137 108
English Pages 334 [333] Year 2021
Determiners and Quantifiers
Syntax & Semantics Series Editor Keir Moulton (University of Toronto, Canada)
Editorial Board Judith Aissen (University of California, Santa Cruz) – Peter Culicover (The Ohio State University) – Elisabet Engdahl (University of Gothenburg) – Janet Fodor (City University of New York) – Erhard Hinrichs (University of Tubingen) – Paul M. Postal (Scarsdale, New York) – Barbara H. Partee (University of Massachusetts) William A. Ladusaw (University of California, Santa Cruz) – Manfred Krifka (Humboldt-Universität zu Berlin) – Pauline Jacobson (Brown University)
volume 44
The titles published in this series are listed at brill.com/sas
Determiners and Quantifiers Functions, Variation, and Change
Edited by
Chiara Gianollo Klaus von Heusinger Maria Napoli
leiden | boston
Library of Congress Cataloging-in-Publication Data Names: Gianollo, Chiara, editor. | Heusinger, Klaus von, editor. | Napoli, Maria, 1977- editor. Title: Determiners and quantifiers : functions, variation, and change / edited by Chiara Gianollo, Klaus von Heusinger, Maria Napoli. Description: Leiden ; Boston : Brill, 2021. | Series: Syntax & semantics, 00924563 ; 44 | Includes bibliographical references and index. Identifiers: lccn 2021045759 (print) | lccn 2021045760 (ebook) | isbn 9789004473317 (hardback) | isbn 9789004473324 (ebook) Subjects: lcsh: Grammar, Comparative and general–Determiners. | Grammar, Comparative and general–Quantifiers. | Reference (Linguistics) | Romance languages–Determiners. | Romance languages–Quantifiers. | Germanic languages–Determiners. | Germanic languages–Quantifiers. Classification: lcc p299.d48 d38 2021 (print) | lcc p299.d48 (ebook) | ddc 415– dc23/eng/20211105 lc record available at https://lccn.loc.gov/2021045759 lc ebook record available at https://lccn.loc.gov/2021045760
Typeface for the Latin, Greek, and Cyrillic scripts: “Brill”. See and download: brill.com/brill‑typeface. issn 0092-4563 isbn 978-90-04-47331-7 (hardback) isbn 978-90-04-47332-4 (e-book) Copyright 2022 by Chiara Gianollo, Klaus von Heusinger and Maria Napoli. Published by Koninklijke Brill nv, Leiden, The Netherlands, except where stated otherwise. Koninklijke Brill nv incorporates the imprints Brill, Brill Nijhoff, Brill Hotei, Brill Schöningh, Brill Fink, Brill mentis, Vandenhoeck & Ruprecht, Böhlau Verlag and V&R Unipress. Koninklijke Brill nv reserves the right to protect this publication against unauthorized use. Requests for re-use and/or translations must be addressed to Koninklijke Brill nv via brill.com or copyright.com. This book is printed on acid-free paper and produced in a sustainable manner.
Contents Preface vii List of Tables viii Notes on Contributors Abbreviations xi
ix
1 Reference and Quantification in Nominal Phrases: The Current Landscape and the Way Ahead 1 Chiara Gianollo, Klaus von Heusinger and Maria Napoli 2 Definite Plural Generics in English: Evidence from De-adjectival Nominalization 29 Artemis Alexiadou 3 Quantification and Classification in Romance Plural Indefinites: From Number to Seinsart? 55 Mario Squartini 4 Topics and the Interpretation of Referential Null Subjects Manuel Leonetti 5 Specificity and Questions of Specification Edgar Onea 6 Being Bare: A Survey of Quantifier Positions Cecilia Poletto
94
130
186
7 Indefinites as Fossils: The Case of wh-based Free Choice Maria Aloni
214
8 The Evaluative Meaning of the Indefinite qualunque in (Old) Italian Olga Kellert 9 Bare and Indefinite Nominal Predicates in the History of German Svetlana Petrova Index of Subjects
319
246
285
Preface This volume explores the interface between morphosyntax and semanticspragmatics in the domain of referential and quantificational nominal expressions. We present synchronic and diachronic case studies from Romance and Germanic languages, with the aim of empirically testing, on the basis of comparative data, the most recent theoretical developments in the analysis of reference and quantification. The contributions collected in this volume are the results of intensive discussion and exchange among scholars involved in the network ‘RED—Referential expressions in discourse’ (https://red.uni‑koeln.de/). The authors presented preliminary versions of their papers during a workshop that took place in Bologna in June 2017 and was jointly organized by the University of Cologne and the University of Bologna. We thank both institutions for their organizational and financial support, as well as the German Research Foundation for financial support for the conference and the volume as part of the SFB 1252 Prominence in Language (Project-ID 281511265). We also gratefully acknowledge the Mobility Grant sponsored by the University of Cologne that initiated the international cooperation between the editors of this volume. Special thanks go also to Nicola Grandi, who was the local co-organizer. We are also grateful to Frederike Weeber and Antonia Braun for their scientific and organizational support at various stages of the project. The volume as a whole profited from the insights of many experts in the field. Many people deserve thanks for their help: the authors, for contributing their expertise and bearing with the complex editorial schedule; Martin Becker and Elisabeth Stark, the invited discussants that took part in the workshop and greatly contributed towards sharpening the papers’ arguments and conclusions; the anonymous reviewers, for their insightful and constructive criticism; the series editor for Syntax and Semantics, Keir Moulton, who provided very helpful and constructive comments, and Elisa Perotti at Brill, for believing in this project and guiding it towards completion. Bologna, Cologne and Vercelli, May 2021
Tables 2.1 3.1 5.1 8.1 8.2 8.3 8.4 9.1 9.2 9.3 9.4 9.5 9.6 9.7
Article distribution 30 Romance indefinite determiners (mass / count) 77 The semantic types 145 Distribution of postnominal and prenominal qualunque in CORIS 260 Distribution of postnominal qualunque in MIDIA 269 Distribution of prenominal qualunque in MIDIA 269 Semantic interpretation of [UN N Qln] in DiaCORIS 273 Dialectal distribution of BNPs and INPs in MHG and ENHG in the database 290 Distribution of BNPs and INPs in ENHG 292 Distribution of BNPs vs INPs in MHG 295 Single vs. conjoint/disjoint role nouns in MHG 298 Effect of subject referentiality on BNPs vs. INPs in MHG 301 The presence of a determiner in modified proper names in REA 311 The presence of a determiner in modified proper names in REM 312
Notes on Contributors Artemis Alexiadou Institut für Anglistik und Amerikanistik Humboldt-Universität zu Berlin & Leibniz-Zentrum Allgemeine Sprachwissenschaft—ZAS, Berlin [email protected] Maria Aloni Afdeling Filosofie (Logic and Language) Faculteit der Geesteswetenschappen & Institute for Logic, Language and Computation Universiteit van Amsterdam [email protected] Chiara Gianollo Dipartimento di Filologia Classica e Italianistica Università di Bologna [email protected] Klaus von Heusinger Institut für deutsche Sprache und Literatur i Universität zu Köln [email protected] Olga Kellert Seminar für Romanische Philologie Georg-August-Universität Göttingen [email protected] Manuel Leonetti Departamento de Lengua Española y Teoría de la Literatura Universidad Complutense de Madrid [email protected] Maria Napoli Dipartimento di Studi Umanistici Università del Piemonte Orientale [email protected]
x Edgar Onea Institut für Germanistik Universität Graz [email protected] Svetlana Petrova Fakultät für Geistes- und Kulturwissenschaften Fachgruppe Germanistik/Linguistik Bergische Universität Wuppertal [email protected] Cecilia Poletto Institut für Romanische Sprachen und Literaturen Goethe-Universität Frankfurt am Main & Dipartimento di Studi Linguistici e Letterari Università degli Studi di Padova [email protected] Mario Squartini Dipartimento di Studi Umanistici Università degli Studi di Torino [email protected]
notes on contributors
Abbreviations The abbreviations in the glosses that follow the Leipzig Glossing Rules are not listed below. Adj BNC BNP BP BQ CLD COCA CORIS CP D Det DeReKo DiaCORIS DO DOM DP DPG EI ENHG Eval Excl exh FC FCI fem FnhdC FQ FR ger HNC Ind INP IO IP
adjective British National Corpus bare noun phrase bare plural bare quantifier clitic left dislocation Corpus of Contemporary American English Corpus di Italiano Scritto Complementizer Phrase determiner determiner German Reference Corpus Corpus Diacronico di Italiano Scritto direct object differential object marking Determiner Phrase definite plural generic epistemic indefinite Early New High German evaluative meaning exclusiveness operator exhaustification operator free choice free choice interpretation feminine Bonner Frühneuhochdeutschkorpus floating quantifier free relative gerund Hungarian National Corpus indiscriminative meaning indefinite noun phrase indirect object Inflection Phrase
xii LF masc MHG MIDIA MLG MWB N NP NS NSL Num NumP O OE OHG OVI PDG PoS PP Pred PrP Q QDet RCI REA REM REN RM RRC S TP UFC UFR V VP wh-
abbreviations Logical Form masculine Middle High German Morfologia dell’Italiano in Diacronia (database) Middle Low German Middle High German dictionary noun Noun Phrase null subject null subject languages number Number Phrase object Old English Old High German Opera del Vocabolario Italiano (database) present day standard German part of speech Prepositional Phrase predicate Predicate Phrase question question determiner random choice interpretation Referenzkorpus Altdeutsch Referenzkorpus Mittelhochdeutsch Referenzkorpus Mittelniederdeutsch/Niederrheinisch relativized minimality reduced relative clause subject Tense Phrase universal free choice universal free relative verb Verb Phrase interrogative
chapter 1
Reference and Quantification in Nominal Phrases: The Current Landscape and the Way Ahead Chiara Gianollo, Klaus von Heusinger and Maria Napoli
1
Introduction
The thematic volume ‘Determiners and quantifiers: Functions, variation, and change’ explores the interface between morphosyntax and semantics-pragmatics in the domain of referential and quantificational nominal expressions. We present case studies from Romance and Germanic languages, dealing with both synchronic and diachronic aspects. Our aim is to empirically test, on the basis of comparative data, the most recent theoretical developments in the analysis of reference and quantification and to identify focal points for future research. Formal syntax and semantics have long abandoned the idea that referential and quantificational expressions may receive a uniform characterization (say, ‘Determiner Phrase’ as concerns syntax and ‘Generalized Quantifier’ as concerns semantics), and have explored a number of dimensions of variation. This work, however, has not yet managed to reach a general, consensual framework delimiting in a principled way the dimensions of cross-linguistic variation at the interface between meaning and form (for recent representative attempts in this direction see the volumes edited by Stark, Leiss and Abraham 2007, Cabredo Hofherr and Zribi-Hertz 2014, Aguilar-Guevara, Pozas Loyo and Vázquez-Rojas Maldonado 2019, Balogh, Latrouite and Van Valin 2020). A further gap in our knowledge is represented by the scarcity of diachronic studies exploring the validity of certain theoretical proposals on the basis of the actual developmental tendencies of referential and quantificational expressions in documented historical stages of languages. Current theoretical work has developed models that envisage a more transparent mapping between meaning and form in nominal phrases, but also in determiners and quantifier words themselves, thanks to a more detailed study of word-internal compositionality, as well as of diachronic processes affecting it. This work has substantially benefited from a renewed attention to conditions on language use, which has led to a closer investigation of pragmatic factors interacting with semantics and of the behavior of nominal phrases in discourse. © c. gianollo, k. von heusinger, m. napoli, 2022 | doi:10.1163/9789004473324_002 This is an open access chapter distributed under the terms of the cc by-nc-nd 4.0 license.
2
gianollo, von heusinger and napoli
This volume brings together experts in syntax, semantics, and pragmatics in order to cooperatively work on filling the gaps highlighted above, and to systematically pursue an improved integration of syntax, semantics, and pragmatics. The chapters in this collection are motivated by the same fundamental research questions and by the methodological focus on synchronic variation and diachronic change as a source of crucial evidence to validate our theoretical models of the interfaces between the structural and the interpretative component. Each chapter contains a contribution to the theoretical analysis of the phenomenon at stake, informed by the contemporary discussion in formal syntactic and semantic frameworks, as well as by typological generalizations. The chapters variously contribute to three fundamental areas in the study of reference and quantification: reference and quantification between morphosyntax and interpretation; reference and quantification in discourse; reference and quantification in diachrony. In this introductory contribution, we single out some outstanding issues for each of these areas, and we provide an overview of the insights emerging from the studies collected in this volume. Before we focus on the three areas listed above, we will start with a more general section on interface issues between morphosyntax and semantics in the domain of reference and quantification.
2
Interface Issues
A first obstacle to overcome in an integrated approach to the morphosyntax and the interpretation of referential and quantificational expressions is a terminological one (on which see also Panagiotidis 2018). As is often the case, terminological mismatches between (sub)disciplines are the symptom of deeper differences in the way of conceptualizing phenomena and research questions, which are scrutinized in this volume. In theoretical syntax, the classification of elements contributing referential or quantificational functions is fundamentally based on their distribution, that is, mainly on their patterns of positioning with respect to the nominal nucleus and of co-occurrence with other functional elements of the lexicon. Most generative work on the nominal domain rests on the so-called DP hypothesis (Szabolcsi 1983, 1987, Abney 1987, Horrocks and Stavrou 1987, Stowell 1989, 1991, Longobardi 1994): nominal phrases are headed by determiners; that is, functional items of the lexicon project their own structure, the Determiner Phrase, which embeds the noun’s projection NP. The DP hypothesis is rooted in the attempt of X-bar-theoretic syntax to explain syntactic structures by means of one general template that applies across categories: according to this view,
reference and quantification in nominal phrases
3
CPs (full clauses) are structurally parallel to DPs (nominals) in having a lexical core that is embedded into nested projections of grammatical categories (the so-called ‘extended functional projections’, since Grimshaw’s seminal work in the 1990s, published as Grimshaw 2005: chapter 1).1 The DP hypothesis has a number of ramifications on which there has always been less consensus in the literature; they are discussed at various points in the present collection. The universality of a DP projection has been questioned, given the existence of languages without articles. The absence of articles has been argued to correlate with other syntactic properties pointing to the absence of a DP projection (Bošković 2008, 2012). In a more semantically-driven perspective, a semantic parameter has been proposed, according to which in some languages NPs would be inherently predicates, and need D to turn them into arguments, whereas in other languages NPs would be inherently of an argumental type (referential names of kinds), with no need for a D projection (Chierchia 1998). Related to this cross-linguistic issue, a further point concerns the analysis of determinerless nominal phrases (bare nouns) in languages that do have articles (see Alexiadou, this volume), which has led to the much-discussed assumption of empty determiners, that is, syntactically and semantically present but phonetically null Ds (Longobardi 1994, 2001). The debate on bare nouns, in turn, branches out into the study of noun incorporation and pseudo-incorporation (on which see Dayal 2011, Borik and Gehrke 2015) and the study of the principles governing the alternation between bare and non-bare forms (for instance, with partitives or generics, as amply discussed in this volume; see also Kabatek and Wall 2013 and recently Ihsane 2021). A separate aspect of the DP hypothesis that is a source of substantial differences across syntactic analyses concerns the way in which semantic-pragmatic concepts are argued to be mirrored by the syntax of nominal phrases: as also happens with clausal syntax (on which see the discussion in Leonetti, this volume), more elaborate treatments tend to enrich the functional structure in order to syntactically represent all meaningful elements, comprising information-structural notions, whereas more reductionist analyses restrict the number of syntactically encoded categories and operations. 1 This parallelism has recently been questioned by Chomsky, Gallego and Ott (2019) for theoretical reasons connected to the understanding of syntactic structure building under a Bare Phrase Structure approach, which, more in general, are leading to re-thinking the notion of projection itself (on which see Borer 2005: chapter 1). Another line of criticism addresses asymmetries between DP s and CP s in terms of selection and featural dependencies: for opposing positions in the debate surrounding these aspects see Bruening, Dinh and Kim (2018) and Larson (2019).
4
gianollo, von heusinger and napoli
Thus, depending on the theoretical framework, proposals vary as to how rich, in terms of categorial projections, the functional extended structure of nominals is assumed to be, and as to how universal this structure is (a longstanding debate summarized in Alexiadou, Haegeman and Stavrou 2007). On the basis of distributional properties, in some generative treatments, a class of determiners is distinguished from a class of quantifiers (Shlonsky 1991, Stowell 1991, Cardinaletti and Giusti 1991, Giusti 1991).2 Accordingly, different projections (and different positions within such projections) are posited for these elements: an article is considered to be the head of the DP projection, whereas quantifiers are analyzed as heads or specifiers of a hierarchically higher Quantifier Phrase QP, accounting this way for their possible cooccurrence, as in allQ theD students.3 Matthewson (2001) shows that in the Salish language St’át’imcets QPs systematically take DP s (of type e), and not NP s (of type ⟨e,t⟩), as their complement. In an attempt to reach cross-linguistic uniformity for the denotation of quantifiers, she proposes that also in structures of English like all cats the quantifier takes a DP with a phonetically empty head as a complement (allQ ØD cats). When we move to semantic theories, we see that the attention obviously shifts towards the interpretative component, traditionally with less attention to the distributional patterns within the nominal phrase. In this perspective, the distinction between determiners and quantifiers that is routinely made in syntactic analyses does not easily match the categorization of functional items based on their semantic properties. The main distinction made from a semantic perspective is, rather, one between referential and quantificational expressions. In original formulations, such as e.g. Stowell (1991), the syntactic differentiation between DPs and QPs was meant to match this distinction: the idea, based on Higginbotham’s (1985) proposal that determiners are needed in order to turn nominal predicates into arguments, was that DP s can be used as referential arguments of type e, whereas QPs are those nominal phrases that do not appear in argument position in the logical structure, since they have to undergo quantifier raising (Generalized Quantifiers of type ⟨⟨e,t⟩,t⟩; see Barwise and Cooper 1981, Keenan and Stavi 1986). However, the class of semantically quan-
2 Hence, the term ‘determiner’ becomes ambiguous between a broader reading, encompassing quantifiers as a subclass, and a narrower reading, excluding quantifiers as a separate category. 3 Note here that in syntactic jargon the term ‘quantifier’ refers to the item of the functional lexicon (e.g. every), while in semantic jargon it refers to the denotation of the entire QP, that is, of the nominal expression containing the functional item and its restriction (generalized quantifiers like e.g. nothing, every cat).
reference and quantification in nominal phrases
5
tificational expressions is broader than the class of nominal phrases for which a QP projection is routinely posited in the syntax and, conversely, the class of referential expressions does not coincide with the class of nominal phrases containing the DP projection: in current practice there are often pervasive mismatches between syntactic categories and semantic types.4 This mismatch is deeply entrenched in the conceptualization of the definite and indefinite article. In syntactic representations, definite and indefinite articles have often been treated alike and considered heads of the DP projection. This analysis is entailed by syntactic approaches that see D as the position that hosts a binary feature [± definite] (e.g. Lyons 1999) and by analyses that require D to be overtly filled in some languages and treat the indefinite article as an appropriate filler (e.g. Longobardi 1994). From a semantic point of view, however, definite and indefinite articles differ substantially in their semantic function, mirrored in terms of their semantic types. The definite article is a function from a (singleton) set to the element of that set, that is, it is defined in terms of the type ⟨⟨e,t⟩,e⟩ and yields a referential reading of the DP (i.e., type e). The indefinite article, on the other hand, under a quantificational analysis, denotes a relation between two sets (a function from sets to sets of sets (⟨⟨e,t⟩⟩,⟨⟨e,t⟩,t⟩)). Following this analysis, an indefinite nominal phrase is of type ⟨⟨e,t⟩,t⟩, i.e., the type of a QP.5 Indeed, since the beginning of the DP hypothesis, various proposals have been advanced in order to distinguish also syntactically between the definite and the indefinite article. Abney (1987: 222), for instance, discusses some advantages of interpreting the English indefinite article a as something other than a determiner, recalling Perlmutter’s (1970) proposal to treat it as the reduced form of the cardinal numeral one (see also Kayne 1994: 124). Lyons (1999: 89–95) highlights the typological rarity of ‘real indefinite articles’, understood as encoding the [- Def] feature in D, and supports an analysis of English
4 The extent and nature of these mismatches, besides being dependent on the specific syntactic framework adopted, also vary across semantic theories. For instance, Montague’s (1973) model avoids them by positing a uniform semantic denotation for all nominal expressions in terms of generalized quantifiers (see Williams 1983 and Higginbotham 1985 for early counterarguments). Russell (1905) treats definite descriptions as quantificational expressions, as further discussed in the main text. Discourse Representation Theory (since Kamp 1981) and File Change Semantics (Heim 1982) opt for a non-quantificational representation of indefinite nominals. 5 The quantificational analysis is only one possible semantic approach to the indefinite article, as we will see below in this section and in Section 4. See also the discussion of type flexibility with the definite and the indefinite article and of possible type-shifting operations in Partee (1987).
6
gianollo, von heusinger and napoli
a as a cardinality term. In current syntactic research, various models have been proposed in which definite and indefinite articles are considered to be firstmerged in the structure at different points (see Zamparelli 2000, Borer 2005 and Klockmann 2020 also for a useful overview of past proposals). Famously, there is a controversy surrounding the semantic type of definite descriptions and there exist quantificational treatments of definite descriptions, which could result in a more uniform syntax-semantics mapping. Russell (1905) assumes that definite descriptions like the king of France are quantifier phrases, i.e. of type ⟨⟨e,t⟩,t⟩. This non-referential type allows to interpret the sentence The king of France is bald as false, and it allows to model scope interactions. However, Strawson (1950) takes definite descriptions as referential expressions of type e with a uniqueness presupposition, and assumes that the sentence above causes a presupposition failure. While this controversy is not yet resolved, a majority of semantic analyses follow Strawson, matching the syntactic analysis of definite descriptions as DP s. Moreover, various observations on contexts with non-prototypical functions of articles cast doubts on the possibility of attributing them a uniform semantic and syntactic characterization. For example, not all instances of what is morphosyntactically a definite article ultimately yield a referential expression. So-called weak definites are syntactically DPs, but semantically do not imply global uniqueness and behave more like indefinite nominal phrases (Carlson et al. 2006, Schwarz 2009; see also Borik and Gehrke 2015 for the semantic parallels with pseudo-incorporation). In some languages, singular and plural definite articles receive generic interpretation (Longobardi 1994, 2001, Borik and Espinal 2012, Barton, Kolb and Kupisch 2015). As for indefinite articles, their semantic import differs (and, consequently, receives different representations, in ways that vary across frameworks) depending on a number of factors, in particular the surrounding linguistic environment and the competition with alternative indefinite expressions in a given language. The most stable semantic contrast we find cross-linguistically is that between a specific (referential) and non-specific (existential) interpretation of a nominal phrase containing an indefinite article (see Fodor and Sag 1982, Diesing 1992, von Heusinger 2002, 2011 and Schwarzschild 2002).6
6 Proposals to encode these differences in the syntactic component can be found among others in Zamparelli (2000) and Ihsane (2008).
reference and quantification in nominal phrases
3
7
Reference and Quantification between Morphosyntax and Interpretation: Articles
Articles represent a particularly thorny domain for the morphosyntax-semantics interface: in this collection, they are discussed from multiple angles in various contributions. Alexiadou’s contribution is dedicated to the role of the definite article in plural generic nominals, based on a comparison between Germanic and Romance. Squartini’s chapter deals with the system of Romance indefinite articles. Petrova presents a diachronic study on the role of indefinite articles in German nominal predicates. The chapter by Alexiadou, ‘Definite plural generics in English: evidence from de-adjectival nominalization’, provides a fresh look on the outstanding issue of plural nominal phrases with a generic reading. She discusses differences and similarities between Romance and English, with a focus on the article + adjective combinations like the poor. Since Carlson (1977), a lot of attention has been devoted to the interpretation of bare plurals in English, and to the cross-linguistic difference with respect to Romance, where (plural) bare nouns are impossible in some languages and severely restricted in others. In the specific case of plural kind and generic interpretation, Spanish and Italian invariantly need the definite article (It. I cani sono intelligenti vs. Engl. Dogs are clever), whereas in English an analogous structure (The dogs are clever) is interpreted as denoting a contextually given maximal set. This cross-linguistic dichotomy breaks down once cases are considered where plural definites can be used with a generic meaning also in English. Among those we find anaphoric generics (Saurischian Bipeds … The saurischians …), instances in dialectal varieties and previous historical stages of English (on which see also Section 6), and, crucially for Alexiadou’s analysis, article + adjective combinations (the poor but also the youngs), where the definite article is in fact the only option to obtain a generic reading. Alexiadou argues for a competition scenario involving definite plural generics and bare plurals in Present Day English, whereby the constrained appearance of definite plural generics is due to the presence of an alternative formal realization for genericity that is structurally simpler: bare nouns, following Borer (2005), are simpler because they lack the quantity phrase #P (with a counting, individualizing function). Where this bare alternative realization is absent, i.e., in the case of adjectival nominalizations, the definite article’s potential to contribute a generic meaning is unconstrained, and English converges with Romance. The difference between English and French in the overall availability of the barenoun strategy is connected to the different properties of number marking in the two languages (on which see Section 6).
8
gianollo, von heusinger and napoli
According to Alexiadou’s analysis, in definite generics the quantity phrase #P is projected: the definite article merges there and raises to D. This has semantic consequences: Alexiadou argues that the generic meaning expressed by plural definites differs semantically from the genericity expressed via bare nouns, since the definite article contributes maximality, as in a prototypical definite reading. Genericity is thus obtained by creating a maximal sum of individuals that is not contextually constrained. The fact that different formal ways to obtain a generic meaning correspond to semantically distinct shades of genericity is a point highlighted also by Leonetti (1999: 870–882), in surveying the various forms of generic nominals in Spanish. Moving to indefinite articles, Squartini’s specific focus in his chapter, ‘Quantification and classification in Romance plural indefinites: from Number to Seinsart?’, is on the paradigm of Romance indefinite articles and, especially, on the differing realizations of plural indefinites. In order to develop his account, Squartini provides a comprehensive overview of the intricate relations between classificational aspects and number marking in French, Italian and Spanish, which connects well also with Alexiadou’s discussion of the syntactic role of number marking. Squartini’s considerations on the role of number marking are also backed up by the analysis of the Piedmontese koiné, an ItaloRomance variety with a particularly impoverished number morphology. Classificational distinctions in the denotation of nominal phrases concern qualitative properties of nominals such as the count-mass distinction (what Rijkhoff 2002 terms Seinsarten). Different conceptualizations of masses and pluralities in the literature lead to different understandings of the relation between the two, at the interface between cognition and grammar: once masses are understood not as homogeneous substances but as inherently composed of portions, their intrinsic cumulativity draws them nearer to plurals. This opens up new perspectives of analysis for determiner systems, and in particular for the paradigm of indefinite articles. The opposition observed in French between indefiniteness marking in singular count (un) and mass nouns (de + number marking) hints at the role of classificational aspects in the article system.7 The issue is, then, whether classificational aspects have a role in the plural as well. On the basis of the assumption of a strict form-function correlation, whereby morphological relatedness is equivalent to semantic relatedness, it has indeed been proposed for Romance that the different forms for the plural indefinite article (e.g. French and Italian 7 Squartini follows Cardinaletti and Giusti (2018) in considering the so-called partitive article as an element of the D projection. See the contributions in Ihsane (2021) for a comprehensive perspective on the debate concerning this issue.
reference and quantification in nominal phrases
9
‘partitive’ article de- vs. Spanish morphological plural of un-) may correspond to different classificational meanings of the plural (“the way in which referents are combined together”, in Squartini’s phrasing): the plural of the partitive would have a mass-like interpretation, creating homogeneous aggregates, while the plural of un- would have a ‘collective’ meaning, crucially with non-distributive readings. Squartini rejects this view on the basis of both empirical data and theoretical considerations. From his perspective, while different classificational aspects may have played a role in the grammaticalization process, a deep similarity in the resulting paradigmatic structure of contemporary Romance indefinite determiners emerges, which goes beyond the more superficial morphosyntactic differences: the governing principle is the preservation of a formal distinction between singular and plural, independently of the strategies employed by each language. Mass and plural are kept distinct in each language and the classificational aspects are always mediated by number, meaning that plural articles never have a genuinely classificational function in Italian, Spanish and French. Nonetheless, one can observe grammaticalization phenomena in which classification is a more prominent shaping factor (on which see further Section 6). Indefinite articles are also at the core of Petrova’s contribution ‘Bare and indefinite nominal predicates in the history of German’. In her study, however, the attention shifts from argumental uses to predicative ones. The appearance of (indefinite or definite) articles in predicative expressions is hard to explain by the approaches, seen in Section 2, that consider D as a category that turns a predicate into an argument (see Williams 1983: 424 for the assumption that predicate nominals “are syntactically identical to referential noun phrases” although they have a different logical type). While it is technically possible to obtain the right semantic type by postulating type-shifting operations (Partee 1987), the functional motivation for the alternation between the presence and the absence of the article is much debated, also in consideration of the differing conditions to which this alternation is subject in various languages, and at different diachronic stages. Furthermore, early treatments (Stowell 1989, 1991) already remarked that not all nouns behave in the same way with respect to their acceptability as bare predicates in various constructions: nouns for professions (doctor) or appointments (president) stand out both intra- and crosslinguistically for their readiness to be used as bare predicates (see Zamparelli 2008 for a comparative study of bare predicate nominals in Romance). Petrova takes these empirical facts as her starting point and investigates the factors allowing for bare and ‘indefinite’ nominal predicates with copula verbs at various diachronic stages of German, and also across dialectal varieties.
10
gianollo, von heusinger and napoli
We will present the aspects that specifically pertain to the diachronic import of the phenomenon in Section 6. For now, let us comment on how Petrova’s diachronic study of the formal alternation ties in with her synchronic analysis. Semantically, she shows how uniqueness strongly correlates to the availability of the bare realization at all historical stages of German; what changes, due to language-external reasons, is the inventory of nouns belonging to the class of individual-identifying nouns denoting a unique function (‘role nouns’, such as ‘bishop’, ‘treasurer’, ‘judge’, as opposed to property-identifying nouns such as ‘hero’, ‘giant’, ‘idiot’). Syntactically, she proposes that the predicative phrase is a DP also when it is articleless, and that role nouns, in virtue of their individual-identifying function, behave as proper names (under Longobardi’s 1994 analysis): they are able to raise to the position of an empty determiner [D e] (in languages that have evidence for such an empty category), accounting this way for the absence of a determiner in the D position, which ends up being occupied by the noun. According to Petrova, this model not only accounts for the behavior of predicative noun phrases in the oldest stages of German, but can also explain the situation observed in Present Day German, with a gradually increasing number of nouns conceptualized as role nouns. In this respect, a further factor, this time language-internal, that is subject to diachronic change emerges from the study of non-standard varieties: the almost complete loss of predicative bare nouns in some Upper German dialects is a consequence of the loss of the possibility of licensing [D e] and, as a result, the loss of N-to-D movement. This makes the presence of a determiner with predicative role nouns obligatory.
4
Reference and Quantification between Morphosyntax and Interpretation: Indefinites
The rich variety in shape and function of indefinite determiners in natural languages has attracted the attention of formal linguists, comparatively and typologically oriented scholars, and historical linguists alike. The fundamental questions concern, on the one hand, the functional drive behind such a remarkable pool of options to express existential quantification: just to give an example, Haspelmath’s (1997) semantic map of English indefinites counts four main series of ‘specialized’ indefinite pronouns (some, any, no, wh-ever), to which also multi-word combinations like a certain, besides the indefinite article, can be added. On the other hand, a pervasive multifunctionality of indefinite determiners is observable in many languages: just to mention some much-discussed examples, any notoriously has a free-choice or a negative-polarity interpreta-
reference and quantification in nominal phrases
11
tion, the indefinite article can have specific and non-specific readings, and so on. In our collection, four contributions investigate indefinites from various angles. One major aspect of the debate is how to best characterize indefinites as formal semantic objects, encompassing also their discourse contribution (on which see further Section 5). In recent years, frameworks like Alternative Semantics and Inquisitive Semantics have shaken up the terms of the discussion, by adding new formats for representing indefinites to the more traditionally discussed ones (Generalized Quantifiers, DRT-style variables, choice functions). In this collection, Aloni shows how an analysis of free-choice indefinites in terms of Alternative Semantics is able to insightfully account not only for their synchronic distribution, but also for their diachronic evolution. Onea presents a novel theory of the discourse contribution of special indefinites, the ‘erotetic theory of indefinites’, which is based on Inquisitive Semantics and on the discourse view elaborated in the Question under Discussion approach. The Alternative Semantics analysis developed by Aloni in her contribution ‘Indefinites as fossils: the case of wh-based free choice’ is based on the hypothesis, stemming from Kratzer and Shimoyama’s (2002) work, that indefinites closely resemble questions in their potential to evoke propositional alternatives: an indefinite like someone, when composed into a proposition, generates a set of alternative propositions (e.g., for someone called, propositions like x called, y called, z called, etc.). This set of alternatives interacts with abstract operators in the clause, yielding different semantic contributions depending on the operator quantifying over the alternatives. For instance, the alternatives generated by someone in an episodic sentence will be bound by an operator with existential quantificational force. The existence of different specialized indefinite lexemes would then be accounted for by assuming that these specialized forms encode the necessity for a given indefinite to be associated with an operator of a certain kind in order to be felicitous. For instance, negatively marked indefinites will encode the need for a negative operator to license them. Aloni explores in particular the conditions imposed on the surrounding semantic context by free-choice indefinites like Spanish cualquier and Dutch wie dan ook; moreover, she investigates how their enriched meaning contribution emerged historically, finding a link between the contemporary conditions of use and the original morphology and distribution (see further Section 6). Onea’s contribution ‘Specificity and questions of specification’ provides a conceptually different perspective, which is based on his discourse-oriented ‘erotetic theory of indefinites’. According to this theory an indefinite makes three different contributions: i) it raises a question that needs an answer, ii) the
12
gianollo, von heusinger and napoli
lexical contribution of special indefinites makes a comment on how this question can be answered (e.g. indicates that the speaker knows the answer); iii) these comments on the answer determine the scopal properties of the indefinite. For example, the German special indefinite ein gewisser N (‘a certain N’) raises the specificational question which N? The contribution of the specialized indefinite is that the speaker does know the answer, and this has an effect on the wide scope behavior of such indefinites (see Section 5 and the more explicit elaboration of the whole theory in Onea 2016). According to Onea, specialized indefinites evoke information about who can resolve the question raised by them. Aloni, on the other side, rather focuses on the semantic interaction of indefinites with other operators. These different perspectives highlight the numerous functions of indefinites (e.g. someone vs. a certain) and the need to account for them at different levels. A further perspective on indefinites, more focused on the interface with syntax (both at the clausal level and word-internally), is explored by Kellert’s and Poletto’s contributions. Kellert’s study, ‘The evaluative meaning of the indefinite qualunque in (Old) Italian’, investigates the historical evolution of a freechoice indefinite, similarly to Aloni. Kellert, however, concentrates on a particular aspect of this evolution: the emergence of a further, evaluative reading of the free-choice indefinite, which in certain contexts is interpreted as ‘ordinary, average’, as in una ragazza qualunque ‘an ordinary girl’. This kind of reading has been observed for free-choice indefinites also in other languages, but it has never been investigated diachronically before. First of all, Kellert defines the structural and interpretative prerequisites for the evaluative reading, showing that the potential ambiguity between a pure free-choice, a random-choice and an evaluative interpretation of qualunque is in fact resolved in context thanks to a number of principled differences in syntactic behavior (the evaluative reading strongly correlates with a postnominal, not determiner-like position for the indefinite) and in semantic-pragmatic requirements (the occurrence in episodic contexts and its ability to modify a nominal phrase introduced by a demonstrative). Secondly, the author shows how the diachronic emergence of the peculiar evaluative reading is dependent on the interplay with surrounding semantic operators, which are syntactically represented, notably negation and focus: the ‘ordinary’ reading emerges from the conventionalization of scalar inferences emerging in a focused negative context (‘not just any’ = ‘ordinary’, see further Section 6). A special attention to the interplay between the quantifier’s internal and external syntax is at the core of Poletto’s study of the distribution of bare quantifiers at various historical stages of Italian (‘Being bare: a survey of quantifier positions’). Bare quantifiers are pronominally used quantificational elements
reference and quantification in nominal phrases
13
like Italian tutto / tutti ‘all’, molto ‘much’ / molti ‘many’, niente ‘nothing’. According to Poletto, they have a different internal syntactic makeup with respect to full quantified nominal phrases. Namely, the QP does not take a NP as its complement, but a Classifier Phrase headed by an abstract classifier-like sortal noun (PERSON, THING, PLACE, TIME, WAY). Some grammaticalization processes are argued to provide evidence for the presence of phonetically realized counterparts of the abstract classifiers, as in the case of -ente ‘being, thing’ in Italian niente ‘nothing’, which is not etymologically transparent for contemporary speakers but would parallel more transparent elements like -thing and -body in e.g. nothing, everybody. This analysis has profound consequences for our understanding of so-called extended projections, that is, of the architecture of functional categories that compose the spine of DP s and QP s. In mainstream theory, they take a lexical category as a complement, and are inserted in order to add inflectional morphemes and perform information-structural operations. In Poletto’s view, inspired by Kayne’s work (see in particular a number of essays collected in Kayne 2005), also the core of the extended projection may be represented by a (potentially unpronounced) functional morpheme. The internal makeup has an effect on the external syntax of these elements, since bare quantifiers reach positions in the clause that are not available to full quantifier phrases. This differential distribution is well known from French, where bare tout ‘all’ and rien ‘nothing’ in the direct-object function surface to the left of the past participle. Poletto investigates the distribution of the Italian analogous in Old Italian, where the position of the participle is comparable to that of Modern French, and in contemporary Italian, where the participle is higher in the clausal spine but a differential distribution for bare quantifiers is still detectable. She elaborates on the original proposal by Beghelli and Stowell (1997), who convincingly demonstrate that English quantifiers can access additional dedicated positions in the clause, and shows that a special positioning for bare quantifiers, higher than the position for normal direct and indirect objects, can be detected in Italian as well.
5
Reference and Quantification in Discourse
Reference and quantification are primarily investigated in the context of a sentence—both for syntactic as well as semantic reasons. Sentences are the basic unit for syntactic theorizing and models, hence models beyond the sentence boundary are often fuzzy and not in the core interest of syntacticians. For semanticists, the differences in the referential and quantificational structure
14
gianollo, von heusinger and napoli
can be best accounted for as differences in truth conditions or propositions —both are sentence bound. However, research in discourse anaphoric relations—starting with the donkey sentences of Geach (1962), E-type pronouns (Evans 1977), discourse referents (Karttunen 1976), and dynamic semantics as envisaged by Kamp (1981) and Heim (1982)—made obvious that anaphoric relations cannot be accounted for without the information provided by and inferred from context. More recently, the notion of questions appears to be crucial for understanding the discourse-semantic meaning of sentences and their referential expressions (Roberts 1996, Ciardelli, Groenendijk and Roelofsen 2019, Onea and Zimmermann 2019). The chapters by Leonetti and Onea provide original contributions to the importance of discourse pragmatic principles for the interpretation of referential and quantificational expressions. Leonetti focuses on the interaction of syntactic, information structural and coherence principles for defining the licensing conditions of null subjects in Italian and Spanish. Onea models the function of indefinite noun phrases similarly to that of questions. He analyzes different types of indefinite pronouns as raising a question and then making different types of comments towards the content of the question. Both contributions show that referential and quantificational structures clearly extend beyond the sentence boundary and are important parameters for text structure. Leonetti addresses the interaction between syntactic restrictions, information structure and discourse pragmatic coherence principles in his contribution ‘Topics and the interpretation of referential null subjects’. Languages like Italian and Spanish allow for null subjects, while others like English or German do not. This is taken to be a central syntactic characteristic, which determines a class of so-called null subject languages (NSL s). Leonetti carefully develops an argument that shows that a purely syntactic theory of null subjects cannot account for the empirical data. He first presents a comprehensive set of syntactic data that are generally understood as showing that a null subject is anaphorically linked to (or licensed by) a topical antecedent. A topical expression, typically a subject, can license a null subject in a subsequent clause, while non-topical arguments must be taken up by personal pronouns or descriptive definite nominal phrases. Leonetti sketches then a recent syntactic theory of null subjects that is based on this set of widely accepted data: Frascarelli (2007, 2018) assumes an extension of the left periphery of the sentence and argues that a null subject is linked to a null topic phrase in this left periphery, which by itself must be licensed by a topical expression in the previous text. The null topic phrase can license the null subject and the question of licensing and interpretation of null subjects follows from syntactic principles. At a first glance, this
reference and quantification in nominal phrases
15
theory seems to integrate information structural categories (topic) into the syntactic layout of the sentence and therefore links sentence structure to discourse structure. However, Leonetti argues against this kind of theories with two main arguments. First, he shows that Frascarelli’s theory has internal technical problems, as it does not state how to restrict the licensing of null topic phrases in the left periphery, which are the licensers of null subjects. But the theory also has an essential conceptual fault, as Leonetti argues: the theory neglects that null subjects can also be licensed by prominent non-topical antecedents, which create a higher coherence. In order to show this, Leonetti returns to the central empirical observation that seems to show that only topics (generally identical to subjects) license null subjects. He convincingly provides, for each of the data points, parallel examples that show that prominent non-topical constituents can license null subjects if they establish a more coherent discourse than the topical constituent. Prominence is defined as the relative ranking of arguments in the sense of Centering Theory (see Walker, Joshi and Prince 1998). There are several parameters that contribute to the prominence of an argument (see von Heusinger and Schumacher 2019 for a recent overview). Leonetti assumes that prominence depends on accessibility, i.e. syntactic structure (subject, subordinated clauses), and on information structure (topic, focus). Prominence is the main predictor (the central condition) for licensing null subjects, as it is for personal pronouns in non-null-subject languages. However, the independent semantic-pragmatic condition of coherence can overwrite this effect and license null subjects that are not prominent but contribute to a higher coherence of the discourse. Thus, Leonetti generalizes the interpretation of null subjects to broader principles of anaphora resolution, and he convincingly shows that this can only be modelled by the interaction of syntactic structure, information structure and general discourse pragmatic principles such as coherence. Onea’s contribution ‘Specificity and questions of specification’ complements the discussion of the role of discourse in licensing null subjects. Onea investigates different types of indefinite determiners such as a certain by embedding their semantics into a representation based on question semantics, a theory he terms ‘erotetic theory of indefiniteness’. As introduced in Section 4, he assumes that an indefinite nominal phrase in a sentence like Ashanti saw a certain professor makes three kinds of contribution to the discourse: Firstly, all indefinites raise a specificational question, for our example: Which professor did Ashanti see? Secondly, the particular indefinite determiner makes a specific contribution, namely a comment on that question: Some salient agent knows the answer to that question. This salient agent is typically the speaker
16
gianollo, von heusinger and napoli
of the sentence. Thirdly, this comment on the question together with the regular assertion yields a contribution like Ashanti saw a professor and I know which professor. Onea can also show that the type of comment can be used to control the scopal properties of indefinite expressions. After introducing the semantic representation underlying the erotetic theory of indefiniteness he shows how this mechanism can capture even very subtle generalizations distinguishing different indefinite determiners. As test cases he analyzes the function of German ein gewisser vs. ein bestimmter (both ‘a certain’), which differ in the way they select the salient agent that can identify the referent. Ein gewisser always selects the speaker of that sentence, while ein bestimmter can also select another salient agent in the discourse, as Ebert, Ebert and Hinterwimmer (2013) have convincingly shown. Onea contributes new observations to this theory and develops a slightly different approach. He shows that both ein bestimmter and ein gewisser always take scope over at least one intensional operator, thus expressing specificity in the very original understanding (Ioup 1977). He further modifies the theory of Ebert, Ebert and Hinterwimmer (2013) by assuming that the salient discourse agent answers the specificational question (which professor?), rather than only identifying the referent without taking the noun phrase description into account. Thus, he can model the differences between the two indefinite determiners by assuming that distinct discourse agents can answer the relevant question. This view on indefinite reference is fundamentally discourse based and shows that reference and quantification can only be properly understood in a discourse model. Interestingly, the discourse use of the indefinite can be shown to systematically correlate with the morphological presence, in its complex form, of an element that can also function as a discourse (modal or answer) particle, synchronically or diachronically (e.g. German bestimmt ‘certainly’, and other crosslinguistic parallels discussed in the paper).
6
Reference and Quantification in Diachrony
Diachronic research on reference and quantification has widely investigated how to interpret their role and, when relevant, their interplay in language change. The papers collected in this volume address this issue enriching the discussion with new perspectives on both articles (including partitives) and indefinites. We shall discuss the two dimensions in turn. It is well known that, as concerns the development of indefinite articles, the complex question arises as to whether quantificational properties are lost
reference and quantification in nominal phrases
17
when referential ones are acquired8—an issue which is of central importance in diachronic models as Givón’s (1981) and Heine’s (1997), among others. In his seminal contribution, Givón (1981: 50–51) proposed a “gradual scale of indefiniteness” which is composed of three (implicational) stages: Quantification → referentiality/denotation → genericity/connotation On the basis of the cline depicted above,9 the first step of grammaticalization of indefiniteness implies the change of the numeral ‘one’ into an indefinite marker through a process of semantic bleaching: when quantification is bleached out, the new indefinite marker is used to denote referential (that is, specific) indefinite nominals. Non-referentiality (including generics and predicatives) corresponds to the third and final stage of the grammaticalization cline, in which the indefinite marker extends its scope to all indefinite uses. In the present collection, special attention is devoted to the grammaticalization of determiner systems, as well as to the grammaticalization of indefinites. As for determiners, a general suggestion emerging from the studies collected here is how the notion of ‘cline’ itself may be rethought if we consider the complexity of factors influencing the development and evolution of determiners. The contribution by Petrova provides a diachronic explanation for the use of predicative bare nouns in German which goes beyond traditional models of the grammaticalization of indefiniteness. An aspect that is at the core of Squartini’s study is the issue of the grammaticalization paths from which different determiners arose in Romance languages. More specifically, he investigates to which extent classificational features of indefinite determiners, besides quantificational and referential ones, play a role in grammaticalization clines and how these features interact with morphosyntactic properties of number, which turns out to be another crucial factor. At many points Squartini refers to the well-known interplay between the robustness of number morphology and the availability of bare nouns with an indefinite interpretation. The interplay with number also emerges in Alexiadou’s chapter in connection to the expression of genericity by means of bare nouns, comparing English with French. Alexiadou shares with Squartini the discussion on how the morphological expression of number may shape the evolution of determiners and the distribution of articles
8 See Lyons (1999: 89–95) for an overview; more recently, Frajzyngier (2011: 632–635). 9 On the notion of grammaticalization cline see Hopper and Traugott (2003: 6), who defined it as “a metaphor for the empirical observation that cross-linguistically forms tend to undergo the same kinds of changes”, generally through different, related and (hypothetically) unidirectional stages (an issue discussed also in Aloni’s study: see below).
