, as reflected by the English translation. Yet the Italian non ‘not’ sentence is compatible with both (non truth-conditionally equivalent) readings.
322â•… Notes to pages 152–158 (i) Volli farlo prima che le mie forze (non) me lo consentissero I.wished to.do.it before that my energies (not) to.me it would.allow ‘I wished to do it before my energies allowed/did not allow me to’ For n-words, an appropriate context is represented by hypotheticals. The two English sentences in (ii) are clearly non-truth-functionally equivalent; by contrast, the Italian sentence in (iii) is ambiguous between (iia) and (iib)€– at least for speakers (like the author) who easily allow nessuno to be licensed by the conditional. (ii) a. If nobody arrives, tell me b. If anybody arrives, tell me (iii) Se nessuno arriva, dimmelo (subito) if N-one arrives tell.me.it (straight.away) It is also worth stressing that the conclusion that neither the n-clitic nor n-words are intrinsically negative is far from unique in the literature. Thus, for Â�Rooryck (2009, and references quoted there), ‘the paradox of negation in languages like bon usage French and Catalan then can be formulated as follows:€how can the combination of two elements that do not intrinsically express negation nevertheless lead to the expression of negation?’. Here we do not push the answer to this ‘paradox’ any further than is stated in the text; in other words, we assimilate the interpretive content of the n-clitic to that of n-argumental/adverbial phrases, treating all of them as negative polarity items; the negation operator (corresponding to the logical negation, or contrary negation in the sense of Horn (1989)) is introduced at LF to license all of these elements. The conjecture is that other polarity operators (nonveridical operators in the sense of Giannakidou (1998)) can license polarity items in the various non-negative contexts. Rooryck’s (2009) own answer to the ‘paradox’ takes the n-clitic to be ‘an NPI-licensing operator expressing contrary negation’, again in the sense of Horn (1989), ‘selecting the smallest possible value in a set of alternatives’ (i.e. the null set at the limit); the set of alternatives is provided by negative polarity items, construed as ‘partially ordered sets including the null set’. What seems to remain constant is that the n-clitic does not correspond to the logical negation embodied by the functional category Neg. Further interpretive matters are beyond the scope of the present chapter. 8. This does not correspond to a grammaticality judgement elicited from our speaker(s), but rather to our reconstruction of the gap found in the data, provided for ease of exposition. 9. If we backtrack to sentential negation adverbs, we may wonder whether the incompatibility with left periphery arguments or adverbs holds of them as well. An interesting case in point could be Müstair in (i). The adverbial negation nylja is found in the pre-infinitival position in (ia)€– actually an example of a negative imperative€– in accordance with what was independently observed in chapter 3 for Piedmontese varieties (cf. section 3.2). As shown in (ia), in this position it can co-occur with the lexically identical negative polarity argument nylja ‘nothing’ in postverbal position. However, (iib) shows a different possibility, namely a simple occurrence of nylja in pre-infinitival position. We conclude that (iib) displays an instance of Q-float of argumental nylja– of the type analysed by Kayne (1975) for French under the title of
Notes to pages 158–181â•… 323 L€— tous. This creates a local configuration of mutual exclusion with adverbial nylja, yielding the simple occurrence of the argument. (i)â•… Müstair (Grisons) a. nylja far nylja not do nothing ‘Don’t do anything!’ b. nylja tor nothing take ‘Don’t take anything!’
5
The middle-passive voice:€evidence from Albanian
1. Work on the Albanian middle-passive voice by Kallulli (2006), while finely disentangling the various bound readings associated with the Albanian forms, does not address the unification with the generic (impersonal) readings€– nor the question (crucial here) of the varying morpholexical instantiations of the relevant meaning cluster. In particular, Kallulli (2006) adopts the view that middle-passive morphology corresponds to a valency reduction executed by means of feature deletion on the v node responsible for introducing the external argument. Thus the passive is derived through suppression of a [+activity] feature from underlying activity structures, the anticausative through suppression of a [+cause] feature from underlying causative structures and the reflexive through suppression of [+intent] feature in underlying agentive structures. We will not enter into the finer points of the proposal; for instance the discussion in the text excludes the possibility that reflexives are non-agentive. What we are interested in is, as before, the fact that the various interpretations are simply encoded in the syntax. The fact that the mapping between the interpretation/syntax and its morphological instantiations becomes completely opaque is obviously not even perceived as a cost to be paid for this. The very same points that we are making here with respect to Kallulli’s analysis are discussed by Roussou (2008) and Manzini, Roussou and Savoia (to appear) in relation to analyses of the middle-passive voice in Greek (see section 5.3 below). For instance, Alexiadou and Anagnostopoulou (2004) argue that the different readings of the Greek middle-passive are syntactically encoded through the projection of different v heads; a passive has a VoiceP above VP with no specifier, while an anticausative has a Result vP embedding the passive VoiceP structure. The point is that the projection of different structures is necessitated by the availability of different readings with no connection to their morphological instantiation(s). Thus, the interface of syntax with morphology becomes opaque, again without this ostensibly being perceived as a cost at all. 2. (21b) is part of the larger question of deponent verbs, i.e. those verbs that are associated with a morphologically middle-passive inflection while having no active counterparts. The deponency phenomenon is present to some extent in any system which has a productive middle-passive voice. Thus, Italian has unaccusative predicates formed with si that have no transitive counterparts, as in (ia–a’); it also has
324â•… Notes to pages 181–183 predicates formed with si that are straight transitives€– nor does si correspond to an independently attested dative or other argument, as in (ib–b’). (i)â•… a. a’. b. b’.â•…
*Gianniâ•… ha arrabbiatoâ•… Lucia/ se stesso G hasâ•… angered L/ himself ‘John has angered Lucia/himself’ Gianni si è arrabbiato G M is angered ‘John has become angry’ *Gianni lo ha dimenticato a Lucia/ se stesso G it has forgotten to L /himself ‘John has made Lucia/ himself forget it’ Gianni se lo è dimenticato G M it is forgotten ‘John has forgotten it’
In effect (i) are the deponents of the Italian middle-passive voice. Following Embick (2000), we would have to say that while the middle-passive forms considered in the text are associated with a voice feature, say [pass], associated with the v functional projection of the verb, the forms in (i) are associated with a [pass] feature associated with the verb root. The objection that we have to this type of approach is that at best it annotates the relevant differences without explaining anything about them. Therefore, despite the dismissal of such ideas by Embick (2000), we tentatively maintain that (i) as well as Albanian (21b) in the text are bona fide instances of middle-passive voice at LF, with the si (or its various counterparts) instantiating a variable in the argument structure of the relevant verbs. This is an internal argument in (ia’) and a second internal argument in (ib’)€– despite the lack of active instantiations of such argument structures. 3. Movement from the object to the subject position could in turn be made compatible with a bona fide clitic treatment of si€– but at a cost. Thus, a base-generated si could coexist with movement from the object position of the EPP argument, as proposed by Dobrovie-Sorin (1998). Further elaborating on this, one could entertain the idea that si is base-generated as the head of a so-called big DP, whose Spec is the ‘doubling’ moved DP. Such extra assumptions are simply unnecessary here. 4. Another case in point is the Surselva variety of Mustér, where the impersonal is ins ‘one’. Note that ins is connected etymologically with unus ‘one’, while French on is connected with homo ‘man’ (along the lines of the German impersonal man). (i)â•… Mustér (Grisons) a. ins drma bain lQu one sleeps well there ‘One sleeps well there’ b. kwl vz-ins adina pr la vias that.one sees-one always in the street ‘One always sees him on the street’
Notes to pages 186–214â•… 325 5. Greek, unlike Albanian, allows for perfective aspect in the present tense as well. The present perfective forms in both the middle-passive and the active voice are characterized as ‘dependent’ ones (Holton, Mackridge and Philippaki-Warburton 1997), since they cannot occur independently in (finite) clauses, but have to be preceded by one of the modal particles (‘subjunctive’ na, ‘future’ tha or ‘hortative’ as), a hypothetical or temporal conjunction (an ‘if’, otan ‘when’, etc.) or a free relative pronoun (e.g. opjos ‘whoever’). 6. In a previous version of this chapter (Manzini and Savoia 2008c), the evidence provided by Roussou (2008) concerning the mutual exclusion of the -th middlepassive morphology and the specialized middle-passive person endings, based on the idea that -th represents a morphological-level counterpart to the si/u clitic, led us to exclude the possibility of identifying the -h morphology of Albanian with the latter€ – which we now take to be too hasty a conclusion. By contrast, the line of analysis pursued here is briefly suggested by Manzini, Roussou and Savoia (to appear), though this work is largely concerned with the interpretation of morphological-level ambiguities of various kinds (syncretisms, etc.), and only partially devoted to the question of varying morphological instantiations of the same interpretation(s). 6
The auxiliary:€have/be alternations in the perfect
1. Note that, given the existence of a person split, the reader may reasonably wonder about the full paradigms for the varieties discussed in connection with transitivity/ voice. There is no person split involved in any language mentioned except where explicitly noted. 2. As for the mapping of bisentential structures like (14) to a perfect interpretation, it is compatible with characterizations found in the literature according to which the perfect, at least in English, roughly denotes a present state arrived at as a consequence of a concluded event, hence a past by implication. For instance, Parsons (1990) assigns to sentences such as John has left a semantics like the following:€there is an event e of ‘leaving’ whose theme is ‘John’ and the state following e, CS(e) holds at the moment of utterance. As a further step, the French/Italian present perfect can take on the meaning of a perfective simple past, as is also well known from the literature (Giorgi and Pianesi 1998). 3. Cf. fn. 14. 4. The deeper level of explanation for this split, in terms of the real primitives of the computational component/LF interface, remains to be explored. A natural way to go is to connect the anchoring to the referential properties€– cf. the concluding remarks section 6.5. 5. Manzini and Savoia (2007) invoke essentially the same constraint to yield the canonical EEA distribution, in the following terms:€ ‘taking essere … configures a particularly simple grammar where, in the absence of any additional constraints, it is the same matrix verb, namely the copula essere, that appears in all contexts where a nominal predicative complement is selected’ as opposed to ‘the more complex grammar, with essere and avere alternating’, as copula and auxiliary respectively. ‘The simpler grammar in the 1st and 2nd person, where the participle is embedded