18
gianollo, von heusinger and napoli
and bare nouns. The investigation of the diachronic processes through which bare nominals may become specialized for indefinite readings, eventually competing with articles, is another common thread in these three contributions to the volume, which confirm the importance of different diachronic forces at stake. We shall now deal with these issues in more detail. Petrova’s contribution challenges the traditional explanation of bare indefinites as an instance of incomplete grammaticalization provided by Givón’s (1981) model. Languages with an indefinite article in which, nonetheless, bare nouns occur in predicative sentences should be considered as representing stage 2 in Givón’s scale, namely, an intermediate stage. How can we account, then, for the alternation between bare nouns and indefinite nouns in predicative sentences? As seen in Section 3, Petrova’s paper deals with this issue focusing on German noun phrases which occur as complements of copula verbs like sein ‘be’ and werden ‘become’. Petrova reconstructs the diachronic paths which led to the situation of variation proper to present day standard German: here, as described in previous literature, the indefinite determiner is omitted in this type of predicative construction if a noun belongs to the category of ‘role nouns’ (Class A), giving rise to a bare variant, whereas it is expressed if a noun belongs to the category of ‘class nouns’ (Class B), although with many differences in the distribution when one looks at some contemporary dialects. Through a corpus-based analysis of both quantitative and qualitative nature, Petrova investigates the behavior of nouns of Class A in predicative constructions within two distinct stages of German, with a particular regard to their dialectal distribution. In the first part of the paper, she shows how in both Early New High German and Middle High German Class A alternates between the bare variant and the variant with an indefinite determiner. Factors playing a role in the distribution are, in particular, the nature of appointment or position denoted and the specific social domain to which such an appointment/position belongs, the property of uniqueness of the denoted institutionalized roles (differently from present day standard German) and, in Middle High German, the non-referential status of the subject of the predicative sentence: uniqueness and non-referentiality favour the drop of the determiner. While this remains constant over time, what changes is the inventory of social roles corresponding to institutionalized activities, on the basis of extra-linguistic reasons. In conclusion, according to Petrova, the differences between the standard language and the dialects in the distribution of bare nouns and nouns introduced by an indefinite determiner in predicative sentences is not a matter of different degrees of grammaticalization, but is the consequence of a change driven by extra-linguistic factors and of a different syntactic feature in the language (presence or absence of empty determiners, as pointed out in Section 3).
reference and quantification in nominal phrases
19
Special attention is paid to the historical development of indefiniteness also in the chapter by Squartini, who analyzes which forms and meanings the plural of the indefinite article may take in its development. As mentioned above, the quantificational properties of the indefinite article in the singular are expected in view of its cross-linguistically frequent diachronic source, the numeral ‘one’. However, even more complicated is the situation in the plural of the indefinite article. According to some authors, the classificational dimension is also relevant there, a point that—as seen in Section 3 on the basis of Squartini’s discussion—is more controversial in its interpretation. An important role in the discussion is played by the etymological source of the determiners themselves: different ways of cognitively apprehending plurality can explain the different diachronic sources for plural indefinite determiners. According to Herslund (2012), the ‘partitive’ form grammaticalized by some Romance languages is a determiner for mass also in the plural, whereas languages that use the plural of the numeral ‘one’ as the plural form of the indefinite determiner start its grammaticalization from uses where the form marks ‘collectives’ (sets of discrete entities), as can be observed in Old French. As Squartini puts it, two different evolutionary patterns are possible, which, in their historical starting point, reflect differences in the cognitive apprehension of referents: aggregates pattern formally either together with substances, or together with bounded entities. What Squartini’s analysis also highlights for Romance is a form of “parasitism of plurals on singulars”, in the sense that all indefinite plural determiners are the pluralized version of singular determiners; this also applies to zero marking, because, in the absence of an explicit determiner, plurality is expressed only through inflectional endings on the noun. In his examination of the interaction of number and classificational aspects in Romance article systems, Squartini also discusses number marking on partitive articles taking the typological cline hypothesized by Bossong (2016) as a starting point. Squartini presents data from a lesser studied variety of Piedmontese, which, like some Occitan varieties, has ‘bare partitives’, that is, forms of de with no further morphological exponents. In Bossong (2016) bare partitives are interpreted as the third, namely intermediate, stage on the scale representing the diachronic process which leads to the obligatory use of the partitive article, as in Modern French. The scale would schematically look as follows (adapted from Bossong 2016: 69): 1. absence of any kind of partitive → 2. minimal form of partitive → 3. use of bare partitive (with no article) → 4. use of partitive + definite article → 5. obligatory use of inflected partitive in almost all syntactic contexts
20
gianollo, von heusinger and napoli
Of particular interest from Squartini’s perspective is the fact that in the analyzed Piedmontese variety, for the class of regular masculine nouns the lack of number marking on the partitive is accompanied by absence of number marking on the noun as well. In this case, according to the author’s analysis, a neutralization of the formal distinction between mass and count nouns takes place, leading to a system modelled by what he defines as “the cognitive drive equating the apprehension of substances and aggregates”. The opposite tendency is mirrored by the development of determiners in which plurals are clearly distinguished from masses, as for the Italian new multi-word expression (tutta) una serie di ‘a (whole) series of’, which arose from the grammaticalization of the noun serie, following in principle the same diachronic path from quantifier to determiner as the indefinite article un ‘a’ (although they show differences in their degree of grammaticalization). The fact that (tutta) una serie di is restricted to plurals inherently having a classificational value, since they refer to a set of bounded elements, and that number is not dismissed—on the contrary, it continues to be codified through inflectional endings on nouns— is interpreted by Squartini as an incipient grammaticalization of classification in plurals, independently of the category of number. This latter point connects to the diachronic role of number marking, which comes up in Alexiadou. Her chapter contains discussion of changing conditions on the generic interpretation of nominal phrases. First, she outlines the change affecting formally definite plural noun phrases in the history of English. Diachrony shows that the typological divide between languages that use plural bare nominals to express genericity and languages that adopt, instead, the definite article can be bridged by language change: Old English was much more similar to Romance in using plural nominal phrases with the definite article to express generic or kind reference.10 In addition, plural bare nouns were a possible alternative strategy. Moreover, earlier stages of English, which had richer number inflection, were similar to Modern German in allowing generic readings also for singular adjectival nominalizations: English the riche could be interpreted as ‘the rich person’, as German der Reiche. According to Allen (2010), the loss of this feature is connected to the loss of number inflection, which led the speakers to reinterpret the determiner + adjective sequence as plural by reason of the greater frequency of plural adjectival nominalizations in the language. Developments in the morphological expression of number are argued by Alexiadou to be responsible also for the second phenomenon of change
10
Alexiadou also discusses further aspects (syntax of possessives and adjectives) that show that the syntax of the Old English nominal phrase was in general closer to Romance.
reference and quantification in nominal phrases
21
that she considers, by moving to the history of French. In Old French, both bare plurals and definite plurals were available strategies to express genericity, similarly to the Old English situation seen above. The trajectory of the change is, however, different in the two languages: in Modern French definite plural generics are retained, while the bare noun strategy is lost. In fact, bare nouns in French are lost completely, a gradual change that is diachronically connected to the loss of number morphology on nouns, which, as expected under Delfitto and Schroten’s (1991) hypothesis, requires the presence of determiners to mark number in the DP. As mentioned at the beginning of this section, this volume shows how also the domain of indefinites, besides articles, can provide insights into the motivation and the development of grammaticalization phenomena. A first clear observation emerging from the contributions dealing with the diachrony of indefinites (Aloni, Kellert) is that synchronic morphological complexity is frequent in indefinites and hints to a grammaticalization scenario. Both Aloni and Kellert, moreover, show how specialized forms of indefinites acquire their meaning contribution by integrating an originally optional pragmatic inference into the conventional meaning. Aloni studies the diachrony of free-choice indefinites in Spanish and Dutch and shows how the historical data can help back up theoretical analyses. Specifically, for her case studies she argues that the association with the two abstract operators licensing the free-choice indefinite in its synchronic uses (the exhaustification operator exh and the universal quantifier [∀]) can be traced back to the source of grammaticalization. The exh operator represents an inheritance from the original wh-morphology (cual- ‘which’ in Spanish cualquier and wie ‘who’ in Dutch wie dan ook), whereas universal quantification over propositional alternatives [∀] finds its origin in earlier universal readings of the construction (free relative clause or ‘no matter’ construction, analyzed as an unconditional) ultimately yielding the indefinite form. A further original contribution of Aloni’s study consists in detailing the stages of the grammaticalization process, where the syntactic aspects intertwine with the semantic ones. From the syntactic point of view, the grammaticalization consists in the loss of the autonomous clausal status of the free relative construction, which becomes part of the main clause. Semantically, the analysis proposes a path through which an originally conversational implicature becomes a conventional implicature and is later reanalyzed as part of the core conventional semantics. Crucially, the grammaticalization does not appear to be completely unidirectional, as is generally conceived of in grammaticalization clines, at least for some stages: the ‘no matter’ use is identified as the initial step of the grammaticalization of the Dutch free-choice indefi-
22
gianollo, von heusinger and napoli
nite, whereas it is the last step of the process in the case of the Spanish form. However, Aloni discusses how it might be plausible, on the basis of theoretical considerations, to conjecture that a ‘no matter’ use existed at earlier, undocumented stages of Spanish, subsequently disappeared and then emerged again more recently. This scenario opens up interesting perspectives for future comparative diachronic studies. Interestingly, Kellert’s analysis of the diachronic development of Italian free-choice indefinite qualunque shows some important similarities to Aloni’s cases, witnessing to the usefulness of comparative approaches. Kellert dedicates her attention to a particular evaluative reading of qualunque, where the free-choice inference is lost, and the indefinite has a non-determiner-like use with the meaning ‘ordinary’. The author shows how this use diachronically emerges from the conventionalization of an originally optional implicature in contexts where qualunque is found in negated and focused nominal phrases. In such contexts, the alternatives evoked by the indefinite are ranked according to a criterion of ‘specialness’, that is strikingness with respect to a pragmatically established value. The conventionalization of this meaning component, whose underlying mechanism is parallel to Aloni’s cases, goes hand in hand with the fixation of the post-nominal positioning for qualunque. A further similarity with Aloni’s case studies is represented by the diachronic role of the relative clause involved in the construction at the source of the diachronic process: data from Old Italian show that the post-nominal indefinite is originally part of a relative clause containing a subjunctive copula, which is later reduced, with the indefinite itself taking over a part of the construction’s modal value, according to the author’s analysis. To conclude, consistently with the Invited Inferencing Theory of Semantic Change (Traugott 1999, Traugott and Dasher 2002), in which ‘invited inferences’ arising from contextual occurrences in the process of interaction between speaker and hearer are viewed as enabling grammaticalization, in both Aloni’s and Kellert’s chapters the context-dependency of change is recognized and a correlation is established between specific uses of a given form and the conventionalization of a new function.
7
Conclusion
In this introductory chapter, we have presented the main common research questions underlying the various contributions collected in this volume and we have summarized their main general results. We hope to have shown how theoretical and empirical considerations are inextricably tied and feed each other in
reference and quantification in nominal phrases
23
this collection, as well as how the cross-linguistic and the diachronic perspectives can cooperate and be instrumental towards a better understanding of the complex systemic effects underlying the grammar of determiners and quantifiers. The presentation of the main issues discussed in this collection had also the goal of defending an integrated approach to grammar, in which syntactic and semantic hypotheses consistently go hand in hand in the theoretical modeling, and are always accompanied by pragmatic considerations. The latter are of paramount importance in accounting for the shape and the division of labor among modules of linguistic competence, from the choice between indefinite forms to the licensing and interpretation of null subjects. We saw the centrality of semantic and pragmatic mechanisms also in a diachronic perspective, in terms of the paradigmatic competition between forms but also in terms of the generation of systematic inferences that are subject to grammaticalization processes. In all these respects, the way ahead was paved by influential work in the past decades, which the contributions in this volume discuss and expand. As many chapters show, a particularly promising direction of research is represented by the combined diachronic and synchronic study of word-internal syntax and semantic compositionality, since morphologically complex forms (such as partitives or indefinites) can provide a hint not just to their historical origin, but also to their interrelations with other elements in the synchronic system.
References Abney, Steven. 1987. The English noun phrase in its sentential aspect. Cambridge, MA: MIT PhD dissertation. Aguilar-Guevara, Ana, Julia Pozas Loyo & Violeta Vázquez-Rojas Maldonado (eds.). 2019. Definiteness across languages. Berlin: Language Science Press. Alexiadou, Artemis, Liliane Haegeman & Melita Stavrou. 2007. Noun phrase in the generative perspective. Berlin: de Gruyter. Allen, Cynthia. 2010. Substantival adjectives in the history of English and the nature of syntactic change. In Rachel Hendery & Jennifer Hendricks (eds.), Grammatical change: Theory and description, 9–25. Canberra: Pacific Linguistics. Balogh, Kata, Anja Latrouite & Robert D. Van Valin Jr. (eds.). 2020. Nominal anchoring: Specificity, definiteness and article systems across languages. Berlin: Language Science Press. Barton, Dagmar, Nadine Kolb & Tanja Kupisch. 2015. Definite article use with generic reference in German: An empirical study. Zeitschrift für Sprachwissenschaft 34(2). 147–173.
24
gianollo, von heusinger and napoli
Barwise, Jon & Robin Cooper. 1981. Generalized Quantifiers and natural language. Linguistics and Philosophy 4. 159–219. Beghelli, Filippo & Tim Stowell. 1997. Distributivity and negation: The syntax of each and every. In Anna Szabolcsi (ed.), Ways of scope taking, 71–107. Dordrecht: Kluwer. Borer, Hagit. 2005. In name only. Structuring sense, volume 1. Oxford: Oxford University Press. Borik, Olga & Berit Gehrke (eds.). 2015. The syntax and semantics of pseudo-incorporation. Leiden: Brill. Borik, Olga & M. Teresa Espinal. 2012. On definite kinds. Recherches linguistiques de Vincennes 41. 123–146. Bošković, Željko. 2008. What will you have, DP or NP? In Emily Elfner & Martin Walkow (eds.), Proceedings of NELS 37, 101–114. Amherst, MA: GLSA. Bošković, Željko. 2012. On NPs and clauses. In Günther Grewendorf & Thomas Ede Zimmermann (eds.), Discourse and grammar: From sentence types to lexical categories, 179–245. Berlin: de Gruyter. Bossong, Georg. 2016. Classifications. In Adam Ledgeway & Martin Maiden (eds.), The Oxford guide to the Romance languages, 63–72. Oxford: Oxford University Press. Bruening, Benjamin, Xuyen Dinh & Lan Kim. 2018. Selection, idioms, and the structure of nominal phrases with and without classifiers. Glossa: A Journal of General Linguistics 3(1, 42). 1–46. Cabredo Hofherr, Patricia & Anne Zribi-Hertz (eds.). 2014. Crosslinguistic studies on noun phrase structure and reference. Leiden: Brill. Cardinaletti, Anna & Giuliana Giusti. 1991. Partitive ne and the QP hypothesis. A case study. University of Venice Working Papers in Linguistics 1. 1–19. Cardinaletti, Anna & Giuliana Giusti. 2018. Indefinite determiners: Variation and optionality in Italo-Romance. In Diego Pescarini & Roberta D’Alessandro (eds.), Advances in Italian dialectology, 135–161. Leiden: Brill. Carlson, Greg. 1977. A unified analysis of the English bare plural. Linguistics and Philosophy 1. 413–457. Carlson, Greg, Rachel Sussman, Natalie Klein & Michael Tanenhaus. 2006. Weak definite noun phrases. In Christopher Davis, Amy Rose Deal & Youri Zabbal (eds.), Proceedings of NELS 36, 179–196. Amherst, MA: GLSA. Chierchia, Gennaro. 1998. Reference to kinds across languages. Natural Language Semantics 6. 339–405. Chomsky, Noam, Ángel Gallego & Dennis Ott. 2019. Generative Grammar and the faculty of language: Insights, questions, and challenges. Catalan Journal of Linguistics, 229–261. Special Issue 2019. Ciardelli, Ivano, Jeroen Groenendijk & Floris Roelofsen. 2019. Inquisitive semantics. Oxford: Oxford University Press.
reference and quantification in nominal phrases
25
Dayal, Veneeta. 2011. Hindi pseudo-incorporation. Natural Language and Linguistic Theory 29. 123–167. Delfitto, Denis & Jan Schroten. 1991. Bare plurals and the number affix in DP. Probus 3. 155–185. Diesing, Molly. 1992. Indefinites. Cambridge, MA: MIT Press. Ebert, Christian, Cornelia Ebert & Stefan Hinterwimmer. 2013. The interpretation of the German specificity markers bestimmt and gewiss. In Cornelia Ebert & Stefan Hinterwimmer (eds.), Different kinds of specificity across languages, 31–74. Berlin: Springer. Evans, Gareth. 1977. Pronouns, quantifiers, relative clauses. Canadian Journal of Philosophy 7(4). 777–797. Fodor, Janet & Ivan Sag. 1982. Referential and quantificational indefinites. Linguistics and Philosophy 5. 355–398. Frajzyngier, Zygmunt. 2011. Grammaticalization of reference system. In Bernd Heine & Heiko Narrog (eds.), The Oxford handbook of grammaticalization, 625–635. Oxford: Oxford University Press. Frascarelli, Mara. 2007. Subjects, topics, and the interpretation of referential pro. An interface approach to the linking of (null) pronouns. Natural Language and Linguistic Theory 25. 691–734. Frascarelli, Mara. 2018. The interpretation of pro in consistent and partial NS languages: A comparative interface analysis. In Federica Cognola & Jan Casalicchio (eds.), Null subjects in Generative Grammar, 211–239. Oxford: Oxford University Press. Geach, Peter. 1962. Reference and generality. An examination of some medieval and modern theories. Ithaca, NY: Cornell University Press. Giusti, Giuliana. 1991. The categorial status of quantified nominals. Linguistische Berichte 136. 438–452. Givón, Talmy. 1981. On the development of the numeral ‘one’ as an indefinite marker. Folia Linguistica Historica 2. 35–53. Grimshaw, Jane. 2005. Words and structure. Stanford: CSLI. Haspelmath, Martin. 1997. Indefinite pronouns. Oxford: Oxford University Press. Heim, Irene. 1982. The semantics of definite and indefinite noun phrases. Amherst, MA: University of Massachusetts PhD dissertation. Heine, Bernd. 1997. Cognitive foundations of grammar. Oxford: Oxford University Press. Herslund, Michael. 2012. Grammaticalisation and the internal logic of the indefinite article. Folia Linguistica 46(2). 341–357. von Heusinger, Klaus. 2002. Specificity and definiteness in sentence and discourse structure. Journal of Semantics 19(3). 245–274. von Heusinger, Klaus. 2011. Specificity. In Klaus von Heusinger, Claudia Maienborn & Paul Portner (eds.), Semantics. An international handbook of natural language meaning, volume 2, 1025–1058. Berlin: de Gruyter.
26
gianollo, von heusinger and napoli
von Heusinger, Klaus & Petra B. Schumacher. 2019. Discourse prominence: Definition and application. Journal of Pragmatics 154. 117–127. Higginbotham, James. 1985. On semantics. Linguistic Inquiry 16. 547–593. Hopper, Paul & Elizabeth Closs Traugott. 2003. Grammaticalization. Cambridge: Cambridge University Press. Horrocks, Geoffrey & Melita Stavrou. 1987. Bounding theory and Greek syntax: Evidence from wh-movement in NP. Journal of Linguistics 23. 79–108. Ihsane, Tabea. 2008. The layered DP: Form and meaning of French indefinites. Amsterdam: Benjamins. Ihsane, Tabea (ed.). 2021. Disentangling bare nouns and nominals introduced by a partitive article. Leiden: Brill. Ioup, Georgette. 1977. Specificity and the interpretation of quantifiers. Linguistics and Philosophy 1. 233–245. Kabatek, Johannes & Albert Wall (eds.). 2013. New perspectives on bare noun phrases in Romance and beyond. Amsterdam: Benjamins. Kamp, Hans. 1981. A theory of truth and semantic representation. In Jeroen Groenendijk, Theo M.V. Janssen & Martin Stokhof (eds.), Formal methods in the study of language. Proceedings of the Third Amsterdam Colloquium, 277–322. Amsterdam: Mathematical Center. Reprinted 2013 in Klaus von Heusinger & Alice ter Meulen (eds.), The dynamics of meaning and interpretation. Selected papers of Hans Kamp, 329–369. Leiden: Brill. Karttunen, Lauri. 1976. Discourse referents. In James D. McCawley (ed.), Notes from the linguistic underground (Syntax and Semantics 6), 363–385. New York: Academic Press. Kayne, Richard. 1994. The antisymmetry of syntax. Cambridge, MA: MIT Press. Kayne, Richard. 2005. Movement and silence. Oxford: Oxford University Press. Keenan, Edward & Jonathan Stavi. 1986. A semantic characterization of natural language determiners. Linguistics and Philosophy 9(3). 253–326. Klockmann, Heidi. 2020. The article a(n) in English quantifying expressions: A default marker of cardinality. Glossa: A Journal of General Linguistics 5(1, 85). 1–31. Kratzer, Angelika & Junko Shimoyama. 2002. Indeterminate pronouns: The view from Japanese. In Yukio Otsu (ed.), Proceedings of the Third Tokyo Conference on Psycholinguistics, 1–25. Tokyo: Hituzi Syobo. Larson, Richard. 2019. The DP hypothesis and (a)symmetries between DP and CP. Linguistic Analysis 42(3–4). 507–548. Leonetti, Manuel. 1999. El artículo. In Ignacio Bosque & Violeta Demonte (eds.), Gramática descriptiva de la lengua española, volume 1, 787–890. Madrid: Espasa. Longobardi, Giuseppe. 1994. Reference and proper names: A theory of N movement in syntax and Logical Form. Linguistic Inquiry 25(4). 609–665. Longobardi, Giuseppe. 2001. How comparative is semantics? A unified parametric theory of bare nouns and proper names. Natural Language Semantics 9. 335–369.
reference and quantification in nominal phrases
27
Lyons, Christopher. 1999. Definiteness. Cambridge: Cambridge University Press. Matthewson, Lisa. 2001. Quantification and the nature of crosslinguistic variation. Natural Language Semantics 9. 145–189. Montague, Richard. 1973. The proper treatment of quantification in ordinary English. In Jaakko Hintikka, Patrick Suppes & Julius M.E. Moravcsik (eds.), Approaches to natural language: Proceedings of the 1970 Stanford Workshop on Grammar and Semantics, 221–242. Dordrecht: Reidel. Onea, Edgar. 2016. Potential questions at the semantics-pragmatics interface. Leiden: Brill. Onea, Edgar & Malte Zimmermann. 2019. Questions in discourse: An overview. In Klaus von Heusinger, Edgar Onea & Malte Zimmermann (eds.), Questions in discourse, volume 1: Semantics, 5–117. Leiden: Brill. Panagiotidis, Phoevos. 2018. Determiners. Oxford research encyclopedia of linguistics. Oxford: Oxford University Press. Partee, Barbara H. 1987. Noun phrase interpretation and type-shifting principles. In Jeroen Groenendijk, Dick de Jongh & Martin Stokhof (eds.), Studies in Discourse Representation Theory and the theory of Generalized Quantifiers, 115–143. Dordrecht: FORIS. Perlmutter, David M. 1970. On the article in English. In Manfred Bierwisch & Karl Erich Heidolph (eds.), Progress in linguistics: A collection of papers, 233–248. The Hague: Mouton. Rijkhoff, Jan. 2002. The noun phrase. Oxford: Oxford University Press. Roberts, Craige. 1996. Information structure in discourse: Towards an integrated formal theory of pragmatics. In Jae-Hak Yoon & Andreas Kathol (eds.), Papers in semantics (Working Papers in Linguistics 49), 91–136. The Ohio State University. Reprinted 2012 in Semantics & Pragmatics 5. 1–69. Russell, Bertrand. 1905. On denoting. Mind 14. 479–493. Schwarz, Florian. 2009. Two types of definites in natural language. Amherst, MA: University of Massachusetts PhD dissertation. Schwarzschild, Roger. 2002. Singleton indefinites. Journal of Semantics 19. 289–314. Shlonsky, Ur. 1991. Quantifiers as functional heads: A study of Quantifier Float in Hebrew. Lingua 84. 159–180. Stark, Elisabeth, Elisabeth Leiss & Werner Abraham (eds.). 2007. Nominal determination. Typology, context constraints, and historical emergence. Amsterdam: Benjamins. Stowell, Tim. 1989. Subjects, specifiers, and X-Bar Theory. In Mark Baltin & Anthony Kroch (eds.), Alternative conceptions of phrase structure, 232–262. Chicago: University of Chicago Press. Stowell, Tim. 1991. Determiners in NP and DP. In Katherine Leffel & Denis Bouchard (eds.), Views on phrase structure, 37–56. Berlin: Springer. Strawson, Peter F. 1950. On referring. Mind 59. 320–344.
28
gianollo, von heusinger and napoli
Szabolcsi, Anna. 1983. The possessor that ran away from home. The Linguistic Review 3. 89–102. Szabolcsi, Anna. 1987. Functional categories in the noun phrase. In István Kenesei (ed.), Approaches to Hungarian 2, 167–190. Szeged: JATE. Traugott, Elizabeth Closs. 1999. The role of pragmatics in a theory of semantic change. In Jef Verschueren (ed.), Pragmatics in 1998: Selected papers from the 6th International Pragmatics Conference, volume 2, 93–102. Antwerp: International Pragmatics Association. Traugott, Elizabeth Closs & Richard B. Dasher. 2002. Regularity in semantic change. Cambridge: Cambridge University Press. Walker, Marilyn A., Aravind K. Joshi & Ellen F. Prince. 1998. Centering in naturally occurring discourse: An overview. In Marylin A. Walker, Aravind K. Joshi & Ellen F. Prince (eds.), Centering Theory in discourse, 1–28. Oxford: Oxford University Press. Williams, Edwin. 1983. Semantic vs. syntactic categories. Linguistics and Philosophy 6. 423–446. Zamparelli, Roberto. 2000. Layers in the Determiner Phrase. New York: Garland. Zamparelli, Roberto. 2008. Bare predicate nominals in Romance languages. In Hendrik Høeg Müller & Alex Klinge (eds.), Essays on nominal determination: From morphology to discourse management, 101–130. Amsterdam: Benjamins.
chapter 2
Definite Plural Generics in English: Evidence from De-adjectival Nominalization Artemis Alexiadou
1
Introduction
As is well known, when it comes to the expression of genericity with plural noun phrases, languages split into two groups (see Carlson 1977, Krifka et al. 1995, Farkas and de Swart 2007 for an overview among many others). On the one hand, languages like English use bare plurals to express generic generalizations and kind reference, (1). On the other hand, Romance languages use definite plurals in these contexts, illustrated in (2) with a French example. The data in (1) and (2) are from Farkas and de Swart (2007: 1659). (1) Dogs are dangerous when they are hungry. (2) Les chiens sont dangereux quand ils ont faim. the dogs are dangerous when they have hunger ‘Dogs are dangerous when they are hungry.’ Earlier literature has pointed out that when an article is present, as in (3), plural noun phrases in English get a specific interpretation. Thus, Longobardi (1994: 653) states that “English never tolerates the use of the definite article with plural and mass generics […]”. Krifka et al. (1995) point out that in English specificity is encoded by definite noun phrases, while genericity is encoded by bare noun phrases (with count nouns). In the same vein, Chierchia (1998) argues that that the definite article lexicalizes specificity in English. Dayal (2004) offers some refinements of Chierchia’s proposal, by highlighting the role of number on indefinite readings of kind-level terms. (3) The dogs are dangerous. Table 2.1 below summarizes the distribution of articles and types of interpretations they receive in English and French, adapted from Krifka et al. (1995: 68), see also Lyons (1999) for related remarks:
© Artemis Alexiadou, 2022 | doi:10.1163/9789004473324_003
30 table 2.1
alexiadou Article distribution
Language Bare English French
Semantics Definite
Dodos are extinct. generic *Dodos sont énteints. –
Semantics
#The dodos are extinct. – Les dodos sont énteints. generic
Farkas and de Swart (2007) relativize this view and argue that there is in fact competition between bare plurals and definite forms for the expression of genericity. They discuss two contexts which show an unexpected neutralization of the contrast in Table 2.1, namely pseudo-generics and so-called anaphoric generics. Focusing on the second case, Farkas and de Swart show that in contexts where kind reference is anaphoric, definite plural generics are possible even in English. Encyclopedia entries provide one such context: they introduce the kind which is then familiar and as a result a definite plural is used to refer to it, as illustrated in (4), from Farkas and de Swart (2007: 1674): (4) Saurischian Bipeds—The saurischians were the first of the two great groups to assume prominence. […] From certain of these forms, the saurischians were certainly derived. (Encyclopaedia Britannica, 1972: 456) More recently, Acton (2019) offers some further examples of definite noun phrases which receive a generic interpretation. He shows that the-plurals, i.e. definite plural generics (henceforth DPGs), can be used in contexts where bare plurals (henceforth BPs) are expected. However, the presence of the definite article adds a social layer of meaning, as shown in the contrast in (5) and (6). Acton argues that the presence of the determiner establishes a distance between the speaker and the referent of the noun phrase. Specifically, the-plural in (5) is used “to talk about all or typical members of a group of individuals; it tends to depict that group as a monolith of which the speaker is not a member and to an extent that using a bare plural (BP) does not.” (Acton 2019: 37–38). (5) The Americans do love cars! (6) Americans do love cars! According to Acton, DPGs are marked, as they default to a specific meaning, but not impossible for use with a generic interpretation. Crone and Frank
definite plural generics in english
31
(2016) report experimental results, which suggest that indeed native speakers can judge definite plurals as generic. According to these authors, generic interpretations of DPGs can be viewed as the result of failure to “ground expressions as referring to particular entities or events”, Crone and Frank (2016: 846). On the basis of these observations, we can thus conclude that DPG s are indeed an option that the English grammar can employ, next to the less marked BPs. As suggested by the aforementioned authors, this leads to a competition between definite and bare forms for the expression of genericity.1 In view of the fact that BPs are the canonical realization of genericity, the contexts in which DPGs are used are restricted and come, perhaps, with an additional flavor of meaning. The fact that DPGs are possible in dialects of English further supports this, as often dialects reflect more canonical uses of what is judged as non-canonical/marked in the standard grammar. (7) is an Irish English example from Filppula (1999: 57), and (8) is a Jamaican English example from Sand (2003: 419). While (7) could be viewed as involving language interference from Irish, language contact cannot generalize across contact situations, as also noted in Sand (2003).2 (7) He didn’t like the women. (8) The men are suddenly finding their egalitarian conscience. The fact that DPGs are allowed albeit restrictively in Present Day English might not come as a surprise in view of the fact that, according to Mustanoja (1960), in Old English (OE), DPGs were possible; it is only during the Middle English period that BPs gain ground steadily, see also Crisma and Pintzuk (2016). Specifically, Mustanoja (1960: 253) notes that “in the majority of OE instances the generic plural occurs with the definite article”. This is shown in (9a), where the use of the definite article would be impossible in Modern English. In
1 Farkas and de Swart (2007) offer an analysis in terms of optimality theory. While I will rely on the core intuition behind their proposal, I will follow Acton (2019) in treating this type of competition in terms of structural complexity. 2 Rehn (2016) shows that Irish English shows identical properties to a variety of German, namely Alemannic, where the determiner is required in generic contexts, unlike Standard German. (i) Do bräuchdsch heit d’Gommistiefel. Da bräuchtest du heute Ø Gummistiefel ‘You would need wellies today.’ This implies that what we find in varieties of English cannot be a contact phenomenon.
32
alexiadou
Middle English further such examples can be found, (9b), from Mustanoja (1960; source: the Ormulum, 360): (9) a. swa feor norÞ saw Þa hwœlhuntan firrest faraÞ as far north as the whalers ever go b. Þatt preostess unnderrfanngenn the priests receive Van Linden and Davidse (2012) find DPGs as late as the Late Modern English period, see (10), their (14), dating from (1879):3 (10) The flowers should keep their honey for the insects. If indeed DPGs are an option that the English grammar has, but speakers default to BPs for the expression of genericity, we would expect them, as showed in Acton (2019), to use DPGs when they want to convey an additional layer of meaning or the context allows it. In cases, however, where there is no other alternative, i.e. no BP form possible, we expect them to only use DPG s. This paper will consider one such case, namely the distribution of D + Adj combinations such as the poor. Combinations of this type are generally acknowledged to have a generic interpretation, and, as will be discussed in Section 2, they reproduce the behavior of Romance DPG s. This has been observed in Kaluza (1981), Longobardi (1994), Chierchia (1998), Lyons (1999), Carlson (2011). While Barton, Kolb and Kupisch (2015) suggest that such usages might point to a grammaticalization of the English article into the generic function, I will argue that this is the one environment where the function of the definite article in English as being part of a generic DP shines through. The paper makes the following contributions; first, it further supports the claim that DPGs are not ungrammatical in English, they are just restricted in distribution. But when BPs cannot be formed as a competing structure, DPGs are the only possible structure to express genericity, as expected from competition-based accounts such as Farkas and de Swart (2007) and Acton (2019). Second, it proposes a syntactic analysis of different types of DPG s adopting Borer (2005). Building on insights in Farkas and de Swart (2007) as 3 In Older as well as in Modern Scots, an English dialect which is considered to be the descendant of Northumbrian OE, the definite article is used in plural contexts, see (i), from Hiles (2008), who claims that Scots preserved the OE usage. (i) the comwnis of Ingland rais aganis thare wikkit king.
definite plural generics in english
33
well as Acton (2019), it will offer a structural analysis of DPG s in general and D + Adj combinations in particular that will account for the similarities between DPGs and D + Adj combinations and explain the differences between DPG s and BPs. BPs, following Borer (2005), are structurally less complex than DPG s. DPGs are also slightly semantically different from BP s: unlike BP s they denote the maximal set of individuals, as argued for in Borik and Espinal (2015). While both DPGs and BPs are in principle possible in a variety of contexts, D + Adj combinations lack an alternative less complex structure to yield a similar interpretation, i.e. they are the only possible structure to express genericity. The paper is structured as follows. In Section 2, I will first discuss D + Adj combinations such as the poor. I will show that such combinations are similar to Romance DPGs and also to marked English DPG s seen in this introduction. In Section 3, I will turn to the structure of the English DP and the layers that are relevant to build DPGs in general and D + Adj combinations in particular. In Section 4, I will address the question of how the Present Day English situation emerged. I will briefly discuss the diachronic development that leads to the DPG vs. BP competition in Present Day English as opposed to Romance languages, discussing French in some detail. In Section 5, I offer some concluding remarks.
2
D + Adj Combinations and Genericity
D + Adj combinations such as the poor have recently emerged as focus of research in the literature on semantics and the morpho-syntax interface of noun phrases, as they constitute a fruitful ground to discuss how natural languages use inflection or derivation to denote entities. See McNally and de Swart (2015), Glass (2013), Alexiadou and Iordăchioaia (2014) among others for discussion. In this paper, I concentrate on examples such as (11). Glass (2013) uses the label individuated type to refer to such cases, which can be paraphrased as adjective + people:4 (11) Too many rich/few rich are unwilling to share.
4 Glass (2013) also discusses examples such as (i), which receive a mass interpretation. A visible morpho-syntactic difference between (12) and (i) is the presence of plural in (12) as opposed to singular verbal agreement in (i): (i) Too much sweet is bad for you.
34
alexiadou
While one might take (11) to be a marginal construction, and available with some adjectives only, Glass (2013) shows, on the basis of corpus data, that it is possible with all classes of adjectives. Some of her examples are given in (12), involving age (12a), value (12b) and speed (12c) adjectives respectively: (12) a. Designing lighting for the elderly requires special consideration. b. The beautiful aren’t simply judged as smarter and more talented. c. The slow are left behind. The generality of this process is also confirmed by Smith’s (2005) detailed British National corpus (BNC) study. As can be seen in (11)–(12), the D + Adj unit agrees in plural form with the verb, suggesting that it is a plural noun phrase. In some cases, illustrated in (13), the adjectives can surface with plural morphology, see also Smith (2005), Glass (2013):5 (13) a. To be down with the youngs. b. Occasionial unhappies were picked up by the police. c. Before the hairies took over. 5 Smith (2005) acknowledges that there is a group of plural forms that seem to have a more lexicalized meaning, shown in (i); note also that riches exist, meaning a large amount of money or possessions or good things. Borer and Roy (2010) also discuss similar such examples. (i) a. earlies = shift/produce b. hards = hard times c. jollies = good times Importantly, however, examples such as the ones in (13) have what Glass calls the individuated reading. By contrast, the forms in (i) bear idiosyncratic nominal interpretations and should not be treated on a par with the examples in (13). Acquaviva (2008), Alexiadou (2011) and others argue that the type of plural in (i) is a nominalizer, i.e. it is not the type of plurality that creates an individuated reading. Thus, while plural on faithfuls really signals the individuated reading, plural on riches does not. In German, all these types, i.e. singular individuated, plural individuated, and lexical plural are morphologically distinct (the goodind vs. the goods; der Gute vs. die Guten vs. die Güter), see also Aschenbrenner (2014). Moreover, note that while e.g. young and youngs both exist, some adjectives do not seem to tolerate plural. I found two instances of plurality on poor in COCA, and there seem to be none in BNC, see (ii): (ii) a. … pulses are used as a part of daily food consumption by rich and poors and their consumption is increasing day by day b. I don’t oppose it because I want to step on the necks of the poors, as you could say. The first example is particularly interesting, as it shows that the plural form can be conjoined with a non-plural form, suggesting that the speaker judges them as equivalent. Finally, a further question that can be raised is whether there is a difference in interpretation between the crazy vs. the crazies. It is not clear to me that this is the case. In view of the fact that co-ordination between plural and non-plural marked adjectives is possible, we may conclude that they are equivalent in interpretation.
definite plural generics in english
35
From the point of view of interpretation, Longobardi (1994), Chierchia (1998), and Giannakidou and Stavrou (1999) all argue that the rich denotes a kind, it means rich people.6 Longobardi (1994: 644) states that the D + Adj are DPGs, as they reproduce the behavior of definite generics in Romance on several counts. First, for the generic reading the definite determiner is required. Second, the bare form is at best marginal and whenever interpreted at all, it seems to have an existential interpretation, as shown in (14), which is not available with DPGs. This means that BPs are not an option to express genericity in this case: (14) a. ??There were young everywhere. b. *Young were everywhere. There is a further similarity between Romance and our D + Adj examples. As has been reported for French and Italian in Heycock and Zamparelli (2003), the determiner can be left out in case of co-ordination of bare nouns, (15). This also holds for our English examples, see (16) from Smith (2005): (15) Tigri e pantere sono in pericolo di estinzione. tigers and panthers are in danger of extinction ‘Tigers and panthers are endangered.’ (16) a. Young and old alike enjoy the rides. b. Rich and poor were wedged together. In Heycock and Zamparelli’s analysis, co-ordination dispenses with the obligatory presence of a determiner. In this respect, co-ordinated BP s in Romance behave similarly to co-ordinated bare adjectives in English. I thus conclude that D + Adj combinations are instances of DPG s and that the presence of a determiner is obligatory in these contexts. (17) shows how D+ Adj combinations fare in comparison to DPGs, see Table 2.1.
6 However, Glass (2013) points out that D + Adj combinations do not always denote a kind, as they can appear with episodic predicates: (i) The young cried and clung to their mothers. Glass (2013) furthermore points out that D + Adj can appear with non-kind selecting determiners, for instance, some, many, and possessives. This will be explained in terms of the structure in (22) to be proposed in Section 3.2. Nevertheless, Glass also states that kind interpretations are the preferred ones, so I will focus on these.
36
(17)
alexiadou
Bare
Semantics Definite
Dodos are extinct. generic *Young were everywhere. –
Semantics
#The dodos are extinct. – The slow are left behind. generic
The important difference between dodos and the slow is the obligatory presence of the determiner in the latter. By contrast, the dodos is not strictly speaking ungrammatical, but restricted in distribution. In the next Section, I will turn to the structure of DPGs in general and the syntax of D + Adj units in particular and offer an explanation of this distribution.
3
The Syntax of Generic D + Adj Combinations
3.1 Distinguishing between DPGs and BPs Structurally Before turning to an analysis of our D + Adj examples, some remarks on the syntax of DPs and generic noun phrases are in order. I follow Longobardi (1994) and assume that only DPs can occupy argument positions in English and in Romance. This means that both BPs and DPGs contain a D layer, the difference being that D remains unrealized in the canonical case in English, while it is always overtly realized in Romance, cf. Chierchia (1998) and Mathieu (2009). Borer (2005) as well as Longobardi (1994) take the structure of generic noun phrases to be identical in English and Romance though the two analyses differ in implementation.7 (18) is the structure of the DP I assume in this paper, adopting and adapting Borer’s (2005) proposal for the internal structure of the noun phrase: (18)
DP D
#P DivP plural
nP
7 In Longobardi’s system the English vs. Romance contrast with respect to generic interpretations is explained as follows: while there is an overt D (expletive) in Romance, in English
definite plural generics in english
37
In (18), D is the locus of definiteness and is typically associated with the presence of definite articles. Following Borer (2005), plural morphology in English introduces division of an undivided space; this takes place in DivP, but see Bale and Barner (2009) for some criticism. Borer argues at length that the plural in English is actually a classifier and not a function from individuals. The creation of individuals comes through counters in #P. In other words, cardinals or quantifiers such as some, which are located in #P, lead to the creation of individuals. Unlike Borer (2005), I assume that there is also an nP layer, which nominalizes acategorial roots, but nothing hinges on this particular assumption. For English, Borer argues that BPs are less complex than definite plurals, in that they lack #P, thus they do not denote individuals. Borer’s argument is based on the well-known contrast in (19), from Borer (2005: 120–121): (19) a. Kim ate apples for an hour/*in an hour. b. Kim ate three apples in an hour. Borer points out that the interpretation of (19a) is consistent with a reading according to which no single apple was eaten. Moreover, bare plurals fail to give rise to telic interpretations, unlike plurals accompanied with numerals in (19b). As both examples contain plural nouns, Borer concludes that it is the presence of counters that leads to a telic interpretation and the existence of individuals. BPs are treated as non-quantity expressions, while definite noun phrases are quantity expressions. As Borer argues, in the presence of a definite determiner, #P always projects. This is so as the definite determiner merges with #. From this perspective, it is first realized under # in (18), the counting projection, and then raises to D.8 The question that arises is whether English BP s are identical to DPG s in Romance. It appears that there are several arguments that English BP s are semantically and syntactically different from Romance DPG s. Borik and Espinal (2015) discuss the following argument: if English BP s and Romance DPGs were identical in all respects, we would expect them to behave alike in all environments. As is well-known, English BPs may have existential readings next to generic interpretations, (20a), from Borik and Espinal (2015: 212). The fact that Romance DPGs never receive existential interpretations in subject positions, unlike BPs in English, suggests to these authors that they are preN-to-D movement takes place at LF in order for the generic interpretation to obtain. Borer (2005) allows for D to remain empty in such cases and receive range by a generic quantifier. 8 Borer also argues that in the case of singular noun phrases, the definite determiner may be merged in Div.
38
alexiadou
sumably not identical semantically and syntactically. They illustrate this with Spanish examples, but the same point can be made for French in (20b) from Gonzalez and Mihoc (2017), a language that uses partitive noun phrases for the existential interpretation. A similar observation can be made on the basis of DPGs and BPs in object position, from Robinson (2005: 119). Robinson argues that (20d) is infelicitous in French because of the maximality associated with the definite determiner in object position. Again, here the partitive would be the felicitous choice. This suggests two things: that French DPG s cannot be equated with English BPs and that the determiner in DPG s in Romance cannot be expletive. (20) a. Lions are destroying my garden. b. Des chiens aboient. of the dogs bark ‘Some dogs bark.’ c. Pigs eat apples. d. #Les cochons mangent les pommes. the-pl pigs eat-3pl the-pl apples ‘The pigs eat the apples.’ Borik and Espinal (2015) propose an analysis, according to which Romance DPGs denote a maximal set of individuals. Specifically, the definite plural makes reference to every member of the set. For Borik and Espinal, this is the result of the combination of plural number, which makes reference to individuals, and the definite article which brings about maximality. This is similar to what Acton (2019) proposes for English the-plurals, which are DPG s, and unlike English BPs, pick out collections of individuals.9 BP s do not allow, however, an interpretation in terms of well-defined collection of individuals. In what follows, I build on these insights and combine them with Borer’s analysis of noun phrases. The function Borik and Espinal attribute to Number is taken over in Borer’s system by counters. As already mentioned, Borer proposed that definite articles can be viewed as counters. Following Borik and Espinal, I will assume that DPGs across languages result from a combination
9 This, Acton (2019) argues, allows the possibility of viewing well defined individuals as a unit and then infer who is seen as part of the group or not.
definite plural generics in english
39
of D and reference to individuals. This is achieved by the presence of an overt determiner that is generated first in #P and then moves to D. Following Borer, I assume that the core difference between the DPG s and BP s is the absence of #P from the structure of BPs, not the D layer, which turns NP s into arguments. It is precisely the presence of #P that leads to the creation of individuals. The absence of #P from the structure of BPs explains why numerals are out with generic NPs. Romance DPGs, by contrast, contain #P, also realized by the article and thus denote the maximal set of individuals. In fact, as I will argue in Section 4, the definite article in Romance realizes also DivP and via head-tohead movement through # surfaces in D. English BP s lack #P and allow for D to remain empty, a point I will briefly turn to in Section 4. 3.2 The Syntax of DPGs Involving Adjectives Let us now turn to the syntax of our D + Adj examples.10 I showed in Section 2 that these are similar to Romance DPGs. Glass (2013) argues in detail that the DP the creative denotes the maximal set of creative individuals in a particular context. As Glass notes, this then creates a need to access this set of individuals linguistically. I propose to map the syntax of D + Adj combinations to the syntax of DPGs. As individuals are needed, I propose to create them syntactically. Importantly, D + Adj combinations are nominalizations of adjectives. Support for this comes from the following areas. As we have seen, adjectives occur with plural morphology. This is also supported by Smith’s (2005) findings on the basis of BNC corpus data (where examples such as the crazies, the faithfuls are provided in addition to the ones discussed in the previous section).
10
It should be noted that the appropriate analysis of D + Adj combinations has been controversially discussed in the literature. These are basically the two families of approaches, see also Marchand (1969) and Aschenbrenner (2014): on the one hand, there are analyses that assume nominalization/substantivization of an adjective, see Giannakidou and Stavrou (1999), Glass (2013), without considering their internal nominal structure. Perhaps Sleeman’s (2013) discussion of Dutch D + Adj combinations is an exception. On the other hand, there are several analyses that assume that we have true attributive adjectives, which modify a null N (related to Pullum’s (1975) notion of people-deletion), similar to cases of ellipsis. This is the line adopted in Lobeck (1995), Kester (1996), Longobardi (1994), Chierchia (1998), Olsen (1988), den Dikken (2001), Haumann (2003), Yamamura (2012), McNally and de Swart (2015), Borer and Roy (2010), Günther (2013, 2018) among others. Günther (2018) notes that the use of one seems to relax the requirement on plurality, as it allows for singular and indefinite noun phrases. For instance, she discusses strings such as he is a tough one to turn down, which she takes as evidence for the ellipsis analysis. However, the singular examples with one as well as the plural forms of one + Adj she discusses are not generic, the noun phrases refer to a specific person or group of people in a given context.
40
alexiadou
Assuming that plural morphology attaches to nominal forms only, pluralization supports a nominalization analysis. A second piece of evidence comes from the observation that they can be modified by an adjective. Assuming, as is standard, that adjectives modify nouns, this is also evidence in favor of nominalization, see (21) from Smith’s (2005) corpus data: (21) a. old faithfuls b. young marrieds Building on Ferrari (2005), Glass (2013) and Sleeman (2013), I take it that our D+ Adj examples involve nominalization. For the purposes of this paper is not crucial whether the nominalization applies to a categorized adjective or a bare root. As Olsen (1988), Glass (2013) and Günther (2018) have shown, some adjectival properties are still visible: comparative and even superlative forms may also appear in the D+ Adj construction, e.g., the poorer. I will thus assume that what is nominalized is an adjective. Crucially, as in Giannakidou and Stavrou (1999), a property turns into an individual. (22) illustrates the structure proposed: (22)
DP the
#P the
DivP
plural
nP aP rich
In (22), D is where the determiner surfaces. In view of the fact that plural morphology is often seen, a Div layer is also included. The definite article is first merged under #, as explained in Section 3.1, and creates reference to individuals.11 As only nominal forms can be pluralized and be modified by adjectives, an nP is part of the structure. n combines with Div and plurality appears on 11
Rehn (2016) also pointed out that since this reading brings about quantification, the determiner should be introduced in #. This would explain, as Rehn observes, the impossibility
definite plural generics in english
41
the noun. The layers of structure compositionally then build the interpretation of D + Adj combinations, i.e. the rich denotes the maximal set of individuals instantiating the property. (22) allows us to explain the obligatory presence of the determiner in the structure, following Borer’s analysis of the in quantity expressions. The plural-only interpretation of these noun phrases in English is unlike what we find in German (23), see also Günther (2018) for discussion, or even earlier stages of English, where singular forms could have generic interpretations, see the Middle English example in (24), from Aschenbrenner (2014: 168). The counterpart of (23) was also found in Old English, as we will see below: (23) der Gute / ein Guter the good-sg / a good-sg (24) The povere is bore as is the riche. the poor is born as is the rich ‘The poor person is born the same way as the rich person.’ The plural only interpretation seems to be a result of diachronic change. As Allen (2010) reports, a systematic change happened in the history of English, as singular nominalized forms were common bearing a generic interpretation.12 Allen shows that during the Old English period, 78% of nominalized adjectives are plural definites in Early West Saxon, while 66 % of them are plural definites in Late West Saxon. She thus concludes that “the modern restrictions reflect usage patterns which go back more than a thousand years.” (Allen 2010: 19). Allen furthermore points out that the nominalized adjectives that are no longer possible in Modern English are the ones that had a low frequency in earlier stages of the language. By comparing D + Adj combinations in Early West
12
of overt numerals e.g. *the five rich. In fact, Smith (2005) says that in her corpus there are no instances of e.g. three or four dead, only three or four of the dead. According to Olsen (1988), the generic reading of these nominalizations is incompatible with numerals. While BP s do not allow numerals, as they lack #P, DPG s do not allow numerals, as the article is generated in #P. Moreover, since D + Adj combinations contain #P, we can explain why some and many are also licit, see footnote 6. A note on the form of the nominalized adjectives in OE is in order. As these appear after the determiner, they bear the weak adjectival endings, which however are identical with the paradigm of the weak noun, the n-declension (Hogg and Fulk 2011). As Aschenbrenner (2014) notes, OE adjectives and nouns were in fact indistinguishable. She further cites forms from earlier stages of English with overt plural morphology, for example, poveren ‘poor-pl’. This all supports the nominalization account.