326â•… Notes to pages 214–224 under essere without interpretive restrictions, corresponds to the independence of speaker, hearer reference from the event structure. The more complex grammar, with essere and avere alternating … is associated with elements whose reference in deictic space is obtained through their anchoring in the event structure’. In other words, languages have a choice as to how far down the scale they implement (21). This may go some way towards explaining another problem hardly touched on here, namely why certain patterns are more robustly attested than others. 6. In the Albanian (Arbëresh) variety of Portocannone in (15), the perfect middleÂ�passive is formed with the u clitic, which combines with the kam ‘have’ auxiliary. In this respect, Portocannone behaves like the Romance varieties in which participial sentences are systematically selected by have (e.g. Verbicaro in (9)). 7. For Reinhart and Siloni (2005), the contrast in (30)–(31) is part of a larger parameter concerning whether arity operations apply in the syntax or in the lexicon. Thus, the fact that in Italian si can correspond to a dative, as in (31), means that voice is determined in the syntax; the fact that in Albanian the counterpart to (31) is just an active sentence, as in (32), means that voice is determined in the lexicon. However, given the spread of morphosyntactic realizations and interpretations connected to the middle-passive voice, it is hard to see in what sense voice could be lexical in Albanian. 8. The evidence is complicated somewhat by the fact that several Romance languages, e.g. Spanish, Southern Italian dialects, have the estar (Spanish) ‘lit:€to stay’ type in contexts which correspond to copular contexts in English (or Italian); however, other copular contexts have be. The relevant distinction appears to be between stage-level predicates (estar) and individual-level ones (ser). 9. In several Romance varieties, e.g. Spanish and Southern Italian dialects, the possession verb is of the type of tener (Spanish) ‘lit:€to hold’; however, have is attested as the necessity modal. 10. Legendre (2007) takes up again the evidence of Sorace (2000) providing a hierarchy of several verbal properties (displacement, telicity, etc.) ranked as a function of the property of blocking have; the auxiliary selection pattern of a language is a result of the ranking of the constraint *E (blocking be) within this hierarchy. We have no insight to offer on the ASH as an empirical generalization, nor on the categories of telicity, agentivity and others that are deemed to enter into it. When it comes to building such categories into the syntax, however, we note that much of the discussion in this work is predicated on the general idea that interpretive categories should be construed at the LF interface and not imported into the syntax for reasons of Interface Uniformity (Culicover and Jackendoff 2005) or for other reasons. 11. Kayne (2008b) attempts a different approach, in particular to the variation between Italian and French anticausatives. He argues that anticausative verbs are formed with a silent causative head. He assumes that in French, the perfect participle of this causative verb (like its overt counterpart) does not agree; if be requires an agreeing participle, this forces selection of have with the relevant class of verbs. In Italian, the participle of these causative verb sagrees, and be is selected. The problem is that, as we stressed in section 6.1, auxiliary selection and perfect participle agreement are independent phenomena. A key observation is that the assumption that
Notes to pages 224–240â•… 327 be implies participle agreement does not even hold in Romance languages, since in Montebello Ionico/Saline Ioniche, as in (18), in which have and be are in free variation, the participle remains non-agreeing with be as well. Furthermore, even if the implication held in Romance languages, it is not clear why it would not hold in Germanic languages (Kayne 2008a:€fn. 24) or in Albanian. 12. Another parameter of variation that we will disregard here concerns auxiliary alternations according to tense/aspect/mood (cf. Manzini and Savoia (2007), also McFadden (2007) for a survey, Cennamo (2002) and Ledgeway (2003b) for different views of Old Neapolitan). 13. Going back to the previous discussion of Kayne (2008b), we do not expect there to be an implicational relationship from perfect participle agreement to auxiliary selection (see the discussion in section 6.1 and in fn. 11)€ – nor, to take another recent proposal, do we expect that auxiliary selection will depend on the null-subject parameter (D’Alessandro and Roberts 2010). Specifically, D’Alessandro and Roberts propose that ‘If a language has person-driven auxiliary selection, that language must be a null-subject language’. However, the evidence concerning Molina in (39) and Briga in (41) shows that Northern Italian varieties can have person splits in auxiliary selection, even though they are subject-clitic languages. Two responses are possible to this objection€ – and both of them seem problematic to us. First, subject-clitic languages could be null-subject languages (Rizzi 1986); but note that French has never been taken to be€– and the overall variation picture within Romance establishes the continuity between French and Northern Italian. A second answer could be that the generalization only holds for auxiliary selection according to person of the simple type in section 6.1, excluding alternations according to both person and transitivity/voice. In fact, nothing in D’Alessandro and Roberts (2010) leads us to believe that they take their theory to apply to cases like Molina or Briga. But if so, this is a further problem for them. 14. In fact, on the basis of existing morphophonological evidence, one may want to argue that, on the contrary, be is derived from have. The interesting thing is that the be auxiliary in (44a–a’) is actually the result of combining the have auxiliary in (44b) with the s morpheme. Manzini and Savoia (2005) suggest that the latter could be identified with the si middle-passive clitic. 7
The noun (phrase):€agreement, case and definiteness in an Albanian variety
1. See Savoia (2008) on the microvariation between the nominal systems of Vena itself and two other Arbëresh varieties, Ginestra and Casalvecchio. If the Vena language fairly systematically reflects the conditions of the standard, the two other varieties show properties more typical of languages in contact. The case system of the Gheg variety of Shkodër, largely coinciding with the standard, is considered by Manzini and Savoia (to appear). 2. The ablative ending -t survives only in a limited number of constructs, illustrated in (i), which for present purposes can be treated as idioms. Manzini and Savoia (to appear) deal with their counterpart in the variety of Shkodër, in which the ablative is more productive.
328â•… Notes to pages 240–256 (i) vstit vazda-t∫ / rs-t∫ / gra-t∫ a garment girls-abl/ men-abl / women-pl ‘a girls’/ women’s/ men’s garment’ 3. An interesting issue that we do not address here concerns eventive nouns. The potential problem is that the genitive in this case lexicalizes not the possessor, which has a notoriously loose relation with the head noun, but what appears to be the internal argument of the eventive noun, which has a much stricter relation to it, as illustrated for the ne clitic of Italian in (i). (i) Ne ho disapprovato il furto of.it I.have disapproved the theft ‘I disapproved the theft of it’ We provisionally maintain the same characterization for the genitives in (i) as we have provided here for partitives/possessors. The reason is that in the languages that we analyse, all of them receive the same morphosyntactic treatment. It looks like the genitive is the all-purpose shape of a nominal complement€– though its interpretation is restricted when it satisfies certain argument slots of eventive nouns. It is implicit in the analysis of Romance in section 7.1 (and more explicit in Manzini and Savoia 2005, 2008a) that nouns are ergative structures of some sort, with the inflection and the determiner satisfying the (true) internal argument of the predicative base and at the same time the D (EPP) property of the noun phrase. On these grounds, we may expect that accusative and nominative are exclusively sentential cases, so that noun phrases are restricted to the oblique (genitive, by-phrase, locative, etc.). 4. An alternative view of the accusative inflection is also possible, under which –n() of Albanian (cf. –m of Latin in Chapter 8) is not so much a specialized nominal class morpheme as an operator-like element. Such a characterization brings it closer to the consonantal inflections of the oblique and of the nominative as well (cf. the discussion of Latin –s in Chapter 8). In this alternative view, the –n() inflection could be construed as introducing a CAUSE-type relation between the argument to which it attaches and another, superordinate argument (the causer, agent, etc.) within the sentential domain. In other words, in accusative marking languages the internal argument of the predicate would not be introduced directly but through an operator (the so-called accusative) akin to Chomsky’s (1995) v. We leave any further discussion of these ideas for future work. 5. A reinterpretation of case in terms of the basic syntactic categories N, D, etc. is proposed by Pesetsky (2008). However, while we construe N, D etc. categorizations as specifying the intrinsic properties of the so-called case morphemes, Pesetsky seems to have (roughly) case-assigning specifications (N for partitive, V for accusative etc.) in mind. 6. There is a tricky point involved here. We invoke the D closure for the definite reading of nominative singulars lexicalized only by a nominal class vowel€– while for indefinites we say that the relevant closure is provided by an indefinite quantifier. But indefinite EPP arguments also agree with a D inflection, leading one to question how this is compatible with the definiteness of the latter. We simply assume that the D
Notes to pages 256–283â•… 329 inflection can take scope over the lexical subject€– but if this results in an ill-formed structure (as in the case of indefinites), then it simply acts as a pronominal double of sorts. In fact, in a language like Italian, in which definite determiners are independently lexicalized, the D inflection of the verb alternates between doubling the lexical Â�subject and actually providing the EPP closure of the sentence, for the simple reason that Italian is a null-subject language. In other words, we are not introducing any ad hoc device here. 7. In recent work, in particular Cuervo (2003) for Spanish, based on Pylkkänen (2002), the dative interpretation is mediated by the postulation of Applicative heads. We will return to this proposal in chapter 8 in connection with Romanian, which also shows the genitive/dative syncretism and has the same structure as Albanian for the embedding of genitives. 8. In the sentence, multiple D positions correspond, for instance, to the multiple occurrences of subject clitics observed in Romance languages (cf. here chapter 4). 9. Another interesting question posed by Albanian concerns the fact that with possessives, the kinship term appears in the indefinite form. This phenomenon is reminiscent of the mutual exclusion between possessives and definiteness in languages like French or English. We may suggest that 1st and 2nd person possessives in Albanian are sufficient to lexicalize definiteness properties for the inflectional domain of the noun phrase otherwise lexicalized by the definite inflection€– therefore excluding the latter. With non-kinship terms, possessives appear in postnominal position, as do adjectives and genitives. 8
(Definite) denotation and case in Romance:€history and variation
1. Fillmore (1968) had already argued in favour of the equivalence of NP-k structures with P-NP ones, in both instances yielding a KP phrase. 2. Recall from the discussion in chapter 7 that ‘sentential’ scope amounts to scope of the Q(⊆) operator over the elementary event (the verb plus its internal argument(s)). Indeed, the Q(⊆) operator to which we reduce the dative denotes a superset-of relation between the argument to which the dative morphology attaches and the internal argument of the superordinate predicate, excluding the external argument (see also chapter 3 on the notion of elementary event, roughly corresponding to the VP of Chomsky (1995), as opposed to its vP). 3. Strictly speaking, then, the discussion in the text sketches an account for the syncretism of dative with locative, namely with one of the interpretations associated with the traditional ablative, leaving out the other main interpretation as instrumental. Similarly, the ablative of the Gheg Albanian variety of Shkodër, considered by Manzini and Savoia (to appear), essentially has a locative value. Providing a precise characterization for instrumental is beyond the scope of the present work. Note that, on the basis of the discussion in the text, we expect genitive/locative syncretisms as well. In effect, it is well known that with a subset of lexical bases of the I, II class denoting cities/towns and small islands, the locative is formed with the -i morphology of the genitive.