42
alexiadou
Saxon, Late West Saxon, inflection-rich Early Middle English, and inflectionpoor Early Middle English, Allen stresses that the definite plural is the most frequent formation in all periods. Allen thus concludes that the loss of inflection led speakers to interpret the D + Adj combinations as plural, since this was the most frequent construction throughout periods of English. On this analysis, what we have is a case of a plural definite generic construction, which is otherwise not the default option in Present Day English. On the basis of the structure in (25), DPGs involve the presence of both a definite article in D, and #, and plurality in Div. As a result, (25) is interpreted similar to Romance DPGs, and the realization of the determiner does not introduce an additional layer of meaning, as was the case in Acton’s (2019) examples. This explains why, similar to Romance DPGs, they refer to the maximal set of individuals. (25) [DP the [#P the count [DivP plural [nP [aP [√rich ]]]]]] There are thus two structures for the expression of genericity in English, illustrated in (26), which differ in semantic and syntactic complexity. Importantly, (26a) competes with (26b) in examples such as (5–6), e.g. The Americans vs. Americans. By contrast, the rich does not have a competing structure, it has only the one in (26a), similar to Romance DPGs: (26) a. [DP the [#P the count [DivP plural [nP]]]] b. [DP [DivP plural [nP]]]
DPG s BP s
Acton (2019), building on Katzir (2007), argues that while (26b) is less marked than (26a), as it is less structurally complex, it is not less informative. Given the availability of the less complex BPs, DPGs are thus marked in generic uses. This leads to the creation of the distancing effect of the-plurals, as there is a less complex alternative. By contrast, Acton notes, in languages where a determiner is required for genericity, we should not expect to find such an effect as there is no less complex alternative. This is the case in Romance. Importantly, such an effect is also absent from English DPGs involving adjectival nominalizations, as similar to Romance there is basically no less complex alternative structure that speakers may use. In other words, there is no competition between (26a) and (26b) in the case of deadjectival nominalizations, which only have the structure in (26a). This now raises the following questions: how did the competition between the two structures emerge in Present Day English and why is it absent in Romance? I will show in the next section that this is a result of diachronic
definite plural generics in english
43
change. Specifically, while (26a–b) were both available in Old French, (26b) is no longer available in Modern French. As will also be discussed in the next section, this relates to changes affecting the realization of plurality in the language.
4
English vs. Romance DPs and Genericity
Epstein (1995) argued that (26b) became the canonical realization of genericity in English due to what he labels “generalization of zero”.13 By contrast, in French the reverse development is observed, which led to the establishment of (26a) as the only structure available for genericity. As shown in (27), from Yvain 478–479, cited in Becker (2013), the zero article was common in Old French in generic contexts, while the definite article must be used in Modern French, see also Mathieu (2009): (27) Tant i fuis que j’oi venir Chevaliers. so often there I be that I hear come knights ‘I was always there when I heard the arrival of knights.’ We have already seen in (9) that in OE DPGs were possible. Crisma (2011: 185) reports, however, that generic interpretations with BP s in Old English cannot be excluded. A similar situation is reported by Mathieu (2009) for Old French: BPs and DPGs are both possible in Old French. Carlier (2001) shows that, when one looks at a translation of a Latin text in Old French and Modern French, it can be seen that in Old French BPs appear 46,4% of the time, while this figure reduces to 18,4% in Modern French. While Carlier reports on articleless NP s independent of interpretation, Mathieu (2009), focusing on BP s with a generic interpretation, notes that in Old French the presence of a determiner is conditioned by two properties: focus, and phonological/metric considerations. When dealing with earlier stages of a language, the question is whether these have a DP layer or not. OE did have such a layer. The definite determiner in English, as in many other languages, started out as a demonstrative. Crisma (2011) shows that the demonstrative had developed into a definite article by the late 9th century in OE, see also Wood (2003), and Allen (2006, 2007).
13
A further domain where this zero generalization process is observed is discussed in Epstein (1995). Epstein notes that in Old English the presence of the article was obligatory with titles, e.g. ce kyning Ælfred ‘the king Alfred’. In Present Day English titles are not introduced by an article, an article is present only in the context of modification.
44
alexiadou
Evidence for this comes from the following set of facts. As has been discussed in Longobardi (1994), in Italian, when an adjective precedes the proper name, the article becomes obligatory. (28) a. Ho incontrato (*il) Gianni. have-1sg met the Gianni ‘I met Gianni.’ b. Ho incontrato (*il) vecchio Gianni. have-1sg met the old Gianni ‘I met the old Gianni.’ In OE, normally proper names are used without determiners. According to Crisma (2011), however, when an adjective is present the rate of determiners goes up to 94,3%. This distribution seems not so different from the facts that Longobardi (1994) discusses for Modern Italian, the difference being that while in Italian N can raise to D, when no D is present, this is no longer possible in English. Thus, both OE and Old French have DPGs as well as BP s, but BP s are no longer available in French. According to Carlier (2007) and Mathieu (2009), definite determiners become obligatory in French following the loss of plural marking on nouns.14 Specifically, Carlier notes that in French, the weakening of the plural affix starts from the 13th century and finds its end in the 15th century. The loss of plural marking on the noun suggests that there is no formal difference between singular non-count nouns and plural count nouns. A new article, the partitive is introduced, which conveys the meaning of unspecified quantity, and thus is in opposition with the indefinite article derived from the numeral one (Carlier 2007: 32). The development described in Carlier suggests that French in particular grammaticalizes the distinction singular vs. plural only in the determiner system, and not on the noun. As Carlier explicitly states (Carlier 2007: 37), we see uniform expression of number in prenominal position by a free morpheme, see Pomino (2016).15
14
15
The situation seems similar in other Romance languages as well, see Stark (2008) and references therein. Stark (2002) reports the presence of bare plurals in a generic function in Old Italian. Stark (2008) entertains the hypothesis that the whole re-arrangement of the Romance article system relates to the loss of declension classes, gender and number marking, thus articles take over the role of nominal inflection markers. Mathieu (2009) also puts forth a formal analysis of this change in French, according to which the determiner system emerged due to the loss of phi-features on nouns.
definite plural generics in english
45
By contrast, the development of the English plural morphology took a different path. Wood (2003) and Ackles (1997) provide arguments that the functional categories associated with number morphology and counting (DivP and #P in our terminology) emerged late in OE, while the DP layer was there earlier on. In particular, Ackles suggests that these emerged through the transition from Old to Middle English. The evidence he presents comes from a variety of domains, among other things the fact that Old English lacked an indefinite article, did not distinguish between mass and count quantifiers (for example much vs. many occur with both types of nouns), and from the case marking on nouns accompanied by numerals. Specifically, nouns after numerals inflected for genitive case, e.g. sume hundred scipa ‘some hundred ships-gen.pl’. The relationship is no longer one of an N-N dependency in Present Day English, suggesting that a reanalysis has taken place. This description seems to suggest that the counting function of the English determiner developed later. In other words, it could very well be that in OE (26b) was the structure for both BPs and DPGs, and it is (26a) that emerged later, following the establishment of counters. If that is the case, then presumably in OE the definite article merged directly in D. By contrast, Mathieu (2009) argues that Old French determiners merge in #P, i.e. are counters, meaning that Old French determiners could be analyzed in terms of the structure in (26a). Thus, in Old French both (26a) and (26b) existed, which explains the reason why in certain contexts only (26a) was possible and had a focalizing effect. On the basis of these observations, (26a), is available in both English and Romance, while (26b) is available in English only, both structures repeated below. (26b) was possible in Old French next to (26a), while (26a) emerged in English once the counting system was in place. In Present Day English, the definite determiner lexicalizes the DP as well as the #P, while in Present Day French, the definite determiner lexicalizes the DP, #P and DivP.16 In English DivP and nP are lexicalized by plural and nominal forms respectively, which combine in one unit. While English D may remain empty in the case of BP s, D must be realized in French, as according to Mathieu (2009), it is the locus of phi-features and not n. In English, n is the locus of phi-features not D.17
16 17
This enables us to accommodate Pomino’s (2016) analysis, who postulates two loci for French plurals. Several authors have argued that plural number is on D in Modern Romance, but on N in English (Bouchard 2002). According to Bouchard (2002), D-languages such as French allow N-ellipsis and disallow bare nouns, while N-languages such as English show the reverse behavior.
46
alexiadou
(26) a. [DP [#P count [DivP plural [nP]]]] b. [DP [DivP plural [nP]]]
English/French DPG s English BP s
The comparison of these two languages supports the view that whenever determiners express phi-features, they are obligatorily realized, as Giusti (2001) argued in detail.
5
Conclusion
In this paper, I provided further support for the claim that DPG s are not illicit in English, but they are restricted in distribution. In particular, focusing on D + Adj combinations, I proposed that these indeed are associated with a structure which does not compete with another one for the expression of genericity, thus the determiner is obligatorily realized. Importantly, in contexts, where BPs cannot be used, as in our D + Adj examples, DPG s are the only possible structure available for the expression of genericity, as expected from the discussion in Farkas and de Swart (2007) and Acton (2019). I also briefly commented on the diachronic development that led to generalization of zero determiners as opposed to Romance. The picture that emerges raises the question of whether we can address the variation observed in German for the expression of genericity along the same lines (see Barton, Kolb and Kupisch 2015 for a recent discussion). In German, both DPGs and BPs are used to express genericity. Acton (2019) discusses Swedish and Dutch, which behave similarly to English: these are languages that use BPs as the default form for generic readings, and when a determiner is present it creates a distancing effect, e.g., Dutch Belgen vs. De Belgen (Belgians vs. The Belgians). It remains to be seen if the presence of a determiner creates a distancing effect in German as well; if that were the case, then it would suggest that (26a–b) compete for the expression of genericity in German and the more complex structure has a marked meaning.
Appendix: The Syntax of OE DP In this appendix, I will show that the OE DP is similar to the Modern Romance DP, as perhaps expected in view of the remarks made in Section 4.18 A first piece
18
I am grateful to Giuseppe Longobardi for directing my attention to these issues.
definite plural generics in english
47
of evidence for the similarities between Old English and Modern Romance comes from the structure of possessives. As is well-known, in languages such as Italian possessive adjectives co-occur with the definite article, while this is no longer possible in Modern English: (1) a. la sua casa the his home (Cardinaletti 1998: 18) b. *the his home Cardinaletti (1998) labels Italian type possessives weak, while she classifies the English type as clitics. This distinction corresponds to a difference between adjectival and head possessives. Alexiadou (2004), building on Demske (2001), looks at the structure of possessives in Old English and shows that possessives in Old English behave like their Modern Romance counterparts. Consider the example in (2), where the possessor co-occurs with the demonstrative, from Demske (2001: 190): (2) þes min gefea is gefylled. this my joy is extinct ‘My joy is extinct.’ Demske (2001) provides a series of arguments that the possessive in Old English is adjectival, as is the case in Modern Italian. Importantly, as in other Modern Germanic languages, the appearance of weak and strong adjectival inflection correlates with the (in)definiteness of the respective determiner, see (3), from Demske (2001: 191): (3) a. [on þone gren-an weald] into the green wood-masc.sg.acc ‘into the green wood’ (Genesis 287.835) b. [sum-ne adlig-ne mannan] some noble man-masc.sg.acc ‘some noble man’ (Ælfrics Homilies ii.31 295.239) Wood (2003) observes that there is a difference between the earlier and the newer version of Gregory’s dialogue: the determiner is missing in the later version, while it is present in the earlier version. This is illustrated in the examples below, taken from Wood’s work (chapter 4):
48
alexiadou
(4) a. min þæt ungesælige mod my that unfortunate spirit (GD:C 4.9) b. min ungesælige mod ‘that unfortunate spirit of mine’ (GD:H 4.9) Wood thus concludes that (4a) is the old pattern, while (4b) seems to be the new pattern. We can then propose the structure in (5) for old pattern possessors, see Alexiadou (2004): (5) [DP the [FP his[np [n [NP wisdom]]]] Finally, Fischer (2001) and Haumann (2010) provide a detailed discussion of the placement of adjectives in OE, and convincingly show that these had similar properties to their Modern Romance counterparts. In OE, unlike the situation in Modern English, adjectives could appear in both pre-nominal and post-nominal position. According to Fischer and Haumann, the two positions differ in the following aspects, shown in (6), from Haumann (2010: 66):
(6) Prenominal position strong/weak inflection given information individual level reading non-restrictive reading
Postnominal position strong inflection new information stage level reading restrictive reading
Inflection aside, OE postnominal adjectives pattern with their Romance counterparts (see Alexiadou 2001, Cinque 2010 among others for a discussion of the behavior of postnominal adjectives in Romance). While the properties associated with the postnominal position are still observable in Modern English with a limited number of adjectives, the pattern was general in Old English, see (7), from Haumann (2010: 63), her (27b) and (28a) respectively: (7) a. Forhwon ne recst þu us þone hwitan⟨WK⟩ hlaf loaf why not give you us the white ‘Why do you not give us the white loaf (of bread)?’ (Bede 2, 5.112.9)
definite plural generics in english
49
b. Hinguar ure cyning cene⟨STR⟩ and sigefæst⟨STR⟩ Hingwar our king bold and victorious ‘Our bold and victorious king Hingwar’ (ÆLS (Edmund) 48) As Haumann (2010) observes, the adjective in prenominal position refers to a permanent property of the noun, while the adjective in post-nominal position refers to a temporary one. This is reminiscent of Modern Italian, where prenominal adjectives are unambiguously individual level, while post-nominal ones can have stage level interpretations, as shown in (8), from Cinque (2010: 7): (8) a. Le invisibili stelle di Andromeda sono molto distanti. the invisible stars of Andromeda are very far ‘Andromeda’s stars, which are generally invisible, are very far.’ b. Le stelle invisibili di Andromeda sono molto distanti. the stars invisible of Andromeda are very far ‘Andromeda’s stars which are generally visible, but which happen to be invisible now, are very far.’ In conclusion, the above supports the hypothesis that the syntax of OE DP s shows interesting similarities to that of Modern Romance.
Acknowledgements I am indebted to one anonymous reviewer, the editors of this volume and the series editor for their comments and remarks, which greatly improved this contribution. Many thanks to audiences at Humboldt University of Berlin, and participants in the Roots iv Workshop at New York University in July 2015, the Masterclass Language Variation in Action hosted by the Dutch Academy of Sciences in February 2016, Workshop on Reference at the University of Cologne in May 2016 and the Dimensions of D workshop at the University of Rochester in September 2016 for discussions. The DFG grant AL 554/8–1 and the Collaborative Research Center 1412 Register, 416591334, are hereby acknowledged.
50
alexiadou
References Ackles, Nancy M. 1997. Historical syntax of the English article in relation to the count/noncount distinction. Washington: University of Washington PhD dissertation. Acton, Eric K. 2019. Pragmatics and the social life of the English definite article. Language 95. 37–65. Acquaviva, Paolo. 2008. Lexical plurals. Oxford: Oxford University Press. Alexiadou, Artemis. 2001. Functional structure in nominals: Nominalization and ergativity. Amsterdam: Benjamins. Alexiadou, Artemis. 2004. On the development of possessive determiners: Consequences for DP structure. In Eric Fuß & Carola Trips (eds.), Diachronic clues to synchronic grammar, 31–58. Amsterdam: Benjamins. Alexiadou, Artemis. 2011. Plural mass nouns and the morpho-syntax of number. Proceedings of WCCFL 28. 33–41. Alexiadou, Artemis & Gianina Iordăchioaia. 2014. Two syntactic strategies to derive deadjectival nominals. Acta Universitatis Wratislaviensis 3585. 67–85. Allen, Cynthia. 2006. Possessives and determiners in Old English. In Terttu Nevalainen, Juhani Klemola & Mikko Laitinen (eds.), Types of variation: Diachronic, dialectal and typological interfaces, 149–170. Amsterdam: Benjamins. Allen, Cynthia. 2007. Variation in the NP/DP in Old English: Determiner and possessive combinations. In Annie Zaenen, Jane Simpson, Tracy Holloway King, Jane Grimshaw & Joan Maling (eds.), Architectures, rules, and preferences: Variations on themes by Joan W. Bresnan, 3–20. Stanford: CSLI Publications. Allen, Cynthia. 2010. Substantival adjectives in the history of English and the nature of syntactic change. In Rachel Hendery & Jennifer Hendricks (eds.), Grammatical change: Theory and description, 9–25. Canberra: Pacific Linguistics. Aschenbrenner, Anne. 2014. Adjectives as nouns, mainly as attested in Boethius translations from Old to Modern English and in Modern German. München: Utz. Bale, Alan C. & David Barner. 2009. The interpretation of functional heads: Using comparatives to explore the mass/count distinction. Journal of Semantics 26. 217–252. Barton, Dagmar, Nadine Kolb & Tanja Kupisch. 2015. Definite article use with generic reference in German: An empirical study. Zeitschrift für Sprachwissenschaft 34. 147– 173. Becker, Martin. 2013. Competing zones: The zero article in Old French and its loss in the history of the French language. In Sofiana Chiriacescu (ed.), Proceedings of the vi Nereus International Workshop ‘Theoretical implications of the syntax-semantics interface in Romance’, 1–18. Konstanz: Fachbereich Sprachwissenschaft, Universität Konstanz. Borer, Hagit. 2005. In name only. Structuring sense, volume 1. Oxford: Oxford University Press.
definite plural generics in english
51
Borer, Hagit & Isabelle Roy. 2010. The name of the adjective. In Patricia Cabredo Hoffher & Ora Matushansky (eds.), Adjectives: Formal analyses in syntax and semantics, 85– 114. Amsterdam: Benjamins. Borik, Olga & M. Teresa Espinal. 2015. Reference to kinds and other generic expressions in Spanish: definiteness and number. The Linguistic Review 32. 167–225. Bouchard, Denis. 2002. Adjectives, number and interfaces: Why languages vary. Amsterdam: Elsevier. Cardinaletti, Anna. 1998. On the deficient/strong opposition in possessive systems. In Artemis Alexiadou & Chris Wilder (eds.), Possessors, predicates and movement in the DP, 17–53. Amsterdam: Benjamins. Carlier, Anne. 2001. La genèse de l’article un. Langue française 130. 65–88. Carlier, Anne. 2007. From preposition to article: The grammaticalization of the French article. Studies in Language 31. 1–49. Carlson, Gregory N. 1977. Reference to kinds in English. Amherst, MA: University of Massachusetts PhD dissertation. Carlson, Gregory N. 2011. Genericity. In Klaus von Heusinger, Claudia Maienborn & Paul Portner (eds.), Semantics: An international handbook of natural language meaning, vol. 2, 1153–1185. Berlin: de Gruyter. Chierchia, Gennaro. 1998. Reference to kinds across languages. Natural Language Semantics 6. 339–405. Cinque, Guglielmo. 2010. The syntax of adjectives: A comparative study. Cambridge, MA: MIT Press. Crisma, Paola. 2011. The emergence of the definite article in English: A contact-induced language change? In Harry Perridon & Petra Sleeman (eds.), The noun phrase in Germanic and Romance, 175–192. Amsterdam: Benjamins. Crisma, Paola & Susan Pintzuk. 2016. An from Old to Middle English. In Sten Vikner, Henrik Jørgensen & Elly van Gelderen (eds.), Let us have articles betwixt us: Papers in historical and comparative linguistics in honour of Johanna L. Wood, 161–184. Aarhus: Aarhus University. Crone, Phil & Michael C. Frank. 2016. Inferring generic meaning from pragmatic reference failure. In Anna Papafragou, Daniel Grodner, Daniel Mirman & John Trueswell (eds.), Proceedings of the 38th Annual Conference of the Cognitive Science Society, 842–847. Austin, TX: Cognitive Science Society. Dayal, Veneeta. 2004. Number marking and (in)definiteness in kind terms. Linguistics and Philosophy 27. 393–450. Demske, Ulrike. 2001 Merkmale und Relationen: Diachrone Studien zur Nominalphrase des Deutschen. Berlin: de Gruyter. Dikken, Marcel den. 2001. “Pluringulars”, pronouns and quirky agreement. The Linguistic Review 18. 19–41. Encyclopaedia Britannica, 14th edn. 1972. London: Encyclopaedia Britannica.
52
alexiadou
Epstein, Richard. 1995. The later stages in the development of the definite article: Evidence from French. In Henning Andersen (ed.) Historical Linguistics 1993: Selected papers from the 11th International Conference on Historical Linguistics, Los Angeles, 16–20 August 1993, 159–175. Amsterdam: Benjamins. Farkas, Donka F. & Henriette de Swart. 2007. Article choice in plural generics. Lingua 117. 1657–1676. Ferrari, Franca. 2005. A syntactic analysis of the nominal systems of Italian and Luganda: How nouns can be formed in the syntax. New York: New York University PhD dissertation. Filppula, Markku. 1999. The grammar of Irish English: Language in Hibernian style. London: Routledge. Fischer, Olga. 2001. The position of the adjective in (old) English from an iconic perspective. In Olga Fischer & Max Nänny (eds.), The motivated sign: Iconicity in language and literature 2, 249–276. Amsterdam: Benjamins. Giannakidou, Anastasia, & Melita Stavrou. 1999. Nominalization and ellipsis in the Greek DP. The Linguistic Review 16. 295–331. Giusti, Giuliana. 2001. The birth of a functional category: From Latin ille to the Romance article and personal pronoun. In Guglielmo Cinque & Giampaolo Salvi (eds.), Current studies in Italian syntax: Essays offered to Lorenzo Renzi, 157–171. Amsterdam: North Holland. Gonzalez, Aurore & Teodora Mihoc. 2017. Bare nominals in French and Romanian. Paper presented at LSRL 47, University of Delaware. Glass, Lelia. 2013. Deriving the two readings of English determiner + adjective. Proceedings of Sinn und Bedeutung 18. 164–181. Günther, Christine. 2013. The elliptical noun phrase in English: Structure and use. New York: Routledge. Günther, Christine. 2018. The rich, the poor, the obvious: Arguing for an ellipsis analysis of adjectives used as nouns. In Alex Ho-Cheong Leung & Wim van de Wurff (eds.), The noun phrase in English: past and present, 77–112. Amsterdam: Benjamins. Haumann, Dagmar. 2003. The postnominal ‘and adjective’ construction in Old English. English Language and Linguistics 7. 57–83. Haumann, Dagmar. 2010. Adnominal adjectives in Old English. English Language and Linguistics 14. 53–81. Heycock, Caroline & Roberto Zamparelli. 2003. Coordinated bare definites. Linguistic Inquiry 34. 443–469. Hiles, Laura. 2008. The use of the definite article in Scots English and its relation to Gaelic, Middle and Old English. Tampere: University of Tampere MA thesis. Hogg, Richard & R.D. Fulk. 2011. A grammar of Old English: Vol. 2 Morphology. WileyOnline Library. https://onlinelibrary.wiley.com/doi/pdf/10.1002/9781444327472. Kaluza, Henryk. 1981. The use of article in contemporary English. Heidelberg: Groos.
definite plural generics in english
53
Katzir, Roni. 2007. Structurally defined alternatives. Linguistics and Philosophy 30. 669– 690. Kester, Ellen-Petra. 1996. The nature of adjectival inflection. Utrecht: LOT PhD dissertation. Krifka, Manfred, Francis Jeffry Pelletier, Gregory N. Carlson, Alice ter Meulen, Godehard Link & Gennaro Chierchia. 1995. Genericity: An introduction. In Gregory N. Carlson & Francis Jeffry Pelletier (eds.), The Generic Book, 1–124. Chicago & London: The University of Chicago Press. Linden, An van & Kristin Davidse. 2012. The role of the accessibility of the subject in the development of adjectival complementation from Old English to Presentday English. In Bettelou Los, María José López-Couso & Anneli Meurman-Solin (eds.), Information structure and syntactic change, 199–227. Oxford: Oxford University Press. Lobeck, Anne. 1995. Ellipsis: Functional heads, licensing, and identification. Oxford: Oxford University Press. Longobardi, Giuseppe. 1994. Reference and proper names: A theory of N-movement in syntax and logical form. Linguistic Inquiry 25. 609–665. Lyons, Christopher. 1999. Definiteness. Cambridge: Cambridge University Press. Marchand, Hans. 1969. The categories and types of Present Day English word formation: A synchronic-diachronic approach. München: C.H. Beck’sche Verlag. McNally, Louise & Henriette de Swart. 2015. Reference to and via properties: The view from Dutch. Linguistics and Philosophy 38. 315–362. Mathieu, Eric. 2009. From local blocking to Cyclic Agree: The role and meaning of determiners in the history of French. In Jila Ghomeshi, Ileana Paul & Martina Wiltschko (eds.), Determiners: Variation and universals, 123–157. Amsterdam: Benjamins. Mustanoja, Tauno F. 1960. A Middle English syntax. Helsinki: Société Néophilologique. Olsen, Susan. 1988. Das „substantivierte“ Adjektiv im Deutschen und Englischen: Attribuierung vs. syntaktische „Substantivierung“. Folia Linguistica 12. 337–372. Pullum, Geoff. 1975. People deletion in English. Ohio State University Working Papers in Linguistics 14. 95–101. Pomino, Natascha. 2016. On the clitic status of the plural marker in phonic French. In Susann Fischer & Mario Navarro (eds.), Proceedings of the vii Nereus International Workshop: ‘Clitic doubling and other issues of the syntax/semantic interface in Romance DPs’, 105–130. Arbeitspapier 128. Konstanz: Fachbereich Sprachwissenschaft, Universität Konstanz. Rehn, Alexandra. 2016. The definite article and its position in the structure of the DP: Evidence from Irish English and Alemannic. Caderno de Squibs: Temas em estudos formais da linguagem 2. 54–66.
54
alexiadou
Robinson, Heather Merle. 2005. Unexpected (in)definiteness: plural generic expressions in Romance. New Jersey: The State University of New Jersey PhD dissertation. Sand, Andrea. 2003. The definite article in Irish English and other contact varieties of English. In Hildegard L.C. Tristram (ed.), The Celtic Englishes iii, 413–430. Heidelberg: Winter. Sleeman, Petra. 2013. Deadjectival human nouns: Conversion, nominal ellipsis or mixed categories. Linguistica 8. 159–180. Smith, Christine Anne. 2005. La substantivation des adjectifs en anglais contemporain. Paris: Université Paris-Sorbonne doctoral dissertation. Stark, Elisabeth. 2002. Indefiniteness and specificity in Old Italian texts. Journal of Semantics 19. 315–332. Stark, Elisabeth. 2008. Typological correlations in nominal determination in Romance. In Hendrik H. Müller & Alex Klinge (eds.), Essays on nominal determination: From morphology to discourse management, 45–63. Amsterdam: Benjamins. Wood, Johanna. 2003. Definiteness and number: Determiner phrase and number phrase in the history of English. Arizona: Arizona State University PhD dissertation. Yamamura Shuto. 2012. A minimalist approach to ellipsis in the history of English. Nagoya: University of Nagoya PhD dissertation.
chapter 3
Quantification and Classification in Romance Plural Indefinites: From Number to Seinsart? Mario Squartini
1
Introduction
Due to the grammaticalization of determiners with partitive etymology (“partitive” articles), some Romance languages have developed a system of indefinite markers whose interpretation as “classifiers” was originally suggested by Herslund (1998, 2003a, 2003b). If the classificational potential (Stark 2005, 2007) of these forms is clear in the paradigmatic opposition between the indefinite singular determiner that marks single bounded entities (French un livre ‘a book’) vs the partitive form that occurs with mass and abstract nouns (de l’ eau ‘some water’), how should we interpret plural indefinites marked by a plural form of the same “partitive” article (des livres ‘some books’)? Can the occurrence of these plural forms be conceived in classificational terms, thus suggesting an interpretation of plurality in which plurals and masses are semantically equated (Chierchia 1998a)? These questions will be elaborated in what follows by showing that, in fact, what Romance data confirm is the composite nature of plurality (possibly bi-featural, see Cowper and Hall 2014), in which plurals can be interpreted either as unbounded aggregates or as additive sets of single bounded entities ([Discrete] in Cowper and Hall’s 2014: 69 terminology). I will describe these oppositions by considering them as realizations of what Rijkhoff (2002) calls Seinsart, a semantic category that denotes the different conceptualizations of nominal referents in terms of their internal homogeneity and shape as bounded / unbounded entities. This perspective is compatible with Herslund’s (2012) conclusions on the different role of classification and quantification in the French forms with partitive origin as opposed to the Spanish system, in which partitive articles do not occur. However, instead of concentrating on formal differences, my analysis will highlight some fundamental similarities within the general Romance paradigm in order to grasp what these similarities tell us as far as the interactions of the classificational properties of the Noun Phrase and the morphosyntactic category expressed by number are concerned. To investigate this point I will consider plurals within the comprehensive paradigm of indefinite mark-
© Mario Squartini, 2022 | doi:10.1163/9789004473324_004
56
squartini
ers, thus systematically charting the whole set of morphosyntactic forms used to denote indefinite single bounded units, mass nouns and plural denotations in French and Spanish, as well as in Italian, which is an intermediate case (Stark 2007) between the other languages. Not only will my analysis confirm the complex interplay of quantification and classification and their interfaces with referential properties, but it will also specify that the role of classificational properties in French, Spanish and Italian plural indefinite determiners, if present, is always mediated via number. In this respect, my crucial empirical point will be that genuine classificational markers referring to plural denotations can only be found by considering either multi-word quantifiers / determiners generalized throughout the Romance area or the special behavior of indefinite determiners of Romance dialects in which plural indefinites tend to reduce the requirement of number marking within the whole morphosyntactic domain of the Noun Phrase.
2
Quantification, Classification and Reference
Apart from indexing referential specificity and signaling various degrees of discourse activation and textual prominence (von Heusinger 2011), indefinite determiners have quantificational properties covering a varied set of dimensions. On the one hand, they act as morphosyntactic correlates providing existential reading and quantificational force to nouns (There was a man with four wives) but, on the other hand, they act as markers of genericity (A gentleman opens doors for ladies, Krifka 1988; see also Greenberg’s 2003: 3 “quantificational genericity”). These diverse functions, which not only belong to the ‘grammar of reference’ but also to a ‘grammar of quantification’, are confirmed by the crosslinguistically varied and unstable boundary between indefinite determiners and quantifiers, as in the recurrent grammaticalization path from the numeral ‘one’ to the indefinite article (Givón 1981 and Dryer 2005), which is also a general ingredient of the Romance system discussed below. Similarly complex is the interplay of determiners with number, a morphosyntactic dimension in which quantification plays a definitional role. From a micro-typological point of view, Stark (2005, 2007, 2008a, 2008b) paid particular attention to the relationship between number and the whole morphosyntactic arrangement of the Romance Noun Phrase. In particular, she investigated the correlation between the obligatory status of determiners and the degree of morphological robustness / transparency of inflectional markers of number on nouns. Stark (2007, 2008b: 53–55) moved a step further by showing that, in these typological dynamics, gender might also play a role, which implies that,
quantification and classification in romance plural indefinites
57
along with a quantificational dimension connected to number, classificational aspects should be taken into account. In this perspective, Stark (2007, 2008a, 2008b) follows Herslund’s (1998, 2003a, 2003b, 2012; see also Wilmet 2008) original suggestion to reappraise Romance indefinite determiners by highlighting their role as “classifiers” sensitive to qualitative properties of the Noun Phrase, in which the opposition between French un livre ‘a book’ and the partitive article de l’eau / du pain ‘(some) water / (some) bread’ marks “the conceptually important distinction between a single, delimited referent and a non-delimited substance” (Stark 2007: 21). As mentioned above, what remains still unsettled in these analyses is the relationship to plurality, especially due to the fact that the French “partitive” article has a plural counterpart (sg du pain ‘(some) bread’ vs pl des livres ‘(some) books’). The plural marker (des in des livres) “is morphologically related to the mass singular” (Herslund 2012: 354), a formal arrangement that opens up the issue whether a classificational interpretation should be extended to plurals as well. This point can be derived from Herslund’s (2003b, 2008, 2012) own arguments, especially when he observes that in languages like French “plurals are homogeneous (mass nouns)” (Herslund 2012: 348), which, in his view, explains why the French plural indefinite determiners des can be “analyzed as a determiner of a mass NP” (Herslund 2012: 354). This amounts to saying that plurals can be interpreted within the same classificational system proposed by Herslund (2003b, 2008, 2012) in depicting the opposition between singular units (un livre ‘a book’) and masses (du lait ‘milk’). This part of Herslund’s (2003b, 2008, 2012) analysis is not entirely followed by Stark (2007), who rather insists on the nature of partitive plurals as “markers of specificity”, thus balancing the functional load in favor of the grammar of reference rather than extending the scope of a Romance grammar of classification. As is already apparent, the Romance indefinite determiners with partitive etymology (they derive from the grammaticalization of the prepositional forms de / di ‘of’, implying an original partitive reading, see Carlier and Lamiroy 2014) are at the crossroads of different subparts of grammar, in which quantification and classification interface with reference. The balance between these diverse sub-grammars will be the major point discussed in what follows, with particular focus on the relationship between the grammar of classification and the morphosyntax of number. The discussion will be developed by also considering its consequences with respect to a comprehensive understanding of plurality intended as a multifarious semantic category that has complex and not bi-unique relations to the marking of number as a morphosyntactic dimension. If one goes beyond the simplistic view of number as a purely quantificational category (see sem-
58
squartini
inal discussions in Jespersen 1924 and further elaboration in Corbett 2000), the “considerable semantic complexity” (Storch and Dimmendaal 2014; see also Acquaviva 2008) corresponding to what is morphosyntactically expressed by number (see Cowper and Hall 2012, 2014 for a formalist account) poses problems in distinguishing quantificational vs qualificational / classificational dimensions. This entangled nature is particularly apparent when comparing the prototypical quantificational opposition single vs multiple referents (book / books) with other classificational oppositions cognitively and semantically based on different categorizations of denotational qualities (Rijkhoff’s 2002 Seinsarten), such as those distinguishing single bounded units (a book), unbounded “substances” (water) and “aggregates” (Jackendoff 1991), i.e. unbounded sets of separable individual items (books / piles of books). The relationship between quantification and classification is also a point at stake in Chierchia’s (1998a) analysis of mass nouns, where classification is, in a sense, subsumed as a secondary effect of quantification. Even though their internal atoms are too vague to be counted (Chierchia 2010), mass nouns are made of parts and are therefore interpreted by Chierchia (1998a) as “inherently plural”, which explains their morphosyntactic incompatibility with plural markers (being inherently plural, they cannot be doubly pluralized!). The “Inherent Plurality Hypothesis” has disruptive effects on the traditional dichotomy count vs mass, which is considered “naïve” (Doron and Müller 2014; see also Rothstein 2010) due to its excessive faith on the homogeneity of masses / substances as opposed to atomicity of plurals (Link 1983). Homogeneity seems in fact to be incompatible with internally structured but morphosyntactically uncountable nouns such as furniture and a conceptual distinction between the uncountable change and its countable counterpart coins is tenable only with difficulty (Italian soldi ‘money’ is plural, as well as its Danish counterpart penge). On top of that, even genuinely homogeneous substances like water, blood or soup, despite their “cumulative” nature that makes them typically nonatomic, can be “divisively”1 quantified as in two portions of soup (Landman 2016: 1), which shows their nature as “sets of portions” (Rothstein 2010, 2017). Thus, masses end up having a lattice structure comparable to plurals (Heycock and Zamparelli 2005), which is eventually due to the intrinsic cumulativity of both plurals and masses. A closer look at Romance determiners confirms all these suggestions on the relationship between plurality and mass nouns (Chierchia 1998a, Delfitto and
1 For a discussion on the interaction between the feature [cumulative] / [divisive] and the combination of these two dimensions in defining prototypical uncountables see Deal (2017).
quantification and classification in romance plural indefinites 59
Schroten 1991, Zamparelli 2008 and Cardinaletti and Giusti 2018), but on the other hand, a comprehensive view of the whole paradigmatic system of indefinite determiners indicates that plurals are in fact at the crossroads between quantification and classification (Borer 2005: 93 and Cowper and Hall 2012, 2014), which explains the complex shape taken by the Romance system. In presenting this system, I will follow Herslund’s and Stark’s comparative analyses of three major Romance languages (French, Italian and Spanish). Unlike their accounts, in which they mostly focused on the differences shown by the indefinite determiners of the three languages, I will be more interested in the overall paradigmatic organization, which, eventually, will lead me to minimize differences and highlight similarities. In so doing, I will deliberately follow a much more traditional structuralist perspective in which what counts is the shape of the paradigm itself, irrespective of the formal mechanisms used by single languages. As opposed to Herslund and Stark, who have more insisted on the special treatment of mass nouns, my focus on plurals suggests I reverse the order of presentation by starting from the opposition between plural and singular indefinite determiners (Section 3), which will then be compared to the determiners occurring with mass interpretations (Section 4). The synoptic analysis of the whole grammatical system of indefinite determiners in the three languages represents the bulk of my discussion (Section 5), where I will focus on the general structure of the system, eventually deriving from it all the interpretative consequences for a better understanding of plurals in their complex nature, in which both number and classification seem to play a role. However, my final part (Section 6) will show that in this rush to classification attention should also be drawn to the grammaticalization of new determiners (multi-word forms with quantificational content, see Herslund 2003b). Significant insights can also be achieved by extending the empirical coverage in a variationist perspective, thus considering the whole Romance speaking area, in which the interplay of number and classification might be rearranged in a morphosyntactic distribution that foregrounds classification. Generally speaking, my observations on the interplay between number and classification will confirm the importance of elaborating on a model in which functional-semantic notions (plurality, plural and mass denotations) are kept neatly distinguished from morphosyntactic categories (number) with respect to which semantics maintains non-biunique relationships.
60 3
squartini
The Paradigmatic Opposition Singular vs Plural: Differences and Similarities in Romance
3.1 French, Italian and Spanish An uncontroversial point in which French, Italian and Spanish show the greatest consistency is the obligatory requirement of indefinite determiners for those singular nouns that, semantically, have countable interpretations as bounded entities, and, syntactically, occupy slots like the argumental positions filled by the direct objects in the following French (1a), Italian (1b) and Spanish (1c) examples: (1) a. Aujourd’hui j’achète *(un) livre. b. Oggi compro *(un) libro. c. Hoy compro *(un) libro. ‘Today I buy a book.’ On a formal side, the three languages in (1) are also quite consistent in having grammaticalized what, at least etymologically, is the same marker, a form derived from the numeral ‘one’. Looking instead at the data of plural nouns in (2), the distribution of parentheses and stars highlights a much more varied picture. (2) a. Aujourd’hui j’achète *(des) livres. b. Oggi compro (dei) libri. c. Hoy compro (unos) libros. ‘Today I buy books.’ The main differential point is the distinctive divide that opposes French (2a) to Italian (2b) and Spanish (2c). While French is internally consistent in construing a paradigm that mandatorily requires a morphosyntactic slot in the area of the Determiner Phrase to be phonetically spelled out in (1a) as well as in (2a), Italian and Spanish have an obligatory singular determiner in (1b) and (1c), while they do admit bare plurals in (2b) and (2c), respectively. However, the distribution with respect to French does not show rigid complementarity, for the morphosyntactically null positions preceding bare plurals can be filled and, as particularly highlighted by Herslund (2012), fillers can have different origins. Italian dei etymologically corresponds to the same marker found in French, which is an occurrence of the so-called “partitive” article2 (Carlier 2 Chierchia (1997: 74) uses the term “bare partitives” to refer to these structures as opposed to
quantification and classification in romance plural indefinites
61
2007 and Carlier and Lamiroy 2014), a form of prepositional origin, which is agglutinated to what in Italian is, more transparently than in French,3 a plural of the definite article (Vincent 2017: 730).4 Spanish indefinite plural arguments can also have an indefinite determiner marked for plurality (unos), but there is no partitive etymology via a prepositional form and the plural marker is not an agglutinated form of an independent article, being rather a transparent and totally consistent (Stark 2007) inflectional marker of plurality. Thus, Spanish unos is the inflectional plural of an allomorphic stem of the indefinite article (uno-), here homophonous with its etymological trigger, the numeral ‘one’ (uno in Spanish), a point that, as we will see below, is crucial in Herslund’s (2012) perspective.5 3.2 Some Caveats: Semantics and Syntax In commenting on these data it must be stressed from the beginning that the pattern shown in (1–2) is meant as a very general paradigm measuring out all
“full partitives”, i.e. constructions headed by full-fledged quantifiers (Some of the bottles are broken). 3 In fact, it would not be easy to tell what a French transparent plural marker should be like, for inflectional markers of plurality tend to be eliminated throughout the nominal system of phonic French, which represents the “invariant” type of Romance number marking (Maiden 2016: 697). French final -s (livre-s) are not audible and only occur in special sandhi conditions triggered by vowels and glides in initial positions of the following word, a phenomenon traditionally called liaison. Thus, in the indefinite determiner des [de] plural interpretation is systematically supported by the regular occurrence of the “sigmatic” markers that occur as cliticized forms (Pomino 2016) in liaison contexts ([dez] arbres ‘(some) trees’). A paradigmatic comparison to the consistent behavior of the plural definite article les [le] / [lez] also confirms a morphotactic interpretation in terms of transparency. 4 The Italian plural partitive article shares the same allomorphic distribution as the definite article. Allomorphy is phonologically dependent on the syllabic onset of the following word (i libri / dei libri ‘the books / some books’, gli studenti / degli studenti ‘the students / some students’), which confirms that what happens here is agglutination of the definite article to the prepositional marker rather than an inflectional form of the partitive article (see Cardinaletti and Giusti’s 2015 analysis of a comparable agglutinative behavior of the Italian distal demonstrative ‘that’ in its plural allomorphic forms quei / quegli). 5 If added to this subtypological survey (Stark 2008b: 54), Romanian would provide a pattern very similar to Spanish and Italian from a distributional point of view (optional bare plurals), even though the form of the indefinite plural determiner used in direct cases is etymologically very different (Romanian niște is an indefinite element derived from Latin nescio quid, Stan 2016: 298). Actually, Romanian also displays a plural indefinite derived from cardinal ‘one’ (see also Vincent 2017: 741 and fn. 18 below) occurring as the oblique case form (casa unor prieteni ‘the house of some (unor = dative / genitive plural) friends’), whose direct case form (unii) also exists but is restricted to full partitive interpretations (some out of a group), at least when occurring as an adnominal modifier (Giurgea 2013: 116).
62
squartini
possible distributions admitted by the grammatical system (Vincent 2017: 732– 733). Thus, parentheses must not be intended as free optionality (Cardinaletti and Giusti 2018, 2020), as many variational factors play a role (for an analysis of variability in Italian, also see the detailed description in Korzen 1996 and Maiden and Robustelli 2000: 76). Semantics is especially crucial in allowing null determiners in (2), where bare plurals can only be licensed if the property denoted by the nominals does not have the full referential force of an entity and is rather interpreted as denoting “kinds of things”.6 All this has significant consequences on various semantic dimensions, including telicity of the predicate (Cardinaletti and Giusti 2018) and, as further discussed in Section 4 below, scope interpretations, which will confirm the multifarious interplay between the grammar of quantification / classification and the grammar of reference. Apart from semantics, parallel caveats have to be posited with respect to syntactic positions. First of all, the paradigm depicted in (1–2) only covers argumental NPs and cannot be extended as such to other syntactic positions. In particular, it is very different from the distribution of non-argumental NP s (either NPs depending on prepositions or independent NP s with adverbial functions). But argumental status is not a sufficient condition to license bare forms, whose distribution is less constrained in post-verbal positions, as is in fact the case with the direct objects in (2), whereas in preverbal position bare plurals (Suñer 1982) show more fine-grained distinctions connected to the nature of the predicate (unaccusative, unergative) and to collocation with modifiers (adjectival modification on the bare noun favors preverbal position). Moreover, there is a complex interplay between syntactic constraints and information structure, very possibly connected to the topical role of sentence-initial positions (Laca 1996, 2013 and Leonetti 2013). However, since my main focus here is on the structure of the paradigm displayed by indefinite determiners in the three languages, irrespectively of influence of other factors linked to the syntactic status of nouns and topicality, all these differential parameters that constrain free optionality, albeit extremely interesting, will not be considered. My account will not only disregard various factors constraining free optionality where parentheses are admitted (Spanish and Italian plurals). In order to concentrate the focus of my elaboration on prototypical cases of internal objects, I will also avoid reference to more special phenomena, in which null determiners would be admitted where parentheses are barred in (1–2) (singular indefinites in the three languages and plural indefinite determiners in French). 6 Not simply Carlsonian “kinds”, for Spanish bare plurals cannot express genericity intended as denotation of a whole class of individuals (Laca 1996, 1999 and Leonetti 2012: 287; but see also Longobardi 2001).
quantification and classification in romance plural indefinites 63
This is most prominently the case of coordination, where the admittance of null determiners apparently circumvents even French rigid configurationality (Märzhäuser 2013). But coordination should be considered as a very special syntactic configuration and semantic interpretation (Heycock and Zamparelli 2003, 2005: 236–237), as is confirmed by its application across the board, even in singular forms (a point on which the differences among French, Italian and Spanish fade away). Another special behavior that will be disregarded affects singular forms and has a much more restricted scope, having to do with the possibility of noncoordinated bare singulars in special intensional contexts within the scope of negation (Sp. No encuentro película que me guste ‘I don’t find a film to my taste’, Wall and Kabatek 2013: 3) as well as structures with ‘have’ verbs (Sp. Ana lleva sombrero ‘(lit.) A wears hat’), which have been interpreted by Espinal and McNally (2011) and Espinal (2010, 2013) as constructions very different from the entity denoting operation that characterizes the occurrence of the nominal ‘a book’ in (1). According to Espinal and McNally (2011) these bare nouns denote “properties of kinds”, which are kept distinct from the denotational operation triggered by bare plurals and indefinites in (2), which in Espinal’s terms should be considered as “properties of objects”. In this interpretation Romance bare plurals occupy an intermediate position in the referentiality hierarchy spanning the denotational space from totally referential entities to non-referential properties of kinds. Following this terminology and trying to fix a cut-off point on this continuum of referentiality, in what follows I will consider the morphosyntactic strategies of indefinite determination for argumental nominals that denote either entities or properties of these entities (“properties of objects” in Espinal’s and “kinds of things” in Laca’s terminology; see also Dobrovie-Sorin and Laca 2003 and Dobrovie-Sorin 2021), while excluding generic reference to “properties of kinds”.7
7 On top of all these caveats, we should all be reminded that Herslund’s and Stark’s models cover some major Romance languages but exclude Brazilian Portuguese, where the distribution of bare nouns is rather exceptional, if seen in a Romance perspective. Brazilian Portuguese admits bare singular nouns in contexts whose interpretation in terms of referentiality is debated between not specific (on the basis of narrower scope with respect to other operators, see Müller 2002) and reference to kinds (Pires de Oliveira and Rothstein 2013). However, Brazilian definitely goes beyond the restricted use as properties of kinds of Spanish bare singular complements of have-verbs and, like Spanish bare plurals, crucially interacts with information structure. This special behavior of Brazilian Portuguese will not be considered in what follows, thus restricting the empirical coverage to those Romance languages in which a sentence corresponding to Brazilian Portuguese Ontem, Maria leu tese ‘(lit.) Yesterday Mary
64
squartini
3.3 Different “Partitive” Forms As to the partitive determiners that I will describe in what follows I must add the proviso that their status as “articled” forms, which is particularly apparent in Italian due to the agglutination of the prepositional element and the definite article (as in the plural form de + i), cannot be generalized, not only because agglutination is much less transparent in French, but also because in the Romance (Piedmontese) dialectal data discussed in Section 6, I will also consider bare8 forms of the prepositional partitive element with various allomorphs (ëd / ’d / dë) that contain neither agglutinated articles nor number morphology. In Bossong’s (2016: 69–70) typological (and geolinguistic) interpretation of Romance partitives, bare forms of the determiner, which are mostly documented in Occitan,9 are considered as intermediate steps of a continuum spanning the typological cline that goes from total absence of partitives in Portuguese, Spanish, Romanian (and Rhaeto-Romance, which is interesting from a geolinguistic point of view) to the “extreme endpoint” represented by the obligatory character of partitive articles in French. In the perspective of my work, it must be considered that their status as determiners is still debated. Stark (2016: 144) elaborates on their still “prepositional” nature by suggesting a syntactic interpretation as a “pseudo-partitive construction with an empty head”, while Cardinaletti and Giusti (2018) definitely posit them within the DP. Following Cardinaletti and Giusti (2018), I will consider them as determiners due to the fact that they share the same paradigmatic distribution as the other Romance partitive determiners mentioned so far, even though they are not “articled” in any sense.10
8 9
10
read thesis’ (Pires de Oliveira and Rothstein 2013: 217) is ungrammatical. This implies that, apart from singulars and plurals morphosyntactically introduced by determiners, I will only consider bare plurals, but not bare singulars. Here, bare should not be intended in Chierchia’s (1997) sense (see fn. 2) as forms having lost their original partitive function, but as a form devoid of the article. These bare determiners not only occur in Occitan, Gascon and Catalan, but also in Sardinian, where they obligatorily collocate with pronominal adverbs derived from Latin inde ‘thence’ (Sardinian: Bi nd’aìat de zente ‘There were many people’). Note, moreover, that bare partitive determiners seem to be possible in Spanish as well (Había de todo en esa tienda ‘There was everything in that store’, Con las lluvias, salieron de esos animalitos ‘With the rains, (some of) those little animals got out’, Treviño 2010 and Leonetti 2013: 293). From a syntactic point of view, it should also be stressed that assuming a paradigmatic approach to the determiners showing up in the syntagmatic distribution presented in (1) and (2) is not tantamount to saying that they necessarily occupy the same syntactic slots or that their overall distributions are totally comparable. In fact, they do have significant syntactic divergences if one considers their whole distribution. In this respect the
quantification and classification in romance plural indefinites 65
3.4 A Preliminary Sketch: Forms and Functions Turning now to a preliminary interpretation of the data, it seems virtually impossible to provide an analysis of the distributional paradigm in (1–2) without taking into account the apparent balance between obligatory status of determiners and availability of inflectional markers of plurality on nouns. This uncontroversial observation justifies why the literature on the topic has mostly elaborated a typological subdivision of different Romance subtypes. The correlation is especially stressed by Stark (2007: 52), who elaborates on previous literature (see Delfitto and Schroten 1991 and Schroten 2001 as specific references on this point). Within Romance, French represents a very special behavior in which final inflectional endings are phonetically absent in most nouns (see fn. 3 above), while inflectional markers are robustly maintained in Spanish (libros) and Italian (libr-i). From a functionalist point of view such an inflectional divide can be compensated by mandatory determiners, which explicitly mark the number opposition between indefinite singulars (un livre) and indefinite plurals (des livres).11 Conversely, Spanish and Italian bare plurals can be interpreted as a consequence of the stability of inflectional morphology in nouns, which, in a very functionalist interpretation of the division of labor between different structural components, do not obligatorily require number marking on the determiner and therefore admit bare nouns. Stark (2007, 2008b) elaborates further on these typological subdivisions by drawing attention not only to the parameter presence vs absence of inflectional morphology on nouns but also to the degree of morpho(syn)tactic transparency / bi-unique correspondence between inflectional forms and the functional expression of plurality. Italian and Spanish do in fact differ as far as the bi-unique nature of form
11
grammaticalized form of the cardinal number ‘one’ now occurring as the Romance indefinite determiner un has unique distributional properties that set it apart from the other indefinite markers. Most notably, the very fact that a French homophonous form of the indefinite determiner un occurs as a pronominal quantifier ( J’ai vu un livre ‘I saw a book’ / J’ en ai vu un ‘I saw one (of the books)’), makes its syntax radically different from a partitive determiner ( J’ ai vu des livres ‘I saw books’ / *J’ en ai vu des ‘I saw some (of them)’). Despite these divergences, what keeps all these forms together is their common distribution for they all occur as prenominal determiners, which makes it feasible to include them within the same paradigmatic account. The particularly robust frequency of sandhi inflectional -s in phonosyntactic connections between determiners and determined nouns (le[z] arbres ‘the trees’, de[z] arbres ‘(some) trees’, see also fn. 3) confirms the interpretation of determiners as a primary locus where number distinctions are spelled out, even though the tendency to overextend inserted [z] in other conditions raises the problem whether these sandhi-triggered inflectional markers (clitics in Pomino 2016) should be better interpreted as inflections prefixed to the noun (c’ est quoi, comme [z]arbres? ‘what sort of trees are they?’, Morin and Kaye 1982: 321, 324).