330â•… Notes to pages 303–309 4. In a subset of masculine singular adjectives and participles, the attributive alternant without -s differs from the predicative alternant with -s in the quality of the stressed vowel, cf. bien vs. bu-s ‘good’, but also ni@f vs. nf-s ‘new’, aviert ‘open’ vs. aviart-s. 5. The possessive, which behaves like an adjective, also alternates between sigmatic forms in copular contexts like (18c) and non-sigmatic forms when embedded under a definite determiner, as in (21b). An interesting issue is raised by examples like (20c) in which the asigmatic form of the possessive is in complementary distribution with the article. The mutual exclusion of determiners and articles in Romance varieties is discussed by Manzini and Savoia (2005). There we argue that the combination of the possessive with the noun itself introduces a type of D closure with kinship terms in particular (cf. also Longobardi 1994, 1996). Evidently, the same type of D closure excludes the –s quantificational inflection in examples like (20c). 6. While it is in principle possible to account for lack of agreement of the verb with the postposed subject by saying that it agrees with the expletive instead, as proposed by Chomsky (1995), in the work cited in the text we have argued against this analysis on empirical, as well as theoretical grounds. 7. It will be noticed that in (a) we have used a pre-auxiliary context, in (b) a context preceding a lexical verb and in (c) an enclitic context; this was done just to display the various morphologies in their fuller form.
References
Acquaviva, Paolo 1994. ‘The representation of operator-variable dependencies in sentential negation’, Studia Linguistica 48:€91–132. Adger, David and Quer, Joseph 2001. ‘The syntax and semantics of unselected embedded questions’, Language 77:€107–33. Adger, David and Ramchand, Gillian 2005. ‘Merge and move:€wh–dependencies revisited’, Linguistic Inquiry 36:€161–93. Alexiadou, Artemis and Anagnostopoulou, Elena 2004. ‘Voice morphology in the causative-inchoative alternation:€ evidence for a non-unified structural analysis of unaccusatives’, in A. Alexiadou, E. Anagnostopoulou and M. Everaert (eds.), The Unaccusativity Puzzle. Explorations of the Syntax-Lexicon Interface. Oxford Â�University Press, pp. 114–36. Anagnostopoulou, Elena 2003. ‘Participles and voice’, in A. Alexiadou, M. Rathert and A. von Stechow (eds.), Perfect Explorations. Berlin:€ Mouton De Gruyter, pp.1–36. Arsenijevic, Boban 2009. ‘Clausal complementation as relativization’, Lingua 119: 39–50. Bastiaansen, Marcel C., Oostenveld, Robert, Jensen, Ole and Hagoort, Peter 2008. ‘I see what you mean:€theta-power increases are involved in the retrieval of lexical semantic information’, Brain and Language 106:€15–28. Baker, Marc 2001. The Atoms of Language. New York:€Basic Books. Bayer, Josef 2006. ‘Nothing/nichts as negative polarity survivors?’, in H.-M. Gärtner, S. Beck, R. Eckardt, R. Musan and B. Stiebels (eds.), Between 40 and 60 Puzzles for Krifka:€a web festschrift for Manfred Krifka. Berlin:€Zentrum für Allgemeine Sprachwissenschaft, Typologie und Universalienforschung (ZAS), pp. 5–30. Belletti, Adriana 1988. ‘The case of unaccusatives’, Linguistic Inquiry 19:€1–35. â•… 1990. Generalized Verb Movement. Turin:€Rosenberg & Sellier. â•… 2000. ‘Speculations on the possible source of expletive negation in Italian comparative clauses’, in G. Cinque and G. Salvi (eds.), Current Studies in Italian Syntax. Essays offered to Lorenzo Renzi. Amsterdam:€North Holland, pp. 19–37. â•… 2004. ‘Aspects of the Low IP Area’, in L. Rizzi (ed.), The Structure of IP and CP:€The Cartography of Syntactic Structures, Volume II. Oxford University Press, pp. 16–51. â•… 2008. ‘Answering strategies:€new information subjects and the nature of clefts’, in A. Belletti, Structures and Strategies. New York:€Routledge, pp. 242–65.
331
332â•… References Belletti, Adriana and Rizzi, Luigi 1981. ‘The syntax of ne:€some theoretical implications’, The Linguistic Review 1:€117–54. Belvin, Robert and den Dikken, Marcel 1997. ‘There, happens, to, be, have’, Lingua 101:€151–83. Benincà, Paola 2001. ‘On the position of Topic and Focus in the left periphery’, in G. Cinque and G. Salvi (eds.), Current Studies in Italian Syntax. Amsterdam:€North Holland, pp. 39–64. â•… 2006. ‘A detailed map of the left periphery of medieval Romance’, in R. Zanuttini, H. Campos, E. Herberger and P. Portner (eds), Negation, Tense, and Clausal Architecture:€ Crosslinguistic Investigations. Georgetown University Press, pp. 53–86. Benincà, Paola and Poletto, Cecilia 2004. ‘Topic, focus and V2:€defining the CP subÂ� layers’, in L. Rizzi (ed.), The Structure of IP and CP. The Cartography of Syntactic Structures. Volume II. Oxford University Press, pp. 52–75. â•… 2001. Alternation according to person in Italo-Romance’, in L. Brinton (ed.), Historical Linguistics 1999. Selected Â�Papers from the 14th International Conference on Historical Linguistics. Amsterdam:€Benjamins, pp. 63–74. â•… 2003. ‘Auxiliary selection and the semantics of unaccusativity’, Lingua 114: 447–71. Benveniste, Émile 1966. Problèmes de linguistique générale. Paris:€Gallimard. Besten den, Hans 1984. ‘On the interaction of root transformations and lexical deletive rules’, in W. Abraham (ed.), On the Formal Syntax of West-Germania, Amsterdam:€Benjamins, pp. 47–131. Bhatt, Rajesh and Pancheva, Roumyana 2006. ‘Conditionals’, in M. Everaert and H. van Riemsdijk (eds.), The Blackwell Companion to Syntax. Volume 1. Oxford:€Blackwell, pp. 638–87. Bobaljik, Jonathan 2008. ‘Missing persons:€ a case study in morphological universals’, The Linguistic Review, special issue Examples of Linguistic Universals 25: 203–30. Brody, Michael 2003. Towards an Elegant Syntax. London:€Routledge. Brunetti, Lisa 2004. A Unification of Focus. Padua:€Unipress. Brunot, Ferdinand and Bruneau, Charles 1969. Précis de grammaire historique de la langue française. Paris:€Masson et Cie. Burzio, Luigi 1986. Italian Syntax. Dordrecht:€Kluwer. Caha, Pavel 2009. The Nanosyntax of Case. Ph.D. Dissertation, University of Tromsoe. Calabrese, Andrea 1998. ‘Some remarks on the Latin case system and its development in Romance’, in J. Lema and E. Trevino (eds.), Theoretical Advances on Romance Languages. Amsterdam:€John Benjamins, pp. 71–126. â•… 2008. ‘On absolute and contextual syncretism:€remarks on the structure of paradigms and on how to derive it’, in A. Nevins and A. Bachrach (eds.), The Bases of Inflectional Identity. Oxford University Press, pp. 156–205. Camaj, Martin 1984. Albanian Grammar. Wiesbaden:€Harrassowitz. Caponigro, Ivano and Polinsky, Maria 2008. ‘Everything is relative:€ evidence from Northwest Caucasian’, paper presented at the GLOW 31 Colloquium, University of Newcastle.
Referencesâ•… 333 Cennamo, Michela 2001. ‘L’inaccusatività in alcune varietà campane, teorie e dati a confronto’, in F. Albano Leoni, E. Stenta Krosbakken, R. Sornicola and C. Stromboli (eds.), Dati empirici e teorie linguistiche. Atti del XXXIII congresso internazionale di studi della SLI. Rome:€Bulzoni, pp. 427–53. â•… 2002. ‘La selezione degli ausiliari perfettivi in napoletano antico:€fenomeno sintattico o sintattico-semantico?’, Archivio Glottologico Italiano 87:€175–222. â•… 2008. ‘The rise and development of analytic perfects in Italo-Romance’, in T. Eythórsson (ed.), Grammatical Change and Linguistic Theory:€the Rosendal Papers. Amsterdam:€Benjamins, 115–42. â•… 2010. ‘Perfective auxiliaries in the pluperfect in some southern Italian dialects’, in R. D’Alessandro, A. Ledgeway and I. Roberts (eds.), Syntactic Variation:€the Dialects of Italy. Cambridge University Press, 210–24. Cennamo, Michela and Sorace, Antonella 2007. ‘Unaccusativity at the syntax-lexicon interface:€evidence from Paduan’, in R. Aranovich (ed.), Split Auxiliary Systems:€a Cross–linguistic Perspective. Amsterdam:€Benjamins, pp. 65–100. Cheng, Lisa 1991. On the Typology of Wh–questions. Ph.D. Dissertation, MIT, Cambridge, Mass. Chierchia, Gennaro 1995. ‘Impersonal subjects’, in E. Bach, E. Jellinek, A. Kratz and B. H. Partee (eds.), Quantification in Natural Languages. Dordrecht:€ Kluwer, pp. 107–43. â•… 1998. ‘Reference to kinds across language’, Natural Language Â�Semantics 6:€339– 405. â•… 2004. ‘A semantics for unaccusatives and its syntactic consequences’, in A. Alexiadou, E. Anagnostopoulou and M. Everaert (eds.), The Unaccusativity Puzzle: Â�Explorations of the Syntax-Lexicon Interface. Oxford University Press, pp. 22–59. Chomsky, Noam 1957. Syntactic Structures. The Hague:€Mouton. â•… 1965. Aspects of the Theory of Syntax. Cambridge, Mass.:€MIT Press. â•… 1973. ‘Conditions on transformations’, in S. Anderson and P. Kiparsky (eds.),â•› A Festschrift for Morris Halle. New York:€Holt, Rinehart and Winston, pp. 232–86. â•… 1975 [1955]. The Logical Structure of Linguistic Theory. New York:€Plenum. â•… 1981. Lectures on Government and Binding. Dordrecht:€Foris. â•… 1986. Knowledge of Language, New York:€Praeger. â•… 1995. The Minimalist Program, Cambridge Mass.:€MIT Press. â•… 2000a. ‘Minimalist inquiries:€ the framework’, in R. Martin, D. Michaels and J. Uriagereka (eds.), Step by Step. Cambridge, Mass.:€MIT Press, pp. 89–155. â•… 2000b. New Horizons in the Study of Language and Mind. Cambridge University Press. â•… 2001. ‘Derivation by phase’, in M. Kenstowicz (ed.), Ken Hale:€A Life in Language. Cambridge, Mass.:€MIT Press, pp. 1–52. â•… 2002. On Nature and Language. Cambridge University Press. â•… 2005. ‘Three factors in language design’, Linguistic Inquiry 36:€1–22. â•… 2008. ‘On Phases’, in R. Freidin, C. Otero and M.-L. Zubizarreta (eds.), Foundational Issues in Linguistic Theory. Essays in Honor of Jean-Roger Vergnaud. Cambridge, Mass.:€MIT Press, pp. 133–66. Chomsky, Noam and Lasnik, Howard 1977. ‘Filters and control’, Linguistic Inquiry 8:€425–504.