66
squartini
/ function correlations in number inflections is concerned. While the Spanish final inflectional marker of plurality (libro / libro-s) consistently covers the whole nominal system, Italian marks plurality through a set of morphs that do not unambiguously correlate with a given value within the category number. In this respect, Stark points out the occurrence of homophonous inflectional morphemes such as Italian -e, which can be both feminine (la ret-e ‘the net’) and masculine singular (il can-e ‘the dog’) as well as feminine plural (le case ‘the houses’), depending on declensional classes. This typological difference is crucial in Stark’s attempt to explaining the intermediate position of Italian between French, where a determiner with partitive etymology is obligatory, and Spanish, which instead does not have grammaticalized partitive determiners. In the intermediate position occupied by Italian, partitive determiners do exist but are not obligatory. A formal account of this synchronic typology is also possible (Stark 2008a) and, from a diachronic perspective, these data can be inserted within the morphosyntactic diachronic dynamics from Latin to Romance (Ledgeway 2012, ch. 4) as another clue of a tendency towards a more configurational setting displayed by French, where the functional area of the Noun Phrase is explicitly spelled out by filling slots within the D(eterminer) P(hrase), while Italian and Spanish are less configurational in admitting zeros in formal marking of functional heads. Whatever approach one should prefer, the crucial point here is that the “striking differences between French, Italian and Spanish” (Stark 2007: 52) only affect the morphosyntactic arrangement, while the structural system is based on the same general principle of maintaining a differential feature between singulars and plurals in argumental positions. The different arrangements will be compared synoptically in Table 3.1, which is discussed at length in Section 5.2 below, but the main point can be grasped by considering the different distribution of the same features within the Noun Phrase. In French the distinction singular / plural is guaranteed solely by the determiners, which are therefore obligatory, while maintenance of inflectional endings allows Italian and Spanish to recruit zeros as determiners. Seen in this very general perspective, the varied morphosyntax of Romance determiners superficially spells out a structural system that is identically based on the functional pressure to keep a formal distinction between singular entities and plural denotations. From this point of view my position differs from Herslund’s and Stark’s, basically because we have different research objectives. They are mostly interested in the formal mechanisms used by the three languages and therefore they capitalize on differences by interpreting them as connected to other differential features. In their (sub)typological perspective a major point is finding out which of the three lan-
quantification and classification in romance plural indefinites 67
guages represents the “mixed” or intermediate element. As mentioned above, in Stark’s analysis it is Italian that is intermediate between French obligatory partitive-like determiners and Spanish bare plurals. As will be shown below, in Herslund’s general treatment of classificational properties it is Spanish that turns out to be “mixed”, due to the quantifier-like form of its plural indefinite determiners (unos). In my perspective, these and other differences including the problem of the intermediate typological element will instead be factored out in order to reach a comprehensive view showing how the grammatical paradigm is shaped, whence a more general interpretation of the relationship between classificational properties and number can be grasped. In this respect, the data analyzed so far allow me to reach the intermediate conclusion that the system of the three languages has basically the same shape, for they all have formal mechanisms capable of distinguishing between reference to a set made of a single bounded unit (a book) and sets of more than one object, i.e. sets of sets (see Heycock and Zamparelli 2005 and Link 1983) of atomic objects (books). Now, the crucial point is to verify whether this paradigmatic distinction can be interpreted in classificational terms, or has to be thought of as a quantificational notion within a more traditional interpretation of number. Put in other terms, the point is whether by using Romance plurals not only are we saying that we have more than one referent, but we are also presenting the way in which referents are combined together. In this respect it is now crucial to cast plurals within the general background of the whole system, where, instead, it is the singular form of the so-called partitive article that has mostly attracted attention as grammatical evidence demonstrating the existence of Romance classifiers.
4
Romance Classifiers: Singulars and Plurals or Singulars Only?
4.1 Herslund’s Focus on Forms As mentioned in Section 2, the crucial trigger of Herslund’s classificational interpretation of French and Italian determiners is the observation that the singular form of the so-called partitive systematically refers to mass denotata in French (3a) and Italian (3b).12 Despite incipient grammaticalization of an
12
The same distribution applies not only to mass but also to abstract nouns, which is consistent with a classificational interpretation, due to the intrinsic lack of boundaries that characterizes the interpretation of abstract concepts (‘patience’ and ‘love’ are not intrinsically bounded, they can only be measured but not quantified through number: a lot of patience, much love). However, the distribution with abstract nouns should be handled
68
squartini
Old Spanish partitive form with masses in a very specialized referential meaning (Gerards and Stark 2021), modern Spanish has no form of partitive article, only having zero markers for mass nouns (3c): (3) a. Aujourd’hui je mange *(de la) viande. b. Oggi mangio (della) carne. c. Hoy como carne. ‘Today I eat meat.’ Herslund (2003b) and Stark (2007) coincide in highlighting the role of the partitive article with mass nouns as the crucial element qualifying the Romance languages that do possess it as languages with “classifiers” (Herslund 1998, 2003b, 2012) or languages having a “countability system” (Stark 2007) expressed by the Noun Phrase through morphosyntactic markers signaling how the intrinsic Seinsart (Rijkhoff 2002) of the noun is linguistically “apprehended” (Seiler 1986, Stark 2007: 56). Starting from these uncontroversial observations Herslund (2012) elaborates further in order to build a comprehensive classificational interpretation. His main argument runs as follows: consistently with the behavior of singulars, where the opposition indefinite (1) vs partitive article (3) codifies a classificational distinction between single bounded entities and mass nouns, the Italian and French markers for plural indefinites (2) are intrinsically classificational, basically because they share the same partitive form, which is “morphologically related to the mass singular” (Herslund 2012: 354). As already mentioned, this position is functionally rooted in the assumption that, in his view, “plurals are homogeneous (mass nouns)” (Herslund 2012: 348) and therefore the plural indefinite article is “a determiner of a mass NP” (Herslund 2012: 354). Through these observations a rigid bi-unique correlation is envisaged between forms and functions, with partitive forms regularly associated with a classificational meaning (generally mass, for plurals are equated to mass). As a consequence, the Spanish indefinite plural unos cannot be interpreted as representative of the same classificational properties proposed by Herslund (2012) for French and Italian partitives. with special care in this general picture (see also Carlier and Lamiroy 2014), for Italian abstract nouns seem to be more recalcitrant to being determined by a partitive (It. ??avere della pazienza ‘have patience’) than Italian masses (It. mangiare della carne ‘eat meat’), whereas both contexts are equally determined by a partitive in French: avoir de la patience / manger de la viande. Heycock and Zamparelli (2005: 211, fn. 10) also indicate a special semantic behavior (admitting “split reading” in coordination) with abstract nouns in Italian (differently from French).
quantification and classification in romance plural indefinites 69
4.2 Spanish unos: A “Collective” Marker? Following this strong focus on forms, Herslund (2012) proceeds further in his classificational perspective when he elaborates on the different forms occurring in Spanish (unos) vs. French / Italian (des / dei). In his view, selecting a given form is indicative of different classificational interpretations. Being inflectionally connected to un, Spanish unos maintains the same quantificational nature, which delivers a plural that functions as a marker of “collectives”, a term glossed by Herslund (2012: 348) as denoting “structured sets of entities”. Attention to the classificational nature of unos is also the main reason why in Herslund’s typology the “mixed” type is represented by Spanish, where, unlike purely “classifier-languages” (modern French and Italian), a quantifier (unos) with special classificational meaning (collective) occurs as a form of plural determination. In Herslund’s (2008: 36, 2012: 348–353) analysis, Spanish collectives are prototypically apparent when unos denotes natural sets (unos dientes deslumbrantes ‘a set of shining teeth’), natural pairs (unas manos largas ‘a pair of long hands’), series (unos golpes ‘a series of knocks’), ad hoc sets (unos palos ‘a combination of poles’) also including pluralia tantum (unas tenazas ‘a pair of tongs’), which intrinsically denote either pairs or sets. Particularly significant as empirical support to Herslund’s collective interpretation is the observation that Old French (see also Buridant 2000: 115–117 and Herslund 2003a) used to display a now obsolete plural form of the indefinite article (masc uns / fem unes), whose distributional pattern typically correlates with Herslund’s (2012: 348) set interpretation (unes armes ‘an armour’, uns dens ‘teeth’, uns eschés ‘a set of chessmen’, uns degrez ‘a staircase’). These diachronic data are indeed suggestive of a possible correlation between grammaticalization of indefinite determiners and classificational properties activated in interpreting nouns. It has repeatedly been observed that in Romance medieval varieties, especially in Old French (Carlier 2001, Herslund 2003a) but also Old Occitan (Jensen 1994: 77) and Spanish (Vincent 2017: 741; for the development of this form in Spanish see also Pountain 2019: 7–8, 10–11), the plural of the indefinite article, which is inflectionally produced from singular un, shows preferential usage in contexts where plural nouns denote sets of discrete objects (what Herslund 2012 calls “collectives”). However, if we go back to Spanish unos and we want to describe the modern system from a synchronic point of view, it is difficult to interpret all occurrences of unos as collectives. Consider the following Spanish example in which bare and “articled” plurals collocate in the same context with the same noun:
70
squartini
(4) Tomaban churros, unos churros bastante asépticos, pero churros al fin. (Laca and Tasmowski 1994: 34) ‘They were having / used to have churros (‘deep-fried dough sticks’), not very tasty churros, but in the end still churros.’ In two of its occurrences in (4) the noun churros is not articled for it denotes non-specific instantiation of a kind. When instead reference is made to a more specific type of the same thing or to a referent made contextually specific in a given situation (‘not very tasty churros’), the noun is morphosyntactically modified by the plural form of the indefinite article unos. Obviously, Herslund (2012) is not unaware of the import of specificity but he argues that it can be “subsumed as a particular instance of the collective interpretation” (Herslund 2012: 350). In this perspective he goes beyond the restricted limits of the argumental objects analyzed so far by also including predicative uses of nouns. In so doing, he reappraises Laca and Tasmowski-De Ryck’s (1994) interpretation of the Spanish indefinite article unos occurring in a predicative function, as in definitional statements (Las ballenas son unos mamíferos de gran tamaño que se parecen a los peces ‘Whales are a kind of mammals of big size that resemble fish’), where the subject is predicated as belonging to a given type. Forming part of a special type is interpreted by Herslund (2012: 350) as insertion into a given set of individuals, which is how he proposes a connection between his general classificational proposal and type interpretation triggered by Spanish indefinite plural unos. 4.3 Spanish unos and Italian dei: Classification and Referentiality As is apparent from this discussion, we are facing here a new point, which calls for a better understanding of the boundary between reference and classification. My interpretation here is that Herslund (2012) seems to be superimposing a grammar of classification over what basically is a grammar of reference. Albeit suggestive of new perspectives, I consider Herslund’s proposal not tenable if the whole set of data is taken into account. If referentiality (specificity, in this case) were so dependent on classification and Herslund’s (2012) focus on form should be rigidly applied, one would expect that the Italian determiner dei (a plural equated to mass interpretation, in Herslund’s perspective) and the Spanish collective unos had different distributions in terms of specificity, as inevitably expected, in Herslund’s interpretation, due to their different forms. But, this is not the case, as demonstrated by the following Italian translation of the Spanish churros-example above (Spanish churros have been culturally domesticized by using Italian gianduiotti ‘nut-chocolates’):
quantification and classification in romance plural indefinites
71
(5) Mangiavano gianduiotti, dei gianduiotti abbastanza asettici, ma in fondo sempre gianduiotti. ‘They were having / used to have gianduiotti, not very tasty gianduiotti, but in the end still gianduiotti.’ Not differently from Spanish unos, Italian dei signals a specific interpretation in the second token of the noun gianduiotti in (5). This means that, irrespective of their different etymologies, both Spanish unos and Italian dei end up as markers of specificity. This is in fact the solution adopted by Stark (2007), as is clear when she observes that “the morphological plural of the partitive article is not functionally a partitive, but rather a normal indefinite plural article” (Stark 2007: 60, fn. 16). Thus, French des (as well as Italian dei) should be “functionally and distributionally” intended as “the indefinite plural article”, which “does not indicate non-countability, but rather specificity” (Stark 2007: 52 and 52, fn. 4).13 When considering plural indefinite articles as markers of specificity, we should always bear in mind that indefinite articles admit but do not require a specific interpretation, which is only triggered within a wider scope configuration with respect to other sentential operators (as usual, negation is the test here). The Italian indefinite plural dei demonstrates the typical scopal ambiguity under negation that generally characterizes indefinite determiners (Chierchia 1997: 91, Cardinaletti and Giusti 2016: 60): (6) Alla riunione non ho visto degli studenti. ‘At the meeting I haven’t seen some of the students (wide scope) / I haven’t seen any students (narrow scope).’
13
Stark’s conclusion that singular partitives, which are interpreted as classificational markers, should be separated from plural partitives, which instead belong to the same referential system as singular indefinites, is independently supported by Cardinaletti and Giusti’s (2016: 60, 2018) suggestion to keep a distinction between the Italian singular vs plural partitive articles also in terms of referentiality as can be measured through scope effects. Cardinaletti and Giusti (2016) observe that a wide scope interpretation seems not to be admitted with mass reading of singular nouns (but on this point see now Giusti 2021: 293), contrary to the scopal ambiguity of plural indefinite articles (see ex. (6) below). Cardinaletti and Giusti (2018, 2020) also underlined that the distribution of scopal effects can be extremely varied even with plural partitives if one takes into account the whole variational set of Italo-Romance languages, with Central varieties turning out to be particularly uncomfortable with narrow scope reading of plural partitives.
72
squartini
On the contrary, Italian bare plurals, just like their Spanish counterparts (Laca 2013: 98), only admit narrow scope (‘no students attended the meeting’): (7) Alla riunione non ho visto studenti. ‘I haven’t seen students at the meeting.’ Thus, what is meant by Stark when she concludes that plural indefinite determiners express specificity is not a bi-unique correlation between the occurrence of articles and specific interpretation (otherwise one would not justify the admittance of narrow non-specific reading with dei (6) in Italian), but more generally the fact that “articled” plurals, unlike zeros, admit a specific interpretation, showing the same Carlsonian ambiguity found in indefinite quantifiers. The very fact that plural indefinites share the same scopal ambiguity as singular indefinites is, I think, an empirical point difficult to reconcile with Herslund’s (2012) classificational interpretation. As seen above, the bulk of his analysis relies on the fact that, being partitive forms, French and Italian plural indefinites should also be considered homogeneous in classificational terms. But homogeneity is exactly the opposite meaning with respect to the intrinsic bounded interpretation that is associated to singular indefinites (It. un libro ‘a book’), which makes it extremely difficult to explain why singulars and plurals, despite their divarication in classificational terms, do share the same scopal interpretations. What these controversies ultimately show is that the grammar of classification and the grammar of reference are independent dimensions that might have multifarious relations in the system of Romance determiners but, still, should be kept separate.14 Otherwise, if classification and reference
14
A comparably entangled coexistence of reference and classification seems to characterize Spanish Differential Object Marking (i.e. the prepositional form a un secretario in Busco a un secretario ‘I am looking for a (certain) secretary’), whose areal distribution (DOM is highly grammaticalized in Spanish and Romanian) has suggested a correlation with the absence of partitive articles (Körner 1987), whence a classificational interpretation of DOM (Stark 2008b: 55–60). However, this is again a case in which what prevails is a cluster of syntactic function (object), semantic properties (affectedness and animacy), discourse topicality and prominence (von Heusinger and Kaiser 2005; see also Schurr 2021) and referentiality (definite and specific referents) having testable effects with respect to wide / narrow scope readings (López 2016). All these ingredients necessarily require highly individualized (Leonetti 2004: 80) and consequently “contoured” (Stark 2008b: 58) referents. But this classificational interpretation might be considered as a side-effect of referentiality rather than the primary content of DOM. Otherwise, one should conclude that, when DOM is absent with inanimate / unaffected objects (Busco un libro), the referent denoted by the SN ‘a book’ is to be intended as an uncountable (“not contoured”) object. As to Körner’s (1987) areal observations, Schurr’s (2021: 75) reappraisal shows that “mutual
quantification and classification in romance plural indefinites
73
were the same category, we should expect any form expressing classificational atomicity (single bounded entities) to be intrinsically specific without any scopal ambiguities. However, showing that Italian dei and Spanish unos may occur in comparable contexts with comparable referential effects in terms of specificity is not tantamount to concluding that they express exactly the same cluster of properties in the interplay between reference, quantification and classification (López-Palma 2007: 236, fn. 1). Actually, the interpretation of Spanish unos in wide / narrow scopes is debated (Gutiérrez-Rexach 2001: 146 and Martí 2008: 10, fn. 1), but apparently wide scope tends to prevail (Martí 2008: 10), which might be an interesting difference with respect to the scopal ambiguity of Italian dei. This simply indicates that the position of unos should be singled out with respect to other Romance indefinite plurals as far as referentiality is concerned, but does not require it to be interpreted in classificational terms. From the point of view of referentiality, unos seems to occupy an intermediate position between Italian dei, which shows the typical scopal ambiguity of other indefinite determiners, and full-fledged indefinite quantifiers like Spanish algunos ‘some’, whose more specific nature is confirmed by the contrast with unos in expressing “partitive specificity” (Enç 1991) triggered by referents selected from a contextually given set (#unos de ellos vs. algunos de ellos ‘some of them’, Leonetti 2012: 294), despite the fact that unos generally admits nondiscourse linked partitive constructions (He leído unos de los mejores libros de lingüística ‘I have read some of the best linguistics books’, Gutiérrez-Rexach 2001: 127). Comparable problems are posed by the interpretation of Spanish unos as a trigger of group-interpretation, i.e. a collectivizer, which is obviously extremely relevant in Herslund’s (2012) classificational perspective based on the idea of interpreting unos as a “collective” marker. The non-distributive interpretation of unos (*Unas llaves abren una puerta cada una ‘Some keys open a door each’) has been pointed out as a sign of the group interpretation intrinsically connected to unos (Gutiérrez-Rexach 2001: 119–120), which distinguishes it from algunos, a point also explicitly exploited in classificational terms by Herslund (2012: 348–349). However, López-Palma (2007: 245–246) has shown the instability of this anti-distributive effect and the possibility of circumventing it by making the referent more specific (Unas llaves que compré ayer abren cada una una puerta distinta del coche ‘Some keys I bought yesterday open each a exclusion” between partitive articles and DOM is not confirmed, at least in its strongest version, for partitives “do co-occur” with “incipient DOM”. This suggests further research on a possible relationship between “incipient DOM” and non-obligatory partitives, which, in fact, is what occurs in standard Italian.
74
squartini
different door of the car’). This confirms that what semantically characterizes unos has basically to do with referentiality (a stricter requirement on specificity). 4.4 Herslund’s and Stark’s Classificational Interpretations of Plurality All in all, independently from the exact quantificational / referential interpretation of unos, what remains not tenable in Herslund’s account is the interpretation in classificational terms, i.e. the idea that unos has different classificational properties opposed to those of Romance indefinite determiners with partitive origin. The differences between unos and partitive plurals should instead be interpreted within an interface of reference / quantification / classification, not by superimposing a comprehensive grammar of classification over the grammar of reference. But in this complex interplay there is another major ingredient that should not be forgotten: the morphosyntactic category traditionally called number. All forms discussed so far (French / Italian partitives des / dei and Spanish unos) are plurals in a morphosyntactic sense, for they codify one of the possible values of number as a grammatical category. This poses the problem of establishing the correct mapping between the morphosyntax of number and plurality or plural denotation as a semantic notion. In Herslund’s hypothesis, the connection between plurality and the plural forms of the indefinite determiners seems to be bi-uniquely intended as a one-to-one relationship based on the forms of the indefinite markers. Spanish unos triggers a collective interpretation due to its “quantifier-like” form, while French and Italian plural partitives are intended as equal to masses.15 On the contrary, Stark (2007: 54, fn. 7) has a different interpretation of the classificational nature of the plurality expressed by partitives when she notes that in languages characterized by the grammaticalization of countability plurals are “always understood as additive”, i.e. as denoting a collection of discrete objects. In her analysis, the consistent additive interpretation in Romance plurals is a significant innovation with respect to Latin, where, instead, plurality also admits non-additive readings with various classificational interpretations, including intensification ( frigora caloresque ‘an intense heat and cold’, Stark 2008b: 50). In Stark’s (2007, 2008a, 15
As a consequence of this equation, Herslund (2012: 348) is forced to conclude that “the plural of una casa ‘a house’ in Spanish is not unas casas, but casas” due to the idea that the essence of plurality is to be intended as coinciding with the homogeneity of masses, whereas unas casas would trigger the special collectivizing effect connected to the intrinsic nature of unas as a quantifier. However, if seen from the point of view of the grammar of reference, the interpretation is exactly the opposite, as demonstrated by Leonetti’s (1999: 842) conclusion that it is unas casas the real plural of una casa, due to their common specific interpretation and (partial?) scopal consistency.
quantification and classification in romance plural indefinites
75
2008b) general view, the diachronic evolution from Latin “also non-additive plurals” to Romance “only additive plurals” consistently supports the assumption that the classificational potential that Latin could also express through inflectional categories (gender and number) was gradually and variedly lost in Romance and therefore replaced by determiners with classificational potential. The different pace shown by Romance languages in this process in which classificational functions have been re-grammaticalized is explained, according to Stark, by considering the different degree of resistance / transparency of inflectional morphology expressing gender and number. However, apart from these typological consequences, what is apparent is that Stark and Herslund seem to show different interpretations on how plurality should be intended in classificational terms. In what follows I will suggest that these discrepancies can be solved by considering that what the Romance data taken as a whole tell us is exactly the fact that plurality is a multifarious notion, a point that is not only demonstrated by comparing Latin and Romance (as suggested by Stark), but is also apparent if we comprehensively consider the whole set of Romance data.
5
The Multifarious Nature of Plurality
5.1 The Interface between Cognition and Grammar Even though the strict form-function correlations proposed by Herslund in his classificational interpretation of plural determiners were not supported by the data analyzed above, one cannot deny that the different forms of Romance plural indefinite determiners belong to those empirical facts that a linguist is expected to interpret somehow. My point on this is that Herslund’s interpretation, in which a rigid bi-unique correlation is posited (Spanish indefinite plurals with unos are “collectives”, whereas French and Italian indefinite plurals are interpreted as homogeneous masses), albeit not empirically confirmed (at least not on a synchronic perspective), should be re-launched at a different “interfacial” level. I suggest that what we are observing here is again another interface, which, however, is not an internal relationship among different subcomponents of grammar (grammar of classification, grammar of reference and grammar of quantification). This is rather an external interface between cognitive apprehension of individuals and what is codified in grammar. Chierchia (2010: 101–102) explicitly addresses the possible distinction between a cognitive level, in which substances and objects are apprehended, and a linguistic level, in which grammars variedly encode that distinction. But I think that the same point should be extended to the semantic notion covered by the term plurality and its connections to number as a morphosyntactic category. In this per-
76
squartini
spective, the variety of formal solutions adopted by Romance grammars generally confirms the “considerable semantic complexity” that has to be forced within the morphosyntax of number (Storch and Dimmendaal 2014) and the multifarious nature of plurality (Acquaviva 2008). In this cognitive sense, I think that Herslund is definitely right in pointing out the different solutions adopted by French / Italian vs. Spanish, but the crucial point is that these intraRomance differences are only indicative of the general driving forces maneuvering behind the grammatical system, which does not necessarily imply that the synchronic arrangement shown in a given system directly codifies what can be distinguished at a cognitive level. Thus, the diachronic origins of the different forms (Spanish pluralizes un-, whereas French and Italian pluralize the partitive forms) can be conceived as representing different grammatical solutions to the cognitive problem of apprehending the pluralization of individual objects. 5.2 Two Opposite Cognitive Patterns As is clear from what seen above, plural denotations can be conceived in (at least) two very different interpretations, which correspond to what Herslund (2012) calls collectives vs masses. Following Jackendoff’s (1991) terminology, this interface between cognition and grammar can be summarized in two different patterns: (a) mass & count pl = substances + aggregates (b) count sg & pl = single individual + aggregates Pattern (a), represented by French and Italian partitive articles, demonstrates that the plural nature of aggregates (unbounded sets of single individuals) can be cognitively conceived as similar to masses (“substances”), whereas in pattern (b) the cognitive similarity that “inspires” grammar is the discrete nature of the single individuals combined in a set. In this pattern, the form denoting indefinite sets is the plural of the singular indefinite article. The different distributions of these two patterns can be synoptically captured by looking at the summary in Table 3.1. The distribution of bold and italics in Table 3.1 depicts the tension between the two different patterns indicated above: pattern a (mass & count pl) is represented by the forms in italics (French and Italian partitives), whereas pattern b (count sg & pl) is in bold (Spanish un / unos). The chart is now an opportunity to frame form/function correlations also within the general interpretation elaborated by Chierchia (1998a) with respect to the relationship between masses and plurals. Seen very simply, the point is whether nouns like roses and wine can be interpreted as different realizations of the same form of cognitive apprehension of referents. In Chierchia’s (1998a: 57) influential reappraisal of
quantification and classification in romance plural indefinites table 3.1
Romance indefinite determiners (mass / count)
French
sg pl
77
Italian
Spanish
mass
count
mass
count
mass
count
du
un des
ø / del
un ø / dei
ø
un ø / unos
the dichotomy mass / count any semantic difference between roses and wine is swept away on the basis of the provocative tenet that “mass nouns are inherently plural”, which is somehow reversed and indirectly echoed by Herslund’s (2012: 343) semantic deduction that “the plural is in fact a way of making heterogeneous nouns homogeneous”. The data charted above show that, as suggested by Herslund (2012), this cannot be the whole story, for plurality turns out to be a composite category. Contrary to Chierchia (1998a), Cowper and Hall (2014: 69) propose a bi-featural analysis of plurality: in their interpretation the feature shared by plurals and masses is [Non-Atomic], but plurality also has a [Discrete] value, which is responsible for the “divisive” nature of plurals as “sets of sets” (Borer 2005: 127 and Gerards and Stark 2021: 109). These bi-featural nature corresponds to the two different ways of categorizing the relationship between plurals and masses that comes out from Table 3.1: either the solution printed in italics, where masses do group together with plurals (thus confirming Chierchia’s 1998a account), or the solution printed in bold, where instead plurals behave very differently from masses, being therefore driven by the “additive” interpretation mentioned in Stark (2007: 54, fn. 7) and by the feature [Discrete] implemented by Cowper and Hall (2014: 69). This suggests that wine and roses do not necessarily correlate (pattern b), even though they can (pattern a), which confirms the intrinsic heterogeneity of plurality as a cognitive mold from which grammar derives.16 However, this is only the cognitive side that indirectly interfaces with grammar.
16
Chierchia (1998b: 73) is sensitive to the possibility of keeping a distinction between wine and roses, when he observes that: “Quantity and group, in other words, seem to be nearly synonyms. They merely differ in that the argument of group is presupposed to be plural, while that of quantity is not”. However, on this specific point (Chierchia 1998b: 73) adds that “For the purpose of the present paper, I will ignore this difference and treat them as synonymous”.
78
squartini
Recognizing these cognitive forces is not tantamount to assuming that different patterns grammatically express different classificational properties within the paradigmatic structure of a given linguistic system. In order to detect what grammar synchronically codifies we have to look at the structure of the paradigm, thus departing from Herslund’s strict focus on forms. Thus, what counts more is not which form fills a cell but the number of cells that are kept distinguished in any single language. From this point of view, I can go back to my initial point on the profound similarity of the three languages. Unlike Herslund’s and Stark’s research on the intra-Romance subtypological parameters, my structuralist reappraisal of the data urged me to foreground what the three languages have in common. If one looks at the shape of the paradigm and at the number of cells, what appears is a robust solidarity in opposing single bounded unities to mass interpretations of nouns either by a dedicated overt marker (the singular partitive in French and Italian) or through zeros. The special nature of un in the three languages is a point that stands out as particularly clear: it grammatically codes a singular bounded object (one of Rijkhoff’s nominal classes or Seinsarten) opposing it to masses and plurals. But the distribution of cells also shows that the three languages keep a distinction between masses and plurals, which is consistently guaranteed by the morphosyntax of number: it is a plural marker (either on the determiner or on the noun or on both) that keeps masses and plurals distinguished and, in this respect, the etymological form of the determiner (either un- or a partitive form) does not play any role, for what counts more is that the languages here considered all have a tripartite paradigm, in which determiners interact with the morphosyntax of number. To me, this is the most crucial point that can be derived from the data charted in Table 3.1, which immediately calls for an interpretation of the interaction between classificational properties and number. I will elaborate further on this point below, but first I have to discuss a caveat linked to the interpretation of zeros, for my interpretation of Table 3.1 only holds if one admits that “bareness itself is a means of determination in its recurrent association with indefinite semantics” (Schurr 2021: 2), which implies that zeros count, a point on which Herslund and Stark appear to differ. In concluding his general assessment of the Romance system, Herslund (2012: 355) explicitly recognizes that zeros count when he observes that “[a] bare noun signals the mass reading”, just like “an NP with the indefinite article the entity reading”. This solution fits well within Herslund’s general classification, in which the “pure” classificational type represented by French and Italian is distinguished from the “mixed” type represented by Spanish.17 In order to be mixed Spanish needs to 17
Herslund’s stance is very different in interpreting German zeros with masses and plurals,
quantification and classification in romance plural indefinites 79
have, along with quantifiers un / unos, a form expressing homogeneous masses, which are in fact marked by zero articles. Stark (2007: 66) concludes instead that, unlike French and Italian, Spanish has “no ‘classification potential’ inside the indefinite determination system”. This is consistent with her account in which Spanish represents a separate type of morphosyntactic arrangement still partially maintaining Latin classificational potential through gender (Spanish pronouns have a neuter form required with abstract antecedent, as Stark (2007: 57) reminds us, and a neuter form of the definite article that occurs with adjectives: lo bueno ‘the good part of it’ vs. masculine el bueno ‘the good / goodlooking man’).18 However, if one adopts a paradigmatic perspective focusing on the shape of the whole grammatical system, Herslund’s assumption that zeros count is more consistent. Albeit lacking an explicit marker, the system is shaped so as to guarantee an oppositional distinction between single bounded units and mass nouns.19 It should be noted that by inserting bare nominals within
18
19
where he considers that countability distinctions are already defined as lexical features and cannot be adjusted by means of a classifier at the Noun Phrase level. In this perspective, Herslund treats German as a language in which classificational properties directly derive from the denotational nature of lexical items like Zucker ‘sugar’ and Wasser ‘water’, which can only occur as bare nouns. French sucre and eau are instead lexically underspecified, thus requiring a morphosyntactic classifier. But what about Spanish azúcar and agua? Why are German zeros interpreted lexically (semantic underspecification) whereas Spanish zeros count as classifiers? In discussing these points, one should also consider how they might be reconciled with Chierchia’s (1998b) typological approach in which Germanic languages admit both forms of generations of names as “arguments” (directly from the lexicon) as well as “predicates” morphosyntactically accommodated by means of determiners (see fn. 20 below). In Stark (2008b), the role of partial gender-maintenance of a peripheral Western Romance language like Spanish is nicely mirrored by an Eastern, but equally peripheral, Romance language (Romanian), where a tripartite system of controller genders (Loporcaro, Faraoni and Gardani 2014: 3) significantly coexists with the absence of partitive articles and the occurrence of bare plurals and plural indefinites etymologically derived from Latin unus ‘one’, on which, however, see also fn. 5 above. In a sense, Stark (2007), too, seems to admit that Spanish zeros should count as classifiers not only when she remarks that zero can be included “as a possible null determiner” assuming an underlying syntactic structure similar to overt determiners (Stark 2007: 51, fn. 3, 2008b: 46, fn. 1), but also when she observes that a Spanish zero can trigger an unbounded interpretation (“massification” in Heycock and Zamparelli’s terms 2005: 222) out of a prototypically bounded noun like ‘eagle’ (¿Has comido águila? ‘Have you eaten eagle?’), that is morphosyntactically massified by a French partitive classifier (As-tu mangé de l’ aigle?). Thus, it is the morphosyntactic arrangement of the Noun Phrase that triggers what might be considered as a form of “coercion” at the interpretative level, which confirms that Spanish and French share the same grammatical system, even though not the same grammatical tools.
80
squartini
the general paradigm I am using “bareness” as a “pre-theoretic characterization of linguistic surface material” (Wall and Kabatek 2013: 9) without necessarily implying empty slots in a Determiner Phrase (see also Schurr 2021: 2). I am simply contrasting the structural opposition between bare nominals and nominals having a determiner.20 5.3 Cognitive Patterns vs Grammatical Expressions If zeros count, the three languages have a tripartite paradigm opposing a bounded singular, a mass classifier (either through zero or a partitive article) and a plural form. But the very fact that the pattern is always tripartite21 should warn us to be cautious in interpreting it as suggesting a similarity between mass nouns and plurals. If they were really so similar, we would not expect a tripartite system, but a bipartite one, in which mass nouns should be formally conflated together with plurals, which is not the case in any of the systems we are describing here. Moreover, if one focuses on the bottom row dedicated to plurals, it will be observed that, despite all the differences pointed out by Herslund (2012), there is instead a general tendency that can be informally summarized as a form of parasitism of plurals on singulars. The fact is that all indefinite plurals are pluralized forms of determiners occurring with singular nouns. This
20
21
This pre-theoretic assumption is consistent with formal accounts such as Cardinaletti and Giusti’s (2018), who include zeros as representatives of the “same syntactic structure” also expressed by partitive determiners within the Determiner Phrase. Null determiners were also posited in Chierchia (1998b), where Romance nouns are generated as “predicates” and require determiners in order to be syntactically well-formed, which is typologically different from languages like Chinese, in which nouns are somehow lexically generated as “argumental” names of kinds immediately projectable as bare nouns in syntax. This implies that, in Chierchia’s (1998b: 355) analysis, Romance postverbal bare plurals, when possible (Spanish and Italian), require null determiners, which can only be licensed in special syntactic conditions like those guaranteed by the head-complement configuration of postverbal arguments. Among the syntactic mechanisms generating Romance bare plurals, one should also consider the possibility of syntactic incorporation (see DobrovieSorin and Laca 2003 on Spanish argumental bare plurals as “incorporated nouns”). Obviously the tripartite shape of the paradigm lies on the assumption that the grammar of classification / quantification is independent from the grammar of reference. Consistently with this assumption, Table 3.1 does not chart any opposition in terms of specificity. Otherwise bare plurals, which are restricted to non-specific interpretation (see §4 above), could not be conflated together within the same cell with “articled” plurals, which also acquire specific reading. The reader is referred to Cardinaletti and Giusti (2016, 2018) for a paradigmatic perspective in which the role of specificity is foregrounded. As a consequence of this focus on specificity, their paradigm of “indefinites” is also interestingly widened by adding definite articles with mass nouns, which share the same scopal effects as genuine indefinites in many Italo-Romance varieties (and the Italian standard variety as well).
quantification and classification in romance plural indefinites
81
applies across the board in the three languages here considered. They do differ in selecting which of the singular forms is parasitized by plural inflectional markers but it is always a plural marker that codifies the form. In this respect, there are no salient differences among the forms here considered: the plural of the partitive article that occurs in French and Italian and the plural form of the singular indefinite article (Spanish unos). Crucially, the parasitism of plurality over singulars also applies to zeros, in which the determiner slot is void and plurality is only expressed by plural inflectional endings on the noun. I know that what I am observing here sounds like a truism, which might be bafflingly summarized as “plurals are plural”, but in fact the perspective can be inverted by rephrasing the truism as follows: there is no dedicated form that conveys the special classificational meaning expressed by plurals, which implies that classification is always expressed via number. The question now is whether this has to be so or is due to the language sample and the data used so far, a point that will be elaborated in the following section.
6
Can Number Be Independent from Seinsart? Crucial Data from Romance Varieties
The elaboration of the data presented in the previous section has highlighted a crucial point in the relationship between the quantificational import of number and the possibility of interpreting plurals in classificational terms. The analysis has shown that number always prevails over classification, since “plurals are first of all plurals” and their potential classificational import is subsumed under the grammatical category traditionally called number. As suggested by Stark (2008b: 55), number might have a double face, a (morpho-)syntactic face represented by the feature [+ plur(al)] and a semantic face with classificational interpretation (‘semantic plural = count’). But, in any case, this classificational meaning is always subsumed under number, and, in fact, the Romance data analysed so far demonstrate that the switch to plurality is made through number, with the only proviso that in French the morphological relationship between the singular partitives masc du / fem de la and their ambigender plural des is much less transparent than in Italian and Spanish (see above fn. 3). Having downplayed the role of classification with plural indefinites, the natural question now is to wonder whether this result is due to the data / languages preliminarily chosen, i.e, is an empirical question, or is more generally due to an overarching structural constraint that makes it impossible to foreground classification in Romance plurals. Discussing this point will tell us whether the problem of a classificational interpretation of plurals can only be posited as
82
squartini
a theoretical hypothesis with no empirical correlation. To address this point we should pose another question: What should a system in which plurals synchronically express classification be like? In principle, I think that one should admit two different empirical arrangements as grammatical expressions of the classificational nature of plurals. The first and more obvious solution requires number to be totally absent and not expressed at all, which would allow classification expressed by other possible morphosyntactic means to prevail. The other possibility does not drop plural inflectional endings throughout, rather requiring a plural indefinite determiner that should not be the plural form of a singular determiner but a dedicated determiner designed to highlight the classificational nature of plurals. Note that the plural determiners considered so far are all plural forms of a singular determiner, but in principle this is not obligatorily the case. Actually, Romance languages do provide us with data that might go in these empirical directions, thus demonstrating that classification can be grammatically foregrounded in plural nouns. To check this point we have to go back to the typological assessment provided by Bossong (2016) in depicting the full range of possible morphosyntactic configurations covered by Romance partitive indefinite determiners (see Sections 3 and 4 above). Between the extreme poles of languages with no partitives at all (e.g. Spanish) and languages that have fully deployed the grammaticalization of partitive determiners (French), Bossong (2016) posits the intermediate continuum especially represented by Occitan varieties in which the partitive form is bare: it is the simple preposition derived from Latin de without any article-like form agglutinated to it and therefore unmarked as far as number (singular / plural) opposition is concerned. An “invariable de” (Stark and Gerards 2021) also occurs in standard French, where, however, its distribution is linked to scopal effect under negation (see Dobrovie-Sorin 2021 for a recent reappraisal), while in Occitan and FrancoProvençal it also occurs independently from polarity. Geolinguistic distribution and sociolinguistic pressures produce an extremely varied picture in the whole area (Sauzet 2012, 2014, Kristol 2014 and Stark and Gerards 2021), which includes varieties close to Occitan and Franco-Provençal in North-Western Italy. In particular, some dialects of Piedmont display a system of indefinite determiners that, in its general design, is directly comparable to the French and Italian distribution seen above. As in French, nouns tend to have lost inflectional endings, thus becoming interchangeable as singulars and plurals and, similarly to Italian, there are Piedmontese dialects in which the partitive has an articled form. However, the crucial point is that in the same area there are also varieties, in particular the koiné centred in the urban area of Turin (Berruto 1974: 57, Bonato 2004, Cerruti and Regis 2014, Miola 2017: 157, Cerruti and Regis
quantification and classification in romance plural indefinites 83
2020), where it is not only the noun but also the partitive determiner that displays no number marking. This is exemplified by the “bare partitive” marker ’d in (8): (8) a. a les ’d lìber ‘S/he reads books.’ b. a les ën lìber ‘S/he reads a book.’ c. a mangia ’d pan ‘S/he eats bread.’ As in oral French, the Piedmontese masculine noun is unmarked for number: the same form lìber occurs as a plural (8a) as well as a singular (8b) noun. But unlike French, the determiner ’d in (8a) is also unmarked for number in this variety of Piedmontese. On the one hand, the morphosyntactic structure made of ‘bare Determiner’ + ‘invariable Noun’ in (8a) can be contrasted to the singular form (8b) of the indefinite article ën (another instance of the grammaticalization of ‘one’ as an indefinite determiner), which shows that it is the opposition ën vs ’d that solely marks singular vs plural readings of countable nouns. In a sense, this is what happens in French, where the singular Noun Phrase un livre ‘a book’ is opposed to plural des livres ‘some books’. But a French partitive would be morphologically marked for plurality in cases like (8a), which triggers the contrastive opposition between plural countable nouns des livres ‘some books’ and singular mass nouns (du pain ‘some bread’) depicted in Table 3.1 above. On the contrary, in comparing Piedmontese plural (8a) and mass (8c) nouns, no morphological marking occurs. This distribution displays a complete neutralization of the opposition plural countable nouns and mass, both having exactly the same morphosyntactic structure ‘bare Determiner + invariable Noun’. To avoid misunderstandings, it is important to stress that this behavior only covers a subpart of the morphosyntax of indefinite Noun Phrases. Only regular masculine nouns are invariable in this variety. Unlike French, feminine nouns have different singular vs plural forms, which indicates that the tendency for the expression of number to be lost does not cover the whole grammatical system. Thus, I am not saying that number is not marked at all in these varieties. Apart from feminine nouns and some irregular masculine nouns, number is still robustly rooted within these systems through person agreement between Noun Phrases and Verb Phrases. What I am observing is simply a tendency for number to be lost within the Noun Phrase for a large subset of nouns (regular masculine nouns). Very generally, this tendency shows the independence between two phenomena (the grammaticalization of partitives as indefinite
84
squartini
determiners and the morphosyntactic expression of number) that are traditionally considered strictly related. More crucially from my perspective, I think that this morphosyntactic arrangement is what is needed to demonstrate that classification, acting as an independent grammatical category, belongs to the structural forces shaping the morphosyntactic system. The formal arrangement of the Piedmontese koiné confirms that the cognitive drive equating the apprehension of substances and aggregates is at work here, as also suggested by Chierchia (1998a) at a general semantic level and defended by Herslund (2012) in his classificational system of the Romance system. However, it is only when number is lost that it can be proven that the cognitive drive has actually entered the grammatical system, requiring a genuine grammaticalized marker of classification that does not interfere with number, which seems to be in line with generalizations on the tendency of number marking and classificational devices to be mutually exclusive across languages (Chierchia 1998b, Borer 2005 and Cooper and Hall 2012: 27). In the field of tension that organizes the apprehension of referents, another possible force is the tendency to interpret plurals as additive sets of discrete elements, as is prototypically the case in Indoeuropean languages (Stark 2007, 2008b and Gerards and Stark 2021). In order to demonstrate that this form of cognitive apprehension might also impact on the grammatical system we would need a form that, instead of neutralizing mass and count interpretations, tends in the opposite direction by displaying a dedicated marker that signals this special additive interpretation of plurality. This can be done by clearly marking the distinction with respect to masses. A dedicated marker of countability for plural nouns is in fact what we find if we look at multi-word quantifiers used as determiners. Some of them (take for instance Italian un sacco di ‘a lot of’) show a neutralization between mass and count interpretations: (9) a. Ho comprato un sacco di miele. ‘I bought a lot of honey.’ b. Ho comprato un sacco di libri. ‘I bought a lot of books.’ But the rich factory of new multi-word determiners also delivers markers specialized for count nouns, as is the case when the multi-word is based on the grammaticalization of a noun like ‘series’, which intrinsically denotes set of discrete referents. Thus, unlike un sacco di, una serie di ‘a series of’ and its “omnially” modified variant tutta una serie di ‘a whole series of’ are restricted to count interpretations (only plurals are admitted):
quantification and classification in romance plural indefinites 85
(10) a. *Ho comprato (tutta) una serie di miele. ‘I bought some honey.’ b. Ho comprato (tutta) una serie di libri. ‘I bought some books.’ A quantitative study on corpora (Onesti and Squartini 2007) has already demonstrated that (tutta) una serie di tends to prevail over other multi-word competitors in a grammaticalization race that now can be more clearly accommodated within a general tendency of the Romance system of determiners to reshape the structural relationship among grammatical categories in a competition between classification and number. Since it is a dedicated marker to count interpretation and is not the plural inflectional form of any elsewhere existing indefinite article, tutta una serie di is a good candidate to grammaticalize classification with a dedicated form, in which number can still be expressed on the noun (It. libr-i in (10b) is a plural) but not in the form of the determiner itself. Obviously, a multi-word determiner has not reached the same degree of grammaticalization as a pure determiner. It still retains the properties of a quantifier, being still admitted in answering a question focused on quantity (Quanti libri? ‘How many books?’ Una serie! ‘Many!’), but in any case it shows that the Romance system is ready to fully grammaticalize a form dedicated to classification and independent from number. In my view, its independence from number derives from the uncontroversial empirical observation that, in itself, tutta una serie di is not the inflectional plural of any other existing determiner and in terms of markedness it has the same impact as the indefinite article un ‘a’, which marks single bounded entities through a special form dedicated to that function. Not differently from the singular un, tutta una serie di is also a special form, which is dedicated to the function of marking a set of bounded entities. All in all, it seems that Herslund (2003b, 2012) was basically right in pointing to plurals as possible classificational elements. But his data only demonstrated classificational properties of plurals when morphosyntactic number is present, i.e. classification mediated via number. Instead, the data on the neutralization of number in koiné Piedmontese as well as the rich reservoir of multiword quantificational determiners demonstrate that classification is ready to be grammaticalized in Romance plurals in terms of the equation plurals = mass that has been suggested in formal semantics (Chierchia 1998a).
86 7
squartini
Conclusion
In a sense, the observations put together above represent another episode in the long quest for the real plural of an indefinite article (Lyons 1977). What is the genuine plural of indefinite articles parallel to English a(n)? From a Romance perspective, the problem is made more complicated by the occurrence of plural indefinites with different etymologies pointing to different classificational interpretations, whose interactions with number have been focused above. In this respect, the list of possible candidates as genuine plurals of the indefinite article might have a new entry now: the multi-word construction tutta una serie di, which shares various features with the singular indefinite article derived from ‘one’. Despite their different degrees of grammaticalization, they have both started off as quantifiers and they both have a marked classificational impact, being forms dedicated to bounded entities. This observation requires further research on the interplay of entrenched determiners and new multiwords in the field of tension between classification and number (a point that was also originally suggested by Herslund 2003b). For the time being, I have tried to contribute to the spirit of this book by showing how the grammar of quantification and the grammar of reference might interact in a complex and multifarious morphosyntactic arrangement, in which one should not overlook the connection to the grammar of classification. In this respect, the behavior of Piedmontese bare determiners has shown that classification per se may play a role that cumulatively involves plurals and masses and is codified independently from number.
Acknowledgements The structure of the argument in this chapter, in which I devote several pages to presenting caveats, disclaimers, and thorough evaluation of possible counterarguments, can be traced back to the fruitful discussions we had in the thriving scientific atmosphere created by Chiara Gianollo, Klaus von Heusinger and Maria Napoli during the two workshops (Cologne, 2016; Bologna, 2017) that, step by step, created the terrain for this publication. The active role played by Elisabeth Stark as discussant in Bologna as well as her generous written comments to a pre-conference version have greatly influenced my revisions, which also benefited from the careful comments made by external and internal reviewers on various intermediate drafts. Many thanks to all of them and to the audiences of the meetings. I am also very grateful to Matteo Rivoira for the Piedmontese data (Section 6), whose complex geolin-
quantification and classification in romance plural indefinites 87
guistic distribution has also been fruitfully discussed with Massimo Cerruti, Nicola Duberti, Riccardo Regis and Davide Ricca.
References Acquaviva, Paolo. 2008. Lexical plurals: A morphosemantic approach. Oxford: Oxford University Press. Berruto, Gaetano. 1974. Piemontese. Pisa: Pacini. Bonato, Massimo. 2004. Partitivo e articolo indeterminativo plurale nel piemontese parlato contemporaneo. Rivista Italiana di Dialettologia 28. 175–196. Borer, Hagit. 2005. In name only. Structuring sense, volume 1. Oxford: Oxford University Press. Bossong, Georg. 2016. Classifications. In Adam Ledgeway & Martin Maiden (eds.), The Oxford Guide to the Romance languages, 63–72. Oxford: Oxford University Press. Buridant, Claude. 2000. Grammaire nouvelle de l’ancien français. Paris: SEDES. Cardinaletti, Anna & Giuliana Giusti. 2015. Cartography and optional feature realizations in the nominal expression. In Ur Shlonsky (ed.), Where do we go from here? Chapters in syntactic cartography, 151–172. Oxford: Oxford University Press. Cardinaletti, Anna & Giuliana Giusti. 2016. The syntax of the Italian indefinite determiner dei. Lingua 181. 58–80. Cardinaletti, Anna & Giuliana Giusti. 2018. Indefinite determiners: Variation and optionality in Italo-Romance. In Diego Pescarini & Roberta D’Alessandro (eds.), Advances in Italian dialectology. Sketches of Italo-Romance grammars, 135–161. Leiden: Brill. Cardinaletti, Anna & Giuliana Giusti. 2020. Indefinite determiners in informal Italian: A preliminary analysis. Linguistics 58(3). 679–712. Carlier, Anne. 2001. La genèse de l’article un. Langue Française 130. 65–88. Carlier, Anne. 2007. From preposition to article: The grammaticalization of the French partitive. Studies in Language 31. 1–49. Carlier, Anne & Béatrice Lamiroy. 2014. The grammaticalization of the prepositional partitive in Romance. In Silvia Luraghi & Tuomas Huumo (eds.), Partitive cases and related categories, 477–522. Berlin: de Gruyter. Cerruti, Massimo & Riccardo Regis. 2014. Partitives in Piedmontese: A preliminary overview. Talk given at the international workshop “Partitivity in Romance and beyond”. University of Zurich: 11–13 December 2014. Cerruti, Massimo and Riccardo Regis. 2020. Partitive determiners in Piedmontese: A case of language variation and change in a contact setting. Linguistics 58(3). 651– 677.
88
squartini
Chierchia, Gennaro. 1997. Partitives, reference to kinds and semantic variation. In Aaron Lawson (ed.), Proceedings of SALT 7, 73–98. New York: Cornell University. Chierchia, Gennaro. 1998a. Plurality of mass nouns and the notion of semantic parameter. In Susan Rothstein (ed.), Events and grammar, 53–103. Dordrecht: Kluwer. Chierchia, Gennaro. 1998b. Reference to kinds across languages. Natural Language Semantics 6. 339–405. Chierchia, Gennaro. 2010. Mass nouns, vagueness and semantic variation. Synthese 174. 99–149. Corbett, Greville. 2000. Number. Cambridge: Cambridge University Press. Cowper, Elizabeth & Daniel C. Hall. 2012. Aspects of individuation. In Diane Massam (ed.), Count and mass across languages, 27–53. Oxford: Oxford University Press. Cowper, Elizabeth & Daniel C. Hall. 2014. The features and exponence of nominal number. Lingue e linguaggio 13. 63–82. Deal, Amy Rose. 2017. Countability and semantic distinctions. Natural Language Semantics 25. 125–171. Delfitto, Denis & Jan Schroten. 1991. Bare plurals and the number affix in DP. Probus 3(2). 155–185. Dobrovie-Sorin, Carmen. 2021. Negation, des-indefinites and bare nouns across languages. In Tabea Ihsane (ed.), Disentangling bare nouns and nominals introduced by a partitive article, 187–226. Leiden: Brill. Dobrovie-Sorin, Carmen & Brenda Laca. 2003. Les noms sans déterminants dans les langues romanes. In Danièle Godard & Anne Abeillé (eds.), Les langue romanes. Problèmes de la phrase simple, 235–281. Paris: CNRS. Doron, Edit & Ana Müller. 2014. The cognitive basis of the mass-count distinction: Evidence from bare nouns. In Patricia Cabredo Hofherr & Anne Zribi-Hertz (eds.), Crosslinguistic studies on noun phrase structure and reference, 73–101. Leiden & Boston: Brill. Dryer, Matthew S. 2005. Indefinite articles. In Matthew S. Dryer, David Gil & Bernard Comrie (eds.), The World Atlas of Language Structures, 158–161. Oxford: Oxford University Press. Enç, Murvet. 1991. The semantics of specificity. Linguistic Inquiry 22. 1–26. Espinal, M. Teresa. 2010. Bare nominals in Catalan and Spanish. Their structure and meaning. Lingua 120. 984–1009. Espinal, M. Teresa. 2013. Bare nominals, bare predicates: Properties and related types. In Johannes Kabatek & Albert Wall (eds.), New perspectives on bare noun phrases in Romance and beyond, 63–94. Amsterdam: Benjamins. Espinal, M. Teresa & Louise McNally. 2011. Bare nominals and incorporating verbs in Catalan and Spanish. Journal of Linguistics 47. 87–128. Gerards, David Paul & Elisabeth Stark. 2021. Why “partitive articles” do not exist in (Old)
quantification and classification in romance plural indefinites 89 Spanish. In Tabea Ihsane (ed.), Disentangling bare nouns and nominals introduced by a partitive article, 105–139. Leiden: Brill. Giurgea, Ion. 2013. The syntax of determiners and other functional categories. In Carmen Dobrovie-Sorin & Ion Giurgea (eds.), A reference grammar of Romanian, volume 1. The noun phrase, 97–174. Amsterdam: Benjamins. Giusti, Giuliana. 2021. A protocol for indefinite determiners in Italian and ItaloRomance. In Tabea Ihsane (ed.), Disentangling bare nouns and nominals introduced by a partitive article, 262–300. Leiden: Brill. Givón, Talmy. 1981. On the development of the numeral ‘one’ as an indefinite marker. Folia Linguistica Historica 2. 35–53. Greenberg, Yael. 2003. Manifestations of genericity. New York & London: Routledge. Gutiérrez-Rexach, Javier. 2001. The semantics of Spanish plural existential determiners and the dynamics of judgment types. Probus 13. 113–154. Herslund, Michael. 1998. Le français, langue à classificateurs? In Annick Englebert, Michel Pierrard, Laurence Rosier & Dan van Raemdonck (eds.), La ligne claire. De la linguistique à la grammaire. Mélanges offerts à Marc Wilmet à l’occasion de son 60° anniversaire, 65–73. Paris: Duculot. Herslund, Michael. 2003a. Le pluriel de l’article indéfini en ancien français. In Peter Blumenthal & Jean-Emmanuel Tyvaert (eds.), La cognition dans le temps. Etudes cognitives dans le champ historique des langues et des textes 75–84. Tübingen: Niemeyer. Herslund, Michael. 2003b. Articles et classificateurs. Cahiers Ferdinand de Saussure 56. 21–33. Herslund, Michael. 2008. Articles, definite and indefinite. In Henrik Høeg Müller & Alex Klinge (eds.), Essays on nominal determination: from morphology to discourse management, 27–43. Amsterdam: Benjamins. Herslund, Michael. 2012. Grammaticalisation and the internal logic of the indefinite article. Folia linguistica 46(2). 341–357. von Heusinger, Klaus. 2011. Specificity, referentiality and discourse prominence: German indefinite demonstratives. In Ingo Reich, Eva Horch & Dennis Pauly (eds.), Proceedings of Sinn und Bedeutung 15, 9–30. Saarbrücken: Universaar—University of Saarland Press. von Heusinger, Klaus & Georg Kaiser. 2005. Animacy, specificity and definiteness in Spanish. In Klaus von Heusinger & Georg Kaiser (eds.), Proceedings of the workshop “Semantic and syntactic aspects of specificity in Romance languages”, 67–101. Arbeitspapier 113. Fachbereich Sprachwissenschaft, Universität Konstanz. Heycock, Caroline & Roberto Zamparelli. 2003. Coordinated bare definites. Linguistic Inquiry 34. 443–469. Heycock, Caroline & Roberto Zamparelli. 2005. Friends and colleagues: Pluralities, coordination, and the structure of DP. Natural Language Semantics 13. 201–270. Jackendoff, Ray. 1991. Parts and boundaries. Cognition 41. 9–45.
90
squartini
Jensen, Frede. 1994. Syntaxe de l’ancien occitan. Tübingen: Niemeyer. Jespersen, Otto. 1924. The philosophy of grammar. London: Allen & Unwin. Körner, Karl-Hermann. 1987. Korrelative Sprachtypologie. Die zwei Typen romanischer Syntax. Stuttgart: Steiner. Korzen, Iørn. 1996. L’articolo italiano fra concetto ed entità, i–ii. København: Museum Tusculanum Press. Krifka, Manfred. 1988. The relational theory of genericity. In Manfred Krifka (ed.), Genericity in natural language. Proceedings of the 1988 Tübingen Conference, 285–312. Tübingen: Seminar für natürlich-sprachliche Systeme der Universität Tübingen. Kristol, Andres. 2014. Les grammaires du francoprovençal: L’expression de la partitivité. Quelques leçons du projet ALAVAL. In Actes de la Conférence annuelle sur l’activité scientifique du Centre d’études francoprovençal René Willien, 29–44. Saint-Nicolas (Aoste): Région autonome de la Vallée d’Aoste, Bureau régional pour l’ethnologie et la linguistique. Laca, Brenda. 1996. Acerca de la semántica de los plurales escuetos del español. In Ignacio Bosque (ed.), El sustantivo sin determinación: la ausencia del determinante en la lengua española, 241–268. Madrid: Visor Libros. Laca, Brenda. 1999. Presencia y ausencia de determinante. In Ignacio Bosque & Violeta Demonte (eds.), Gramática descriptiva de la lengua española, 891–928. Madrid: Espasa Calpe. Laca, Brenda. 2013. Spanish bare plurals and topicalization. In Johannes Kabatek & Albert Wall (eds.), New perspectives on bare noun phrases in Romance and beyond, 95–119. Amsterdam: Benjamins. Laca, Brenda & Liliane Tasmowski. 1994. Le pluriel indéfini de l’attribut métaphorique. Linguisticae Investigationes 18. 27–47. Laca, Brenda & Liliane Tasmowski-De Ryck. 1994. Référentialité du pluriel indéfini dans les langues romanes. In Laurent Danon-Boileau & Mary-Annick Morel (eds.), L’indéfini. Faits de Langues 4. 97–104. Landman, Fred. 2016. Iceberg semantics for count nouns and mass nouns: Classifiers, measures and portions. The Baltic International Yearbook of Cognition, Logic and Communication 11. 1–48. Ledgeway, Adam. 2012. From Latin to Romance. Morphosyntactic typology and change. Oxford: Oxford University Press. Leonetti, Manuel. 1999. El artículo. In Ignacio Bosque & Violeta Demonte (eds.), Gramática descriptiva de la lengua española, 787–890. Madrid: Espasa Calpe. Leonetti, Manuel. 2004. Specificity and Differential Object Marking in Spanish. Catalan Journal of Linguistics 3. 75–114. Leonetti, Manuel. 2012. Indefiniteness and specificity. In José Ignacio Hualde, Antxon Olarrea & Erin O’Rourke (eds.), The handbook of Hispanic linguistics, 285–305. Madrid: Espasa.
quantification and classification in romance plural indefinites
91
Leonetti, Manuel. 2013. Information structure and the distribution of Spanish bare plurals. In Johannes Kabatek & Albert Wall (eds.), New perspectives on bare noun phrases in Romance and beyond, 121–155. Amsterdam: Benjamins. Link, Godehard. 1983. The logical analysis of plurals and mass terms: A lattice-theoretical approach. In Rainer Bäuerle, Christoph Schwarze & Arnim von Stechow (eds.), Meaning, use and interpretation of language, 302–323. Berlin: de Gruyter. Longobardi, Giuseppe. 2001. How comparative is semantics? A unified theory of bare nouns and proper names. Natural Language Semantics 9. 335–369. López, Luis. 2016. (In)definiteness, specificity, and Differential Object Marking. In Susann Fischer & Christoph Gabriel (eds.), Manual of grammatical interfaces in Romance, 241–266. Berlin: de Gruyter. López-Palma, Helena. 2007. Plural indefinite descriptions with unos and the interpretation of number. Probus 19. 235–266. Loporcaro, Michele, Vincenzo Faraoni & Francesco Gardani. 2014. The third gender of Old Italian. Diachronica 31. 1–22. Lyons, John. 1977. Semantics, volume 2. Cambridge: Cambridge University Press. Maiden, Martin. 2016. Number. In Adam Ledgeway & Martin Maiden (eds.), The Oxford Guide to the Romance languages, 697–707. Oxford: Oxford University Press. Maiden, Martin & Cecilia Robustelli. 2000. A reference grammar of modern Italian. London: Arnold. Martí, Luisa. 2008. The semantics of plural indefinite noun phrases in Spanish and Portuguese. Natural Language Semantics 16. 1–37. Märzhäuser, Christina. 2013. Coordinated bare nouns in French, Spanish and European Portuguese. In Johannes Kabatek & Albert Wall (eds.), New perspectives on bare noun phrases in Romance and beyond, 283–300. Amsterdam: Benjamins. Miola, Emanuele. 2017. The position of Piedmontese on the Romance grammaticalization cline. Folia linguistica 51. 133–167. Morin, Yves-Charles & Jonathan Kaye. 1982. The syntactic basis for French liaison. Journal of Linguistics 18. 291–330. Müller, Ana. 2002. The semantics of generic quantification in Brazilian Portuguese. Probus 14. 279–298. Onesti, Cristina & Mario Squartini. 2007. Tutta una serie di. Lo studio di un pattern sintagmatico e del suo statuto grammaticale. In Manuel Barbera, Elisa Corino & Cristina Onesti (eds.), Corpora e linguistica in rete, 271–284. Perugia: Guerra. Pires de Oliveira, Roberta & Susan Rothstein. 2013. Information structure and the distribution of Spanish bare singular arguments in Brazilian Portuguese. In Johannes Kabatek & Albert Wall (eds.), New perspectives on bare noun phrases in Romance and beyond, 189–222. Amsterdam: Benjamins. Pomino, Natascha. 2016. On the clitic status of the plural marker in phonic French. In Susann Fischer & Mario Navarro (eds.), Proceedings of the vii Nereus interna-
92
squartini
tional workshop ‘Clitic doubling and other issues of the syntax / semantics interface in Romance DPs’, 105–130. Arbeitspapier 128. Fachbereich Sprachwissenschaft, Universität Konstanz. Pountain, Christopher. 2019. The development of the articles in Castilian. Languages 4. 20. 1–17. Special issue: Dorien Nieuwenhuijsen & Mar Garachana (eds.), Studies in historical linguistics and language change. Grammaticalization, refunctionalization and beyond. Rijkhoff, Jan. 2002. The noun phrase. Oxford: Oxford University Press. Rothstein, Susan. 2010. Counting and the mass-count distinction. Journal of Semantics 27. 343–397. Rothstein, Susan. 2017. Semantics for counting and measuring. Cambridge: Cambridge University Press. Sauzet, Patrick. 2012. Occitan plurals. A case for a morpheme-based morphology. In Sascha Gaglia & Marc-Oliver Hinzelin (eds.), Inflection and word formation in Romance languages, 179–200. Amsterdam: Benjamins. Sauzet, Patrick. 2014. ‘Partitives’, number marking and negation in Occitan. Talk given at the international workshop “Partitivity in Romance and Beyond”. University of Zurich: 11–13 December 2014. Schroten, Jan. 2001. L’absence de determinant en espagnol. In Georges Kleiber, Brenda Laca & Liliane Tasmowski (eds.), Typologie des groupes nominaux, 189–203. Rennes: Presses Universitaires de Rennes. Schurr, Agay. 2021. Bound to be? Bare and partitive-marked noun phrases in Romance languages and the emergence of prominence-conditioned patterns. In Tabea Ihsane (ed.), Disentangling bare nouns and nominals introduced by a partitive article, 54–104. Leiden: Brill. Seiler, Hansjakob. 1986. Apprehension. Language, object, and order. Part iii: The universal dimension of apprehension. Tübingen: Narr. Stan, Camelia. 2016. The article and other determiners. In Gabriela Pană Dindelegan (ed.), The syntax of Old Romanian, 288–303. Oxford: Oxford University Press. Stark, Elisabeth. 2005. Typological correlations in nominal determination in Romance. In Klaus von Heusinger, Georg A. Kaiser & Elisabeth Stark (eds.), Specificity and the evolution / emergence of nominal determination systems in Romance, 123–138. Arbeitspapier 119. Fachbereich Spachwissenschaft, Universität Konstanz. Stark, Elisabeth. 2007. Gender, number, and indefinite articles: About the ‘typological inconsistency’ of Italian. In Elisabeth Stark, Elisabeth Leiss & Werner Abraham (eds.), Nominal determination: Typology, context constraints, and historical emergence, 49–71. Amsterdam: Benjamins. Stark, Elisabeth. 2008a. The role of the plural system in Romance. In Ulrich Detges & Richard Waltereit (eds.), The paradox of grammatical change. Perspectives from Romance, 57–84. Amsterdam: Benjamins.
quantification and classification in romance plural indefinites 93 Stark, Elisabeth. 2008b. Typological correlations in nominal determination in Romance. In Henrik Høeg Müller & Alex Klinge (eds.), Essays on nominal determination: from morphology to discourse management, 45–61. Amsterdam: Benjamins. Stark, Elisabeth. 2016. Nominal morphology and semantics—Where’s gender in Romance? In Susann Fischer & Mario Navarro (eds.), Proceedings of the vii Nereus international workshop ‘Clitic doubling and other issues of the syntax / semantics interface in Romance DPs’, 131–149. Arbeitspapier 128. Fachbereich Sprachwissenschaft, Universität Konstanz. Stark, Elisabeth & David Paul Gerards. 2021. “Partitive articles” in Aosta Valley Francoprovençal—Old questions and new data. In Tabea Ihsane (ed.), Disentangling bare nouns and nominals introduced by a partitive article, 301–334. Leiden: Brill. Storch, Anne & Gerrit J. Dimmendaal. 2014. One size fits all? On the grammar and semantics of singularity and plurality. In Anne Storch & Gerrit J. Dimmendaal (eds.), Number. Constructions and semantics. Case studies from Africa, India, Amazonia & Oceania, 1–33. Amsterdam: Benjamins. Suñer, Margarita. 1982. The syntax and semantics of Spanish presentational sentencetypes. Washington DC: Georgetown University Press. Treviño, Esthela. 2010. Bare partitives in modern Spanish. In Alejandra Capistrán-Garza & Edgar Madrid (eds.), Estudios de lingüística teórica, 49–95. Mexico: Universidad Autónoma Metropolitana / Ediciones del Lirio. Vincent, Nigel. 2017. Determination and quantification. In Andreas Dufter & Elisabeth Stark (eds.), Manual of Romance morphosyntax and syntax, 727–769. Berlin: de Gruyter. Wall, Albert & Johannes Kabatek. 2013. Laying bare nominal determination. An introduction to the present volume. In Johannes Kabatek & Albert Wall (eds.), New perspectives on bare noun phrases in Romance and beyond, 1–33. Amsterdam: Benjamins. Wilmet, Marc. 2008. A stranger in the house: The French article de. In Henrik Høeg Müller & Alex Klinge (eds.), Essays on nominal determination: From morphology to discourse management, 65–78. Amsterdam: Benjamins. Zamparelli, Roberto. 2008. Dei ex-machina: A note on plural / mass indefinite determiners. Studia Linguistica 63. 301–327.
chapter 4
Topics and the Interpretation of Referential Null Subjects Manuel Leonetti
1
Introduction
Most of the advances in contemporary research on null arguments, and in particular null subjects, concern the problem of licensing, i.e. what the conditions are for a language to display null arguments (see the discussion in Biberauer et al. (eds.) 2010, Duguine 2013, 2014, Camacho 2013, Sheehan 2016, a.o.). Since not all languages display null subjects, the problem of licensing must be seen as a strictly grammatical issue.1 A related problem concerns the interpretation of null subjects, i.e. how they are assigned a referential value in languages that do feature them. It is reasonable to assume that the problem of interpretation involves the interaction of the grammatical system and general pragmatic principles, since determining the reference of null subjects is a context dependent task. Thus, how hearers choose the optimal discourse antecedent for a null subject is not a strictly grammatical issue, but rather a matter of interplay between grammar and pragmatics; moreover, this is also true for the interpretation of overt pronouns in discourse (Arnold 2010, Gundel, Hedberg and Zacharski 1993, 2012, Gundel 2010, Kehler 2002, a.o.). According to this view, the central research question in this area should be how the grammar contributes to interpretation, in particular how it constrains possible interpretations of null arguments, given that they apparently encode no instructions for reference assignment. In this paper my aim is to address the problem of interpretation of null subjects (NSs), by investigating how the division of labour between grammar and pragmatics is established. I will not deal with the problem of licensing— though the connection between formal licensing and interpretation is surely worth examining and is still a major issue for many researchers (see Section 4.1 for some discussion). Instead, I will focus on the role of a well-known factor that contributes to determining the interpretation of NS s: topicality of
1 Not necessarily syntactic. See Duguine (2013).
© Manuel Leonetti, 2022 | doi:10.1163/9789004473324_005
topics and the interpretation of referential null subjects
95
the antecedent. The starting point will be the preference of NS s for topical antecedents that has been discussed in the literature from different perspectives (see Frascarelli 2007, 2018, Camacho 2013 for a formal approach, and Di Eugenio 1998, Filiaci 2011, Filiaci, Sorace and Carreiras 2013, Taboada 2008, Taboada and Wiesemann 2010 for a processing approach). If we assume that the grammar of null subject languages (NSLs) encodes some kind of restrictive information-structural condition on the antecedents of NS s, we obtain a plausible way to model the contribution of the grammatical system to interpretation, while still keeping a complementary role for pragmatic inference. In a nutshell, the idea could be that the requirement of a topical antecedent guarantees that a discourse antecedent for the NS has to be identified; in order to satisfy this requirement, general pragmatic principles, like the communicative principle of relevance, should determine the choice of an adequate antecedent and thus specify the optimal reading. This seems to be a simple, reasonable view of the division of labour between grammar and pragmatics: for each construction, the grammatical system sets the constraints on interpretation, and pragmatic inference supplies the contextual assumptions needed to reach a relevant interpretation. However, there are reasons to believe that for NS s the contribution of grammar is not exactly the one just sketched in this picture. I will try to show that this contribution is actually reduced to a minimum,2 and the role of pragmatics, on the other hand, is decisive. More precisely, I intend to show that syntactic accounts like the one put forward in Frascarelli (2007, 2018) and Frascarelli and Jiménez-Fernández (2019), although quite successful and seldom challenged in the tradition of formal linguistics, are ill-oriented, because they are based on an inadequate view of the interaction between grammar and pragmatics. To make clear how the basic grammar-pragmatics distinction is conceived in this paper I will say that I am simply assuming that what is conventionally encoded in the features of lexical items, in rules and in constraints belongs to the grammar, and what is inferred by speakers in the interpretive process belongs to pragmatics (see Ariel 2008). Though the critical review of Frascarelli’s proposal is the main goal of this paper, the paper is also meant to address a whole trend of research in current grammatical theory that is based on the general assumption that every aspect
2 As is well known, in languages like Italian and Spanish the central condition imposed by the grammar on the identification of antecedents is established by subject agreement, i.e. by person and number features. My claim is that the grammar does not encode any further condition on antecedents, like definiteness or topicality. One might think that the mere fact that the phenomenon is limited to subjects reveals another syntactic condition: this is true, but it is related to the licensing problem, not to the interpretation problem.
96
leonetti
of the interpretation of an utterance is represented in its syntactic structure, which, on the one hand, leads to positing quite complex structures, and on the other hand leads to accepting a very strict isomorphism between the syntactic and the interpretive component. Frascarelli’s approach is a paradigmatic example of this line of thought. A few brief remarks will be enough to set the limits of this investigation explicitly. First of all, I will discuss only so-called referential null subjects, thus excluding the case of expletive and arbitrary null elements. Second, I will only analyse data from consistent null subject languages (NSL s) like Spanish and Italian; this excludes other kinds of NSLs from consideration, i.e. partial NSL s like Finnish and Brazilian Portuguese (Holmberg, Nayudu and Sheehan 2009, Frascarelli 2018, Frascarelli and Jiménez-Fernández 2019) and discourse or radical NSLs like Chinese. Third, I only consider third person subjects. In a sense, I concentrate on the familiar data that constitute the core of the literature on NSs. The paper is organized as follows. Section 2 reviews some ideas in the literature on the role of topicality in the interpretation of NS s; these assumptions should provide the context for the discussion in the following sections. Section 3 presents a critical view of the proposal in Frascarelli (2007, 2018) and some related work; the problem with this account is that (a) there is no reason for including a condition on topics in the grammar of NS s, and (b) the often signalled preference for topical antecedents, when it is in force, is simply a pragmatic effect. Section 4 comments on some theoretical consequences and some pending questions, and aims at placing the whole discussion inside a wider context. Section 5 presents some conclusions.
2
The Preference for Topical Antecedents
In the recent literature on NSs it is usually assumed, at least for consistent NSLs like Spanish and Italian, that referent identification in referential NS s is topic-oriented, i.e. it depends on a matching relation between the subject and a preceding topic3 (see Calabrese 1986, Cordin 1988, Grimshaw and Samek-
3 The literature on pronominal anaphora in discourse often mentions a preference for subject antecedents, instead of a preference for topic antecedents. I assume that the tendency to choose subjects as antecedents of pronouns is mostly a consequence of the prominence of subjects as sentential topics, so that ultimately it is the notion of topic that is relevant:
topics and the interpretation of referential null subjects
97
Lodovici 1998, Frascarelli 2007, 2018, Sigurðsson 2011, Camacho 2013, a.o. from a formal perspective; Di Eugenio 1998, Carminati 2002, Alonso-Ovalle et al. 2002, Gelormini-Lezama and Almor 2011, Filiaci 2011, Filiaci, Sorace and Carreiras 2013, Taboada and Wiesemann 2010, Godoy, Weissheimer and Araújo Mafra 2018 from a processing perspective). Here topic must be understood as what the sentence is about (Reinhart 1981). A preference for topic antecedents is in fact noticeable in well-known contrasts in Italian like the one in (1)–(2), from Samek-Lodovici (1996: 31); the NS in the second clause is represented as e, for ‘empty’, to avoid any commitment with respect to the nature of the gap (but see Section 4.1 for some remarks on this issue). (1) Questa mattina, la mostra è stata visitata da this morning the exhibition be.prs.3sg been visit.ptcp by Giannii. Più tardi, e*i ha visitato l’ università. Gianni more late have.prs.3sg visit.ptcp the university ‘This morning, the exhibition was visited by John. Later on, he visited the university.’ visitato la mostra. Più (2) Questa mattina, Giannii ha this morning Gianni have.prs.3sg visit.ptcp the exhibition more tardi, ei ha visitato l’ università. late have.prs.3sg visit.ptcp the university ‘This morning, John visited the exhibition. Later, he visited the university.’ In (1), the DP Gianni is inside a by-phrase and, according to Samek-Lodovici, can hardly be taken as the antecedent of the NS in the second clause,4 whereas the same DP provides an adequate antecedent for the NS in (2), where it appears as the subject of the first clause. As preverbal subjects are naturally interpreted as topics, and by-phrases are not, it is reasonable to conclude that the availability of a topic antecedent is a crucial condition for the licensing of NSs in languages like Italian (see Rizzi 2018: 515 for similar data). This fits in quite well with the well-established correlation between zero forms and
the right generalization should not concern grammatical functions, since there is a natural motivation for topics to be chosen as antecedents, but no clear motivation for the privileged status of subjects. 4 In (1) the intended coreferential reading is marked as ungrammatical, with an asterisk. Here, as in the rest of this section, I am just following the usual modus operandi in the syntactic literature, in particular in Samek-Lodovici (1996). The discussion in Section 4 will make clear that such reading is simply infelicitous or anomalous.
98
leonetti
discourse-given information (Givón 2017). Samek-Lodovici (1996: 29) puts forward the generalization in (3) and shows that it holds for data from Italian, Greek, Hebrew and Chinese. (3) Null subjects must be licensed by topic antecedents. A related fact is the asymmetry between null and overt subjects that can be observed in (4)–(5). In both Italian examples there is coordination between two clauses; the first one includes two potential antecedents, and the second one includes an anaphoric item, either a NS, as in (4), or an overt pronoun, as in (5). (4) Mariai ha salutato Paolaj, e poi ei lj’ ha Maria have.prs.3sg greet.ptcp Paola and then her have.prs.3sg abbracciata. hug.ptcp ‘Mary greeted Paula and then hugged her.’ (5) Mariai ha salutato Paolaj, e poi leij li’ Maria have.prs.3sg greet.ptcp Paola and then she her abbracciata. ha have.prs.3sg hug.ptcp ‘Mary greeted Paula and then she hugged her.’ The examples show that whereas NSs tend to choose topics as antecedents, overt pronouns tend to prefer non-topic antecedents: in (4) the NS, in its strongly preferred reading, refers to Maria—the preverbal subject, which is also the unmarked topic—and in (5) the pronoun lei is preferably understood as referring to Paola. The diverging behaviour of NS s and overt subject pronouns is related to the competition, and the resulting division of labour, between the two forms in NSLs. As null forms are more economical than overt forms and carry less grammatical features, it is expected that they will be only helpful in pointing towards the intended referent when such referent is highly accessible, i.e. when it is maximally salient, and thus easy to identify, for the addressee. Referents that count as topics of previous predications in the discourse are prototypically accessible, which makes them ideal candidates for reference by means of a NS; in general, when referents are not highly accessible, null elements, being deficient or not informative enough, are not adequate for retrieving them, and an overt anaphoric device is usually preferred. Thus, the preference for topical antecedents explains, on the one hand, the interpretations that
topics and the interpretation of referential null subjects
99
speakers tend to assign to NSs, and, on the other hand, the interpretive contrasts between NSs and overt pronouns in NSLs.5 Both aspects were addressed, on the basis of data from Italian, in some seminal papers by Andrea Calabrese (notably, Calabrese 1986) in which he resorted to the term Thema, understood as subject of predication, to characterize the preferred antecedents of NS s in Italian. According to Calabrese (1986: 26–27), the contrast between (6) and (7) shows that the NS in the main clause tends to pick out the topical subject—the Thema—of the preceding adverbial clause as its antecedent, instead of choosing a complement, whereas the overt pronoun lui exhibits the opposite preference and tends to pick out the complement—a non-topical expression—as antecedent. This is the same pattern observed in (4)–(5). picchiato Antonioj, ei/*j era (6) Quando Marioi ha when Mario have.prs.3sg hit.ptcp Antonio be.pst.3sg ubriaco. drunk ‘When Mario hit Antonio, he was drunk.’ (7) Quando Marioi ha picchiato Antonioj, lui?i / j era when Mario have.prs.3sg hit.ptcp Antonio he be.pst.3sg ubriaco. drunk The association of NSs with topical subjects—the “Calabrese effect”, in Rizzi’s terms (Rizzi 2018)—is explained in Calabrese (1986) along the following lines. Calabrese claims that overt pronouns are used only when the mention of their referents is not expected in discourse, which implies that their weak, unstressed or null competitors—i.e. clitic and null pronouns—are used under the opposite condition, i.e. when the occurrence of their referents is expected. This principle, in combination with another principle that states that a sub5 A reviewer rightly points out that, although in principle this generalization holds true across Romance languages, recent research shows that there are intriguing differences concerning the division of labour between null and overt subjects (see Filiaci, Sorace and Carreiras 2013 for a comparison of Italian and Spanish, Torregrossa, Bongartz and Tsimpli 2015 for Italian and Greek, and Dufter 2011 for a comparison of Old and Modern Spanish). This is, in fact, one of the main trends in current research on the interpretation of null subjects. I cannot deal with this issue here, but the data of cross-linguistic variation suggest that we face a multifactorial problem that involves grammar and information structure. Thus, such data could fit much better in an account based on the interaction of syntax and pragmatics like the one I defend here than in a purely syntactic one.
100
leonetti
ject pronoun is expected to have the referent of another subject, accounts for the salient readings of (6)–(7): the NS in (6) takes the expected antecedent, the preverbal subject Mario, and the overt pronoun in (7) takes the less expected antecedent, the object Antonio. The possibility that the NS takes some other discourse topic as antecedent is also predicted by the principles, but is not represented in the examples. Calabrese (1986: 31) reformulates his second principle by substituting the term subject with the term Thema, as in (8), which is in fact an explicit statement of the preference for topical antecedents (notice that NSs are considered as instances of Themas, i.e. as preverbal topical subjects): (8) A pronominal in position of Thema is expected to have a referent of another Thema. This principle leads us to conclude that NSs are a specialized device for marking topic continuity (Givón 1983, 2017), since they contribute to maintaining the discourse topic. One of the strong predictions of the principle is that NS s should be unable to take antecedents in focus. The prediction is actually false, as will become clear later, in Section 3.3, but this does not make Calabrese’s observations less interesting. He was aware of the fact that his principle has to be constrained in some way (Calabrese 1986: 33, Rizzi 2018: 518), by taking into consideration different factors that interact in a complex way, such as the type of subordinate clause in contexts like the one in (6)–(7), and the relative order of main and subordinate clauses.6 I cannot deal with this important issue in this paper, and I will merely stress the value of Calabrese’s findings, without even discussing the nature of generalizations like the principle in (8). Another interesting piece of evidence pointing towards the same direction comes from the contrast in Italian in (9)–(10), from Cordin (1988: 548), which concerns backwards anaphora and the properties of preverbal and postverbal subjects (see also Calabrese 1992: 99). In the examples, the NS appears in an adverbial clause, and its potential antecedent, the DP Gianni, is the subject of the main clause (I discard readings with another possible discourse antecedent, since they are irrelevant here).
6 It has often been noted that the distinction between forward and backward anaphora has remarkable consequences for the interpretation of NS s (see, for example, Tsimpli et al. 2004).
topics and the interpretation of referential null subjects
101
(9) Dopo che ei è arrivato, Giannii ha after that be.prs.3sg arrive.ptcp Gianni have.prs.3sg parlato. speak.ptcp ‘After he arrived, John spoke.’ (10) Dopo che e*i è arrivato, ha parlato after that be.prs.3sg arrive.ptcp have.prs.3sg speak.ptcp Giannii. Gianni The contrast shows that the preverbal subject in (9) is an adequate antecedent for the NS in the subordinate clause, but the postverbal subject in (10) is not. It is commonly accepted that the basic difference between the two positions for the subject is associated with information structure: preverbal subjects tend to be topics, and postverbal ones are usually foci. This again suggests that NS s require topical antecedents. Though the contrast in (9)–(10) is straightforward, it is not clear that it holds systematically in every context with an anaphoric link between a main clause and an adverbial clause, due to the complex interplay of different factors mentioned above. However, the same contrast involving preverbal and postverbal overt subjects as potential antecedents of NS s shows up again in other environments, such as control constructions, including absolute clauses, which suggests that in fact there is something robust and systematic in it. Depending on the chosen analysis, in control structures the NS may have different properties from the ones that NSs display in finite clauses, but in any case it is worth including this case in the discussion; the Italian examples in (11)–(12), from Calabrese (1992: 99), show how only preverbal subjects—and not postverbal ones—are able to trigger control of NS s in gerundival and participial constructions.7 (11) a. ei Passeggiando nel parco, Carloi l’ ha abbracciata. walk.ger in-the park Carlo her have.prs.3sg hug.ptcp ‘Walking in the park, Carlo hugged her.’
7 Camacho (2011, 2013) analyses parallel contrasts in Spanish. In his approach, the NS in absolute constructions is treated as a small pro, i.e. like NS s in finite clauses.
102
leonetti
b. Appena ei uscito di casa, Carloi l’ ha as soon as leave.ptcp of house Carlo her have.prs.3sg vista. see.ptcp ‘As soon as he left the house, Carlo saw her.’ (12) a. e*i Passeggiando nel parco, l’ ha abbracciata walk.ger in-the park her have.prs.3sg hug.ptcp Carloi. Carlo ‘Walking in the park, Carlo hugged her.’ di casa, l’ ha vista b. Appena e*i uscito as soon as leave.ptcp of house her have.prs.3sg see.ptcp Carloi. Carlo ‘As soon as he left the house, Carlo saw her.’ Whatever the analysis of the NSs in control structures may be, the contrast in (11)–(12) looks clearly related to all the previous ones and confirms that there is enough evidence for assuming that NSs in languages like Italian and Spanish show a strong preference for antecedents that are interpreted as topics. Finally, it is also worth mentioning an interesting argument supporting the crucial role of topics, taken from Frascarelli (2007: 715) (see also JiménezFernández 2016). The argument is based on the scopal interaction between indefinite subjects and quantified phrases. Frascarelli points out that in Italian, whereas (13) is scopally ambiguous, (14) is no longer ambiguous, due to the presence of the null subject in the second clause.8 a guardia di ogni angolo. (13) Un poliziotto stava a policeman be.pst.3sg at guard of each corner ‘A policeman guarded each corner.’
8 As one of the reviewers points out, the reason why (14) is not ambiguous like (13) is probably due to general conditions on the inaccessibility of indefinites as antecedents when they are in the scope of other operators. Thus, it is not clear that the contrast actually supports Frascarelli’s analysis.
topics and the interpretation of referential null subjects
103
(14) Un poliziotto stava a guardia di ogni angolo, e a policeman be.pst.3sg at guard of each corner and ‘A (single) policeman guarded each corner and … e fumava in continuazione. smoke.pst.3sg in continuation … was smoking continuously.’ In (13), the indefinite subject un poliziotto may be assigned a wide scope reading —i.e. a specific reading—and also a narrow scope, non-specific one. The reason why the subject un poliziotto can only receive a specific, wide scope reading in (14), and no longer a narrow scope one, is that the wide scope reading is usually associated with the topical status of the DP,9 and this seems to be crucial for the licensing of the NS in the second clause; in the distributive (narrow scope) reading, the indefinite subject is not interpreted as a topic and is not an adequate antecedent for the NS, which is the reason such reading disappears in (14). Briefly, the NS in the second coordinated clause takes the preverbal subject in the first clause as its antecedent, and this forces the overt subject to be interpreted as specific, because indefinite topical DP s receive either a specific reading or a generic one—this last option being discarded here by the context. To sum up, there is evidence for a robust tendency of NS s to establish anaphoric links with topic antecedents. Such links explain how reference is assigned to the null element. It is an open issue what the best strategy is to give an account of the facts. The fundamental question is whether we should integrate the facts into the grammatical system, by means of some syntactic condition, or alternatively account for them by resorting to extra-grammatical principles. Frascarelli (2007, 2018) explicitly argues in favour of the first option. The following section is devoted to presenting and discussing her proposal.
3
A Syntactic Condition on Topics?
3.1 Frascarelli (2007, 2018): Licensing of NS s by an Aboutness Topic In Frascarelli (2007, 2018) and Frascarelli and Jiménez-Fernández (2019) the preference for topic antecedents is considered as a property of the grammar of NSs. This is, in the author’s view, the way in which grammar contributes 9 The assumption that topics receive specific interpretations is in itself incorrect, if related to dislocated topics. I cannot discuss this idea here (but see Leonetti 2014 for some basic remarks).
104
leonetti
to establishing the referent. Frascarelli adopts the classical approach to NS s as empty categories containing pronominal features (pro). The central idea is that the interpretation of a topical/referential pro always depends on a matching relation between the empty category and a specific kind of topic in the left periphery, namely the so-called ‘Aboutness-Shift Topic’ (Frascarelli and Hinterhölzl 2007). This kind of topic provides a value for pro through a feature agreement relation (Agree), a syntactic relation that links pro with the closest ‘Aboutness-Shift Topic’. The intuition behind this approach is that in tensed sentences,10 pro plays the same role as resumptive clitics do in Clitic Left Dislocation (CLD) constructions (for instance, Las brochetas, las serviremos al final, ‘The skewers, we will serve them at the end’ in Spanish). Thus, according to this view, a pro is always bound by a topic, which in turn can be empty in cases of referential continuity. In this way, Frascarelli tries to capture two crucial properties of pro: (i) the need for a discourse antecedent, and (ii) the topichood requirement that constrains the search for that antecedent. The configuration is the one in (15), where the dislocated topic is found in Shift Phrase (the position for Aboutness-Shift Topics) and from that position it binds the pro located in vP: (15) [ShiftP DPi [AgrP [vP proi [vP ] ] ] ] In this approach both the licensing of NSs and their interpretation follow from grammatical principles only; actually, the two facets of the problem are reduced to a single condition on interpretation, reproduced in (16). (16) Frascarelli’s (referential) NS interpretation Let YP be the Aboutness-Shift Topic in the local C(OMP)-domain of an occurrence of pro: then pro—sitting in edge position—obtains the grammatical specification of the features on Y through a matching (Agree) relation. In the same vein, Frascarelli (2007: 31) recasts Chomsky’s (1981) classical Avoid Pronoun Principle as a purely syntactic condition related to information structure: “Avoid strong pronoun, whenever it agrees with the local Aboutness-Shift Topic.”
10
Frascarelli and Hinterhölzl (2007) suggest that different kinds of topics (namely, Aboutness Topics, Familiar Topics and Contrastive Topics) occupy different syntactic positions at the left periphery, following a cartographic approach to sentence structure. I will not
topics and the interpretation of referential null subjects
105
For a proper understanding of Frascarelli’s hypothesis, it is crucial to consider whether the topic antecedent is itself overt or not. When it is implicit, a topic chain has to be formed, with a series of empty copies of the initial overt topic. Topic chains are needed to guarantee both the locality of the identification of pro and topic continuity in discourse. The Italian example in (17), in which Maria is intended as the antecedent of the pro found in the embedded clause, is represented by the structure in (18) (from Frascarelli 2018: 217): (17) Mariai pensa che ei vincerà la gara. Maria believe.prs.3sg that win.fut.3sg the race ‘Mary believes that she will win the race.’ (18) [ShiftP Mariak [AgrSP prok pensa [ che [FamP ⟨Mariak⟩ [AgrSP prok vincerà la gara ]]]]] The representation in (18) shows that the subject of the main clause is itself treated as a dislocated Aboutness-Shift Topic11 that serves as the antecedent of a pro sitting in the canonical subject position (i.e., the specifier of a Subject Agreement Phrase), whereas the pro in the embedded clause is linked not by the subject of the main clause, but by an empty copy of it located in the extended COMP domain of the embedded clause, namely, the position of a Familiar Topic Phrase, which forms a chain with the previous topic. I cannot dwell here on a detailed discussion of this proposal, nor can I go into the role of the Syntax-Phonology interface in the analysis. It is important, however, to take a closer look at the formation of topic chains, given that this process plays a crucial part in Frascarelli’s approach. According to her, a NS can only occur if a dislocated topic has been already introduced in the representation. That begs the question, then, of what happens in cases like the Spanish example in (19), where the subject Juan cannot be a topic but nevertheless is the antecedent of pro in the answer (Frascarelli 2007, fn. 29):
11
discuss this idea here, since it is orthogonal to the point of this paper. In what follows, then, I will merely reproduce Frascarelli’s (2007, 2018) views. In her system, all preverbal subjects in NS languages are analyzed as dislocated constituents (Frascarelli 2007: § 5). Therefore, all categorical (topic/comment) sentences would contain a dislocated subject. See Camacho (2013) for the relation between this hypothesis and the analysis of NS s.
106
leonetti
(19) A: ¿Qué hizo Juani? what do.pst.3sg Juan ‘What did Juan do?’ B: proi Habló con María. speak.pst.3sg with Maria ‘(He) spoke with María’ To account for such cases, Frascarelli is forced to assume that Juan counts in fact as a topic, because it is actually linked to an empty Aboutness-Shift Topic that occurs both in the question and in its answer. Then, it is this empty topic in the answer that binds the pro. The possibility to generate empty topics whenever they are needed without further restrictions is, of course, a very problematic solution. Frascarelli then argues that the fact that an explicit Aboutness Topic always becomes the antecedent of a NS at its right (excluding other possible candidates) indicates that the identification of pro must be, therefore, a local process: this, she claims, supports the idea of inserting empty topics to satisfy the requirement that pro is licensed by a topic. It is worth noting that Aboutness Topics cannot be iterated in the left periphery: only one of them is allowed; hence, if this topic is overt, the possibility of having an implicit topic is no longer available and so the overt one is the only one that can count. The Italian example in (20) (also from Frascarelli 2007: § 6.2) illustrates the case of a postverbal subject—clearly the focus of its clause—that is nevertheless the antecedent of pro. parlato Leoi, proi ha (20) Quando ha when have.prs.3sg speak.ptcp Leo have.prs.3sg convinto tutti. convince.ptcp all ‘When Leo spoke, he convinced everyone.’ Frascarelli does not consider this case as a counterargument to her proposal. On the contrary, she claims that postverbal subjects can be coreferential with empty local topics, which are in turn responsible for the identification of pro. A topic can be silent even in its first occurrence, as the head of a topic chain. Therefore, the representation of (20) would contain an initial empty topic in the local COMP domain, coindexed with both the postverbal subject and pro. In this way, Frascarelli argues, the licensing conditions for pro can be maintained. To motivate her claim, Frascarelli invokes again the effects of inserting an overt topic in the structure, as the extrasentential DP Marco in (21):
topics and the interpretation of referential null subjects
107
(21) Marcoj, quando ha parlato Leoi, proj/*i ha Marco when have.prs.3sg speak.ptcp Leo have.prs.3sg convinto tutti. convince.ptcp all ‘Marco, when Leo spoke, convinced everyone.’ If a new, explicit topic is introduced, it is this topic that identifies pro, not the postverbal subject: in (21) coreference is only possible with Marco (notice that the subject of the main clause in (21) has to be implicit). According to Frascarelli, the fact that coreference with another topic is excluded in (20) implies that the DP Leo must be present in the local COMP domain as a silent Aboutness-Shift Topic. Thus, an overt topic always has precedence as an antecedent and guarantees that pro is bound in its local domain; this is supposed to be evidence for the syntactic relation between the topic and the NS. Notice that this proposal leads to assuming that in (20) Leo is, at the same time, focal and coindexed with a topic, which does not sound plausible. Frascarelli’s arguments seem not convincing at this point, and sentences like (20) look like real counterexamples to her proposal (see Section 3.2.1 for discussion). One of the predictions of Frascarelli’s hypothesis should be that the subject of a thetic sentence, being non-topical, cannot be the antecedent of pro. This prediction, however, does not seem to be borne out. Consider the Spanish examples in (22) (see Rizzi 2018: 519 for a confirmation of the same fact on the basis of Italian data): saliendo el sol. Y e ya calienta. (22) a. Está be.prs.3sg go-out.ger the sun and already warm.prs.3sg ‘The sun is rising. And it already warms the air.’ b. Entra Quique en el campo. e Sustituye a Jiménez. enter.prs.3sg Quique in the field replace.prs.3sg to Jiménez ‘Quique enters the field. He is replacing Jiménez.’ The NS of the second sentence in (22a) takes the postverbal subject in the first sentence as its antecedent; however, there is no topic in the sentence (with the exception of a possible spatiotemporal or stage topic, which is not relevant to the current discussion). The same goes for (22b). As a reviewer observes, in this case Frascarelli would probably resort to the same analysis provided for (20), with a null topic. If that is the case, her strategy seems not only unconstrained, but completely unrelated to the real informational articulation of the sentence.