334â•… References Cinque, Guglielmo 1990. Types of A’-Dependencies. Cambridge, Mass.:€MIT Press. â•… 1995. Italian Syntax and Universal Grammar. Cambridge University Press. â•… 1999. Adverbs and Functional Heads:€a Cross-linguistic Perspective. Oxford University Press. â•… 2006. Restructuring and Functional Heads:€the Cartography of Syntactic Structures Volume 4. Oxford University Press. Cinque, Guglielmo and Rizzi, Luigi 2008. ‘The cartography of syntactic structures’, Studies in Linguistics 2, University of Siena, pp. 42–58. Cocchi, Gloria 1995. La selezione dell’ausiliare. Padua:€Unipress. Cocchi, Gloria and Poletto, Cecilia 2002. ‘Complementizer deletion in Florentine’, in C. Beyssade, R. Bok-Bennema, F. Drijkoningen and P. Monachesi (eds.),â•› Romance Languages and Linguistic Theory 2000. Amsterdam:€Benjamins, pp. 57–76. Cornilescu, Alexandra 1995. ‘Rumanian genitive constructions’, in G. Cinque and G. Giusti (eds.), Advances in Rumanian Linguistics. Amsterdam:€ Benjamins, pp. 1–54. Cruschina, Silvio 2010 ‘Fronting as focalization in Sicilian’, in R. D’Alessandro, A. Ledgeway and I. Roberts (eds.), Syntactic Variation. The Dialects of Italy. Cambridge University Press, pp. 247–60. Cuervo, María C. 2003. Datives at Large. Ph.D. Dissertation, MIT, Cambridge, Mass. Culicover, Peter and Jackendoff, Ray 2005. Simpler Syntax. Oxford University Press. â•… 2006. ‘The simpler syntax hypothesis’, Trends in Cognitive Sciences 10:€413–18. D’Alessandro, Roberta and Ledgeway, Adam 2010. ‘At the C-T boundary:€investigating Abruzzese complementation’, Lingua 120:€2040–60. D’Alessandro, Roberta and Roberts, Ian 2010. ‘Past participle agreement in Abruzzese:€ split auxiliary selection and the null-subject parameter’, Natural Language and Linguistic Theory 28:€41–72. Damonte, Federico 2006. ‘Complementatori e complementi congiuntivi in alcuni dialetti sardi’, Quaderni di lavoro dell’ASIt 6:€71–95. Davidson, Donald 1997 [1968]. ‘On saying that’, in P. Ludlow (ed.), Readings in the Philosophy of Language, Cambridge, Mass.:€MIT Press, pp. 817–32. Davis, Henry 1999. ‘Subject inflection in Salish’, UBC Working Papers in Linguistics 1, University of British Columbia, pp. 181–241. De Dardel, Robert and Gaeng, Paul 1992. ‘La déclinaison nominale du latin non classique:€essai d’une méthode de synthèse’, Probus 4:€91–125. Dehaene, Stanislas, Izard, Veronique, Pica, Pierre and Spelke, Elizabeth 2006. ‘Core knowledge of geometry in an Amazonian indigene group’, Science 311: 381–4. DeLancey, Scott 1981. ‘An interpretation of split ergativity and related patterns’, Language 57:€626–57. Diaconescu, Constanţa R. and Rivero, María Luisa 2007. ‘An applicative analysis of double object constructions in Romanian’, Probus 19:€209–33. Dimitrova-Vulchanova, Mila and Giusti, Giuliana 1998. ‘Fragments of Balkan nominal structure’, in A. Alexiadou and C. Wilder (eds.), Possessors, Predicates and Movement in the Determiner Phrase. Amsterdam:€Benjamins, pp. 333–60.
Referencesâ•… 335 Dikken, Marcel den 1998. ‘Predicate inversion in DP’, in A. Alexiadou and C. Wilder (eds.),â•› Possessors, Predicates and Movement in the Determiner Phrase. Amsterdam: Benjamins, pp. 177–214. Dixon, Robert M. W. 1994. Ergativity. Cambridge University Press. Dobrovie-Sorin, Carmen 1987. ‘A propos de la structure du groupe nominal en Roumain’, Rivista di Grammatica Generativa 12:€123–52. â•… 1998. ‘Impersonal se constructions in Romance and the Â�passivization of unergatives’, Linguistic Inquiry 29:€399–437. Embick, David 2000. ‘Features, syntax and categories in the Latin perfect’, Linguistic Inquiry 31:€185–230. Epstein, Samuel and Seely, T. Daniel 2006. Derivations in Minimalism. Cambridge Â�University Press. Ernst, Thomas 2002. The Syntax of Adjuncts. Cambridge University Press. Fici, Francesca, Manzini, M. Rita and Savoia, Leonardo M. 2001. ‘Clitics in Macedonian’, in G. Zybatov, U. Junghanns, G. Melhorn and L. Szucsich (eds.), Current Issues in Formal Slavic Linguistics. Frankfurt:€Peter Lang, pp. 148–58. Fillmore, Charles 1968. ‘The case for case’, in E. Bach and R. Harms (eds.), Universals in Linguistic Theory. London:€Holt, Rinehart and Winston, pp. 1–88. Fodor, Jerry 1983. The Modularity of Mind. Cambridge, Mass.:€MIT Press. Formentin, Vittorio 1998. ‘Commento Linguistico’, in Loise De Rosa, Ricordi, Tomo I. Rome:€Salerno Editrice. Frascarelli, Mara and Hinterhölzl, Roland 2007. ‘Types of topics in German and Italian’, in S. Winkler and K. Schwabe (eds.), On Information Structure, Meaning and Form. Amsterdam:€Benjamins, pp. 87–116. Freeze, Ray 1992. ‘Existentials and other locatives’, Language 68:€553–95. Garzonio, Jacopo and Poletto, Cecilia 2009. ‘Quantifiers as negative markers in Italian dialects’, Linguistic Variation Yearbook 9:€127–52. Geis M. 1970. Adverbial Subordinate Clauses in English. Ph.D. Dissertation, MIT, Cambridge, Mass. Giannakidou A. 1998. Polarity Sensitivity as (Non-)veridical Dependency. Amsterdam: Benjamins. Giorgi, Alessandra and Pianesi, Fabio 1998. Tense and Aspect. From Semantics to Morphosyntax. New York:€Oxford University Press. â•… 2004. ‘Complementizer deletion in Italian’, in L. Rizzi (ed.), The Structure of CP and IP:€the Cartography of Syntactic Â�Structures. Volume 2. Oxford University Press, pp. 190–210. Giusti, Giuliana 1995. ‘A unified structural representation of (abstract) case and article’, in H. Haider, S. Olsen and S. Vikner (eds.), Studies in Comparative Germanic Syntax. Dordrecht:€Kluwer, pp. 77–93. â•… 2001. ‘The rise of a functional category from Latin ILLE to the Â�Romance article and personal pronouns’, in G. Cinque and G. Salvi (eds.), Current Studies in Italian Syntax. Amsterdam:€North Holland, pp. 265–86. Giusti, Giuliana and Turano, Giuseppina 2007. ‘Case assignment in the pseudoÂ�partitives of Standard Albanian and Arbëresh:€a case for micro-variation’, Quaderni del Â�Dipartimento di Linguistica dell’Università di Firenze 17:€33–51.
336â•… References Grimshaw, Jane 1979. ‘Complement selection and the lexicon’, Linguistic Inquiry 10:€279–326. Haegeman, Liliane 2007. ‘Operator movement and topicalization in adverbial clauses’, Folia Linguistica 41:€279–325. â•… 2009. ‘The movement derivation of conditional clauses’, ms., University Â� of Ghent. Haegeman, Liliane and Zanuttini, Raffaella 1991. ‘Negative heads and the NEG criterion’, The Linguistic Review 8:€233–51. Haiman, John 1988. ‘Rhaeto-Romance’, in M. Harris and N. Vincent (eds.), The Â�Romance Languages. London:€Routledge, pp. 351–90. Hale, Ken and Keyser, Samuel J. 1993. ‘On argument structure and the lexical expression of syntactic relations’, in K. Hale and S. J. Keyser (eds.), The View from Building 20. Cambridge, Mass.:€MIT Press, pp. 53–109. Halle, Morris and Marantz, Alec 1993. ‘Distributed morphology and the pieces of Â�inflection’. In K. Hale and S. J. Keyser (eds.), The View from Building 20. Cambridge, Mass.:€MIT Press, pp. 111–76. Halle, Morris and Vaux, Bert 1997. ‘Theoretical aspects of Indo-European nominal morphology:€ the nominal declension of Latin and Armenian’, in J. Jasanoff, C. Melchert and L. Olivier (eds.), Mir Curad. A Festschrift in Honor of Calvert Watkins, University of Innsbruck. Harley, Heidi and Ritter, Elizabeth 2002. ‘Person and number in pronouns:€a featuregeometric analysis’, Language 78:€482–526. Harris, James and Halle, Morris 2005. ‘Unexpected plural inflections in Spanish:€reduplication and metathesis’, Linguistic Inquiry 36:€195–222. Haspelmath, Martin 1997. Indefinite Pronouns. Oxford:€Clarendon Press. Hauser, Marc, Chomsky, Noam and Fitch, W. Tecumseh 2002. ‘The faculty of language:€what is it, who has it and how did it evolve?’, Science 298:€1569–79. Heim, Irene 1982. The Semantics of Definite and Indefinite Noun Phrases. GSLA, Dept. of Linguistics, University of Massachussets at Amherst. Hespos, Susan and Spelke, Elizabeth 2004. ‘Conceptual precursors to language’,â•› Nature 430:€453–6. Higginbotham, James 1985. ‘On semantics’, Linguistics Inquiry 16:€547–93. Hoekstra, Teun and Mulder, René 1990. ‘Unergatives as copular verbs:€locational and existential predication’, The Linguistic Review 7:€1–79. Holton, David, Mackridge, Peter and Philippaki-Warburton, Irene 1997. Greek Grammar:€a Comprehensive Grammar of the Greek Language. London:€Routledge [2nd edition, 1999]. Horn, Laurence 1989. A Natural History of Negation. Chicago University Press. Hornstein, Norbert 1999. ‘Movement and control’, Linguistic Inquiry 30:€69–96. Jackendoff, Ray 1972. Semantic Interpretation in Generative Grammar. Cambridge, Mass.:€MIT Press. â•… 1994. Patterns in the Mind:€Language and Human Nature. New York: Basic Books. â•… 2002. Foundations of Language. Oxford University Press. Johnson, Kyle 1991. ‘Object positions’, Natural Language and Linguistic Theory 9, 577–637. Johnston, Jason 1997. Systematic Homonymy and the Structure of Morphological Categories:€ Some Lessons from Paradigm Geometry. Ph.D. Dissertation, University of Sydney.