108
leonetti
After this short and schematic presentation of Frascarelli’s hypothesis, one question that arises is whether the evidence we have for a connection between null subjects and topics is enough to justify a purely syntactic approach. The preference of null subjects for topic antecedents could in fact be explained in an alternative way, as an effect of the need for highly accessible antecedents— topics are discourse-prominent, and thus easily accessible as antecedents—, which is in turn a consequence of the competition and the division of labour between null subjects and strong, overt pronouns. Under this alternative view, the link between null subjects and topics depends on interpretive principles that lie outside of the grammatical system. As indicated above, the issue will be discussed in Section 4. Before considering this option, I will present some arguments against an approach based on topic chains and silent topics. 3.2 Against a Syntactic Approach Frascarelli’s hypothesis represents an explicit attempt to build the preference of NSs for topical antecedents into the grammatical system, by means of a specific syntactic condition. An account along these lines requires exploiting a rich and articulated view of the left periphery in sentences and resorting to a massive display of null elements. It is in principle a legitimate strategy, but certainly not a simple and economic one as far as syntax is concerned. As already mentioned, the opposite stance would consist in avoiding any syntactic condition on topics and explaining the preference for topical antecedents by means of general principles external to the grammatical system, i.e. processing factors and pragmatic inference. Since such principles are independently justified, this second option looks simpler and less costly (see Duguine 2014: 533 for a similar argument against Frascarelli and in favour of argument ellipsis as a unified account of NSs). I believe that there are strong reasons to choose a pragmatic/processing approach to the preference of NSs for topical antecedents, and reject a syntactic analysis based on topic chains. This section is devoted to presenting a series of arguments against this kind of analysis. Before I proceed to lay out such arguments, I would like to comment on a puzzling basic feature of Frascarelli’s proposal. As the reader may have noticed, among the arguments that support the preference of NSs for topical antecedents in Section 2, only one fact is mentioned by Frascarelli: the contrast between (13) and (14), related to the interplay between NSs and quantifier scope. None of the remaining facts concerning grammatical relations, word order, and the competition with overt subject pronouns is included in her discussion, though they represent a significant body of evidence for a dominant tendency of NS s in languages like Italian and Spanish, and have been known for a long period. Thus, it seems
topics and the interpretation of referential null subjects
109
that Frascarelli builds her proposal without paying attention to what looks like the strongest and most visible kind of evidence in favour of a salient role of topicality in the interpretation of NSs. More specifically, she avoids considering the huge amount of empirical evidence collected in research on linguistic processing of NSs, in particular Carminati (2002). This seems to me surprising, because it suggests that we are dealing with two unrelated, independent problems in theoretical syntax and in processing. My view is just the opposite: the problem is one and the same, and evidence from processing and from pragmatically inspired analyses should be integrated in a unified view. In what follows, I take into account all kinds of evidence. The fundamental problem with accounting for the data presented in Section 2 by means of a syntactic constraint is that there are counterexamples for the basic generalizations that show that the preference for topical antecedents is just that, a preference, or a strong tendency, but not a rule of syntax. I assume that a true syntactic condition would not allow for violations in such a natural way. The counterexamples provide us with instances of NS s that take nontopical DPs as antecedents and are still acceptable for native speakers. I review four different cases: NSs with antecedents inside by-phrases, NS s with objects as antecedents, NSs with postverbal subjects as antecedents, and NS s with narrow scope quantified antecedents. 3.2.1 Antecedents Inside By-phrases The contrast in (1)–(2) involves by-phrases as antecedents of NS s: by-phrases are notoriously bad as antecedents. It is true that preverbal subjects make better antecedents than by-phrases, but if the context is conveniently modified, a DP inside a by-phrase can make an adequate antecedent for a NS, as shown in the following Italian examples.12 (23) A: Il guasto è stato riparato da Giannii. the damage be.prs.3sg be.ptcp repair.ptcp by Gianni A: ‘The damage was repaired by John.’ B: ei È proprio bravo, eh? be.prs.3sg really good ah? B: ‘He is really good, isn’t he?’
12
See Samek-Lodovici (1996: 37) for related examples in Italian wh-interrogatives where byphrases can be antecedents of NS s. According to this author, not all Italian speakers would fully accept examples like (23) and (24) with the intended interpretation.
110
leonetti
(24) Siamo stati convinti dalla venditricei. be.prs.1pl be.ptcp convince.ptcp by-the saleswoman ‘We have been persuaded by the saleswoman.’ ei Ci ha spiegato tutto in modo chiarissimo. loc have.prs.3sg explain.ptcp all in way very-clear ‘She explained us everything very clearly.’ In (23) and (24) the by-phrase is not topical. However, it counts as an acceptable antecedent for the NS in the following clause because an additional factor, which was absent in the examples in (1)–(2), becomes relevant: it is the coherence relation of Explanation holding between the two clauses. Whereas in (1)–(2) the coherence relation is one of Narration (see Kehler 2002, Jasinskaja and Karagjosova 2020 for an overview of the role of coherence relations in anaphora), because the two clauses are linked by a relation of temporal succession, in (23) and (24) the coherence relation is Explanation: there is no temporal succession, and the second clause is understood as an explanation of the situation described in the first one. Under these conditions, for reasons that are well beyond the scope of this paper, a NS can be anaphorically linked to a non-topical antecedent. If these remarks are on the right track, we happen to be facing a rather familiar scene in the recent literature on discourse anaphora: on the one hand, we evaluate the impact of two different kinds of factors, i.e. prominence or salience—according to which more prominent expressions, such as topics, being highly accessible, make better antecedents for NSs—versus coherence; on the other hand, we can verify how coherence is able to override prominence and favour an interpretation that contradicts the predictions based on the discourse salience of antecedents exclusively (despite being non-topical, antecedents contained in a by-phrase can be linked to NS s when a coherence relation like Explanation is established). The interplay of prominence and coherence is not an idiosyncratic feature of the search for adequate antecedents with NSs. It is, rather, a pervasive property that has been extensively investigated in the literature on discourse anaphora in several languages (Kehler 2002, Kehler et al. 2008). Besides providing us with a reasonable account for the unexpected facts observed in (23) and (24), these considerations lead us towards certain interesting consequences: if the topical/nontopical nature of a phrase, as a factor determining the retrieval of an antecedent for an anaphoric expression, can be overridden by the need to establish coherence relations, then the preference for topics cannot be a principle of core grammar. It is softer than true syntactic constraints, and must originate from some motivation that is external to the grammatical system.
topics and the interpretation of referential null subjects
111
3.2.2 Objects as Antecedents The same conclusion applies to other basic data that apparently support the role of topic antecedents, such as the general preference of NS s for topical subjects, against objects, as antecedents (see the contrasts in (4)–(5) and (6)– (7)). Here, again, it is possible to find counterexamples. Calabrese (1986: 33) himself notices that in contexts like the one in (25), from Rizzi (2018: 516), the NS in the subordinate clause can take both the subject and the object in the main clause as antecedents, and tries to introduce some additional notion to account for this kind of exceptions; Rizzi (2018) resorts to c-command, more precisely to the basic assumption that a subject pronoun is expected to have the referent of a c-commanding DP: as both Francesca and Maria in (25) ccommand the NS, it can be linked to any of these DP s (intuitively, this means that NSs can be tied to any prominent antecedent, but not necessarily to a topic). (25) Francescai ha fatto notare a Mariaj che proi,j Francesca have.prs.3sg make.ptcp realize.inf to Maria that era molto stanca. be.pst.3sg very tired ‘Francesca made Maria realize that she was very tired.’ Why is the subject/object asymmetry absent in (25), whereas it holds in the case of (6)–(7)? In (6)–(7) (Quando Mario ha picchiato Antonio, pro/lui era ubriaco), the possible antecedents are inside a subordinate adverbial clause and do not c-command the NS, and Calabrese’s principle (8) is in force, thus forcing the anaphoric link with the topical subject and blocking the alternative link with the object. In (25) the two DPs c-command the NS, as already mentioned, and principle (8) seems to be inactive (Rizzi 2018: 516). This solves the problem of the contrast between (25) and (6). Notice that the new landscape emerging does not support a view of the preference for topical antecedents as a grammatical constraint. On the one hand, principle (8), which is a plausible formulation of such preference, must be supplemented by a complementary principle that introduces a new factor, syntactic prominence (realised as c-command); moreover, principle (8) is only relevant when there are no ccommanding antecedents, which heavily reduces its range. As a result, it is no longer the central principle governing the interpretation of NS s; there is no evidence in favour of Frascarelli’s version of the principle. On the other hand, even considering only contexts where principle (8) is in force—for instance, anaphoric links between independent sentences—, it seems that under the appropriate circumstances NSs may choose objects, instead of preverbal sub-
112
leonetti
jects, as antecedents: this is what happens in (26), where a coherence relation of Explanation forces an interpretation of the NS as referring to Arturo. solo Arturoi. ei è il suo migliore (26) Carlo ascolta Carlo listen.prs.3sg only Arturo be.prs.3sg the his best amico. friend ‘Carlo listens to Arturo only. He is his best friend.’ 3.2.3 Postverbal Subjects as Antecedents After reviewing by-phrases and objects as possible antecedents for NS s, now I turn to the case of postverbal subjects. The contrast in (9)–(10) suggests that postverbal subjects, being focal, cannot be proper antecedents of NS s. However, this should exclude an example like (20), which is perfectly acceptable (see Calabrese 1986: fn. 3 and Rizzi 2018: 518 for postverbal subjects in different contexts). Examples (10) and (20), repeated here as (27) and (28), differ in the position of antecedent and NS in the main clause or the subordinate clause, which may have effects on c-command, and in the relative order of the two subjects (in fact, (27) is a case of cataphora or backwards anaphora, but (28) is not). Such factors interact with topic and focus in complex ways. (27) Dopo che e*i è arrivato, ha parlato after that be.prs.3sg arrive.ptcp have.prs.3sg speak.ptcp Giannii. Gianni ‘After he arrived, Gianni spoke.’ parlato Leoi, ei ha (28) Quando ha when have.prs.3sg speak.ptcp Leo have.prs.3sg convinto tutti convince.ptcp all ‘When Leo spoke, he convinced everyone.’ I believe that it is the well-known incompatibility between cataphora and focal antecedents, rather than a specific constraint on the interpretation of NS s, that rules out coreference in (27), together with the parallel examples with gerunds and absolute constructions in (11) and (12). The constraint affects cataphora when a pronominal and an R-expression (for instance, a name) are involved, and the R-expression is in focus. Reinhart (1986: 138–140) gives the following formulation (see also Erteschik-Shir 1997: 78, and Bianchi 2009 for a general discussion):
topics and the interpretation of referential null subjects
113
(29) Topic-antecedent hypothesis for backward anaphora Backward anaphora is possible only if the antecedent is in sentence-topic position. The impossibility of coreference with postverbal subjects in (10), (11) and (12) is, thus, explained by the constraint in (29), which concerns a particular relation between pronouns and antecedents, but is not specifically related to null pronouns. Given this, plus the well-formedness of (20)/(28) under the relevant coindexation, the conclusion is that NSs are perfectly compatible with focal antecedents, at least when such antecedents are clearly salient and there are no competing topical candidates (see also Lubbers Quesada and Blackwell 2009, and Clements and Domínguez 2017 for similar observations). As shown by the counterexamples discussed in this section, there may be a preference of NS s for topical antecedents in a set of contexts, but it is not operating in all contexts, and cannot be taken as a principle of core grammar. It is difficult to see how the data can be accommodated in Frascarelli’s model without resorting to an unconstrained proliferation of null topics in the left periphery. 3.2.4 Antecedents with Narrow Scope Finally, something must be said on Frascarelli’s argument based on wide and narrow scope readings of indefinite subjects and their relation to NS s in (13)– (14). It is not impossible to have NSs linked to indefinite DP s with narrow scope. In the Italian example in (30), adapted from Frascarelli (2007), the NS in the second sentence can take the indefinite subject in the first sentence— qualche studente ‘some student’—as its antecedent, in spite of its non-specific interpretation (I assume that the indefinite subject is interpreted with narrow scope with respect to the modal element parere ‘seem’). (30) Pare che qualche studente abbia archiviato seem.prs.3sg that some student have.prs.sbjv.3sg file.ptcp quel libro. that book ‘Apparently some student filed that book.’ e L’ avrà sicuramente fatto con l’ aiuto della it have.fut.3sg surely do.ptcp with the help of-the bibliotecaria. librarian ‘(S)he surely did it with the help of the librarian.’
114
leonetti
The non-specific indefinite subject is a non-topical antecedent, and still the link with the ensuing NS holds, against the predictions in Frascarelli (2007). It seems that maintaining in the second sentence the same modal context created in the first one by the verb parere ‘seem’ is the crucial factor for the link to survive (see the notion of ‘modal subordination’ in Roberts 1989). The Spanish example in (31) shows the role of modal subordination: the modal context created by the future tense in the first sentence extends to the second sentence and enables the anaphoric connection between the NS and the preceding indefinite subject un agente, despite its non-specific interpretation (JiménezFernández 2016 reaches similar conclusions on Spanish NS s). (31) Un agente recogerá a cada testigo, an agent pick-up.fut.3sg to each witness ‘An agent will pick up each witness … y e lo acompañará al aeropuerto. and him accompany.fut.3sg to-the airport … and take him to the airport.’ The interpretation of NSs in (30) and (31) does not support Frascarelli’s conclusions: it provides new evidence that NSs do not necessarily require topical antecedents. Although the arguments gathered in Section 2 are powerful enough to suggest that NSs clearly tend to choose topics as antecedents, this tendency cannot be modelled as a ‘hard’ grammatical principle. Despite the inadequacies mentioned so far, Frascarelli’s hypothesis has had a remarkable impact on the field. Quite often her views have been simply taken for granted and then integrated into major theories on NS s or applied to new data, with the risk of making established truth out of a series of controversial assumptions. Her contribution must be placed in a wider context that includes related proposals such as the ones in Sigurðsson (2011) and Camacho (2013); a common feature of all these proposals is the formulation of generalizations on discourse phenomena intended as grammatical principles that involve the left periphery in sentential structure. This is not new in the formal tradition, and it is basically misguided, in my opinion. Also Calabrese’s (1986) rules for the interpretation of NSs and pronouns are actually discourse/pragmatic generalizations presented as grammatical principles: they deal with discourse expectations, parallelism and competition between elements displayed along a scale—factors that should be best treated as components of inferential processes that contribute to specifying the propositional content of an utterance.
topics and the interpretation of referential null subjects
115
If, in fact, the alleged grammatical principles suggested in the reviewed literature have to be discarded in favor of extra-grammatical processes—which seems to be more in consonance with the current minimalist framework in generative grammar—, other proposals should be reconsidered too. One significant case is the way Frascarelli’s hypothesis is exploited in Holmberg (2010) to solve a technical problem with the features of finite inflection in NSL s. Also interesting is the role attributed to topics as licensers of NS s in Old French (Ingham 2018), Old English (van Gelderen 2013) and Arabic (Shormani 2017). However, I cannot discuss these extensions here. I will rather concentrate on some general issues in the next section, with the aim of looking back at the central research question I mentioned in the introduction: how grammar contributes to the interpretation of NSs.
4
The Interpretation of Referential Null Subjects: How Much Depends on Grammar?
4.1 Conceptual Problems with the Notion of Topic Before addressing the consequences of the previous discussion for the theoretical debate on the interplay between grammar and pragmatics, in this section I would like to dwell on some conceptual problems related to the use of the notion topic in Frascarelli (2007, 2018). I intend to raise the following points: 1. Two different types of topic must be distinguished, which in this hypothesis are erroneously conflated into a single syntactic device. The idea of having silent copies of left-dislocated phrases is misguided. 2. The management of discourse referents, which is not a strictly syntactic operation, is represented as an effect of left dislocation. 3. The strategy of resorting to null copies of dislocated topics is too powerful, and seems to be unconstrained. 4. Accounting for the interpretation of NSs by means of topic chains is merely stipulative, and blurs the real motivation that lies beyond the frequently noted connection between NS s and topics. Combined with the empirical problems reviewed in the previous section, these conceptual flaws should lead us, in my opinion, to discard an approach based on topic chains. 4.1.1 Topic ≠ Left Dislocation It is probably an effect of viewing the syntax-discourse relation through a cartographic lens—although by no means a logical consequence of it—that sentence topics and discourse topics are systematically represented as instances of (Clitic) Left Dislocation in Frascarelli (2007, 2018). Left Dislocation is cer-
116
leonetti
tainly the most salient grammatical device for topic marking in Romance NSL s. However, not all sentence topics—in particular, preverbal subjects—are to be equated with left dislocated phrases. This has been a debated issue in the literature on preverbal subjects in Romance in the last three decades (see discussion in Sheehan 2006, Villa-García 2012, Rizzi 2018 a.o.). I think that there is a general consensus nowadays in avoiding an identification of sentence topics with dislocated topics in Romance: sentence topics may appear as dislocated expressions, but they may also appear inside IP/TP, either as preverbal subjects or as other preverbal constituents (for instance, fronted indirect objects in constructions like Ad Andrea interessa l’iconografia ‘Andrea is interested in iconography’, adapted from Calabrese 1986). In NSL s, preverbal subjects are predominantly interpreted as topics in categorical predications, and this does not imply that they are syntactically dislocated; a topic interpretation may be imposed by dislocation, but is also typically associated with the preverbal subject position. There are, then, two ways to express the topic. Against this line of thought, Frascarelli assumes that Romance preverbal subjects are syntactically dislocated. This assumption, subjected to severe criticism by Sheehan (2006), Villa-García (2012) and other authors, leads to the unwelcome result of identifying the notions of sentence topic and left-dislocated topic, thus taking a strategy for the expression of marked topics (left dislocation) as the default way of expressing unmarked topics. As will become clearer below, the distinction between marked and unmarked topics is needed for descriptive and theoretical reasons. The basic idea, along the lines of Dobrovie-Sorin (2000), Brunetti (2009) and Leonetti (2013), is that unmarked topics are integrated in the sentential domain—let’s say, in Spec, IP/TP—and their association with topic interpretation is only a default condition: I assume that there is no [aboutness] feature encoded in the preverbal position, and topic interpretation is, rather, pragmatically inferred on the basis of information related to different factors (thematic prominence, first of all, but also lexical aspect and definiteness). The phrase that is the best candidate for aboutness topic raises to the preverbal slot and is interpreted as topic, except when a thetic, all-focus reading is favoured. Unmarked topics are unmarked in the sense of representing the participant that under normal conditions is the most salient candidate to count as an address for information update. In case the speaker wants to mark as topic a phrase that is not the best candidate—i.e. a phrase that would not be taken as topic otherwise—, the unmarked strategy is not available and he has to resort to a marked one, which is Left Dislocation. In a few words, dislocation is selected when the involved phrase is not a natural topic by itself or when a special effect is looked for that cannot be obtained through the unmarked strategy. As special tools are used for special purposes, marked, detached topics may contribute an
topics and the interpretation of referential null subjects
117
additional interpretive flavour, contrastiveness, which can hardly be conveyed with unmarked topics; moreover, and crucially for our purposes here, marked topics are best suited for topic shifting in discourse (Leonetti 2013, 2014, a.o.), whereas unmarked topics are the ideal device for establishing topic continuity. The result is a balanced division of labour between unmarked and marked strategies. Most Romance languages exploit this system. Now, in Frascarelli’s model the basic distinction is lost, and a single notion of topic, the one corresponding to marked topics (Left Dislocation), is used to account for the properties of NSs. This yields a distorted view of the facts: the formal device responsible for introducing marked topics is presented as the licensing factor of the prototypical expression of unmarked topics. If dislocation is associated to topic shifting and NSs are associated to topic continuity, how can dislocation, with the corresponding ‘topic chains’, be presented as the central mechanism for maintaining the same topic? The problem is that topic continuity is typically obtained by means of NSs, and nothing is gained by relating NSs to dislocation. The source of this misconception is the identification of topic with dislocated topic. Finally, it is worth recalling that all the evidence collected in Section 2 to illustrate the preference for topical antecedents concerns unmarked topics, and not left dislocation. In my opinion, the most serious weakness in the topic chains hypothesis is the very idea of having silent copies of left dislocated topics: if the distinction between unmarked and marked topics is assumed, the proposal is quite difficult to accept. Left Dislocation is a marked strategy for introducing topics: as such, it is typically associated with topic shift and contrast—notice that these are just the interpretive values that a NS cannot express. Thus, if this is a correct characterization, there should not be null dislocated topics. Dislocation should always be overt. A marked construction is there to convey marked readings, and it could not carry out its tasks if it were null. On the contrary, unmarked topics can be null. In fact, NSs are a natural manifestation of unmarked topics. Then, they cannot be licensed by dislocation, if they have different properties from dislocated phrases. Topic continuity cannot be made of chains of dislocations. The problem is, once again, the wrong identification of topic and left dislocation. 4.1.2 Discourse Referents In a topic chains account, it is necessary to assume that silent topics are merged to build topic chains and ensure topic continuity across sentences. When a null subject is linked to an antecedent occurring in a focal position (for instance, in Vorrei presentarti Leoi. proi è il mio migliore amico ‘I’d like to introduce Leo
118
leonetti
to you. He is my best friend’, from Frascarelli 2018: 221), a silent topic must be posited in the COMP domain in the second clause to identify pro. Now, it is true that in the example the DP Leo introduces a discourse referent that can become the topic of a subsequent predication, but this does not imply that a dislocated topic has to be merged in the syntax; it rather shows that null subjects may be linked to non-topical antecedents, if the antecedents are salient enough. At the source of the topic chains approach lies a wrong strategy that has serious consequences: it treats a non-syntactic process, the introduction of discourse referents in the semantic representation, as if it were a syntactic operation, merging of left-dislocated topics. The two notions cannot be equated. The management of referents in discourse representation and their contribution to discourse coherence have been thoroughly analysed in Centering Theory (Walker, Joshi and Prince (eds.) 1998) and other recent models; in these theories a processing perspective is taken, and the rules for interpretation are not supposed to be a part of core syntax. Following this line of thought, I assume that the introduction and activation of discourse referents is governed by principles external to syntax. 4.1.3 Null Topics So far it has been made clear that the idea of null topic is an essential component in Frascarelli’s system. The problem is that there is no clear evidence in favour of such silent topics. Their alleged effects can always be explained in some alternative way. Null topics are certainly not required when the apparent antecedents of NSs are in focus, as in (20), (23), (24) and (26): it is enough to abandon the rigid assumption that antecedents must be topics and choose the more flexible and perfectly plausible idea that NS s simply require the most salient, accessible antecedents in the context. In many cases this will in fact single out topical antecedents, but in contexts where no sentence topic is particularly salient, a discourse referent just introduced by a focal DP will make an optimal antecedent for a NS. This is what happens in the examples just mentioned.13 One could think that null topics are still required to ensure that the antecedent–NS link obeys locality in some sense, but, again,
13
The effect of adding an explicit topic in example (21), reproduced here, deserves some attention. The NS has to be coreferent with the overt topic, and cannot be coreferent with the postverbal subject Leo. This is due to the need to connect the topic Marco with some argumental slot in the adjacent sentence: the subject slot in the main clause is the obvious solution. Otherwise, the dislocated topic would be cut off from the predication, as if it were a hanging topic, which is impossible in (i). The requirement of a link for the external topic overrides the possibility of establishing a link with the postverbal subject. Nothing
topics and the interpretation of referential null subjects
119
there is no strong evidence showing that the anaphoric connection has to obey locality in a strict, syntactic sense (see Frascarelli 2018 for some discussion on the role of locality at the interface): the most obvious obstacle is the fact that the connection survives in discourse contexts, among independent sentences. The dubious motivation for null topics is just one side of the problem. The other side is that postulating the existence of topic chains in the syntax is an unconstrained move, apparently ad hoc. There are no obvious answers for important questions like the following ones: under what conditions is it possible to establish silent copies of aboutness topics?14 How is the link between overt referential expressions in discourse and topic chains constrained? How are interpretations established when there is more than one silent topic, or more than a potential antecedent? Introducing a powerful hypothesis without specifying appropriate restrictions on its application results in an undesirable strategy. 4.1.4 A Stipulative Proposal A syntactic approach to the preference for topic antecedents is merely stipulative. Frascarelli argues that in consistent NS languages the interpretation of NSs depends on a matching relation with a specific type of topic. This leaves some questions unanswered: why should there be any relation between NS s and topics? Why can’t NSs be licensed in some other way, maybe unrelated to information structure? Why do certain languages display left dislocations but lack NSs (for instance, French, among Romance languages)? If the proposal were empirically sound, one could still accept it and expect that future research will provide at least partial answers to our questions. But, as the proposal does not cover the data adequately and suffers from conceptual inadequacies, it seems clearly preferable to look for an alternative hypothesis that is able to offer a motivated explanation for the facts. A pragmatic account of the interpretation of NS s can provide this kind of motivated explanation, in a very simple way. In particular, it offers a natural answer for the first question in terms of accessibility (Ariel 1990, Gundel, Hedberg and Zacharski 1993). Briefly, for obvious reasons, NS s are the least
14
here supports the assumption that NS s obligatorily must be licensed by dislocated topics. (i) Marcoj, quando ha parlato Leoi, proj/*i ha convinto tutti. ‘Marco, when Leo spoke, convinced everyone.’ Recall that overt dislocated topics can usually be iterated in Romance: this is one of their defining properties. Thus, it is legitimate to wonder whether null copies of topics can be freely iterated too.
120
leonetti
specified elements in the toolbox for (pro)nominal reference in a NSL. Being devoid of conceptual content, they are unable to access discourse referents whose representation is not already activated in the working memory of the addressee: actually, they can only be used to retrieve given, activated referents that are in focus of attention for the addressee, i.e. maximally salient/accessible antecedents. This arises from the competition between null elements— ellipsis—and other pronominal and lexical DP s that can be located across a range of positions on an Accessibility Scale or a Givenness Hierarchy, and it is the reason why NSs behave like clitics and unstressed pronouns in many respects. The preference of NSs for topical antecedents is a natural consequence of their limitations in retrieving given discourse referents: topics make particularly salient antecedents, and thus, they are chosen instead of other competing candidates, especially when different options are contextually available. This is enough to give a response to the question about NS s and topics, and is at the same time compatible with two issues raised in the previous discussion: one is the role of syntactic prominence as c-command—syntactically prominent antecedents are also cognitively more prominent and thus favoured against other competitors—, and the other is the possibility to link NS s with focal antecedents—if they are accessible because they have been recently mentioned and do not compete with higher rated candidates, the resulting reading should be acceptable. In an approach along these lines, everything is derived from the informational status of NSs with respect to other DP s, without unwarranted stipulations. There seems to be no room for considerations of this kind in Frascarelli’s proposal. 4.2 Why a Pragmatic Approach Is To Be Preferred Once we assume that the problem of interpretation of NS s lies partially outside the limits of grammatical theory, the main question raised in the introduction —how much of the interpretation of NSs depends on the grammar, and how it contributes to such interpretation—acquires a central status. It is initially plausible to think that the grammar contributes just some constraint, i.e. some way to restrict the search for an antecedent by placing a condition on it, instead of fully specifying the final interpretation. It is quite obvious that in consistent NSLs the basic condition imposed by the grammar is the matching relation between the antecedent and the person and number features of verbal inflection; I will have nothing to add on this. Another indirect contribution of the grammar is provided by the system of DP forms that alternate with NS s, that is, the internal articulation of the accessibility scale in each language; notice that the scale does not specify any inherent property of null pronouns: it just shows the relative order of the forms as for accessibility marking. This is surely
topics and the interpretation of referential null subjects
121
relevant, and I will take for granted that the articulation of the scale must be carefully considered, but here I want to focus on specific constraints on the interpretation of NSs. There are two possible candidates for these constraints: one is a condition that states that there must be a topic antecedent, and the other one is a condition that states that the antecedent must be uniquely identifiable, i.e. that the interpretation must be definite. The first option has been discussed at length in Sections 2 and 3 and has been discarded. The grammar does not need to include a condition on topical antecedents, neither along the lines of Frascarelli (2007) nor in a different version. The second option consists in positing a [+definite] feature either in the NS (for instance, in a null pronoun) or in the element that licenses it (verbal inflection), so that definiteness can guide the search for an antecedent. It is reasonable to think of definiteness as a basic condition for interpretation, just because referential/thematic NSs typically display definite readings. I will briefly discuss this point in Section 4.3. Now I would like to go back to the data in Sections 2 and 3 to draw some conclusions on how to account for them, given that postulating a condition on topics does not seem adequate. I will argue for a pragmatic approach. The first fact, or set of facts, that one needs to explain is that NS s show a true preference for topical antecedents, especially when more than one potential antecedent is available, and this is supported by asymmetries between preverbal and postverbal subjects, and by asymmetries between preverbal subjects and complements. The second fact is that this preference is not always in force: NSs may also be linked to non-topical antecedents, and in some cases this is due to the pressure imposed by discourse coherence. The third fact is the relevance of syntactic prominence (c-command) for determining the adequate antecedents. As already shown in Section 4.1.4, a pragmatic account inspired by accessibility theories is able to cover all these empirical requirements: the preference for topical antecedents is nothing more than an effect of the need for accessible antecedents that happens to be particularly noticeable in a series of environments. This implies that the grammar does not impose any condition on topics, as argued above. I am aware that I am not offering a fully worked out theory based on accessibility and pragmatic inference, but that is not among the goals of this paper. The crucial point is that accessibility (together with related notions like discourse prominence or salience; see Jasinskaja et al. 2015, von Heusinger and Schumacher 2019 for an overview of the role of topicality with respect to the prominence structure of discourse) is the key concept.
122
leonetti
A pragmatic account is preferable for a number of reasons. Here are some of them: (a) Pragmatic processes are more or less successful depending on the assumptions that the context makes available. This implies that certain inferences may lead to the optimal interpretation in a particular context, but may be overridden by the salience of some different assumptions in another context. Pragmatic inference is known to be cancellable. The flexibility and the context-dependence typical of pragmatic processes is precisely what we observe in the counterexamples to the original generalization on topics (Section 3.2.1). (b) As noted before, the hypotheses that attempt to build a condition on topical antecedents in the syntax do it in a stipulative fashion, without actually explaining why such condition should exist. The pragmatic perspective, on the contrary, offers a natural motivation for the tendencies observed in the interpretation of NSs. (c) A syntactic representation of the link between NS s and antecedents, such as the one provided in Frascarelli (2007), requires some ad-hoc machinery: null dislocated topics are the clearest example. A pragmatic account does not impose any complication to the syntax, and it is thus preferable for its simplicity. Moreover, it exploits only general principles that are independently justified, and it is the optimal solution on economy grounds. In a model like Duguine’s (2013, 2014), for instance, if ellipsis of nominal arguments is possible in a language, all is needed is that pragmatic inference solves the reference of the elided expression on the basis of the same factors that determine the resolution of other types of ellipsis; the grammar is no longer involved in this task. (d) One interesting advantage of pragmatic accounts is that they allow us to connect the data with the results of recent research on processing of null subjects (Arnold 2010, Filiaci 2011, Di Eugenio 1998, Taboada 2008), and, from a general perspective, to integrate the facts into the wider panorama of psycholinguistic research on the resolution of anaphora, including also overt pronouns. This connection is usually neglected or obscured in syntactic accounts, which is an undesirable effect. (e) The relation of NSs with overt pronouns is usually modelled on the basis of the Avoid Pronoun Principle. Such principle states that null pronouns are preferred over overt pronouns, unless certain special conditions are given. It highlights the effects of the competition between the two kinds of pronominal elements. I contend that every aspect of interpretation derived from the competition between options inside a scale or a paradigm is pragmatically inferred and does not need to be stipu-
topics and the interpretation of referential null subjects
123
lated in the grammatical system. The consequence is that a pragmatic perspective is best suited to account for the relation between NS s and overt pronouns, and, more generally, for an adequate view of pronominal reference in discourse anaphora. Both for null and overt pronouns, the key factor is the interplay of linguistic prominence (accessibility) and discourse coherence. According to this, the Avoid Pronoun Principle is pragmatic and external to grammar. This view of the problem of interpretation of NS s is in consonance with the general hypothesis defended in Duguine (2013, 2014): NS s are best treated a case of argument ellipsis, both in consistent NSL s and in radical NSL s like Chinese. 4.3 Definiteness Let’s suppose that NSs in consistent NSLs are always assigned definite interpretations, as noted above.15 This could be explained by postulating a [+definite] feature encoded in agreement morphology; according to this proposal, the grammar would constrain the search for an antecedent by means of the condition imposed by definiteness—unique identifiability. However, I believe that definiteness as an encoded feature can be dispensed with (against Holmberg 2010 and many others). To ensure that NS s will get definite readings, it is enough to know that they cannot be interpreted if an antecedent is not retrieved. The search for an antecedent guarantees that (at least) one adequate antecedent must be selected, and this requirement equals the effects of encoded definiteness: the referent has to be given information, and has to be properly identifiable. Nothing else is needed. Definite readings are obtained without NSs including any definiteness feature. My conclusion is that, if the grammar neither imposes any condition on the topical status of antecedents nor imposes any condition like definiteness, there is no way in which the grammar constrains the interpretation of NS s (apart from subject agreement). There is argument ellipsis—if the relevant conditions are met—, and the resolution of ellipsis depends on processes which are external to the grammatical system.
15
I will stick to this usual assumption, though it seems to me that certain indefinite interpretations are possible, at least in Spanish, when NS s have bare nominals as antecedents. If this is correct, it could favour a pragmatic approach to the prevailing definiteness of NSs. I leave this issue for future research.
124 5
leonetti
Conclusions
The main issue addressed in this paper is the problem of the interpretation of referential NSs in languages like Spanish and Italian. Interpreting NS s means determining their reference. Assuming that a number of factors, both grammatical and non-grammatical, are involved in the determination of reference, two research questions come to the forefront: 1. How is the division of labour between grammar and pragmatics established in the interpretation of NS s? 2. What is the status of topicality? Here I focussed on the second question, and in particular on a specific proposal about the role of topicality: Frascarelli’s claim that Romance NSs are licensed through an Agree relation with an AboutnessShift Topic in the sentential left periphery. In Frascarelli’s model, the grammar includes the condition that NSs must be linked to an antecedent that occurs as a dislocated topic; thus, the role of topicality in the licensing and interpretation of NSs is set by the grammatical system, and the role of pragmatics in interpretation is not specified. I have tried to argue against this approach along the following lines. First of all, in Section 2 I reviewed significant evidence from the literature on Spanish and Italian supporting the idea that NS s tend to choose topical antecedents. Most of the evidence comes from asymmetries between two sets of antecedents: preverbal subjects, on one hand, and postverbal subjects and complements, on the other. The data show that NS s are preferably linked to preverbal subjects, and it is reasonable to think that this is due to the default topical status of preverbal subjects in NSLs. Once the preference for topical antecedents has been introduced, the relevant question is how to account for it. In Section 3, I presented an overview of the ideas in Frascarelli (2007, 2018) followed by some critical observations: I tried to show that the mentioned preference for topical antecedents is not always in force and cannot be integrated into the grammatical system as a syntactic principle. Finally, Section 4 deals with some of the consequences of my critical review of the topic chains approach. It starts with additional critical remarks on the way the notion of topic is used in this proposal and, in particular, on the idea of null copies of dislocated topics. I conclude that the role of topicality in the interpretation of NSs should not be captured by means of a syntactic constraint, in particular one based on Clitic Left Dislocation. The problem of licensing of NSs and the problem of their interpretation must not be conflated: the grammar should explain how NSs are licensed in a language, but not how their referents/antecedents are chosen and identified, because this does not pertain to the domain of grammar. Another obvious consequence is that the
topics and the interpretation of referential null subjects
125
preference for topical antecedents, together with all other aspects of the identification of referents, must be explained from a pragmatic perspective; I suggest that this can be done relying on available theories of discourse anaphora based on accessibility, and that this view of the problem has a number of advantages, among them the possibility of finding a natural motivation for the way speakers use and interpret NSs. As for the question concerning the division of labour between grammar and pragmatics, my conclusion is that the grammar contributes only minimally to the interpretation of NSs: in languages like Spanish and Italian, its contribution is limited to the features of subject agreement. There is no constraint due to a [definite] feature, and there is no condition on the topical status of antecedents. From a more general viewpoint, and transcending the particular issue of how NSs are linked to their antecedents, the ideas in this paper could be considered as a small contribution to reinforcing the claim that (pro)nominal reference is only minimally determined by syntax (along the lines of previous proposals on the role of pragmatic principles in anaphora resolution, like Reinhart 1986).
Acknowledgements The investigation presented in this paper is included in the research projects “The Semantics-Pragmatics Interface and the Resolution of Interpretive Mismatches” (SPIRIM), funded by the Spanish Ministerio de Economía y Competitividad (FFI2015–63497-P) and “Evidencialidad, subjectivización y perspectivización en las interfaces de la lengua” (PID2019–104405GB-100). Previous versions were presented at the workshop The Grammar of Reference and Quantification: Functions, Variation and Change (RED 2017) (Università di Bologna, June 2017), at the International Workshop on the Interface of Information Structure and Argument Structure (Universidad de Sevilla, October 2017), and at the workshop on Cartography and Explanatory Adequacy (Universitat Autònoma de Barcelona, May 2018). I am grateful to the organizers of RED 2017 in Bologna and to the audiences for stimulating discussion. Thanks are due also to Chiara Gianollo, Klaus von Heusinger, Elisabeth Stark, Cecilia Poletto, Julio VillaGarcía and Victoria Escandell-Vidal, who provided me with invaluable feedback, and to Aoife Ahern for checking my English.
126
leonetti
References Alonso-Ovalle, Luis, Susana Fernández-Solera, Lyn Frazier & Charles Clifton, Jr. 2002. Null vs overt pronouns and the topic-focus articulation in Spanish. Rivista di Linguistica 14(2). 151–169. Ariel, Mira. 1990. Accessing noun phrase antecedents. London: Routledge. Ariel, Mira. 2008. Pragmatics and grammar. Cambridge: Cambridge University Press. Arnold, Jennifer E. 2010. How speakers refer: The role of accessibility. Language and Linguistic Compass 4. 187–203. Bianchi, Valentina. 2009. A note on backward anaphora. Rivista di Grammatica Generativa 34. 3–34. Biberauer, Theresa, Anders Holmberg, Ian Roberts & Michelle Sheehan (eds.). 2010. Parametric variation. Null subjects in Minimalist Theory. Cambridge: Cambridge University Press. Brunetti, Lisa. 2009. On links and tails in Italian. Lingua 119. 756–781. Calabrese, Andrea. 1986. Some properties of the Italian pronominal system: An analysis based on the notion of thema as subject of predication. In Harro Stammerjohan (ed.), Tema—Rema in italiano, 25–36. Tübingen: Gunter Narr Verlag. Calabrese, Andrea. 1992. Some remarks on focus and logical structures in Italian. Harvard Working Papers in Linguistics 1. 19–27. Camacho, José. 2011. Chinese-type pro in a Romance-type null-subject language. Lingua 121(6). 987–1008. Camacho, José. 2013. Null subjects. Cambridge: Cambridge University Press. Carminati, Maria N. 2002. The processing of Italian subject pronouns. Amherst, MA: University of Massachusetts PhD dissertation. Chomsky, Noam. 1981. Lectures on Government and Binding. Dordrecht: Foris. Clements, Maria & Laura Domínguez. 2017. Reexamining the acquisition of null subject pronouns in a second language: Focus on referential and pragmatic constraints. Linguistic Approaches to Bilingualism 7(1). 33–62. Cordin, Patrizia. 1988. I pronomi personali. In Lorenzo Renzi, Giampaolo Salvi & Anna Cardinaletti (eds.), Grande grammatica italiana di consultazione, vol. 1, 535–549. Bologna: Il Mulino. Di Eugenio, Barbara. 1998. Centering in Italian. In Ellen F. Prince, Aravind K. Joshi & Marilyn A. Walker (eds.), Centering Theory in discourse, 115–138. Oxford: Oxford University Press. Dobrovie-Sorin, Carmen. 2000. Le(s) thème(s) entre la syntaxe et la structure de l’information. In Claude Guimier (ed.), La thématisation dans les langues, 169–183. Bern: Peter Lang. Dufter, Andreas. 2011. Sujetos pronominales y estructura informativa en español tardomedieval. In Andreas Dufter & Daniel Jacob (eds.), Syntaxe, structure information-
topics and the interpretation of referential null subjects
127
nelle et organisation du discours dans les langues romanes, 49–78. Frankfurt: Peter Lang. Duguine, Maia. 2013. Null arguments and cross-linguistic variation. A minimalist analysis of pro-drop. University of the Basque Country and Université de Nantes PhD dissertation. Duguine, Maria. 2014. Argument ellipsis: A unitary approach to pro-drop. The Linguistic Review 31(3–4). 515–549. Erteschik-Shir, Nomi. 1997. The dynamics of focus structure. Cambridge: Cambridge University Press. Filiaci, Francesca. 2011. Anaphoric preferences for null and overt subjects in Italian and Spanish: A cross-linguistic comparison. Edinburgh: University of Edinburgh PhD dissertation. Filiaci, Francesca, Antonella Sorace & Manuel Carreiras. 2013. Anaphoric biases of null and overt subjects in Italian and Spanish: A cross-linguistic comparison. Language and Cognitive Processes 29. 825–843. Frascarelli, Mara. 2007. Subjects, topics, and the interpretation of referential pro. An interface approach to the linking of (null) pronouns. Natural Language and Linguistic Theory 25. 691–734. Frascarelli, Mara. 2018. The interpretation of pro in consistent and partial NS languages: A comparative interface analysis. In Federica Cognola & Jan Casalicchio (eds.), Null subjects in generative grammar, vol. 1, 211–239. Oxford: Oxford University Press. Frascarelli, Mara & Roland Hinterhölzl. 2007. Types of topics in German and Italian. In Susanne Winkler & Kerstin Schwabe (eds.), On information structure, meaning and form, 87–116. Amsterdam: Benjamins. Frascarelli, Mara & Ángel Jiménez-Fernández. 2019. Understanding partiality in prodrop languages: An information-structure approach. Syntax 22(2–3). 162–198. van Gelderen, Elly. 2013. Null subjects in Old English. Linguistic Inquiry 44(2). 271–285. Gelormini-Lezama, Carlos & Armit Almor. 2011. Repeated names, overt pronouns, and null pronouns in Spanish. Language and Cognitive Processes 26(3). 437–454. Givón, Talmy. 1983. Topic continuity in discourse: An introduction. In Talmy Givón (ed.), Topic continuity in discourse: A quantitative cross-language study, 1–42. Amsterdam: Benjamins. Givón, Talmy. 2017. The story of zero. Amsterdam: Benjamins. Godoy, Mahayana C., Janaina Weissheimer & Matheus Araújo Mafra. 2018. When grammar meets pragmatics: Subject preference and coherence relations in Brazilian Portuguese pronoun interpretation. Journal of Portuguese Linguistics 17(6). 1–22. Grimshaw, Jane & Vieri Samek-Lodovici. 1998. Optimal subjects and subject universals. In Pilar Barbosa, Danny Fox, Paul Hagstrom, Martha McGinnis & David Pesetsky (eds.), Is the best good enough? Optimality and competition in syntax, 193–219. Cambridge, MA: MIT Press.
128
leonetti
Gundel, Jeanette. 2010. Reference and accessibility from a Givenness Hierarchy perspective. International Review of Pragmatics 2. 148–168. Gundel, Jeanette, Nancy Hedberg & Ron Zacharski. 1993. Cognitive status and the form of referring expressions in discourse. Language 69(2). 274–307. Gundel, Jeanette, Nancy Hedberg & Ron Zacharski. 2012. Underspecification of cognitive status in reference production: Some empirical predictions. Topics in Cognitive Science 4(2). 249–268. von Heusinger, Klaus & Petra B. Schumacher. 2019. Discourse prominence: Definition and application. Journal of Pragmatics 154. 117–127. Holmberg, Anders. 2010. Null subject parameters. In Theresa Biberauer, Anders Holmberg, Ian Roberts & Michelle Sheehan (eds.), Parametric variation. Null subjects in Minimalist Theory, 88–124. Cambridge: Cambridge University Press. Holmberg, Anders, Aarti Nayudu & Michelle Sheehan. 2009. Three partial null-subject languages: A comparison of Brazilian Portuguese, Finnish and Marathi. Studia Linguistica 63(1). 59–97. Ingham, Richard. 2018. Topic, focus and null subjects in Old French. Canadian Journal of Linguistics 63(2). 242–263. Jasinskaja, Katja, Sofiana Chiriacescu, Marta Donazzan, Klaus von Heusinger & Stefan Hinterwimmer. 2015. Prominence in discourse. In Amedeo de Dominicis (ed.), pS– prominenceS: Prominences in linguistics, 134–153. Viterbo: DUSICOM Press. Jasinskaja, Katja & Elena Karagjosova. 2020. Rhetorical relations. In Daniel Gutzmann, Lisa Matthewson, Cécile Meier, Hotze Rullmann & Thomas E. Zimmerman (eds.), The Wiley Blackwell companion to semantics. Oxford: Wiley Blackwell. Jiménez-Fernández, Ángel. 2016. When discourse met null subjects. Borealis 5(2). 173– 189. Kehler, Andrew. 2002. Coherence, reference, and the theory of grammar. Stanford: CSLI. Kehler, Andrew, Laura Kertz, Hannah Rohde & Jeffrey L. Elman. 2008. Coherence and coreference revisited. Journal of Semantics 25(1). 1–44. Leonetti, Manuel. 2013. Information structure and the distribution of Spanish bare plurals. In Johannes Kabatek & Albert Wall (eds.), New perspectives on bare noun phrases in Romance and beyond, 121–155. Amsterdam: Benjamins. Leonetti, Manuel. 2014. On contrastive readings in the interpretation of NPs/DPs. In Sofiana Chiriacescu (ed.), Proceedings of the vi NEREUS International Workshop ‘Theoretical implications of the syntax-semantics interface in Romance’, Arbeitspapier 127, 99–116. Universität Konstanz: Fachbereich Sprachwissenschaft. Lubbers Quesada, Margaret L. & Sarah E. Blackwell. 2009. The L2 acquisition of null and overt Spanish subject pronouns: A pragmatic approach. In Joseph Collentine, Maryellen García, Barbara Lafford & Francisco Marcos Marín (eds.), Selected Proceedings of the 11th Hispanic Linguistic Symposium, 117–130. Sommerville, MA: Cascadilla Proceedings Project.
topics and the interpretation of referential null subjects
129
Reinhart, Tanya. 1981. Pragmatics and linguistics: An analysis of sentence topics. Philosophica 27. 53–94. Reinhart, Tanya. 1986. Center and periphery in the grammar of anaphora. In Barbara Lust (ed.), Studies in the acquisition of anaphora, vol. 1, 123–150. Berlin: Springer. Rizzi, Luigi. 2018. Subjects, topics, and the interpretation of pro. In Roberto Petrosino, Pietro C. Cerrone & Harry van der Hulst (eds.), From sounds to structures: Beyond the Veil of Maya, 510–530. Berlin: de Gruyter. Roberts, Craige. 1989. Modal subordination and pronominal anaphora in discourse. Linguistics and Philosophy 12(6). 683–721. Samek-Lodovici, Vieri. 1996. Constraints on subjects: An optimality-theoretic analysis. New Jersey: Rutgers University PhD dissertation. Sheehan, Michelle. 2006. The EPP and null subjects in Romance. Newcastle upon Tyne: University of Newcastle PhD dissertation. Sheehan, Michelle. 2016. Subjects, null subjects and expletives. In Susann Fischer & Christoph Gabriel (eds.), Manual of grammatical interfaces in Romance, 329–362. Berlin: de Gruyter. Shormani, Mohammed Q. 2017. SVO, (silent) topics and the interpretation of referential pro: A discourse-syntax interface approach. Italian Journal of Linguistics 29(2). 131–188. Sigurðsson, Halldór A. 2011. Conditions on argument drop. Linguistic Inquiry 42(2). 267–304. Taboada, Maite. 2008. Reference, centres and transitions in spoken Spanish. In Jeanette Gundel & Nancy Hedberg (eds.), Reference: Interdisciplinary perspectives, 176–215. Oxford: Oxford University Press. Taboada, Maite & Loreley Wiesemann. 2010. Subjects and topics in conversation. Journal of Pragmatics 42(7). 1816–1828. Torregrossa, Jacopo, Christiane Bongartz & Ianthi M. Tsimpli. 2015. Testing accessibility: A cross-linguistic comparison of the syntax of referring expressions. LSA Annual Meeting of the Linguistic Society of America, Extended Abstracts 2015, vol. 6(29). 1–5. Tsimpli, Ianthi M., Antonella Sorace, Caroline Heycock & Francesca Filiaci. 2004. First language attrition and syntactic subjects: A study of Greek and Italian near-native speakers of English. International Journal of Bilingualism 8(3). 257–277. Villa-García, Julio. 2012. Spanish subjects can be subjects: Acquisitional and empirical evidence. Iberia 4(2). 124–169. Walker, Marilyn A., Aravind K. Joshi & Ellen F. Prince (eds.). 1998. Centering Theory in discourse. Oxford: Oxford University Press.
chapter 5
Specificity and Questions of Specification Edgar Onea
1
Introduction
There is significant morphological and semantic variation in indefinite pronouns and determiners both within and across languages, as extensively reviewed in Haspelmath (1997). While in many languages there are default indefinites, often headed by an indefinite article as in (1), there are also many other morphologically and semantically more marked forms expressing indefiniteness, as shown in the examples in (2). While the latter are oftentimes known under the label of indefinite pronouns, in this paper I prefer the label specialized indefinites for examples as in (2), as opposed to default indefinites, for examples such as (1). The term specialized indefinites permit flexibly referring to specialized indefinite pronouns, NPs or determiners, as shown in (2). (1) Ashanti saw a professor.