Referencesâ•… 337 Jones, Michael 1993. Sardinian Syntax. London:€Routledge. Kallulli, Dalina 2006. ‘A unified analysis of passives, anticausatives and reflexives’, in O. Bonami and P. Cabredo Hofherr (eds.), Empirical Issues in Formal Syntax and Semantics 6, pp. 201–25. Kayne, Richard 1975. French Syntax:€ the Transformational Cycle. Cambridge, Mass.:€MIT Press. â•… 1976. ‘French relative “que”’, in F. Hensey and M. Luján (eds.), Current Studies in Romance Linguistics, Washington, DC:€ Georgetown University Press, pp. 255–99. â•… 1984. Connectedness and Binary Branching. Dordrecht:€Foris. â•… 1991. ‘Romance clitics, verb movement and PRO’, Linguistic Inquiry 22:€647–86. â•… 1993. ‘Toward a modular theory of auxiliary selection’, Studia Linguistica 47:€3–31. â•… 1994. The Antisymmetry of Syntax. Cambridge, Mass.:€MIT Press. â•… 2006. ‘On the parameters and on principles of pronunciation’, in H. Broekhuis, N. Corver, R. Huybregts, U. Kleinhenz and J. Koster (eds.), Organizing Grammar:€ Linguistic Studies in Honor of Henk van Riemsdijk. Berlin:€ Mouton de Gruyter, pp. 289–99. â•… 2008a. ‘Expletives, datives and the tension between morphology and syntax’, in M. T. Biberauer (ed.), The Limits of Syntactic Variation. Amsterdam:€Benjamins, pp. 175–217. â•… 2008b. ‘A note on auxiliary alternations and silent causation’, ms., New York Â�University. â•… 2009. Some silent first person plurals. Merging Features. Oxford Scholarship Online Monographs, pp. 276–93. â•… 2010. ‘Why isn’t this a complementizer?’, in Comparisons and Contrasts. Oxford University Press. Kayne, Richard and Pollock, Jean-Yves 2001. New thoughts on stylistic inversion’, in A. Hulk and J.-Y. Pollock (eds.), Subject Inversion in Romance and the Theory of Universal Grammar. Oxford University Press, pp. 107–62. Kiparsky, Paul 1982. Explanation in Phonology. Dordrecht:€Foris. Kiparsky, Paul and Kiparsky, Carol 1970. ‘Fact’, in M. Bierwish and K. E. Eidolph (eds.), Progress in Linguistics. The Hague:€Mouton, pp. 143–73. Klamer, Marian 2000. ‘How report verbs become quote markers and complementisers’, Lingua 110:€69–98. Kratzer, Angelika 2000. ‘Building statives’, paper delivered at the Berkeley Linguistic Society 26, http://semanticsarchive.net/Archive/GI5MmI0M/kratzer.building. statives.pdf. Labov, William 1994. Principles of Linguistic Change:€Internal Factors. Oxford:€Blackwell. La Fauci, Nunzio and Loporcaro, Michele 1989. ‘Passifs, avancements de l’objet indirect et formes verbales périphrastiques dans le dialecte d’Altamura (Pouilles)’, Rivista di linguistica, 1:€161–96. Laka, Itziar 1990. Negation in Syntax:€ on the Nature of Functional Categories and Projections. Ph.D. Dissertation, MIT, Cambridge, Mass. Larson, Richard 1985. ‘On the syntax of disjunction scope’, Natural Language and Linguistic Theory 3:€217–64.
338â•… References â•… 1987. ‘“Missing prepositions” and the analysis of English free relative clauses’, Linguistic Inquiry 18:€239–66. â•… 1988. ‘On the double object construction’, Linguistic Inquiry 19:€335–91. Lebeaux, David 1984. ‘Anaphoric binding and the definition of PRO’, in C. Jones and P. Sells (eds.), Proceedings of NELS 14, GLSA, University of Massachusetts, Amherst, Mass. Ledgeway, Adam 2003a. ‘Il sistema completivo dei dialetti meridionali:€la doppia serie di complementatori’, Rivista Italiana di Dialettologia 27:€89–147. â•… 2003b. ‘L’estensione dell’ausiliare perfettivo avere nell’antico napoletano:€ Intransitività scissa condizionata da fattori modali’, Archivio Glottologico Italiano 88:€27–71. â•… 2005. ‘Moving through the left periphery:€ the dual complementiser system in the dialects of Southern Italy’, Transactions of the Philological Society 103:€339–96. â•… 2009. ‘Aspetti della sintassi della periferia sinistra del cosentino’, in D. Pescarini (ed.), Studi sui dialetti della Calabria (Quaderni di lavoro ASIt 9), pp. 3–24. Legendre, Geraldine 2007. ‘Optimizing auxiliary selection in Romance’, in R. Aranovich (ed.), Split Auxiliary Systems. Amsterdam:€John Benjamins, pp. 145–80. â•… 2010. ‘A formal typology of person-based auxiliary selection in Italo-Romance’, in R. D’Alessandro, A. Ledgeway and I. Roberts (eds.), Syntactic Variation:€the Dialects of Italy. Cambridge University Press, pp. 86–101. Lenneberg, Eric H. 1967. Biological Foundations of Language. New York:€Wiley. Lewis, David 1975. ‘Adverbs of quantification’, in E. Keenan (ed.), Formal Semantics of Natural Language. Cambridge University Press, pp. 3–15. Lieberman, Philip 1991. Uniquely Human:€the Evolution of Speech, Thought, and Selfless Behaviour. Cambridge, Mass.:€Harvard University Press. Löbner, Sebastian 1989. ‘German schon€– erst€– noch:€an integrated analysis’, Linguistics and Philosopy 12:€167–212. Longobardi, Giuseppe 1992. ‘In defence of the correspondence hypothesis:€ island effects and parasitic constructions in Logical Form’, in J. Huang and R. May (eds.), Logical Structure and Linguistic Theory. Dordrecht:€ Kluwer, pp. 149–96. â•… 1994. ‘Reference and proper names:€a theory of N-movement in syntax and Logical Form’, Linguistic Inquiry 25:€609–65. â•… 1996. ‘The syntax of N-raising:€a minimalist theory’, OTS Working Papers in Linguistics. Utrecht University. â•… 2000. ‘The structure of DPs:€some principles, parameters and problems’, in M. Baltin and C. Collins (eds.), Handbook of Contemporary Syntactic Theory. Oxford: Blackwell. â•… 2001. ‘Formal syntax, diachronic minimalism and etymology:€the history of French chez’, Linguistic Inquiry 32:€275–302. Loporcaro, Michele 1988. Grammatica storica del dialetto di Altamura. Pisa:€Giardini. â•… 2001. ‘La selezione degli ausiliari nei dialetti italiani:€ Dati e teorie’, in F. Albano Leoni, E. Stenta Krosbakken, R. Sornicola and C. Stromboli (eds.), Dati empirici e teorie linguistiche. Atti del XXXIII congresso internazionale di studi della SLI. Rome:€Bulzoni, pp. 455–76. â•… 2007. ‘On triple auxiliation in Romance’, Linguistics 45:€173–222.
Referencesâ•… 339 McCloskey, James 2004. ‘Questions and questioning in a local English’, in R. Zanuttini, H. Campos, E. Herburger and P. Portner (eds.),â•› Negation, Tense and Clausal Architecture:€ Cross–Linguistic Investigations. Georgetown University Press, pp. 87–126. McFadden, Thomas 2007. ‘Auxiliary selection’, Language and Linguistics Compass 1:€674–708. MacSwan, Jeff 2000. ‘The architecture of the bilingual language faculty:€evidence from intrasentential code switching’, Bilingualism:€Language and Cognition 3:€37–54. Manzini, M. Rita 1982. ‘Italian prepositions before infinitives’. MIT Working Papers in Linguistics 3:€115–22. â•… 1986. ‘On Italian si’, in H. Borer (ed.), The Syntax of Pronominal Clitics (Syntax and Semantics vol. 18). New York:€Academic Press, pp. 241–62. â•… 2000. ‘Sentential complementation:€the subjunctive’, in P. Coopmans, M. Everaert and J. Grimshaw (eds.), Lexical Specification and Insertion. Amsterdam:€ John Benjamins, pp. 241–67. â•… 2009. ‘Pro, pro and NP-trace (raising) are interpretations’, in K. Grohmann (ed.), Phase Theory:€Features, Arguments, Interpretations. Amsterdam/Oxford:€Elsevier, pp. 131–80. Manzini, M. Rita and Roussou, Anna 2000. ‘A minimalist theory of A-movement and control’, Lingua 110:€409–47. Manzini, M. Rita, Roussou, Anna and Savoia, Leonardo M. forthcoming ‘Voice morphology in Albanian and Greek’, in S. Özsoy and A. Gürel (eds.), Current Issues in Mediterranean Syntax. Manzini, M. Rita and Savoia, Leonardo M. 1998. ‘Clitics and auxiliary choice in Italian dialects:€their relevance for the Person ergativity split’, Recherches Linguistiques de Vincennes 27:€115–38. â•… 2002. ‘Negative adverbs are neither Adv nor Neg’, in Masako Hirotani (ed.), Proceedings of North East Linguistic Society 32, GLSA, University of Massachusetts, Amherst, pp. 327–46. â•… 2003. ‘The nature of complementizers’, Rivista di Grammatica Generativa 28:€87– 110. â•… 2005. I dialetti italiani e romanci. Morfosintassi generativa, Alessandria:€Edizioni dell’Orso, 3 vols. â•… 2007. A Unification of Morphology and Syntax:€Studies in Romance and Albanian Dialects. London:€Routledge. â•… 2008a. Work Notes on Romance Morphosyntax:€Appunti di morfosintassi romanza, Alessandria, Edizioni dell’Orso. â•… 2008b. ‘Uninterpretable features are incompatible in morphology with other minimalist postulates’, in R. Freidin, C. Otero and M. L. Zubizarreta (eds.), Foundational Issues in Linguistic Theory. Cambridge, Mass.:€MIT Press, pp. 43–72. â•… 2008c. ‘Non-active voice in Albanian:€implications for the theory of movement’, in L. M. Savoia, Studi sulle varietà arbëreshe, Università degli Studi della Calabria, Rende. â•… 2009a. ‘Mesoclisis in the imperative:€ phonology, morphology or syntax?’, in V. Moscati and E. Servidio (eds.), Proceedings XXVIII Incontro di Grammatica Generativa, Studies in Linguistics on line 3, Università di Siena, pp. 51–76.