Default indefinite NP
(2) a. Ashanti saw someone. b. Ashanti saw a certain professor. c. Ashanti saw this professor.
Specialized indefinite pronoun Specialized indefinite NP Specialized indefinite determiner
The meaning of specialized indefinites is known to be more complex than the meaning of default indefinites. As such, specialized indefinites have been a longstanding topic of semantic theory. Ideally, the meaning of specialized indefinite pronouns and determiners should be expressed within a unified semantic framework. Such a framework should provide the technical flexibility to capture the various semantic contributions specialized indefinite determiners and pronouns may have, for example, in terms of scope or speakeridentifiability inferences. Such effects are illustrated in (3). While (3a), containing the default indefinite a dog, is compatible with all three continuations in (3a:i–iii), the variant in (3b) with the specialized indefinite determiner a certain is only compatible with (3b:ii–iii). (3b:i) is not acceptable because this reading involves narrow scope for the indefinite under the intensional verb search, a reading not available for a certain dog.
© Edgar Onea, 2022 | doi:10.1163/9789004473324_006
131
specificity and questions of specification
(3) a. Anais is searching for a dog. i. She really needs a dog. Any dog would do. ii. She lost one of her dogs during a hike tour. iii. Namely Laika, the dog of her mother. b. Anais is searching for a certain dog. i. #She really needs a dog. Any dog would do. ii. She lost one of her dogs during a hike tour. iii. Namely Laika, the dog of her mother. Onea (2016) outlines a general theory according to which specialized indefinite pronouns and determiners grammatically associate with a question of specification and semantically provide a comment on that question. I will dub this theory the erotetic theory of indefiniteness.1 Simplifying somewhat, for (2b), the specificational question commented on by a certain is the one in (4a), the comment introduced by a certain is (4b), and the eventual meaning of the entire assertion reads in English as in (4c). Thereby, the salient agent is usually the speaker. (4) a. Q: Which professor did Ashanti see? b. Some salient agent knows Q. c. Ashanti saw a professor and some salient agent knows which professor Ashanti saw.
Specificational question Comment on question Final paraphrase
In this paper, I further elaborate on the erotetic theory of indefiniteness in two important respects. Firstly, the analysis in Onea (2016) is formulated in a technically demanding version of Inquisitive Semantics (see Ciardelli, Groenendijk and Roelofsen 2019 for a recent overview). In this paper, I clarify that the analysis is conceptually independent of Inquisitive Semantics and could be phrased, in principle, in any of the major theories of semantic composition that capture both questions and assertions alike, e.g., the Structured Meanings approaches (Krifka 2001, cf. also Krifka 2011). Secondly, in Onea (2016), I claim that questions of specification are not only useful to capture the intuitive meaning contribution of specialized indefinites but these questions also control the scopal behavior of indefinites. In this paper, I further substantiate this claim by providing a detailed analysis of the German specialized indefinite determiners ein gewisser and ein bestimmter (roughly: ‘a certain’) following up on the work of 1 The term erotetic goes back to the Ancient Greek ἐρωτητικός meaning ‘pertaining to questioning’. This term is frequently used in the logical literature, e.g., in Wisniewski (1995) and earlier work by Hintikka.
132
onea
Ebert, Ebert and Hinterwimmer (2013). The analysis of ein gewisser and ein bestimmter will lead to several surprising insights about the nature of specific indefinites. Chiefly, I argue that (at least some) specificity markers not only encode comments on questions of specification, but also an interaction with intensional or performative speech act operators at LF, an observation hitherto missed in the literature. For example, bestimmt always modifies an intensional or speech act operator making sure that its semantic contribution is part of the question bestimmt comments on. Hence, for (5), the question commented on by bestimmt is (5a) and not (5b). The difference between (5a) and (5b) may seem to minor if we consider the broader pragmatic interpretation, however the actual claim here is that bestimmt needs an intensional embedding operator or a speech act operator to be interpretable. Therefore, outscoping such an operator is not a pragmatic but a core semantic operation. This will be the core ingredient in the analysis of subtle observations on the scopal properties of bestimmt-indefinites in German. (5) Suleika glaubt, dass ein bestimmter Arzt ihr helfen kann. Suleika believes that a bestimmt doctor her help can ‘Suleika believes that a certain doctor can help her.’ a. Of which doctor does S. believe that he can help her? b. Which doctor can help Suleika? As for the structure for this paper, Section 2 introduces the erotetic theory of indefiniteness, thereby clarifying some issues either left unclear in Onea (2016) and embeds the theory into the literature. Section 3 shows how the framework can handle the two German specialized indefinite determiners mentioned above, spelling out details of the theory. Section 4 summarizes and closes the paper.
2
The Erotetic Theory of Indefiniteness
In this section, I first introduce three phenomena that have been usually associated with indefiniteness, namely a) scope assignment, b) epistemic specificity effects and c) indefinite-interrogative affinity. These phenomena do not exhaust the range of semantic intricacies indefinites are involved in; see von Heusinger (2011) for a comprehensive overview. However, both in terms of the attention they have received in the literature and in terms of cross-linguistic ubiquity, the three phenomena belong to the most important ones. In discussing these phe-
specificity and questions of specification
133
nomena in Section 2.1, I concentrate on the formulation of general questions any theory of indefiniteness needs to answer. In the second step, in Section 2.2, I present and extend the erotetic theory of indefiniteness formulated in Onea (2016), showing how it addresses the questions raised in Section 2.1. 2.1 Three Phenomena and the Questions They Raise 2.1.1 Epistemic Specificity Following Farkas (1994), the term epistemic specificity, conveys that the speaker has a particular individual on her mind when using an indefinite, whereas epistemic non-specificity labels the use of an indefinite without a singular referent on the speaker’s mind. Example (6) shows that this distinction is essentially a pragmatic one, as the very same default indefinite can have both an epistemically specific use, as in (6a), and an epistemically non-specific use, as in (6b). (6) Skylar is dating a professor. a. Namely my best friend, Ashley. b. I wish I knew which one, though. Two immediate questions emerge when discussing epistemic specificity. Firstly, do any specialized indefinites encode epistemic specificity or nonspecificity? Secondly, what exactly does it mean for the speaker to have a particular referent on her mind, i.e., what exactly is epistemic specificity? Concerning the first question, the literature converges on a positive answer. It has been argued, for example, that English a certain, Russian koe- and German gewiss indefinites or Romanian pe-marked indefinite direct objects are markers of epistemic specificity, cf. Breheny (2003), Onea and Geist (2011), Ebert, Ebert and Hinterwimmer (2013), Kamp and Bende-Farkas (2019). One example for Russian koe-indefinites, from Eremina (2012), is given in (7). At the same time, it is by no means clear that these expressions actually signal the same kind of information about the epistemic status of the speaker towards the referents introduced by the indefinite. Similarly, a range of indefinites have been discussed that in some way signal that the speaker has no or even could not possibly have a particular referent on her mind, e.g., Kratzer and Shimoyama (2002) for German irgendein, Fălăuș (2009) for Romanian vreun and Alonso-Ovalle and Menéndez-Benito (2010) for Spanish algún and Alonso-Ovalle and MenéndezBenito (2015) for an overview. Epistemically non-specific indefinites are usually dubbed epistemic indefinites. The distributional properties of various epistemic indefinites are not identical, and they tend to exhibit variation in terms of free choice effects, speaker ignorance regarding the identity of the referent or the usage as polarity items. One example for Spanish algún is given in (8), from Alonso-Ovalle and Menéndez-Benito (2010).
134
onea
(7) Ja nashla koe-chto interesnoje v etoj knige. (Russian) I found KOE-what interesting in this book ‘I saw something interesting in this book (I know exactly what it is and I may tell you later but at this point I don’t what to specify anything)’. (8) María se casó con algún estudiante. Maria REFL married with ALGUN student ‘Maria married some student or other.’
(Spanish)
The literature has explored a number of approaches to what it means to have someone on one’s mind when uttering an indefinite. For example, several scholars have assumed that epistemically specific indefinites should be modelled in terms of a special kind of Skolem function. While standard Skolem functions are existential closed, thus yielding the equivalence in (9a), one can think of such functions as free variables, as in (9b), that can have some more specific way to pick their referents depending on their anchor argument. For example, Yeom (1998) assumes some sort of cognitive contact between the speaker and the referent as identifying the respective function. Especially when there is no scopal interaction with nominal quantifiers, the notion of a Skolem function may not be the handiest tool to capture indefinites, hence sometimes, scholars have used choice functions instead, cf. e.g., Kratzer (1998) or von Heusinger (2002). (9) a. ∀y.P(y) → ∃x.Q(x)∧ R(y, x) = ∃f.[∀z.P(z) → Q(f(z))]∧∀x.P(x) → R(x, f(x)) where f is contextually or lexically specified b. ∀x.P(x) → R(x, f(x)) such that it holds that: ∀z.P(z) → Q(f(z))
Others, following Schwarzschild (2002), suggest that epistemically specific indefinites refer to a discourse referent picked from a singleton set created from the set denoted by the NP and some additional identifying idea (Breheny 2003). Kamp and Bende-Farkas (2019) suggest that epistemically specific indefinites are referentially anchored, again similar to a Skolem function, to some entity in the speaker’s mental representation, and the addressee will construct a mental representation of the same discourse referent in which the unknown representation of the speaker acts as the referential anchor. In other words, the addressee can refer to whomever the speaker had in her mind. Yet another approach suggested by Ebert, Ebert and Hinterwimmer (2012, 2013) is that epistemic specificity may involve having the answer to a specificational question such as the one in (10) for the example in (6). There, of course, a range of pos-
135
specificity and questions of specification
sible complications may arise that are connected to the very interpretation of questions, as discussed in Aloni (2001). For example, the specificational question in (10) can be understood as ranging over alternative individuals on a picture, as in (10a), or over domains of expertise by which professors at a university could be identified, as in (10b). Such different systematic means of identifying individuals are called conceptual covers. It is possible for a speaker to be able to answer a question under one conceptual cover, but not under another. This then gives rise to the possibility that the very same indefinite is epistemically specific and non-specific at the same time, as in (11), from Aloni and Port (2010). (10) Original example: Skylar is dating a professor. Specificational question: Q: Which professor is Skylar dating? epistemic effect: Speaker can answer Q. a. Which one of the professors on this picture? b. Which one of the professors in terms of domain of expertise? (11) Devo incontrare un qualche professore. I-must meet a some professor ‘I must meet a certain professor, but I don’t know who he is.’
(Italian)
Indefinites that signal epistemic non-specificity are usually dubbed epistemic indefinites.2 The main intuition to be captured for such indefinites is the speaker-ignorance inference. One natural test for such speaker-inference involves the unacceptability of a guess-who continuation, as shown in (12) (Italian), from Aloni and Port (2015). (12) Maria ha sposato un qualche professore. #Indovina chi? (Italian) who? Maria has married a qualche professor guess ‘Maria married some professor, the speaker doesn’t know who. #Guess who?’ Some scholars, such as Kratzer and Shimoyama (2002), Aloni and van Rooij (2007), Chierchia (2013), Alonso-Ovalle and Menéndez-Benito (2010), take epistemic effects to be pragmatic side-products of some independent semantic
2 A class of indefinites not discussed here that is usually considered related to epistemic indefinites is the class of free choice indefinites, see Aloni (2022) and Kellert (2022) for a recent analysis.
136
onea
contribution specialized epistemic indefinites may have. For example, AlonsoOvalle and Menéndez-Benito (2010) assume that German irgendein and Spanish algún provide a set of alternatives that ranges over a widened domain. The ignorance inference (or related other inferences such as free choice) is derived from there as pragmatic implicatures. Other theories assume a semantic felicity condition to the extent that the speaker using an epistemic indefinite has no particular referent on her mind, e.g., Giannakidou and Quer (2013). However, just as with epistemically specific indefinites, there are some problems explaining what it means not to have a referent on one’s mind; see Aloni and Port (2015) for discussion. At present, there is no unified, generally accepted theory of epistemic specificity or epistemic indefinites. In fact, given the current literature, one may expect that (despite the similar sounding denominations) epistemic indefinites and epistemic specificity may be semantically unrelated. After all, most influential approaches such as Kratzer and Shimoyama (2002) or Chierchia (2013) and subsequent literature, suggest that epistemic indefinites appear to live mainly at the syntax-pragmatics interface whereas epistemically specific indefinites appear to be mainly situated at the interface between pragmatics and cognitive representation, as in Kamp and Bende-Farkas (2019). In Onea (2016), I have suggested that both types of specialized indefinites may in fact encode comments on specificational questions, but a detailed derivation of the distribution of epistemic indefinites in such a framework was never provided. 2.1.2 Scopal Specificity The term scopal specificity is meant to capture the fact that specific indefinites tend to have wide scope, whereas non-specific indefinites tend to have narrow scope. This seems to be particularly clear with respect to intensional operators, cf. Ioup (1977), Zimmermann (1993, 2006). For example (13), the epistemically specific continuation in (13c) is only compatible with a wide-scope interpretation as paraphrased in (13a), and not with a narrow scope interpretation, as in (13b). (13) Kim is searching for an environmental activist. a. There is an environmental activist, such that Kim is searching for her. b. Kim is searching for some environmental activist or other. c. Namely Dominique, the leader of the college environmental association. Scopal specificity may also involve referentially transparent operators, as in the example (14). Here, a further type of interpretation emerges. The continuation
specificity and questions of specification
137
(14a) amounts to a wide scope interpretation, (14c) to a narrow scope interpretation. (14b) is special because the indefinite has—technically—narrow scope, but is usually analyzed as involving functional wide scope, in which a (wide scope) Skolem function assigns students to teachers. This function is often existentially closed and later identified in the continuation, as discussed for example (9) above, cf. Hintikka (1986). (14) Every teacher praised a student. a. Namely Hadley. ∃x.S(x)∧∀y.T(y) → P(y, x) x=H b. Her most undisciplined student. ∃f.[∀x.T(x) → S(f(x))]∧∀y.T(y) → P(y, f(y)) f = her most undisciplined student c. Ashanti praised Skylar, Kyle praised Ivory, Lashawn praised Chrystal etc. ∀y.T(y) → ∃x.S(x)∧ P(y, x) The existence of functional readings unfortunately blurs the difference between narrow and wide scope indefinites at an empirical level. In particular, (14c) can also be analyzed as functional wide scope: after all the speaker explicitly describes the respective function assigning students to professors by giving a list of pairs. Moreover, plain wide scope readings are—in fact—also a special case of functional wide scope readings: in this case, the function is a constant function. Hence, a great amount of caution is needed in analyzing the scopal properties of indefinites with respect to nominal quantifiers. In this paper, I ignore functional readings nearly entirely and mainly focus on interaction with operators that are non-susceptible to functional readings, such as intensional operators, negation, conditionals, etc. The reason why such operators cannot give rise to functional readings is that they operate on worlds rather than individuals, and there seems to be no Skolemization over world variables in natural language, cf. e.g., Brasoveanu and Farkas (2011) for a brief discussion. Scopal specificity leads to interesting observations when scope islands are considered. Fodor and Sag (1982) observed that indefinite scope is not restricted to syntactic scope islands. This is shown in (15), which has a prominent widest scope reading for the indefinite a certain professor even though the indefinite is located in a scope island, namely the antecedent of a conditional. (15) If a certain professor comes to the party, Kenley will leave. Wide scope reading: There is a professor such that if she comes to the party, Kenley will leave.
138
onea
Assuming that scope islands are boundaries for LF movement, one may rule out the usual quantifier raising approach to semantic scope for indefinites (but see Szabolcsi 2010 and Heim 2011 for somewhat different views). Thus, various other ideas have been explored in the literature. Among others, the literature has suggested: choice functions of various kinds (cf. von Heusinger 1997, 2002, Reinhart 1997, Winter 1997, Kratzer 1998), Skolem functions (cf. Kamp and Bende-Farkas 2001, Endriss 2009), quantifier domain restriction (Schwarzschild 2002), various types of presupposition projection (Jäger 2007, Geurts 2010, Onea 2015) and even an independence friendly logical implementation of existential quantifiers (Brasoveanu and Farkas 2011). In many of these approaches, especially in the choice and Skolem functional ones, there is an (oftentimes) implicit connection between scopal and epistemic specificity, to the extent that the semantic device that drives wide scope can be loaded with epistemic information, e.g., an identifying idea in the sense of Breheny (2003). This would be the case in (14b) for Skolem functions and in (16) for choice functions. In particular, ε is a classical choice function, but can in principle be enriched to choose an element from a set using some heuristic, marked by … in (16). For example, the method by which ε picks students could be based on speaker-acquaintance, i.e., ε could pick the individual from each set, that the speaker is best acquainted with. (16) Every teacher praised a student. ∀x.T(x) → P(x, ε({y|S(y)}))Thereby, ∀A → 𝜖(A) ∈ A∧ method(ε) = … Unfortunately, approaches that derive exceptional scope from any device that seems obviously connected to epistemic specificity have been shown to have technical difficulties. This is particularly clear for choice functional and Skolem functional approaches; see e.g., Chierchia (1998), Schwarz (2001). Moreover, Farkas (1994) suggests that scopal and epistemic effects are independent of each other. In particular, she argued that it is possible to have epistemically non-specific wide scope readings, epistemically specific narrow scope readings (at least in transparent contexts), even though in most examples the notions tend to converge. However, given the complication involving functional readings, it is not so easy to make a clear argument for or against the availability of specific narrow scope readings that are not functional. Hence, the question regarding connection between scopal specificity and epistemic specificity is still an open one. This is particularly interesting because, usually, specialized indefinite determiners do not only mark scope, they also have some epistemic flavor. In fact, Brasoveanu and Farkas (2011) suggest that default indefinites generally have entirely unrestricted scope, even across scope islands, and it is only
specificity and questions of specification
139
specialized indefinite determiners and pronouns that exert control over the scope configuration. However, in doing so, their semantic contribution does not seem to be limited to scope control. In fact, to my knowledge, the only type of indefinites that are in no way connected to epistemic specificity and still have clear impact on scope, are the so-called dependent indefinites, e.g., reduplicated indefinites in Hungarian, cf. Farkas (2002), Brasoveanu and Farkas (2011). Reduplicated indefinites indeed impose narrow scope for the indefinite under a quantifier without making any comment on the epistemic status of the indefinite itself, as shown in (17) from Hungarian, which is compatible both with a pair-list and a functional reading. According to Szabolcsi (2010) and Csirmaz and Szabolcsi (2012), however, such examples should rather be analyzed in terms of event-based distributivity of reduplicated numerals rather than in a theory of indefiniteness. (17) Minden tanár egy-egy majmot látott. every teacher one-one ape.acc saw ‘Every teacher saw an ape each.’
(Hungarian)
Thus, the generalization remains on the table that all specialized indefinites that have some scopal effect also have some epistemic flavor. If this is correct, an attempt to derive scopal effects from epistemic information associated with indefinites seems to be promising: Thus, using comments on specificational questions to derive scopal effects may be a theoretically attractive option, given that alternative approaches using choice or Skolem functions tend to run into technical difficulties. 2.1.3 Indefinite Interrogative Affinity It is well known in the literature, that there is a ubiquitous morphological connection between indefinites and interrogatives known under the label indefinite-interrogative affinity. Examples (18a) and (18b), from German, show that the very same indefinite pronoun can act either as an interrogative or as an indefinite depending on its syntactic position. Moreover, the same morphological stem can be prefixed to yield a different, semantically more marked, indefinite pronoun as shown in (18c). (18) a. Hast du wen getötet? have you wh.acc killed ‘Did you kill anyone?’
140
onea
b. Wen hast du getötet? who.acc have you killed ‘Whom did you kill?’ c. Quinn hat irgend-wen gesehen. Quinn has irgend-who.acc seen ‘Quinn has seen some person or another.’ This morphological connection between interrogatives and indefinites is not limited to German. Ultan (1978) finds that in 77 out of 79 languages indefinite pronouns are either identical or derivationally related to interrogative pronouns and, in a more recent study; Haspelmath (1997) suggests that indefinites are related to interrogatives in 64 out of 100 languages. Further examples are given in (19) (Lakhota/Siouan) and (20) (Kannada/Dravidian) from Bhat (2004: 226), see also the original discussion in van Valin (1992). (19) a. šúka ki táku yaxtáka he dog the taku bite int ‘What did the dog bite?’
(Lakhota)
b. šúka ki táku yaxtáka dog the taku bite ‘The dog bit something.’ (20) a. ra:ju ellige ho:da Raju ellige went ‘Where did Raju go?’
(Kannada)
b. ra:ju ellig-o: ho:da Raju ellige-o went ‘Raju went somewhere.’ Assuming that this morphological relation between indefinites and interrogatives is not accidental, it is natural to ask for a semantic explanation. The question is then: are there good reasons to assume that indefinites are the same as interrogatives or derived from interrogatives, and if so, what follows from this for a theory of indefinites? Not every major theory of indefiniteness has anything interesting to say about this question. In fact, most theories mentioned above do in no way account for the indefinite-interrogative affin-
specificity and questions of specification
141
ity. Theories using choice functions or domain restrictions are cases in point. However, some theories have emerged which view indefinites and interrogatives on a par. In fact, for any major theory of interrogatives, there is a theory of indefinites that implements the same idea. A well-known example is Karttunen (1977), who explicitly develops the semantics for interrogatives from the semantics used for indefinites. For Karttunen, interrogatives are run-ofthe-mill existential quantifiers just like indefinites. To the extent that exceptional scope for indefinites is a problem for theories on which indefinites are plain existential quantifiers, this also applies to Kartunen (1977). However, an independence friendly approach, such as Brasoveanu and Farkas (2011), could carry over to a Karttunen (1977) based interrogative and question semantics, though—admittedly—this line of attack has not been explored in the literature to my knowledge. Kratzer and Shimoyama (2002), Ciardelli, Roelofsen and Theiler (2017) and Charlow (2014) provide various theories of indefinites in which they are underlyingly alternative introducing devices just in the same way as interrogatives are in Hamblin (1973) and subsequent literature. Similarly, the structured meanings approach to interrogatives and questions (Ginzburg 1992, Krifka 2001), according to which wh-words provide the restrictor of a variable, has a natural correspondence to presuppositional theories of indefiniteness such as Abusch (1994), Jäger (2007) and Onea (2015). In the latter type of theories indefinites project a restrictor set over a variable replicating the logic of structured meanings approaches. Thus, all major theories of indefiniteness that can capture exceptional scope (and many who cannot) can be stated without any significant alterations as theories of interrogatives. This is a remarkable observation, because only Hamblin-semantics based theories of indefinites and theories couched within the framework of inquisitive semantics have explicitly made the connection between indefinites and interrogatives to one of their features. 2.1.4 Interim Summary The discussion above shows three important facts. Firstly, specialized indefinite determiners tend to encode some information about the epistemic perspective of the speaker towards their referent. Secondly, indefinites tend to take scope in ways that ignore scope islands, but, at the same time, the scope configuration of specialized indefinite determiners is at least correlated with their epistemic meaning contribution. Finally, there is a deep connection between indefiniteness and interrogatives, that surfaces in virtually all major theories of indefiniteness and questions. However, it is not yet sufficiently understood how specialized indefinite determiners exert scope control and how this connects to the semantic similarity between interrogatives and indefinites. Crucially, to
142
onea
the extent that epistemic effects connect to the epistemic state of the speaker towards a question of specification, and if scope effects associated with specialized indefinite determiners can be corroborated with this epistemic attitude, the indefinite-interrogative affinity may suggest a compositional role of questions of specification in deriving all these facts. 2.2 The Proposal In this section, I lay out a revised version of the proposal in Onea (2016) and thereby eliminate several issues of the original proposal. Part of the reason why this is necessary is that Onea (2016) is couched within a variant of Inquisitive Semantics, similar to Ciardelli, Roelofsen and Theiler (2017); see also Ciardelli, Groenendijk and Roelofsen (2019). This framework is compatible with the compositional theory in Onea (2015) using partial functions for denotations only in a fairly intransparent way, and may, thus, obscure the view on the actual conceptual merits of the theory. Thus, I here provide a simpler and leaner variant of the erotetic theory of indefiniteness.3 At the same time, I correct some assumptions in my older work about how exactly questions commented on by specialized indefinites are targeted by the latter. The proposal contains four essential aspects. Firstly, a compositional theory that shows how indefinites can gain exceptional wide scope. Secondly, the idea that specialized indefinites morphologically incorporate particles that encode comments on specificational questions. Thirdly, a range of possible comments on specificational questions which then impact the eventual scope indefinites will get. And, finally, a pragmatic story on how specialized indefinite determiners may have grammaticalized. We will consider each of these aspects in the following sections. 2.2.1 What Are Indefinites and How They Get Scope The core idea of the compositional system is—in a sense—very similar to the usual treatment of quantification, as described, for example, in Heim and Kratzer (1998). There, quantifiers do not actually combine with predicates. Instead, a predicate can combine with a pronoun-like trace of the quantifier whereas the quantifier itself is moved to some syntactic position from where it is able to bind its trace, as shown in (21). For our purposes, (in addition to the classical idea of syntactic movement) one can conceptualize quantifier movement as a kind of bottom-up compositional delay, i.e., the actual semantic value 3 In discussing the compositional system, I focus on introducing the main ideas and providing an intuitive understanding of the mechanisms involved. I leave some formal definitions implicit. Readers can consult Onea (2015, 2016) for missing details.
specificity and questions of specification
143
of a quantifier is combined with the predicate at a later (or at least different) stage of semantic derivation. (21) a. direct computation: [ Pred [ QDet NP ]1 ] b. delayed computation: [ [ QDet NP ] [1 [ Pred t1 ] ] ] We extend this idea to all referential expressions: a predicate always combines with something like a pronoun, while the descriptive material of the referential expression is compositionally delayed, albeit not by syntactic movement. Instead, when we want to combine a predicate like danced with a referential expression like Ashanti, we combine the predicate with a pronoun and express—figuratively speaking—a ‘hope’ that, eventually, that pronoun will denote the individual named Ashanti. This ‘hope’ is expressed as a partiality condition on the resulting denotation and we name it a referential constraint. In result, if the pronoun will eventually not denote Ashanti, the composition will—at some point—collapse. Consider example (22) for illustration. (22b) standardly says that a pronoun denotes a function from assignments h, i.e., a function from indices to individuals, to some individuals assigned to the respective index by h. As shown in (22a), the proper name Ashanti denotes a partial function from assignments h that make sure that the referential index of Ashanti will be assigned to Ashanti by h. In other words, the only difference between a pronoun and a proper name is that the proper name expresses a condition on assignments that the referent will be the intended one. Apart from this referential constraint, pronouns and proper names are the same. So far, these denotations are at least similar to ideas from dynamic semantics, cf. Kamp (1981/2013), Heim (1982), Groenendijk and Stokhof (1991) and others. What looks more radically different compared to more common compositional theories is the denotation of the predicate in (22c). Here, the first argument is the denotation of a referential expression x and the second one is an assignment h. The result is the statement—put here in terms of simple extensions—that the individual denoted by the referential expression x given the assignment h danced. But there is a twist to this: the second argument needs to be an assignment h that is compatible with x. Mathematically, the subscripted c(h, x) means that h is in the domain of x. For the example in (22d), this boils down to the requirement that h(i) equals Ashanti. (22) a. b. c. d.
⟦Ashantii ⟧ = λhh(i)=A .h(i) ⟦proi ⟧ = λh.h(i) ⟦danced⟧ = λx.λhc(h,x) .D(x(h)) ⟦Ashanti danced⟧ = λhh(i)=A .D(h(i))
144
onea
Thus, predicates combine only with pronouns in the sense that their resulting denotation is ignorant about the intended referents. The descriptive material, e.g., the proper name, is not moved syntactically but rather transformed into a referential constraint, turning the entire denotation partial. In order to turn the result standard again, existential closure, defined in (23), applies. Existential closure turns referential conditions into simple truth-conditional statements conjoined to the derived partial truth conditions and guarantees that there is an assignment satisfying the referential constraints and the truth conditions simultaneously.4 Going back to the metaphor of delayed composition, one could say that existential closure is the place at which the descriptive content of referential expressions enters the truth conditional domain, thus this is the end of the compositional delay and, thus, the counterpart of the landing site of moved quantifiers. (23) a. ⟦∃α⟧ = ∃h.c(h, ⟦α⟧)∧⟦α⟧(h) b. ⟦∃ Ashanti danced⟧ = ∃h.h(i) = A∧ D(h(i))
Before moving on, we can also clarify the semantic types used in this system, in comparison to more classical typing. We do this in Table 5.1. Thereby, a is the type of assignments and I ignore situation/world indices for simplicity. Indefinites are also referential expressions. They only differ from proper names in that their referential constraint is compositionally computed, rather than lexically stored. This is shown in (24). A teacher denotes a pronoun but expresses the referential constraint that, eventually, the value of that pronoun will be a teacher. This referential constraint is turned into truth-conditional content by the existential closure operator. The result is entirely equivalent to the more familiar: ∃x.T(x)∧ D(x). (24) a. b. c. d. e.
⟦INDEFi ⟧ = λP.λhP(λg.g(i))(h) .h(i) ⟦teacher⟧ = λx.λhc(h,x) .T(x(h)) ⟦INDEFi teacher⟧ = λhT(h(i)) .h(i) ⟦INDEFi teacher danced⟧ = λhT(h(i)) .D(h(i)) ⟦∃ INDEFi teacher danced⟧ = ∃h.T(h(i))∧ D(h(i))
Existential closure may not only come with an unselective flavor, as in (23a) but it also may target a specific referential index, a set of indices or all but a spe4 This makes the treatment of referential expressions similar to presupposition projection; see Beaver and Krahmer (2001), the main alternative to this partialization framework.
specificity and questions of specification table 5.1
145
The semantic types
Expression type
Classical type Current type
Proper names, pronouns One place predicates Two place predicates Partial propositions Propositions
e ⟨e, t⟩ ⟨e, ⟨e, t⟩⟩ t
⟨a, e⟩ ⟨⟨a, e⟩, ⟨a, t⟩⟩ ⟨⟨a, e⟩, ⟨⟨a, e⟩, ⟨a, t⟩⟩⟩ ⟨a, t⟩ t
cific index (or all but a set of referential indices). Such operators can be defined following the patterns provided in (25). (25) a. ⟦∃i α⟧ = λhci(h(⟦α⟧)) .∃g.g =ˆi h∧ c(g, ⟦α⟧)∧⟦α⟧(g) b. ⟦∃i,j,k α⟧ = λhci,j,k(h(⟦α⟧)) .∃g.g =i,j,k h∧c(g, ⟦α⟧)∧⟦α⟧(g)
The formula in (24d) might remind the reader of a classical question denotation in structured meanings approaches, as given in (26b) for the question (26a). (26) a. Which teacher danced? b. Structured meanings denotation: λxT(x) .D(x) The formula in (24d) is not exactly a question because it is not incomplete. However, it is general enough to be equally applicable for interrogatives as well. Hence, we can postulate the lexical entries in (27). The only difference between indefinites and interrogatives is that an indefinite co-occurs with an existential closure operator whereas an interrogative co-occurs with a question operator. The question operator can be defined as in (27c), to illustrate the intuition. Applying this operator to a partial proposition with a wh-expression would yield a question, as shown in (27d). (27) a. b. c. d.
⟦whichi ⟧ = λP.λhP(λg.g(i))(h) .h(i) ⟦whichi teacher danced⟧ = λhT(h(i)) .D(h(i)) ⟦?i α⟧ = λX∃g.c(g,⟦α⟧)∧g(i)=X .∃h.h(i) = X∧ci (h, ⟦α⟧)∧⟦α⟧(h) ⟦?i whichi teacher danced⟧ = λXT(X) .D(X)
However, in the formulae in (27c,d), we used X instead of x to signal that the variables have different types. X stands for an individual, i.e., type e, whereas x stands for a function from assignments to individuals, i.e., type ⟨a, e⟩. While
146
onea
this makes clear that we can turn formulae classical at any point, this is also a departure from the overall logic of the compositional system and is likely to cause problems dawn the way. A more consequent solution is the operator in (28a), yielding the result in (28b) after combination with a partial proposition. (28) a. ⟦?i α⟧ = λx.λhc(h,⟦α⟧)∧c(h,x)∧h(i)=x(h) ⟦α⟧(h) b. ⟦?i whichi teacher danced⟧ = λx.λhc(h,x)∧h(i)=x(h)∧T(h(i)) D(h(i))
In addition to its explication within the structured meanings framework for questions, the ? operator can be given in alternative frameworks as well. Onea (2016) provides an alternative operator within the Inquisitive Semantics framework. In Onea (2016), I also discuss the relation between this approach and the theory of Kratzer and Shimoyama (2002). Finally, in Onea (2014) a dynamic version of this theory is discussed, which also captures donkey anaphora.
2.2.2 Specialized Indefinite Determiners Specialized indefinite determiners may be semantically complex expressions grammaticalized into one morphological or syntactic unit. While their morphological structure may vary greatly, I suggest that their semantic structure is usually the following: they contain an element standing for default indefiniteness and a discourse-particle-like element that provides further speaker or addressee oriented (usually) not-at-issue information. We take English a certain as an example in which this type of structure is fairly transparent. There are good reasons not to analyze a certain teacher as [a [certain teacher]], where certain acts like an adjective, but rather as [[a certain] teacher], where a certain forms a semantic unit. Indeed, this is how the literature has usually proceeded, e.g., Hintikka (1986), Farkas (2002), Kratzer (1998) and others. Historically, certain comes from Latin certus, which has an adjectival use claimed to have developed to a determiner usage in Romance languages, cf. Gianollo (2018). However, the lexical meaning of certain is in many respects similar to that of modal particles, and indeed there are both in Latin (certo, certus) and modern Italian as well as in English usages which are modal adverbial or particle like. Some English examples are given in (29), examples from Italian and Latin are provided in (30) and (31) (Italian and Latin Examples and glosses provided by Chiara Gianollo, p.c.), showing modal and answer-particle like usages of certain.5
5 Both examples provide, in addition, good evidence for the high, CP level interpretation of the
specificity and questions of specification
(29) a. b. c. c.
147
I know it for certain. It is certain that she will win. I will certainly come. A: Will you do it? B: Certainly.
(30) a. Certo enim ego vocem hic loquentis modo me certo indeed I:nom voice:acc here speaking:gen just me:acc audire visus sum. hear:inf seem:ptcp.nom be:1sg Plautus—Aulularia 811 ‘It surely seemed as if I heard someone’s voice just now.’ me alienabis numquam quin b. Certe edepol tu that.not certe by.Jove you:nom me:acc take.away:2sg never noster siem our be:1sg Plautus—Amphitruo 399 ‘And by thunder, you shall never do me out of being our family’s servant.’ (31) a. Certo non mi chiederai di venire con te certo not me:dat ask:2sg of come:inf with you ‘Certainly you won’t ask me to come with you.’ b. A: Mi puoi aiutare? me:dat can:2sg help:inf ‘Can you help me?’ B: Certo! certo ‘Sure!’ Similar observations can be made for a variety of specialized indefinite determiners, including Hungarian egy bizonyos, German ein gewisser/bestimmter, which have modal and answer particle-like usages, illustrated with corpus examples in (32) and (33) respectively.6 Example (32b) provides specific eviadverbial certo/certe, as in both cases these expressions precede an emphatic adverb/interjection and the emphatic subject. 6 The Hungarian examples are from the Hungarian national corpus (HNC). The German data are from the German Reference Corpus (DeReKo).
148
onea
dence that bizonyos can in fact be used literally to mean known, and (32c) suggests that bizonyos can embed a question. In the same vein, Russian -nibud indefinites go back to a modal meaning as discussed in Haspelmath (1997). Others directly stem from question-words, such as Russian koe- indefinites. What these seem to have in common is a relatively high position at LF—both within the DP or within a CP. (32) a. […] akkor szent és bizonyos, hogy ki fog végezni. then saint and bizonyos that PRT will finish. ‘… in that case it is most certain that (s)he will exterminate me.’ (MNSZ2: doc: pers_hu_kozmed_017) már bizonyos? b. De mi van, ha az út vége but what is if the road end.gen already bizonyos ‘But what if the end of the road is already known?’ (MNSZ2: doc: lit_er_M_Szabo_Istvan_Elfelejtett_szavak_antologiaja) c. Nem volt mindig bizonyos, hogy ki kicsoda. not was always bizonyos that who who ‘It was not always clear who is who.’ (MNSZ2: doc: lit_hu_dia_Nemes_ Nagy_Agnes___Szo_es_szotlansag) (33) a. Wird Merkel an ihm festhalten? Gewiss. Zunächst. will Merkel at him hold_on gewiss for_now ‘Will Merkel hold on to him? Sure. For now.’ (BRZ05/SEP.11802 Braunschweiger Zeitung, 13.09.2005) b. Hätte mir so etwas auch in Mannheim passieren können? AUX me so something too in Mannheim happen can Bestimmt nicht. bestimmt not ‘Could such a thing have happened to me in Mannheim too? Surely not.’ (M09/MAI.33561 Mannheimer Morgen, 02.05.2009) c. Aber Argentinien gewinnt. Bestimmt. Wahrscheinlich. Hoffentlich. but Argentina wins bestimmt likely hopefully ‘But Argentina will win. Surely. Maybe. Hopefully.’ (SOL14/JUL.01328 Spiegel-Online, 13.07.2014)
specificity and questions of specification
149
Thus, given the close connection between modal particles/discourse particles and specialized indefinite determiners, I suggested in Onea (2016) that once meaning components are grammaticalized to parts of determiners, they no longer stay within the DP at LF but rather appear high in some matrix CP projection that allows them direct access to speaker and addressee parameters of the utterance. Thus, we could analyze a sentence such as (34a) as suggested in (34b). On this analysis, the linear appearance of certain inside the DP is a syntax-semantic mismatch that marks the co-indexing at LF. The main motivation for this analytic decision is not historic or the incidental observation that such determiners have modal or particle like usages, but rather the fact that this approach provides technical possibilities regarding scope control exerted by the specialized indefinite determiner, as discussed in Section 2.2.3. (34) a. A certain teacher danced. b. [… [CP certaini […. [ [ ai teacher ] danced ] ] ] ] With this background, we can finally make explicit, how the meaning contribution of indefinite determiners can be understood in this framework: (35) Specialized indefinite determiners comment on questions: The meaning contribution of specialized indefinite determiners is a notat-issue comment on a question of specification gained, intuitively, by replacing the indefinite with a which-phrase in the host utterance. Such comments on questions of specification can be spelled out in various ways. Some examples are given in (36). However, there is no particular reason why a specialized indefinite determiner would not provide any combination of such comments or have vaguer meaning contributions such as those suggested in (37). (36) a. b. c. d.
The speaker can answer Q. The speaker doesn’t want to answer Q. The speaker doesn’t believe that Q is a reasonable question to ask. The speaker believes that Q is relevant for the discourse tasks at hand.
(37) a. The speaker can answer Q but doesn’t want to answer Q. b. The speaker wants the addressee to keep Q in mind. c. The speaker wants to discourage the addressee from asking Q.
150
onea
While being aware that there are several empirical issues regarding the exact meaning contribution of English a certain, I here adopt for explicitness a simplistic analysis suggested in (38).7 The particle certain signals that the speaker (the Y argument) knows the answer to the question of specification formed from its propositional sister, i.e., p. The semantics of certain thereby incorporate the question operator defined above. For the example in (34), the result is the one in (39). Thereby, I treat the meaning contribution of certain as a presupposition for simplicity. In general, for specialized indefinites, other natural options also lend themselves. For example, some comments on specificational questions are probably expressed on some additional dimension of meaning, in the sense of Potts (2005) (cf. also Gutzmann 2013), depending on the projective behavior of the epistemic inference. Moreover, some of these comments can even be part of the regular—potentially at-issue—meaning of the resulting sentence. (38) ⟦certaini ⟧ = λp.λY : KNOW(Y, ?i p).p
(39) [… [S [ certaini […. [ [ai teacher] danced] ] ] ] ] presupposes: KNOW(cs , λXT(X) .D(X))—the speaker knows which teacher danced8 at issue meaning: λhT(h(i)) .D(h(i))—possible existential closure in the next step. 2.2.3 How Specialized Indefinite Determiners Control Scope Given the architecture of the theory, there are two syntactic places at which specialized indefinite DPs end up interpreted: the default indefinite DP part is interpreted in situ, whereas the particle part is interpreted at a high CP level, usually in the matrix clause. In addition, one could conceive of other perspectival centers within a sentence associating with the particle. For the default indefinite, scope is entirely unrestricted. For widest scope, speech act level existential closure is needed, for any other scope configuration, selective existential closure can be applied at lower levels. The question, thus, arises, how a specialized indefinite can rule out certain unavailable scope configurations. The way in which specialized indefinites work is that the particle part of their structure
7 In light of the analysis of bestimmt and gewiss indefinites in German given in section 3, it is reasonable to expect that both the syntactic and the semantic assumptions about English certain need revisions. However, I have no detailed analysis of English certain to offer in this paper. 8 We assume that the speaker is a contextual parameter noted as cs , following Kaplan (1978).
specificity and questions of specification
151
will always try to comment on a question, as shown above. Usually, this will only succeed if the sister of the particle has a denotation from which a question can be derived. However, if the indefinite has already been existentially closed, i.e., it received scope at a lower position, it will no longer be possible to extract a question at the level of the particle. These configurations are shown in (40). In (40a) a question is available, because the existential closure occurs higher than certain. In (40b) a question is not available, because the existential closure occurred too early. (40) a. [∃i [ S [ certaini [ … [ INDEFi NP …. ] ] ] ] ] ] Question available b. [∃ [ S [ certaini [ … [∃i [… [ INDEFi NP ….] ] ] ] ] ] ] ] Question not available Thus, an indefinite determiner can force widest scope by commenting on a question suggesting, for example, that the speaker knows the answer to the it. After all, if the existential closure were to occur at any lower position, the composition would collapse. Thus, only widest scope remains as an option. An indefinite determiner can rule out widest scope by lexically signaling that there is no question to be created in its syntactic environment. This is arguably what -nibud indefinites do in Russian, as suggested in Onea (2016). In other cases, a specialized indefinite may signal that the speaker cannot answer the question, if there is one, and leave open the possibility that there is no question in the first place—thus combining narrow scope requirements with speaker ignorance, as arguably the case for German irgend-indefinites. Thus, both scope control by epistemic and epistemically specific indefinites appears naturally captured. Onea (2016) remains vague on how the comments trigger scope effects, even if it may appear natural what the results ought to be. In Section 3, I provide some of the missing details by essentially suggesting that particles like German bestimmt/gewiss are actually modifiers of intensional operators that regulate the way in which questions derived from indefinites combine with these operators, thus resulting in high or low scope for the indefinite. 2.2.4 The Pragmatics of Indefinite Determiners The erotetic theory of indefiniteness may be attractive because it combines three aspects that are crucial for our understanding of indefinites: the scope variation of specialized indefinite determiners, the ability of indefinites to receive exceptional wide scope and the indefinite-interrogative affinity. Still, it is a natural question how and why such comments on questions of speci-
152
onea
fications might have grammaticalized. In this section, I provide a pragmatic explanation for the phenomenon. Building on observations already presented in van Kuppevelt (1995), Wisniewski (1995, 2013), Onea and Volodina (2011) and Onea (2016) I argue that, in general, assertions are not only used to answer questions, but they also raise questions in discourse. Questions raised by assertions can be considered possible or likely discourse continuations, whence their name potential questions in Onea (2016). Thereby, one can either overtly ask or mention such a potential question or, especially in monologues or texts, one can address an implicit potential question directly by answering it. In (41), from the movie Animal House, both cases are illustrated. In the first turn, Pinto starts out by saying that there is something he needs to tell his interlocutor. Such a claim naturally leads to the question what it is that he has to say. In the very next sentence, he answers this question without making the question itself explicit. In fact, by answering the implicit question a new question is raised; first it is stated that a lie occurred, and the answer to the question what the lie was about is given in the next sentence. Just a few lines later, Clorette de Pasto says that she lied too. This also naturally raises the question about the topic and content of her lie. This time, she does not answer the question in the same turn, and, little surprisingly, her interlocutor makes this question explicit in the very next turn. Thereby, example (41) exemplifies how an indefinite (something) raises an implicit specificational question. An example in which an indefinite raises an explicit specificational question is given in (42) from the very same movie. (41) Pinto: Clorette: Pinto: Clorette: Pinto:
Before we go any further, there’s something I have to tell you. I lied to you. I’ve never done this before. You’ve never made out with a girl before? No. No, I mean, I’ve never done what I think we’re gonna do in a minute. I sort of did once, but I was drunk … That’s okay, Larry. Neither have I. It’s my first time too. And besides, I lied to you, too. Oh, yeah? What about?
(42) Boon [to Otter]: […] What happened to you? You look grotesque. Otter: Some of the Omegas jumped me and did a little dancing on my face. Bluto: Who was it? Otter: It was Greggie and Douggie … and some of the other Hitler youth.
specificity and questions of specification
153
The way in which indefinites raise questions of specification, as (43a) and (43b), is special as compared to other potential questions assertions raise, as (43c)–(43e). Since default indefinites and interrogatives share their semantic contribution, the question of specification in (43a) is made more salient in the process of computation of the meaning of (43) than the potential questions in (43c–e). Therefore, I call such special potential question raised by indefinite determiners primary potential questions. (43) A student entered the room. a. Which student entered the room? b. Who entered the room? c. What did she do in the room? d. What did she find in the room? e. Why did she enter the room? Crucially, this does not mean that questions of specification must always be addressed or answered in discourse. Sometimes the speaker does not know the answer to the respective question of specification, sometimes it will just not be relevant to answer that question, given the conversational aims of the speaker in the sense of Roberts (2012a, 2012b). However, from the point of view of the addressee, the questions are highly salient. Hence, when using indefinites, the speaker is confronted with a dilemma. Either she may use an indefinite and continue the discourse as desired, thereby risking ignoring the expectations raised in the addressee that the primary potential question will be addressed in future discourse. Or she may immediately provide the necessary comment on this expectation. Such comments could be phrased as in (44a–f). Part of the problem with such comments is that they disrupt the flow of discourse and they require a significant effort from the speaker. Specialized indefinites come in precisely to encode such comments on questions as part of the grammatical system without providing disruption to the flow of discourse. As shown above, the use of a specific indefinite as in (45) is very similar to at least one of the sentences in (44d–f). (44) A student came in. a. A student came in, but I don’t know who it was. b. A student, don’t ask me who it was, came in. c. A student, I couldn’t care less who, came in d. A student came in, and I know who it was. e. A student, you may ask me who it was, came in. f. A student, it is highly relevant what student, came in.
154
onea
(45) A certain student came in. Of course, this does not explain the grammaticalization of individual specialized indefinite determiners and pronouns in detail. However, it seems that it may not be an accident that several markers of specificity stem from or are also used as positive answers to a question, such as certo/certainly and the other examples discussed above. Similarly, the Romanian specificity marker anume can also be used as the equivalent of the particle namely, as shown in (46), which according to Onea and Volodina (2011) marks the answer to a specificational question. On the other hand, markers of non-specificity are sometimes used or related to particles used to negatively answer a question, such as Russian nibud-indefinites, or even the Romanian default plural indefinite article niste from Latin ne scio. (46) a. Am întâlnit un anume student have met a anume student ‘I have met a certain student.’