340â•… References â•… 2009b. ‘Morphology dissolves into syntax:€ infixation and doubling in Romance languages’, Annali online dell’Università di Ferrara. Sezione di Lettere IV, 1:€1–28.n. â•… 2010. ‘Syncretism and suppletivism in clitic systems:€underspecification, silent clitics or neither?’, in R. D’Alessandro, A. Ledgeway and I. Roberts (eds.), Syntactic Variation:€the Dialects of Italy. Cambridge University Press, pp. 86–101. â•… to appear. ‘Reducing “case” to “agreement”:€nominal inflections in the Geg Albanian variety of Shkodër’, Linguistic Variation Yearbook. Manzini, M. Rita and Wexler, Kenneth 1987. ‘Binding theory, parameters and learnability’, Linguistic Inquiry 18:€413–44. Marantz, Alec 1984. On the Nature of Grammatical Relations. Cambridge, Mass.:€MIT Press. â•… 1997. ‘No escape from syntax:€don’t try morphological analysis in the privacy of your own lexicon’, in A. Dimitriadis, L. Sieger, Cl. Surek-Clark and A. Williams (eds.) Proceedings of the 21st Annual Penn Linguistics Colloquium, U. Penn Working Papers in Linguistics 4.2, pp. 201–25. Mensching, Guido and Remberger, Eva-Maria 2010. ‘Focus fronting and the left periphery in Sardinian’, in R. D’Alessandro, A. Ledgeway and I. Roberts (eds.), Syntactic Variation:€the Dialects of Italy. Cambridge University Press, pp. 261–76. Meyer-Lübke, Wilhelm 1899. Grammatik der Romanischen Sprachen. Dritter Band:€Syntax. Leipzig. Michaelis, Laura 1996. ‘On the use and meaning of already’, Linguistics and Philosophy 19:€477–502. Mioto, Carlos 2003. ‘Focalização e quantificação’, Revista Letras 61:€169–89. Moro, Andrea 1997. The Raising of Predicates. Cambridge University Press. Newmeyer, Frederick 2005. ‘On split CPs, uninterpretable features and the “perfectness” of language’, ZAS Papers in Linguistics 35, pp. 399–422. Nichols, Johanna 1992. Linguistic Diversity in Space and Time. Chicago University Press. Nilsen, Øystein 2003. Eliminating Positions, Ph.D. Dissertation, University of Utrecht. Origgi, Gloria and Sperber, Dan 2000. ‘Evolution communication and the proper function of language’, in P. Carruthers and A. Chamberlain (eds.), Evolution and the Human Mind:€Language, Modularity and Social Cognition. Cambridge University Press, pp. 140–69. Paoli, Sandra 2007. ‘The fine structure of the left periphery:€ COMPs and Subjects. Evidence from Romance’, Lingua 117:€1057–79. â•… 2010. ‘In focus:€ an investigation of information and contrastive constructions’, in R. D’Alessandro, A. Ledgeway and I. Roberts (eds.), Syntactic Variation:€the Dialects of Italy. Cambridge University Press, pp. 277–91. Parry, M. Mair 1995. ‘Some observations on the syntax of clitic pronouns in Piedmontese’, in M. Maiden and J. C. Smith (eds.), The Romance Languages and Contemporary Linguistic Theory. Amsterdam:€Benjamins, pp. 133–60. â•… 1997. ‘Preverbal negation and clitic ordering, with particular reference to a group of North-West Italian dialects’, Zeitschrift für romanische Philologie 113: 244–70.
Referencesâ•… 341 Parsons, Terence 1990. Events in the Semantics of English. Cambridge, Mass., MIT Press. Pesetsky, David 1982. Paths and Categories. Ph.D. Dissertation, MIT, Cambridge, Mass. â•… 1985. ‘Morphology and logical form’, Linguistic Inquiry 16: 193–246. â•… 1991. Zero Syntax vol. 2:€Infinitives. Ms., MIT, Cambridge, Mass. â•… 2008. ‘Russian case morphology and the syntactic categories’, paper delivered at the Mediterranean Syntax Meeting€ – II, Bogazici University, Istanbul, 16 Â�October. Pesetsky, David and Torrego, Esther 2007. ‘The syntax of valuation and the interpretability of features’, in S. Karimi, V. Samiian and W. Wilkins (eds.), Phrasal and Clausal Architecture. Amsterdam:€Benjamins, pp. 262–94. Philippaki-Warburton, Irene 1973. ‘Modern Greek verb conjugation:€inflectional morphology in a transformational grammar’, Lingua 32:€193–226. Poletto, Cecilia 2000. The Higher Functional Field. Oxford University Press. â•… 2001. ‘Complementizer deletion and verb movement in standard Italian’, in G. Cinque and G. Salvi (eds.), Current Studies in Italian Syntax. Amsterdam:€North Holland, pp. 265–86. Poletto, Cecilia and Vanelli, Laura 1995. ‘Gli introduttori delle frasi interrogative nei dialetti italiani settentrionali’, in E. Banfi, G. Bonfadini, P. Cordin and M. Iliescu (eds.), Italia settentrionale:€crocevia di idiomi romanzi. Tübingen:€Niemeyer, pp. 145–58. Pollock, Jean-Yves 1989. ‘Verb movement, universal grammar and the structure of IP’, Linguistic Inquiry 20:€365–424. Polinsky, Maria and Potsdam, Eric 2006. ‘Expanding the scope of control and raising’, in B. Davies and S. Dubinsky (eds.), Special Issue on Raising and Control. Syntax 9:€171–9. Postal, Paul 1974. On Raising. Cambridge, Mass.:€MIT Press. Pulvermüller, Friedemann 1999. ‘Words in the brain’s language’, Behavioral and Brain Sciences 22:€253–336. Pylkkänen, Liina 2002. Introducing Arguments. Ph.D. Dissertation, MIT, Cambridge, Mass. Ralli, Angeliki 1988. Eléments de la morphologie du Grec Moderne:€La structure du verbe. Doctoral dissertation, University of Montreal. â•… 2005. Morfologia. Athens:€Patakis. Reinhart, Tanya 1997. Syntactic Effects of Lexical Operations:€Reflexives and Unaccusatives. OTS Working Papers in Linguistics, Utrecht University. Reinhart, Tanya and Siloni, Tali 2005. ‘The Lexicon-Syntax parameter:€reflexivization and other arity operations’, Linguistic Inquiry 36:€389–436. Renzi, Lorenzo 1987. ‘Le développement de l’article en roman’, in C. Buridant (ed.), Romanistique€ – Germanistique:€ une confrontation. University of Â�Strasbourg, pp. 297–317. Rivero, María Luisa 1994. ‘Clause structure and V-movement in the languages of the Balkans’, Natural Language and Linguistic Theory 12:€63–120. Rizzi, Luigi 1982. Issues in Italian Syntax. Dordrecht:€Foris. â•… 1986. ‘On the status of subject clitics in Romance’, in O. Jaeggli and C. Silva-Â�Corvalan (eds.), Studies in Romance Linguistics. Dordrecht:€Foris, pp. 391–419.
342â•… References â•… 1990. Relativized Minimality. Cambridge, Mass.:€MIT Press. â•… 1996. ‘Residual verb second and the wh–criterion’, in A. Belletti and L. Rizzi (eds.), Parameters and Functional Heads. Oxford University Press, pp. 63–90. â•… 1997. ‘The fine structure of the left periphery’, in L. Haegeman (ed.), Elements of Grammar. Dordrecht:€Kluwer, pp. 281–337. â•… 2001. ‘On the position “Int(errogative)” in the left periphery of the clause’, in G. Cinque and G. Salvi (eds.), Current Studies in Italian Syntax. Amsterdam:€North Holland, pp. 287–96. â•… 2004. ‘Locality and left periphery’, in A. Belletti (ed.), Structures and Beyond. Â�Volume 3:€ The Cartography of Syntactic Structures. Oxford University Press, pp. 223–51. Roberts, Ian 1993. ‘The nature of subject clitics in Franco-Provençal Valdotain’, in A. Belletti (ed.), Syntactic Theory and the Dialects of Italy. Turin:€Rosenberg & Sellier, pp. 319–53. Roberts, Ian and Roussou, Anna 2003. Syntactic Change. Cambridge University Press. Rohlfs, Gerhard 1969[1954]. Grammatica storica della lingua italiana e dei suoi dialetti. Sintassi e formazione delle parole. Turin:€Einaudi. â•… 1977. Le guascon. Tübingen:€Niemeyer. Ronjat, Jules 1937. Grammaire Istorique des Parlers Provençaux Modernes. t.III. Montpellier:€Société des Langues Romanes. Rooryck, Johan 2009. ‘A compositional analysis of French negation’, ms., University of Leiden. Rosenbaum, Peter 1967. The Grammar of English Predicate Complement Constructions. Cambridge, Mass.:€MIT Press. Roussou, Anna 1994. The Syntax of Complementizers. Ph.D. Dissertation, University College London. â•… 2008. ‘Voice morphology and ergativity in Modern Greek’. Proceedings of the 8th International Conference on Greek Linguistics (ICGL8), University of Ioannina. â•… 2009. ‘In the mood for control’, in J. Quer (ed.), Lingua 119 (Special Issue on Mood):€1811–36. â•… 2010. ‘Subjects on the edge’, in Ph. Panagiotidis (ed.), The Complementizer Phase. Oxford University Press, pp. 76–116. Saccon, Graziella 1992. VP-internal arguments and locative subjects. Proceedings of NELS 22, GLSA, University of Massachusetts, Amherst, pp. 383–97. Santorini, Beatrice 1989. The Generalization of the Verb-Second Constraint in the Â�History of English. Ph.D. Dissertation, University of Pennsylvania, Philadelphia. Savoia, Leonardo M. 2008. Studi sulle varietà arbёreshe. Università della Calabria, Rende. Savoia, Leonardo M. and Manzini, M. Rita 2007. ‘Variazione sintattica nel costrutto ausiliare Arbёresh:€la variazione come problema teorico’, in Minoranze Linguistiche. Prospettive, strumenti, territori. Rome:€Carocci, pp. 85–102. â•… 2010. ‘Lexicalization of 3rd person object clitics:€clitic enclisis and clitic drop’, in R. D’Alessandro, A. Ledgeway and I. Roberts (eds.), Syntactic Variation:€the Dialects of Italy. Cambridge University Press:€102–18. Schmid, Heinrich 1951/52. ‘Zur Geschichte der rätoromanischen Deklination’, Vox Romanica 12:€21–81.
Referencesâ•… 343 Schwartz, Bonnie and Vikner, Sten 1996. ‘The verb always leaves IP in V2 clauses’, in A. Belletti and L. Rizzi (eds.), Parameters and Functional Heads. Oxford University Press, pp. 11–62. Sgrilli, Paola 1983. Il “Libro di Sidrac” salentino. Pisa:€Pacini. Shlonsky, Ur 1994. ‘Agreement in comp’, The Linguistic Review 11:€351–75. Solano, Francesco 1972. Manuale di lingua albanese. Corigliano Calabro. Sorace, Antonella 2000. ‘Gradients in auxiliary selection with intransitive verbs’, Language 76:€859–90. â•… 2004. ‘Gradience at the lexicon-syntax interface:€evidence from auxiliary selection and implications for unaccusativity’, in A. Alexiadou, E. Anagnostopoulou and M. Everaert (eds.), The Unaccusativity Puzzle:€ Explorations of the Syntax-Lexicon Interface. Oxford University Press, pp. 243–87. Sperber, Dan and Wilson, Deirdre 1986. Relevance. Oxford:€Blackwell. Starke, Michal 2004. ‘On the inexistence of Specifiers and the nature of heads’, in A. Belletti (ed.) Structures and Beyond. Oxford University Press. â•… 2009. ‘Nanosyntax:€a short primer to a new approach to language’, Nordlyd (Tromsoe University Working Papers in Language and Linguistics) 36.1:€1–6. Stowell, Timothy 1981. Origins of Phrase Structure. Ph.D. Dissertation, MIT, Cambridge, Mass. Svenonius, Peter 2002. ‘Subject positions and the placement of adverbials’, in P. Svenonius (ed.), Subjects, Expletives and the EPP. Oxford University Press, pp. 201–42. de Swart, Henriëtte 1993. Adverbs of Quantification. New York:€Garland. Szabolcsi, Anna 1994. ‘The noun phrase’, in K. Kiss and F. Kiefer (eds.), The Syntactic Structure of Hungarian. Syntax and Semantics 27. New York:€Academic Press, pp. 179–274. Taraldsen, K. Tarald 1978. ‘On the NIC, vacuous application, and the that-t Filter’, ms., MIT, Cambridge, Mass. â•… 1990. ‘D-projections and N-projection in Norwegian’, in J. Mascaró and M. Nespor (eds.), Grammar in Progress:€ Glow Essays for Henk van Riemsdijk. Dordrecht :€Foris, pp. 419–31. â•… 2002. ‘The que/qui alternation and the distribution of expletives’, in P. Svenonius (ed.) Subjects, Expletives and the EPP. Oxford University Press, pp. 29–42. Terzi, Arhonto and Wexler, Ken 2002. ‘A-chains and S-homophones in children’s grammar’, in M. Hirotani (ed.) Proceedings of NELS 32, GLSA; University of Massachusetts at Amherst, pp. 519–37. Torres Morais, M. Aparecida and Salles, Heloisa to appear. ‘Parametric change in the grammatical encoding of indirect objects in Brazilian Portuguese’, Probus. Tortora, Christina 2002. ‘Romance enclisis, prepositions, and aspect’, Natural Language and Linguistic Theory 20:€725–57. Trommer, Jochen 2005. ‘Closest c-command in Albanian non-active constructions’, Â�paper presented at Morphosyntax Workshop, University of Vienna. Turano, Giuseppina 2002. ‘On modifiers preceded by the article in Albanian DPs’, Â�University of Venice Working Papers in Linguistics 12:€169–215. â•… 2003. ‘Similarities and differences between standard Albanian and Arbëresh numerals:€A case of micro-parametric variation’, University of Venice Working Papers in Linguistics, 13:€155–77.
344â•… References Vincent, Nigel 1997. ‘The emergence of the D-system in Romance’, in A. van Kemenade and N. Vincent (eds.), Parameters of Morphosyntactic Change. Cambridge University Press, pp. 149–69. Wasow, Thomas 1977. ‘Transformations and the lexicon’, in P. Culicover, T. Wasow and A. Akmajian (eds.), Formal Syntax. New York:€Academic Press, pp. 327–60. Weiss, Helmut 2002. ‘Three types of negation:€a case study in Bavarian’, in S. Barbiers, L. Cornips and S. van dee Kleij (eds.), Syntactic Variation. Meertens Institut Electronic Publications in Linguistics. Whitman, John 2000. ‘Relabelling’, in S. Pintzuk, G.Tsoulas and A. Warner (eds.), Diachronic Syntax:€Models and Mechanisms. Oxford University Press, pp. 220–38. Zamboni, Alberto 1990. ‘Premesse morfologiche e tipologiche del composto italiano “capinera”, “pettirosso”’. In Parallela 4:€Morfologia/ Morphologie. Tübingen:€Â�Niemeyer, pp. 97–109. â•… 1998. ‘Cambiamento di lingua o cambiamento di sistema? Per un Â�bilancio cronologico della transizione’, in J. Herman (ed.), La transizione dal latino alle lingue romanze. Tübingen:€Niemeyer, pp. 99–127. Zanuttini, Raffaella 1997. Negation and Clausal Structure. New York:€Oxford Â�University Press.
Index
ablative, 240, 283 syncretism with dative, 283 accusative, 191, 240, 277 as internal argument, 253 syncretism with nominative, 251, 284 adjective not sensitive to case, 268 preceded by article, 266 adverb aspectual, 80, 94, 108, 109, 110, 111, 118, 119, 125 aspectual/manner, 107 aspectual/quantificational, 102, 121 deictic/quantificational, 125 follows participle, 124 hierarchy, 82 manner, 113, 114, 117, 118, 119, 125 order, 108, 116, 120 position, 116, 120, 126 precedes participle, 124, 125 quantificational, 111, 112, 117, 119, 125 Agliano (Tuscany), 137 agreement, 11 as cheking, 238 as sharing of referential properties, 238 in copular construction, 268 in NP, 269 Aliano (Lucania), 225 Àllai (Sardinia), 36 ambiguity is not necessarily resolved syntactically, 107 anticausative interpretation, 216 vs. reflexive, 216 Applicative, 294, 328 arbitrarization, 182 Arconate (Lombardy), 116 Ardaùli (Sardinia), 65
Arena (Calabria), 20, 51, 64 article and inflection, 287 as K, 276 from case to, 277 postposed, 247, 287 with adjectives, 266, 270 with kinship terms, 242, 270, 271 auxiliary, 10 be independent of transitivity and voice, 201 have independent of transitivity and voice, 201 syncretism of have and be, 228, 232 triple, 227 auxiliary selection (non-)reversible, 203, 222, 223 according to person, 201 according to person and transitivity, 225 according to transitivity, 218 according to voice, 216 in middle-passive, 198 Auxiliary Selection Hierarchy (ASH), 223 Bagolino (Lombardy), 320 base bare, 285 verbal, 190 be as defective predicate, 222 auxiliary and copula, 222 Bedizzano (Tuscany), 138, 141 biolinguistic perspective, 1 biolinguistic program, 7 Briga Novarese (Piedmont), 228, 229, 230 Buonabitacolo (Campania), 227 Buru, 34
345
346â•… Index C, 24 as a verbal position, 17 as position of interrogative modality, 26 domain, 141 Finiteness, 41, 61, 313 Focus, 41 Force, 41, 61, 312 recursion, 28–30 Cantoira (Piedmont), 24, 27, 95, 96, 105, 108, 110, 112, 114, 115 Càrcare (Liguria), 143, 144, 151 cartography, 10, 37–8 Casaccia (Grisons, Val Bregaglia), 314 case, 10, 11 (in)sensitivity of adjectives to, 272 abstract, 299 as a feature, 244 as argument, 269 as denotational properties, 277 as descriptive label, 255 as N, Q and D specifications, 286 as nominal class, 255 as primitive category, 277 as quantificational closure, 290 as referential closure, 294 as Tense, 245 constraints, 298, 300 in the absence of inflections, 256 legible at LF interface, 236 loss of, 276, 290 of clitics, 309, 310 of pronouns, 308 Case Resistance Principle, 30 Casei Gerola (Lombardy), 86 Càsola (Tuscany), 134, 135 Casorezzo (Lombardy), 86, 99, 106, 116 Castellazzo Bormida (Piedmont), 22, 23, 24, 26, 27, 95, 105, 109 Castellinaldo (Piedmont), 124 Castiglione d’Adda (Lombardy), 28, 101, 105 categorization, 4–5 Celle di Bulgheria (Campania), 122 Cerano (Piedmont), 23, 85 chain, 11, 153, 206 in doubling, 154 change, 298, 301 Civate (Lombardy), 77 Civita (Calabria, Arbëresh), 172 clause-typing, 29 clitic doubling, 153
Colledimacine (Abruzzi), 225 Colonnata (Tuscany), 139, 140 comparatives, 62 complementation as relativization, 65 complementizer, 9, 10, 12 and subjunctive, 52–4, 59 as head of noun phrase, 17 as relative pronoun, 18–19 coinciding with demonstrative, 15, 34 coinciding with verb, 35 coinciding with wh–, 15, 55, 58, 59, 61, 62 definite, 55, 57, 61, 63, 65, 71 deletion, 56–9, 316 factive, 32 functional projection of the verb, 14 generalized wh–, 63 indefinite, 55, 61, 63, 65, 69, 71 interrogative, 69 interrogative/hypothetical, 65, 66, 67, 68 matrix, 36 modal, 72 polarity, 71 vs. wh–phrase, 73, complementizer agreement, 314 complex NP islands, 31–2 concept, 6 concepts, 4, 8 geometrical, 5 Corte (Veneto), 40 Cravagliana (Piedmont), 86 cycle, 11 dative, 240 as Q, 252 as superset-of, 283, 293 inherent possession, 251 singular syncretic with non-oblique plural, 311 syncretism with ablative, 283 syncretism with genitive, 250, 265, 280, 283, 288, 293 default, 61 an inadequate notion, 281 definite declension, 239, 290 Dego (Liguria), 145, 146, 156 Determiner, 11, 238, 271 as referential closure, 294
Indexâ•… 347 combines with definite inflection, 246 development of, 290 in AP, 271 in genitive, 288, 291 preadjectival, 271 discourse-anchoring, 90, 136, 138, 141, 210, 309 Distributed Morphology, 3, 7, 12, 137, 142, 229, 236, 245, 246, 260, 281, 296, 297, 299, 300, 301 ECP, 86 enclisis, 85 proclisis alternations, 172 English, 322 EPP argument, 93, 206, 207, 253, 282 as a variable, 75 as D, 238 impersonal, 181 in participial sentences, 216 morphological-level, 75 realized by verb inflection, 74 syntactic-level, 75 verb inflection as, 178 ergativity split according to aspect, 188 according to person, 209 event domain, 82 elementary, 93, 112, 119 structure, 90 event-anchoring, 90, 136, 138, 141, 210 Exceptional Case Marking, 253 factive islands, 32 sentences, 31 faculty of language broad (FLB), 1, 8 narrow (FLN), 1, 8 feature checking, 11 uninterpretable, 11 unvalued, 11 Filattiera (Tuscany), 27 Finale Emilia (Emilia), 84 Firenze (Tuscany), 36 focus, 19, 22, 41, 59, 315 as interpretation, 42 Fontane (Piedmont), 124, 155
Fontanigorda (Liguria), 43, 73 Forno Valle Strona (Piedmont), 204 French, 46, 85 functional category, 4, 5, 7, 11, 12, 235 hierarchy, 12 lexicon, 8 gender as nominal class, 238 genitive, 240, 277 as superset-of, 265, 283, 293 introduced as predication, 264 introduced by an article, 262 scope of, 282 syncretism with dative, 250, 265, 280, 288, 293 syncretism with nominative, 281 under negation, 87 Gizzeria (Calabria), 113 Gjirokastër (Albanian), 160, 161, 163, 164, 165, 166, 167, 168, 173, 190, 197, 198, 200, 208, 216, 217, 218, 221 grammatical categories, 8, 9 functions, 9 grammaticalization, 12, 35–6, 84 Greek, 186, 193 Guglionesi (Molise), 50, 60, 63 hanging topic, 47 have as non-defective predicate, 223 denotes set inclusion, 231 deontic, 222 derived from be, 231 hierarchy as scope, 237 homophony vs. single lexical entry, 87 I domain, 141 –i inflection, 289 as Q closure, 285 distribution, 280 I-language, 1, 6 Impoverishment rule, 282 Inclusiveness, 61
348â•… Index indefinite declension, 239 infinitive followed by negation, 126 languages without, 51 preceded by negation, 121 innateness, 5 instrumental, 283 Interface Uniformity, 6, 176, 178, 188, 218 inversion complex, 45 Italian, 16, 19, 31–2, 36, 39, 42, 43, 47, 52, 56, 57, 65, 68, 71, 72, 74, 76, 88, 125, 158, 182, 196, 197, 219, 221, 222, 223, 308, 315, 316, 321, 323, 328 kinship terms, 242, 246 l morphology as definiteness, 290 as inflection, 307 mutual exclusion, 292 La Pli de Mareo (Alto Adige/ South Tyrol), 20, 84, 316 Làconi (Sardinia), 60, 66, 76, 221 Late Insertion, 229, 230, 236 Latin, 278, 281 left periphery, 41, 45, 59, 314 above complementizer, 21–3 below and above complementizer, 23–5 in noun phrases, 23 under complementizer, 20 lexical base, 238 categories, 4 items, 9 parametrization, 2, 6 terminal, 235, 275, 289, 299 lexicon, 6, 8 functional, 8 functional vs. substantive, 275 LF interface closures at, 259 locative, 283 Luras (Sardinia), 33 Margarita (Piedmont), 97, 98, 110, 112, 114, 115, 123, 127 Marzano Appio (Campania), 113 meaning, 6, 9 Mezzenile (Piedmont), 24, 97, 155
middle-passive, 10 anticausative reading, 179 as defective predicate, 220 as predicate with an open variable, 208 be-participle, 164, 169, 188, 192 clitic, 161 imperfective, 186 impersonal reading, 167, 181 movement analysis, 233 passive reading, 166, 177, 192 perfective, 186 reflexive reading, 165, 177, 180 specialized inflection, 159, 169, 184, 186 split according to aspect, tense, 163, 187 vs. active, 164, 171 with unaccusatives, 167 with unergative verbs, 166 Miglionico (Lucania), 67 Minimalist program, 6, 236, 246, 275, 289, 294, 301 Minimality, 33 minimizer, 84 Modena (Emilia), 39, 44 Molina di Ledro (Trentino), 226 Mombercelli (Piedmont), 102, 103, 105, 108, 109, 110 Montaldo (Piedmont), 83, 97, 104, 106, 108, 109, 111, 123 Montebello Ionico/ Saline Ioniche (Calabria), 209 Montenerodomo (Abruzzi), 50, 64 Morano (Calabria), 314 Morcone (Campania), 202 morphological structure, 12 morphology unification with syntax, 236, 273 Müstair (Grisons), 323 Mustér (Grisons), 1098n, 104, 106 mutual exclusion negation/negative polarity argument, 156 subject/object clitics, 137 Nanosyntax, 299, 301, 302 Neg Criterion, 81 Neg position, 80, 81, 82, 84, 91, 94, 108, 117, 122, 128, 129, 132, 137, 151, 319, 322, 339 negation, 9, 11 negation adverb, 81, 82, 135 and argument structure, 88, 91 and aspectual adverb, 94, 102
Indexâ•… 349 and negative argument, 83, 84, 87 and person split, 91 as a nominal element, 91 as argument, 91 as internal argument, 154 bare noun, 84, 85 complementary distribution with negative arguments and adverbs, 155 interaction with internal argument, 153 introduces a variable, 154 minimizer, 88 nominal, 117 nothing-type, 84 no-type, 98, 115 position, 118, 121 presuppositional vs. non-presuppositional, 94 quantifying over internal argument, 92, 93 negation clitic, 80, 81, 128, 129 and discourse-anchoring, 136 and event-anchoring, 136 and partitive, 151 and person split, 129, 130, 136, 148 and subject clitic, 135, 136, 137 as a negative polarity element, 152 as nominal element, 138, 152 as object clitic, 138 base generated, 130 complementary distribution with accusative, 147 complementary distribution with partitive, 149 distribution, 129, 144 doubling, 130, 139, 145, 153 in C-domain, 143, 148 in I-domain, 143, 148 introduces a variable, 154 two positions of, 133, 141 negative concord, 81, 130, 143, 152, 153 operator, 70, 87, 91, 92, 118, 152 polarity item, 81, 92, 152 polarity item as variable, 87 quantifier, 81 nominal class as denotational element, 277 as internal argument, 252 as sole inflection, 284 in the scope of D/Q, 255, 256, 258, 284 nominative, 191, 240, 277 syncretism with accusative, 251, 284
non-finite verbs, 82 null subject language, 75 Oblique, 260 of its lexical content, 232 Old French, 85, 295, 296, 310 opacity, 298 Optimality Theory, 137 Orsomarso (Calabria), 114 Oviglio (Piedmont), 38, 83, 147, 148, 149, 153 Pamparato (Piedmont), 103, 105, 108, 109, 112, 113 parameter micro- vs. macro-, 234 participle (non-)agreeing, 200, 209 as independent sentence, 123, 203, 204 lacks temporal reference, 123 position, 121 preceded by negation, 121 partitive, 85, 87, 91, 92, 149, 151 selected by negation, 82, 85 passive as chain formation, 177, 178, 179 interpretation, 216 movement analysis, 159 vs. anticausative, impersonal, reflexive, 159 past imperfective, 162, 165, 166, 184, 187 perfective, 162, 166, 169 Paulilàtino (Sardinia), 53, 60, 64, 70, 74 person (ir)reversibility of split, 227 1st plural, 213 2nd plural, 213 discourse-anchored vs. event-anchored, 212 lack of split in the plural, 213 singular vs. plural, 211 split, 82, 90, 209, 210 split under negation, 90 Pescolanciano (Abruzzi), 201 phase, 313 Piverone (Piedmont), 115 pluperfect, 230 plural bare, 86 polarity item, 71 operator, 71
350â•… Index Pomaretto (Piedmont), 95, 105 Portocannone (Molise, Arbëresh), 169, 170, 171, 173, 185, 190, 205 possession as inclusion, 251 possessive construction, 283 pronoun, 271 poverty of stimulus argument, 4 Pozzaglio (Lombardy), 95, 105 Pramollo (Piedmont), 84 predicate projecting referential contents, 237 Premana (Lombardy), 84 prepositions as sentential introducers, 18, 62 as two place predicates, 260 aspectual type categories, 260 license all cases, 259 present, 162, 166, 184, 187 present perfect, 164, 165, 166 Q-floated position, 87 Quarna Sopra (Piedmont), 88 Quarna Sotto (Piedmont), 83, 89, 92, 121 question operator, 57, 69, 70 raising interpretation, 192 readjustment rules, 278, 298 reflexive interpretation, 216 vs. anticausative, 216 relative analysis for conditionals, 69 appositive, 64 clause, 45, 63, 65 complement sentences, 70 restrictive, 64 representational, 6 grammar, 11 rhetorical question, 41 Romanian, 287, 288, 289, 309 Russian, 85 Ruvo di Puglia (Apulia), 72 –s inflection, 284, 295 absent from expletive, 305 as definiteness, 306
as Q, 282, 297, 305 denotational content, 282 S.Angelo Lodigiano (Lombardy), 100, 106 S.Bartolomeo Pesio (Piedmont), 150, 156 S.Benedetto del Tronto (Marche), 202 S.Benedetto Ullano (Calabria, Arbëresh), 172 S.Maria Maggiore (Piedmont), 40 Scuol (Grisons), 20 Secinaro (Abruzzi), 72 selection, 30 s-selection, 313 serial verbs, 34 Shkodёr (Albanian), 1099n, 161, 162, 163, 164, 165, 166, 167, 168, 173, 177, 179, 180, 182, 184, 185, 189, 191, 192 silent categories, 4, 7, 45 Sillano (Tuscany), 132, 152 si-type clitic and person split, 141 as dative, 221 as internal argument, 177, 183 as middle-passive, 207 as pronominal object clitic, 172, 173, 178 as reflexive, 176 as variable, 69, 176 correlates with be, 197 variation, 174, 182 vs. impersonal pronoun, 183 with transitive, 182 Soazza (Grisons), 199, 219 Stroppo/ Macra (Piedmont), 83, 86, 87 subject clitic, 38–40, 43–4 lexical, 38–40, 44, 313 subject clitic inversion, 24–7 subjunctive, 52–4, 57, 70 particle, 52 Subset Principle, 281, 299 Superset Principle, 299 syncretism (non-)accidental, 301 contextual vs. absolute, 296 –t inflection as plural, 258 as superset of, 258 terminal abstract, 275 thematic vowel, 186, 187, 190, 243, 278, 279, 283, 284, 285 theta-theory, 12
Indexâ•… 351 topic, 19, 22, 24, 28, 41, 43, 59 as interpretation, 42 transitive as complex predicate, 93, 208 transitivity, 207, 223 Trecate (Piedmont), 85, 233 Trun (Grisons), 46, 84, 220 unaccusative active, 181 as defective predicate, 220 as elementary event, 93 as predicate with a single argument, 208 unergative as concealed transitive, 167, 208 as incorporating internal argument, 93 uniformity thesis, 2, 5 Urbino (Marche), 307 V2, 313 embedded, 20–1, 24 embedded in questions, 28 vs. complementizer, 58 Vagli di Sopra (Tuscany), 132, 133, 152, 309 variable individual, 64 propositional, 69, 71
variation, 9, 294 language, 2–3 Vastogirardi (Molise), 226 Vella (Grisons), 302, 303, 304, 305, 306 Vena (Calabria, Arbëresh), 241, 242, 243, 259, 263, 264, 266, 267, 268, 270, 272, 273 Verbicaro (Calabria), 201 Viano (Tuscany), 131, 134, 135, 153 Viguzzolo (Piedmont), 26, 28, 100, 101, 105 voice, 207, 223 wh–, 62, 72 argument, 15 determiner, 16 D-linked, 40 introducing exclamative, 16 introducing relative, 16 not intrinsecally interrogative, 68 variable, 63 Wh–Criterion, 39 wh–phrase, 19, 22, 24, 38, 41, 43 word order change, 276 Zernez (Grisons), 315 Zoldo Alto (Veneto), 77