(Romanian)
b. Am întâlnit un student. Anume Silviu. have met a student anume Silviu ‘I have met a student. Namely Silviu.’ 2.2.5 Interim Summary The main advantage of the erotetic theory of indefiniteness is that it provides uniform explanations to three crucial observations about indefinites: scope control by specialized indefinite determiners is explained by capitalizing on the morphological similarity between indefinites and wh-words and at the same time crucially using the information that scope control seems to be connected to some sort of epistemic specificity or non-specificity effect; the notion of commenting on a question of specification seems to provide quite exactly the right theoretical framework to capture these intuitions and comes with useful formal tools as well. Still, one crucial problem of the analysis is that it remains to be seen how exactly it can handle real data and how it competes with alternative approaches in the literature. In the next section I provide one step in addressing this gap.
specificity and questions of specification
3
155
Scope Control by Indefinites: A Case Study
In this section, I discuss the details of how comments on specificational questions can control scope for specialized indefinites using, as examples, the case of bestimmt and gewiss indefinites in German. One reason why these indefinites lend themselves for the analysis is that they are empirically well-described in the literature and there is a widely accepted semantic analysis on the market for both in Ebert, Ebert and Hinterwimmer (2013). Thus, we will first consider the main data, following Ebert, Ebert and Hinterwimmer (2013), consider their analysis of the two particles, and show how this analysis can be further refined using the erotetic theory of indefiniteness. It turns out that the changes needed lead to a fundamentally different approach to what these specialized indefinite determiners do at the syntax-semantics interface, especially regarding their interaction with intensional and speech act operators. At the same time, most of the main intuitions and ideas discussed in Ebert, Ebert and Hinterwimmer (2013) will be kept in the new theory as well. 3.1 Gewiss and Bestimmt Indefinites in German German comes with two quite similar indefinite determiners ein bestimmter and ein gewisser (roughly: ‘a certain’) that both give rise to specificity effects. The main semantic effects they give rise to are exemplified in two corpusexamples in (47). (47a) is only true if there is a pre-determined, particular time-budget such that if it is exceeded, controls will be necessary, i.e., a narrow scope reading for the indefinite is not available; moreover, the speaker most likely knows this pre-determined time-limit. (47b) shows a clear example of epistemic specificity: the very name of the referent under discussion is provided in the same sentence. (47) a. Wird ein bestimmtes Zeitbudget überschritten, ist ein will a bestimmt time-budget surpassed is a Kontrollverfahren vorgesehen. control-routine planed ‘If a certain time budget is surpassed, a control will take place.’ (A97/ JUN.09781 St. Galler Tagblatt, 17.06.1997)
156
onea
b. Damals trat ein gewisser Leutpriester namens W. L. als then stepped a gewiss parish-priest named W. L. as Zeuge auf. witness up ‘At that time, a certain parish priest names W.L. testified.’ (A99/ JAN.02183 St. Galler Tagblatt, 12.01.1999) Both the syntactic distribution and the major semantic effects associated with gewiss vs. bestimmt German indefinites have been discussed in detail in Ebert, Ebert and Hinterwimmer (2013). In what follows I will briefly summarize their findings here, adding some additional observations. For more details, the reader is referred to the original paper. It is also there that other usages of these expressions are discussed and compared to further cross-linguistic specificity markers. The examples used, except for the corpus examples, are either directly from Ebert, Ebert and Hinterwimmer (2013) or similar to them. 3.1.1 Syntax While both bestimmt and gewiss usually follow the indefinite article/determiner and precede the (rest of) the NP, Ebert, Ebert and Hinterwimmer (2013) suggest that they differ in two important respects: gewiss cannot be combined with quantifiers other than indefinites and numerals, whereas bestimmt can appear in quantificational examples such as (48a), but not examples such as (49). Crucially, in cases like (50), however, in which bestimmt is clearly an adjective because it is modified with a PP containing the external argument of the verbal participle bestimmt (‘determined’), it can appear even with such quantifiers. (48) a. Niemand sollte weniger als drei bestimmte CD s von nobody should less than three bestimmt CD s of Madonna besitzen. Madonna own ‘Nobody should own fewer than three CD’s by Madonna’. b. Niemand sollte weniger als drei (*gewisse) CD s von nobody should less than three gewiss CD s of Madonna besitzen. Madonna own ‘Nobody should own fewer than three CD s by Madonna.’
specificity and questions of specification
157
(49) Peter mag die meisten (*bestimmten) CD s von Madonna. Peter likes the most bestimmt CD s of Madonna ‘Peter likes most of the CDs of Madonna.’ (50) Die meisten vom Trainer bestimmten Spieler haben die the most from coach bestimmt-adj players have the Erwartungen erfüllt. expectations fulfilled ‘Most of the players that were chosen by the trainer have fulfilled the expectations.’ From these facts, they conclude that gewiss cannot be treated as an adjective and, thus, is better analyzed as part of a complex determiner. Bestimmt, on the other hand, appears to be Janus-faced to the extent that it can either act as an adjective, as in (50), or as part of a complex determiner, as in (49). I will have nothing to say about the adjectival usage of bestimmt in this paper. 3.1.2 Identifiability While both bestimmt and gewiss indefinites come with some epistemic specificity flavor, Ebert, Ebert and Hinterwimmer (2013) correctly argue that gewiss encodes identifiability by the speaker, whereas bestimmt only encodes identifiability by some other salient agent under some salient conceptual cover (as in Aloni 2001). The crucial data are provided in (51) and (52). In particular, (51) shows that speaker-identifiability is necessary for gewiss but not for bestimmt. (52), on the other hand, is acceptable even in a situation in which nobody on earth knows the code under discussion; still there is some conceptual cover under which this code can be identified (e.g., the one built in by some manufacturer), thus providing evidence for the generalization that the referent only needs to be identifiable under some (and not any) conceptual cover for bestimmt indefinites. (51) a. Gerd sucht schon lange nach einer bestimmten CD. Keine Gerd searches already long for a bestimmt CD no Ahnung nach welcher. idea after which ‘Gerd searches already for a long time after a certain CD. I have no idea which one.’
158
onea
b. Gerd sucht schon lange nach einer gewissen CD. #Keine Gerd searches already long for a gewiss CD no Ahnung nach welcher. idea after which ‘Gerd searches already for a long time after a certain CD. I have no idea which one.’ (52) Diese Tür öffnet sich nur bei einer bestimmten this door opens itself only by a bestimmt Zahlenkombination. code ‘This door opens only with a certain code.’ 3.1.3 Scope The scopal properties of bestimmt and gewiss differ quite substantially. In particular, Ebert, Ebert and Hinterwimmer (2013) argue that gewiss takes widest scope in any semantic configuration including interaction with negation, conditional, intensional operators and nominal quantifiers. I only give one example for illustration here, as this seems to be an uncontroversial point: (53) can only be true if everyone met the same diplomat (known by the speaker).9 (53) Jeder hat einen gewissen Diplomaten getroffen. everyone has a gewiss diplomat met ‘Everyone met a certain diplomat.’ As for bestimmt the situation is more complicated. Ebert, Ebert and Hinterwimmer (2013) argue that it can have both wide and narrow scope with regard to negation, conditionals and nominal quantifiers, but it cannot have narrowest scope if embedded under intensional operators. Their argument is nicely summarized by the example in (54), originally from Endriss (2009). In particular, (54) has three relevant readings: a) there is a horse such that if the kids want to ride it (widest scope), we have a problem; b) if there is a horse such that all the kids want to ride it (below conditional but above nominal quantifier and intensional operator), we have a problem; and c) if for each the children there is a particular horse they want to ride, we have a problem, i.e., the children are picky (below nominal quantifier but above intensional operator). Crucially, there is no reading according to which bestimmt is in the scope of want. 9 Using non-monotonic quantifiers instead of every would not change anything in the resulting conclusion.
specificity and questions of specification
159
(54) Wenn morgen wieder alle Kinder ein bestimmtes Pferd reiten if tomorrow again all children a bestimmt horse ride wollen, haben wir ein Problem. want have we a problem ‘If tomorrow all kids want to ride a certain horse again, we have a problem.’ It is important, however, to clarify that the cases in which bestimmt indefinites take narrow scope under negation and conditionals are not quite as unrestricted as suggested by Ebert, Ebert and Hinterwimmer (2013). In particular, bestimmt indefinites can take scope under negation or conditionals only if they simultaneously outscope some intensional operator in the scope of which they are base-generated. To see this, consider the example (55). It is not possible for bestimmt to get narrow scope in (55a), i.e., the only available reading is where ein bestimmter Mann has wide scope over negation. Accordingly, the reading paraphrased in (55a–i) is acceptable, whereas the reading paraphrased in (55a– ii) does not appear to obtain. Similarly, (55b) in which the narrow scope reading is syntactically enforced, the acceptability is strongly degraded for all consultants. As opposed to this, once an intensional operator enters the picture, as in (55c), the sentence is acceptable with ein bestimmter scoping under negation but above the intensional operator, as paraphrased in (55c–i). It is not possible for ein bestimmter to take scope between negation and a nominal quantifier, as in (55d), which lacks the reading paraphrased in (55d–i). (55) a. Es ist nicht der Fall, dass ein bestimmter Mann tanzt. it is not the case that a bestimmt man dance ‘It is not the case that a certain man is dancing.’ i. There is a man, such that it is not the case that he danced. ii. # It is not the case that there is a man such that he danced. b. ??Kein bestimmter Mann tanzt. neg:indef bestimmt man dance intended: ‘It is not the case that there is a certain dancing man.’ c. Alfred glaubt nicht, dass ein bestimmter Mann tanzt. Alfred believes not that a bestimmt man dances ‘Alfred does not believe that a certain man dances.’ i. It is not the case that there is a man such that Alfred believes that he danced.
160
onea
d. Es ist nicht der Fall, dass genau vier Studenten einen it is not the case that exactly four students a bestimmten Kurs besucht haben. bestimmt course attended have ‘It is not the case that exactly four students attended a certain course.’ i. #It is not the case that there is a course such that exactly four students visited it. In the case of conditionals, a similar argument can be made. We have witnessed for (54) how ein bestimmt NPs can scope under the conditional—but in this example too, there was an intensional operator. Removing it changes the resulting readings very clearly: in (56) the reading paraphrased in (56a) is unavailable, however the reading in (56b) is. (56) Wenn morgen wieder alle Kinder ein bestimmtes Pferd sehen, if tomorrow again all children a bestimmt horse see haben wir ein Problem. have we a problem ‘If tomorrow all kids see a certain horse, we have a problem.’ a. # If there is a horse, such that all children see that horse, we have problem. b. There is a horse such that if all children see that horse, we have problem. Thus, the correct generalization appears to be this: if there is no intensional operator dominating them, bestimmt indefinites take wide scope with respect to negation and conditionals but may end up scoping below nominal quantifiers. If, however, there is one intensional operator dominating the bestimmt indefinite, the bestimmt indefinite will take scope above that intensional operator thereby possibly scoping under any other operator including negation and conditionals. Finally, when there are several intensional operators dominating them, bestimmt indefinites will only need to outscope one of them, but they can outscope up to all intensional operators as well. One important point to note is that the claim that gewiss always takes scope over all operators overtly dominating it needs no further qualification. However, the situation for bestimmt is complicated by the availability of functional readings. As discussed above, a functional reading, exemplified in (57), is a special instance of a narrow scope reading, in which a function variable with wide scope is used to capture the semantic contribution of the indefinite. For our example, the functional variable is shown in (57a), and the contribution of the
specificity and questions of specification
161
second sentence is modelled in (57b) by identifying that function by the functional expression: mother. (57) Jede Professorin liebt eine bestimmte Frau. Und zwar ihre every professor loves a bestimmt woman and although her Mutter. mother ‘Every professor loves a certain woman. Namely her mother.’ a. ∃f.(∀x.P(x) → W(f(x)))∧∀y.P(y) → L(y, f(y)) b. f = λx.ιy.Mother(y, x) Starting from Groenendijk and Stokhof (1984), Hintikka (1986) and others various mechanisms have been proposed to capture functional readings. I will have nothing to say about such readings except for one crucial remark: we can assume wide scope for indefinites in these constructions. This is because the wide scope existential closure operator uses the restrictor material of the indefinite, i.e., the first universal quantification in (57a). Thus, functional wide scope readings can be analyzed as wide scope readings, albeit the indefinite itself is now ranging not over individuals but over functions. If this is correct, the crucial difference between the way bestimmt and gewiss interact with nominal quantifiers is this: both take wide scope over nominal quantifiers, however, gewiss does not operate on Skolem functions, whereas bestimmt has no restriction to this effect. 3.2 The Proposal of Ebert, Ebert and Hinterwimmer (2013) The analysis of Ebert, Ebert and Hinterwimmer (2013) for both gewiss and bestimmt is designed to capture the empirical generalizations they made. As I have somewhat complicated the empirical picture, we can expect some problems for the analysis. Nevertheless, it is important and instructive to go over some of the details of the analysis they provide, partly because they implement remarkably similar intuitions to what I will eventually propose in this paper. Ebert, Ebert and Hinterwimmer (2013) suggest that ein gewisser is a nearly regular existential quantifier, with the crucial difference that it conventionally implicates that a salient agent knows the referent under discussion under some salient conceptual cover. (58) ⟦ein gewiss⟧ = λP.λQ.∃x.[P(x)∧Q(x)] ⋅ Kα (↑n y)
162
onea
The intuitive idea is this: if a conventional implicature always has widest scope, following Potts (2005), at least for all practical purposes, gewiss indefinites will act as if they had widest scope as well. Moreover, since conventional implicatures are (usually) speaker oriented, we expect that speaker identifiability will follow, i.e., α = cS . (Crucially, conceptual covers will play not so much of a role here, as the speaker usually knows which cover she considers). How exactly does the proposal then derive widest scope from these assumptions? By showing that the conventional implicature associated with gewiss indefinites would fail to target the correct variable if the existential quantifier had any other scope except widest scope. For bestimmt, the semantics is quite similar, as shown in (59). The crucial difference is that the knowledge of the referent is now plain at issue content— thus, it follows that bestimmt indefinites have more unrestricted scope; after all any operator now can interact with the existential quantifier in the usual way. Moreover, since α is just a free variable, it can be either the speaker or some other salient individual that is responsible for the identification of the x. (59) ⟦ein bestimmt⟧ = λP.λQ.∃x.[P(x)∧ Kα (↑n x)∧Q(x)]
Importantly, Ebert, Ebert and Hinterwimmer (2013) attempt to capture the difference between bestimmt and gewiss with regard to scopal interaction with negation, suggesting that the former can take narrow scope because the Kα (↑n x) meaning component is plain at issue part of the truth condition (whence it can be negated), whereas in the case of gewiss it is a conventional implicature (whence it cannot be negated). An example is given in (60). Indeed, the predicted truth conditions for the variant of (60) containing bestimmmt are given in (61). (60) Ich habe keine bestimmte/*gewisse Person für die Stelle vorgesehen. I have no bestimmt/gewiss person for the job planned ‘I don’t have a particular person on mind for the job.’ (61) There is no person x, such that I know x in way Y, and I have x on mind for the job. While the proposal of Ebert, Ebert and Hinterwimmer (2013) has a significant number of undisputed virtues, it also suffers from a range of problems. I will discuss some of these problems one by one in the following. The critical considerations will be used to suggest possible improvements that will eventually be incorporated into the proposal in Section 3.3. In the discussion, I will ignore
specificity and questions of specification
163
the issue that the proposal in Ebert, Ebert and Hinterwimmer (2013) does not come well-equipped to handle exceptional scope, as this issue has already been discussed sufficiently above and is in no way specific to the semantics of these two specialized indefinites. Firstly, the identifiability of a variable under some conceptual cover appears to be too weak. Consider first, what Kα (↑n x) means. The definition in (62) suggest that if α knows x under cover n, this means that α has some unique notion of x without any further entailments. For example, this predicts that adding bestimmt or gewiss to any indefinite DP will never be informative, as long as it is warranted that the speaker knows each element of the restrictor in some interesting way (e.g., a (certain) PhD student of mine). In part, Ebert, Ebert and Hinterwimmer (2013) are aware of the problem and suggest that there needs to be some pragmatic connection between the actual issue described by scope and restrictor and the knowledge of the relevant agent; at the same time, the trivial conceptual cover provided by the sentence itself also needs to be ruled out—again using some pragmatic mechanism. While I do not wish to claim that such a pragmatic solution is excluded in principle, I certainly do not see how in detail it leads to the right predictions. (62) Kα (↑n x) = Kα (↑? yn .yn = x)
Reflect, for example, on a relatively simple case like (63a). Even if we only consider wide-scope readings, it is not clear how the proposed theory captures the intended intuition, that the speaker knows which friend of hers danced every day. Given that the speaker knows each friend of hers under some label, adding the predicate Kα (↑n x) will not change the truth conditions of the sentence at all. Moreover, if we slightly modify the example as in (63b); adding bestimmt would only predict that the friend of Max who danced is in the subset of Max’ friends’ the speaker knows—not that the speaker is talking about a particular individual, as would be factually correct. Put more abstractly, the problem is that the relevant property of the referent, e.g., being a friend of Max and having danced, are outside of the scope of the knowledge predicate; hence, the relevant agent can know the discourse referent in some way without being aware that she has the properties attributed to her by the very sentence the indefinite occurs in. Intuitively what we want is that the speaker or some salient agent should not only know who some referent x is, but, for our example, who of her friends danced every day. Arguably, for gewiss in (63c), the situation is less problematic. The reason it is the same person who knows the referent and the properties predicated to it by the sentence; after all, the speaker is committed to the truth of the sentence and, thereby, to believing its content. However, a
164
onea
technical detail remains problematic. Ebert, Ebert and Hinterwimmer (2013) suggest that in some way the variable y, corresponding to the referent introduced by one of Max’ friends still needs to have wide scope. To achieve that, y is argued to be identified with the witness of the existential statement: but the technical mechanism to do just that is not provided in sufficient detail in Ebert, Ebert and Hinterwimmer (2013).10 (63) a. Ein bestimmter Freund von mir hat jeden Tag getanzt. a bestimmt friend of mine has every day danced ‘Every day, a certain friend of mine danced.’ b. Ein bestimmter Freund von Max hat jeden Tag getanzt. a bestimmt friend of Max has every day danced ‘Every day, a certain friend of Max danced.’ c. Ein gewisser Freund von Max hat jeden Tag getanzt. a gewiss friend of Max has every day danced ‘Every day, a certain friend of Max danced.’ Secondly, Ebert, Ebert and Hinterwimmer (2013) forcefully argue that speaker identifiability cannot be part of the meaning of bestimmt, because it is possible to combine the non-specific irgendein with bestimmt. This is illustrated in (64), reminiscent of Ruys (2000): here, irgendein signals that the speaker does not have identifying knowledge about the relative of Michael who would need to die in order for him to inherit, but bestimmt also signals that not any relative would do, i.e., someone (e.g., Michael) would know the rich relative. Again, the intuitive diagnosis seems to be on the right track. However, there are two technical issues. Firstly, if ein bestimmter is a complex determiner amounting to an existential quantifier, as suggested by Ebert, Ebert and Hinterwimmer (2013), it is not obvious how one can compositionally combine the meaning 10
Ebert, Ebert and Hinterwimmer (2013) suggest that Potts (2005) has a solution for a similar problem on which they could capitalize in some way. In particular, example (i) arguably has a reading according to which most elderly people got home early and THEY heard Jackson. If this is true, indeed, there is a question which variable the non-restrictive relative clause comments on. (i) Most elderly, who heard Jackson, got home early. In my exegesis, however, Potts suggests that his theory cannot truly deal with the relevant reading (cf. Potts 2005: 126–127.) and to the extent that some people get this reading, some non-standard mechanism of interpretation is invoked. This then raises the question why gewiss would be a grammaticalized non-standard CI trigger.
specificity and questions of specification
165
of ein bestimmter with irgend(ein): either irgend would need to be an adjective combining with Verwandter or irgend would need to somehow modify a quantificational determiner, a quantifier or a proposition: all of which are at least uncommon options given the literature, cf. Kratzer and Shimoyama (2002). Secondly, the very example in (64) runs into the problems already mentioned: there is a salient agent, namely Michael, who—most likely—will be able to identify his own relatives. Hence, adding bestimmt seems to have no predicted function in this sentence. However, intuitively, the whole point of using bestimmt is to achieve wide scope over the conditional. It is not at all obvious how the theory would predict this. Again, the relevant piece of knowledge does not seem to be identifying relatives under some conceptual cover, but instead, the knowledge of the answer to the question: Who is such that if (s)he dies, Michael would inherit a fortune? (64) Wenn irgendein bestimmter Verwandter von Arne stirbt, erbt er if irgendein bestimmt relative of Arne dies inherits he ein Vermögen. a fortune ‘If a certain relative of Michael dies, he will inherit a fortune.’ Thirdly, it is not obvious what forces bestimmt to not have narrow scope under intensional operators. Ebert, Ebert and Hinterwimmer argue that the arising reading with narrow scope predicted by their theory is not sensible and, thus, pragmatically ruled out; therefore, only a wide scope reading remains. However, the predicted truth conditions are comprehensive—it just so happens that they are not a valid reading. An example is given in (65). Again, the intuitive problem seems to be that knoweldge of the individual, i.e., being able to identify the individual, is not sufficient to trigger scope effects. This is particularly important, because we have seen above, in Section 3.1.3, that scoping below negation and conditionals appears to be mediated by the presence of an intensional operator. The analysis in Ebert, Ebert and Hinterwimmer (2013) simply does not provide the analytical toolkit to capture this generalization (which they, indeed, appeared to have missed.) (65) Bea sucht einen bestimmten Freund von mir. Bea searches a bestimmt friend of mine ‘Bea is searching for a certain friend of mine.’ Predicted narrow scope reading: Bea is satisfied if there is a friend of the speaker such that Bea/the speaker can identify that friend and she finds that friend. True, e.g., if Bea is searching for any friend of the speaker.
166
onea
We now turn to a final problem: Ebert, Ebert and Hinterwimmer (2013) intended to allow the compositional import of bestimmt to be negated and thus allow bestimmt to scope under negation, as discussed around example (60) above. However, I have also shown above that this is an undesirable prediction. In fact, bestimmt can only scope under negation if it simultaneously scopes over an intensional operator; indeed vorsehen (‘have in mind’) in (60) is such an intensional operator the role of which is in no way captured by the theory. 3.3 The New Proposal In this section, I propose a theory of the German specialized indefinite determiners gewiss and bestimmt that improves on the proposal by Ebert, Ebert and Hinterwimmer (2013). In the first step, we briefly recap the main problems that need to be solved. In the second step, in Section 3.3.2, I propose a semantic analysis of gewiss, followed by an analysis of bestimmt in Section 3.3.3. The analysis of bestimmt will introduce performative speech act operators at LF. In Section 3.3.4, I return to gewiss and provide a unified analysis of gewiss and bestimmt, and briefly discuss some loose ends. 3.3.1 Goals The main problematic findings emerged from the discussion of Ebert, Ebert and Hinterwimmer (2013) are: Firstly, Ebert, Ebert and Hinterwimmer (2013) correctly—in my opinion—suggested that the specific meaning contribution of both bestimmt and gewiss is expressing that some agent knows the answer to some question. However, the question they suggested is a pure identificational question regarding the referent introduced by the indefinite. This seemed to be too weak a condition. Instead, we would want to have another question, which is gained by replacing the indefinite with a wh-word within some CP. Secondly, we have seen that the scope of gewiss is generally widest scope, which nicely correlates with the intuition that gewiss literally encodes speaker-identifiability of the referent under discussion. As opposed to this, the scope of bestimmt is more variable, which also correlates with the intuition that the person who may know the answer to the relevant question may be some other salient agent, for example. some individual explicitly mentioned in the sentence. Here, we generally agree with Ebert, Ebert and Hinterwimmer (2013). Thirdly, however, we have seen that the scope of bestimmt is not only variable but also exhibits a peculiar pattern: it seems that bestimmt indefinites have a specific way to interact with intensional operators. In particular, they need to scope over an intensional operator if there is one. This, of course, raises the question, what happens if there is no such operator.
specificity and questions of specification
167
The theory proposed will focus on the issues mentioned above. In the discussion of the semantics of both gewiss and bestimmt, I ignore the conceptual covers discussed in Ebert, Ebert and Hinterwimmer (2013). In essence, I agree with Ebert, Ebert and Hinterwimmer (2013) that conceptual covers are necessary, however, I also agree with Aloni (2001) that they are necessary in general for the semantics of questions. Thus, they are ultimately orthogonal to the semantics of gewiss and bestimmt. Similarly, I ignore functional readings. While I acknowledge that some stipulation is necessary to the extent that gewiss does not range over functional variables, the general problems functional readings come with go beyond the scope of this paper. 3.3.2 The Semantics of gewiss The erotetic theory of indefiniteness suggests that specialized indefinites provide a CP level comment on a question of specification derived from the indefinite. Hence, the actual LF structure involving gewiss-indefinites is not (66a–i) but (66a–ii). The comment introduced by gewiss itself seems to be the one in (66b); i.e., gewiss extracts a question from its propositional sister at LF and suggests that the speaker knows the answer to this question. (66) a. i. [CP… [ … indef gewiss NP … ] ] ii. [CP… [gewissi [ … [ … indefi NP … ] ] ] ] b. ⟦gewissi ⟧c = λp : K(cs , ?i p).p Thereby, we use Ebert, Ebert and Hinterwimmer (2013) idea that gewiss encodes the information that the speaker knows the answer to a question, however it is a different question we consider here, as shown schematically in (67) at an intuitive level. (67) a. Ein gewiss P Q. b. Ebert, Ebert and Hinterwimmer (2013): Who is x? Who is the P who Q?11 c. current proposal: Which P Q? At first glance, the difference may seem immaterial. However, at the technical level, the approaches lead to somewhat different prediction. Under the current proposal, gewiss signals that the speaker knows the answer to the question
11
In this formulation the P who Q has to be understood as a referential term, cf. Donnellan (1966).
168
onea
extracted from the syntactic sister of gewiss. This opens exactly two possibilities: either such a question can be extracted or not. If that question can be extracted, gewiss contributes the inference that the speaker knows the answer to it. If, however, the question cannot be extracted, the derivation will fail. Crucially, the question can only be extracted from p if existential closure has not yet been applied to the referential index of gewiss, i.e., if the indefinite itself has scope above gewiss. In other words, the fact that gewiss attempts to comment on a question of specification guarantees widest scope interpretation for gewiss. The only assumption needed is that if such a question cannot be extracted the computation collapses. Moreover, under this analysis two facts seem to follow naturally. Firstly, the theory naturally captures the grammaticalization path for gewiss-indefinites in German. Gewiss etymologically seems to stem from an old participle of the verb to know/to see (Old Germanic: wizzan) (Kluge 1899: 145); i.e., indefinites with gewiss originally may have been referring to known entities. Secondly, even in present-day German, the discourse particle gewiss can embed questions (especially under negation), as witnessed by (68)—just as other specialized indefinite particles, discussed above in Section 2.2.2. While a unified semantics for the discourse particle gewiss and the specialized indefinite gewiss is not the aim of this paper, the fact that they both semantically take a question as an argument highlights their commonalities. (68) Noch ist nicht gewiss, wer von den Bisherigen über die Klinge still is not gewiss who from the previous:N-pl above the blade springen muss. jump must ‘It is not yet obvious who of the previously mentioned will be sacrificed.’ (A00/AUG.57177 St. Galler Tagblatt, 28.08.2000) Unfortunately, an open issue remains. The proposal does not quite predict widest scope for gewiss-indefinites. Instead, we only predict that gewiss moves to a CP layer, albeit not necessarily to the root CP layer. There are a few ways in which we may attempt to solve the problem. One way would be to syntactically stipulate that gewiss needs to be interpreted within a root environment. For example, we might say that root environments have some specific syntactic projection, e.g., ForceP (cf. Truckenbrodt 2006 for German) that embedded CPs may lack. However, motivating such a syntactic rationale for gewiss to move to ForceP is not entirely trivial. Another approach would be a semantic contribution of gewiss as given in (69b), adapted to a syntactic configuration as in (69a). In this case, the assumption is that the speaker is represented in syntax (e.g., Speas and Tenny 2003) in root environments. Now, gewiss needs
specificity and questions of specification
169
to take an argument which is coreferential with the speaker of the speech act. However, this would not suffice either, because ForceP is not the only environment in which an argument co-referential with the speaker can be found at LF, hence even more elaborate work would need to be invested. (69) a. [CP…[speaker [gewiss [ … [ … indef NP … ] ] ] ] ] b. ⟦gewissi ⟧ = λp.λYy=cS : K(Y, ?i p).p Instead, we will solve this problem by assuming that gewiss directly takes a performative speech act operator as its argument. Hence, gewiss moves to a root CP for purely semantic reasons. How this can happen and why this would be a reasonable idea will be easier to see after the discussion of bestimmt in the next section. 3.3.3 The Semantics of bestimmt There are two core observations that constitute a natural starting point for investigating the meaning of bestimmt. The first one is that—as opposed to gewiss—the agent who knows the answer to the specificational question need not be the speaker, as argued in Ebert, Ebert and Hinterwimmer (2013). Secondly, bestimmt needs to outscope the first intensional operator that dominates it, i.e., it needs to induce a de re reading. The solution to the first problem appears to be a simple modification of the lexical entry so far developed for gewiss. The one presented in (70b), operating on a structure like (70a), appears to be sufficiently sophisticated for a start. (70) a. [CP… [ bestimmt [ … [ … indef NP … ] ] ] ] b. ⟦bestimmti ⟧ = λp : K(Y, ?i p).p where Y is a salient sentient agent. It might seem tempting to phrase the second issue as the question: how can a lexical entry force a de re interpretation with respect to one intensional operator? The most natural solution for such cases would appear to be some mechanism that forces the interpretation of the respective operator relative to the actual world @ instead of the world compositionally available at the respective level of embedding w, the world the respective intensional operator quantify over. However, this will not exactly lead to success in our case, because bestimmt does not require absolute de re readings that are interpreted with respect to the actual world: the only necessary thing is that bestimmtindefinites outscope one intensional operator. The indefinite can well be interpreted within the scope of other intensional operators, negation or conditionals as discussed in Section 3.1.3. Thus, it seems an inevitable conclusion that
170
onea
there must be a specific kind of semantic interaction between bestimmt and intensional operators. Before we can contemplate the nature of the semantic interaction between bestimmt and intensional operators, we need to briefly discuss how default indefinites in general can get narrow or wide scope with respect to an intensional operator in the first place. In standard theories, this issue has been solved in various ways. However, this question has not been discussed in Onea (2015, 2016) for the present compositional system. We will adopt the following strategy: assume that propositional intensional operators like believe take a partial proposition as an argument and pass all referential restrictions therein encountered to the mother context. This is shown in (71), where the constant B is a standard belief operator that operates on propositions; this is the case because the actual argument of B is not a partial propositionp, but p(h) a classical proposition. Thus, if there are several referential expressions within the scope of believe, whether or not their referential restrictions will be passed to the matrix context depends on whether or not the respective referential expressions have witnessed selective existential closure within the scope of believe. If that happened, the respective referential expression only has scope below believe, otherwise it has wide scope and results in a de re/specific reading. For the example (72), indef1 is interpreted with narrow scope, whereas indef2 is interpreted with wide scope. Notice that existential closure will transform a referential constraint propagated all the way up the tree into regular truth conditional content, hence the descriptive content of de re interpreted referential expressions will be interpreted outside the scope of believe and is available for further semantic interactions. (71) ⟦believe⟧ = λp.λx.λhc(h,x)∧c(h,p) .B(x(h), λw.pw (h)) (72) […[∃2 […[believe [∃1 […indef1 …indef2 …]]]]]…]
Hence, we could assume that bestimmt, as defined in (70b) appears immediately above an intensional operator, as in (73), and requires that its referential index allows the extraction of a question. This guarantees that the respective indefinite ends up having wide scope. However, it is not obvious how we could semantically motivate such a movement. After all, the type of the argument of bestimmt will in no way be specific for intensional operators (in other words, bestimmt cannot access believe from its position). Instead, we may assume that bestimmt has a semantic type that forces it to take an intensional operator as an argument. Then, bestimmt does not just move to CP, but, more precisely, it moves to a position at which it can semantically directly interact with an inten-
specificity and questions of specification
171
sional operator. There are two conceivable LF structures which could count as landing sites for bestimmt. These are given in (74). From a syntactic perspective, it seems that (74a) is a more natural position, hence I will go with this structure in what follows.12 The essential job of bestimmt is to make sure there was no existential closure for the respective referential index: this job is already sufficiently performed by the proposed entry in (70b), because if there had been an existential closure below bestimmt the resulting question could not be extracted. This seems to solve the problem how wide scope can be achieved over one intensional operator. (73) [bestimmt1 [believe […indef1 …]]]
(74) a. [believe [bestimmt1 […indef1 …]]] b. [[believe bestimmt1 ] […indef1 …]] A second problem arises, however. Ebert, Ebert and Hinterwimmer (2013) already point out that there are some issues if we want to interpret a question within the domain of an intensional operator. Specifically, if the person who ought to know the answer to the question is the speaker. Hence, a better solution seems to be to impose that the question commented on by bestimmt includes the interpretation of the intensional operator. In the same vein, the agent/experiencer of that intensional operator seems to be a very good candidate for knowing the answer to the respective question, but we will not hard wire this into the semantic of bestimmt, leaving it to general pragmatic principles to select the sentient entity who knows the answer to the question. This idea is implemented in (75). Accordingly, the structure and the result of the derivation for the example in (76) are given in (77): (75) ⟦bestimmti ⟧ = λp.λψ.λxK(Y,?iψ(p)(x)) .ψ(p)(x) tient agent.
12
where Y is a salient sen-
There may be several indefinites with various specialized indefinite determiners in the complement of believe. It is not exactly obvious how we could stack them below believe and keep their semantic types stable. In (i), for example, it is simply not obvious how bestimmt2 can take believe as an argument while having the same semantic type as bestimmt1 (i) [believe [bestimmt1 [bestimmt2 […indef1 …indef2 ….]]]] What might work, however, would be to assume some sort of a silent iterator that makes sure all particles gathered below believe will be jointly interpreted. We will not explore this technical exercise in this paper, however.
172
onea
(76) Greta glaubt, dass ein bestimmter Politiker lügt. Greta believes that a bestimmt politician lies. ‘Greta believes that a certain politician lies.’
(77) a. [∃[Greta [believe [bestimmti [indefi Politician lies]]]]] b. Greta believes that a politician lied; some salient agent knows of which politician Greta believes that he lies.
If we were to hard wire into the semantics that the agent of intensional verbs should know the answer to the specificational question, some intensional verbs such as the German counterparts of forget or being unaware would pose serious problems. Similarly, this may lead to too strong predictions for unembedded usages of bestimmt. Our theory seems to generalize to a range of intensional operators such as non-performative speech act verbs, know, surprise etc. However, some intensional operators are quite different as compared to believe. These include verbs like search or have in mind. Zimmermann (1993) argues convincingly that search usually takes a property as its argument, which naturally captures the narrow-scope, non-specific reading indefinites get as arguments of search. Hence, a sentence like (78) can be interpreted (78a). The way bestimmt was designed above, however, only allows it to combine with intensional predicates that take partial propositions as arguments. Moreover, bestimmt needs a referential index that properties lack. In order to solve this problem, we first consider how Zimmermann (1993) derives wide scope readings for indefinites under search. Since he assumes that search takes a property as its argument, he cannot simply QR the indefinite from the scope of search to get de re readings, as this would not create the necessary property for search to combine with. What he needs to do, in addition, is to create a property out of the trace. Indeed, he achieves this by transforming a variable into its essence, that can be written as in (78b). (78) Lou is searching for an article on colonial history. a. searchv (Lou, λw.a − o − c − hw ) b. ∃x.a − o − c − hv (x)∧ searchv (Lou, λw.λy.y = x) Can our theory replicate this strategy? In the current system, the plain syntactic argument of search is neither a proposition nor a property, but rather a function from partial assignments to individuals, i.e., of type: ⟨a, e⟩, as shown in (79a). If Zimmermann (1993) is correct in requiring that search semantically combines with a property, the question is how a property can be derived from
specificity and questions of specification
173
some denotation such as (79a). Since I do not wish to change the lexical entry for search provided by Zimmermann here, I will use a somewhat unorthodox operator that moves from the assignment-based system to a classical property for purposes of illustration. The operator itself is given in (79b), where Z is of type e. Applying this operator to (79a) yields the result in (79c), which is exactly the classical property of being an article on colonial history. Thus, we can use this strategy to provide two lexical entries for search. One for de re and one for de dicto reading. The two entries are shown in (80a) for de dicto readings and (80b) for de re readings. Now, inserting bestimmt below search will still suffer a type mismatch, as the type of search and believe is not exactly the same. There are several ways to deal with this problem, but for the purposes of this paper we will assume that bestimmt is just polymorphic. It seems that this is a reasonable assumption, as the actual semantic job performed by bestimmt in both cases would remain identical and the result is the desired one: bestimmt can only be combined with search in the de re reading, as otherwise a question cannot not be derived. (79) a. ⟦ani article on colonial history⟧ = λha−o−c−h(h(i)) .h(i) b. λx.λZ.∃h.c(h, x)∧x(h) = Z c. λZ.∃h.a − o − c − h(h(i))∧h(i) = Z
(80) a. ⟦search1 ⟧ = λx.λy.λh.S(y(h), λw.λZ.∃h.c(h, xw )∧ xw (h) = Z) de dicto b. ⟦search2 ⟧ = λx.λy.λhc(h,x) .S(y(h), λw.λZ.xw (h) = Z) de re Consider what predictions the theory makes as it stands. In doing so, we consider cases in which the bestimmt-indefinite is syntactically generated in a very low position and various combinations of operators that may dominate it. I list the most important cases in (81), where, for reasons of space, I only provide the structural elements and not real German sentences. As for (81a), the prediction is what we have discussed above, hence I will have nothing to add to this. As for (81b), the prediction comes from the fact that bestimmt needs to move immediately below the intensional operator. The predicted reading is indeed the only available one. For (81c), we see a certain level of flexibility: bestimmt needs to move to an intensional operator, but if there are two dominating it, it can move either to the lower or the higher one. If it moves to the lower one, the arising reading is the one in (81c–i), giving bestimmt intermediate scope, whereas if it moves to the higher one, we get widest scope for bestimmt as in (81c-ii). More problematic are the predictions in (81d) and (81e). In particular, in (81d) the predicted reading is indeed available, as discussed above in detail. However, there
174
onea
is another possible reading, namely the widest scope reading in (81d-ii), which is not predicted. The situation gets even worse in (81e), in which no intensional operator appears. The prediction is that the sentence has no reading whatsoever. In fact, however, a wide scope and only a wide scope reading for bestimmt exists, as discussed above. (81) a. structure: IOp bestimmt-NP prediction: bestimmt-NP > IOp correct b. structure: IOp Quant/Neg/Cond bestimmt-NP prediction: bestimmt-NP > IOp >Quant/Neg/Cond correct c. structure: IOp1 IOp2 bestimmt-NP correct prediction: i) IOp1 > bestimmt-NP > IOp2 ii) bestimmt-NP > IOp1 > IOp2 correct d. structure: Quant/Neg/Cond IOp bestimmt-NP prediction: Quant/Neg/Cond > bestimmt-NP > IOp available readings: i) Quant/Neg/Cond > bestimmt-NP > IOp ii) bestimmt-NP > Quant/Neg/Cond > IOp e. structure: Quant/Neg/Cond bestimmt-NP prediction: no interpretation available reading: bestimmt-NP > Quant/Neg/Cond In order to solve the remaining problems, I will postulate speech act operators such as assert in the root CP, arguably in ForceP. I will not attempt to discuss their ontological status or semantics here in detail, because such operators cannot operate in the same semantic domain as usual intensional operators (cf. already Szabolcsi 1982 and Krifka 2014), but rather they enact facts in the world. Modeling such facts in the world is beyond the ambitions of the current paper. However, there are two main points I wish to make: Firstly, it is one of the oldest arguments for the inclusion of performative speech act operators into the LF structure that some adverbials/particles modify such operators, as in (82), from Schreiber (1972), which have also a descriptive meaning with overt non performative speech act verbs as in (83). Bestimmt seems to be one more example of the class of expressions with this property. Moreover, the fact that bestimmt can act as an answer particle and it can, as an adverbial, modify speech act verbs, is further evidence that our analysis is on the right track. (82) Honestly, you are wrong. (83) Zaha told you honestly that you are wrong.
specificity and questions of specification
175
Secondly, if one attempts to spell out the meaning of such operators, one will need to represent states in the real world that are affected by speech act operators. This is exactly what Krifka (2014), Cohen and Krifka (2014) and subsequent work13 have proposed. In such proposals, operators like assert use some propositional object (not exactly a proposition) to modify so-called commitment spaces. The idea is that if the speaker asserts a proposition, she rules out commitment states in which she is not actually committed to the truth of that proposition. For Krifka (2014) commitment spaces contain commitment states which are represented as plain indices. This makes speech act operator very similar to intensional operators. The main semantic difference between speech act operators in Krifka’s theory and intensional operators in the current approach concerns the assignment argument, which is missing in Krifka’s analysis of speech act operators. Crucially, however, while Krifka does not discuss cross sentential anaphora, it should be clear that commitment states need to include discourse referents as well, and one natural way to do just that appears to be the addition of assignments to commitment spaces, reminiscent of world-assignment pairs in dynamic semantics. Thus, the type of speech act operators and intensional operators potentially could be unified (even if their workings may be very different). Thus, we can assume the general structure in (84). Then, leaving technicalities aside, bestimmt can do its usual job combining with speech act operators: it simply makes sure that a discourse referent is established, i.e., the resulting contextual assignment will contain the respective referential index and its identifying property. (84) [Ispeaker [ASSERT [bestimmt1 [….indef1 ….]]]]
If understood in such a way, bestimmt not only perfectly fits into the overall architecture of grammar advocated in Krifka (2014) and subsequent research and further empirically motivates such an architecture, but the particular technical spell-out proposed in this paper, according to which bestimmt makes sure that a referential index passes an intensional operator further motivates the necessity to combine classical dynamic semantics with commitment spaces not only as a general device for anaphora but also as a core constituent of the syntax-semantics interface. Acknowledging the missing details, let us establish the predictions of the system. Allowing bestimmt to move to assert, immediately solves the
13
See, e.g., Condoravdi and Lauer (2012) and Lauer (2013) for an alternative view.
176
onea
wrong predictions in (81d–e), since we now have an additional possible landing site for bestimmt, as shown in (85). (85) a. structure: SPact Quant/Neg/Cond IOp bestimmt-NP prediction: SPact > Quant/Neg/Cond > bestimmt-NP > IOp prediction: bestimmt-NP > SPact > Quant/Neg/Cond > IOp b. structure: SPact Quant/Neg/Cond bestimmt-NP prediction: bestimmt-NP > SPact> Quant/Neg/Cond
correct correct correct
However, even if we do not have an exact semantics for assert, it should be clear that the resulting interpretation for the example in (86a), ought to be something along the lines of (86b). The problem with (86b) is that the commitment status of the inference introduced by bestimmt appears to be missing, i.e. we do not know who is committed to the second conjunct. (86) Ein bestimmter Politiker hat gelogen. a certain politician has lied ‘A certain politician has lied.’ a. [∃[I [ASSERT [bestimmti [indefi politician lies]]]]] b. There is a politician such that the speaker is committed to the proposition that “He lied” and some salient agent knows of which politician the speaker is committed to the truth of the proposition that he lied. Fixing this problem goes beyond the ambitions of this paper, however it seems that for any presupposition and conventional implicature we will encounter similar problems: in those cases, the assert operator will not make it true that the speaker is committed to the truth of those propositions. Instead, they must already be part of the mutually accepted commitment space. This last step is needed to capture the intuition that, indeed, the inference introduced by bestimmt is not actually at issue, at least not if it is interpreted at the speech act level, but it still can interact with various operators, e.g., it can be negated if it is computed in association with a lower intensional operator, as observed in Ebert, Ebert and Hinterwimmer (2013). 3.3.4 The General Pattern: Specificity, Questions and Operators This section concludes the discussion of the German specialized indefinite determiners ein bestimmter and ein gewisser. We briefly take stock of our results and show how the semantics of the two particles can be further unified. Moreover, we consider some—potentially—general conclusions about specificity.
specificity and questions of specification
177
The key idea of the analysis of bestimmt was that bestimmt moves to an intensional or speech act operator and inserts a comment on a question of specification at that very syntactic location. At this point I show the meaning contribution of bestimmt only informally in (87). (87) a. … agent bestimmt1 OP p[indef NP1] … b. … there is an NP1 … Agent OP p[he1] AND it is presupposed that someone knows of which NP1 Agent OP p[he1] In particular, the OP in (87) can be a (performative) speech act verb or an intensional verb. The specificational question always includes the attitude of some relevant agent or the speaker towards some proposition. The comment itself is—depending on further factors—either at-issue content or a presupposition; it is regularly a presupposition if bestimmt associates with a performative speech act operator in ForceP. One natural question is whether we can extend this kind of semantic analysis to gewiss? At the end of Section 3.3.2. I raised the question why gewiss would be forced to move to a root CP. One natural answer—in light of our analysis of bestimmt—seems to be this: gewiss also needs to take an intensional operator as an argument, however, as opposed to bestimmt, this operator needs to be a performative speech act operator. Gewiss, then, seems to be a more specialized version of bestimmt. (88) a. [speaker [gewiss1 [SPact p[indef NP1]]]] … b. ⟦gewissi ⟧ = λp.λψSPact λxK(x,?iψ(p)(x)) .ψ(p)(x) b. There is an NP1 such that speaker SPact p[he1] AND speaker knows of which NP1 speaker SPact p[he1] Generalizing the analysis to gewiss is a parsimonious but not a spectacular move. The specialized indefinite determiner which is theoretically more interesting is bestimmt. The question naturally raised at this point is whether bestimmt is a special case within the general domain of specialized indefinite determiners or whether it is in fact a prototypical one. To my knowledge, the literature does not report similar distributional patterns. However, as we have seen with Ebert, Ebert and Hinterwimmer (2013), such distributional facts easily go unnoticed even in remarkably detailed papers. So, I can at least imagine that on closer scrutiny, similar distributional patterns will be found and, thus, similar analyses will be given to other specialized indefinite determiners as well. After all, the very fact that a lot of specialized indefinite determiners appear to have morphological similarity with not only wh-words but also with high adverbials/particles, could be suggestive of similar semantic pat-
178
onea
terns. Potentially, then, specificity could be related to intensionality in a way essentially already envisioned in Ioup (1977) and Zimmermann (1993), i.e., wide scope over intensional (or speech act) operators could be the essence of specificity (or at least of a large chunk of it). Another problem that I mentioned regarding the theory in Ebert, Ebert and Hinterwimmer (2013) is the coocurrence of irgendein and bestimmt, cf. example (64). Since I have no theory of irgendein to offer in this paper, the discussion of this point remains brief, but the point is an important one and cannot be ignored entirely. After all, in other languages co-occurrence of several specialized indefinite pronouns has been observed, in fact some that have quite different epistemic effects. Another example already mentioned in this paper, from Italian, is (11). In most competing theories, specialized indefinite determiners have a local semantic contribution such that they are either determiners themselves or they are items that directly combine with indefinite determiners. In either case, combining several of them is usually difficult. In the present theory, however, all specialized indefinites have in common that they consist of two parts: the default indefinite (article) and a particle-like element. Combining several specialized indefinite determiners now simply boils down to distributing several indefinite determiners along the LF outside the DP itself. Just like stacking several discourse particles in a sentence, this does not lead to compositional problems whatsoever. There are only two constraints. First, their meaning contributions (in terms of the comments they impose on questions of specification) are logically compatible. This is usually not a problem, since the questions they comment upon can differ substantially given different interpretations of questions in terms of conceptual covers. The second constraint is that they be located at LF spots that are compatible with the existential closure of the respective indefinite. Thus, a particle like bestimmt can only appear below the existential closure and irgend (arguably) only above the existential closure. This predicts that in a plain extensional context irgend and bestimmt cannot co-occur within one DP. This prediction is confirmed by the oddity of (89). In (89) bestimmt needs to associate with the performative speech act operator and irgendein would need to appear even higher at LF. However, there is no higher landing site than the speech act operator. Hence, the current approach seems to provide the necessary flexibility to handle the co-occurrence of several specialized indefinite determiners within one DP. A detailed study of such examples, however, is a natural task for further research. (89) #Irgendein bestimmter Professor raucht vor der Tür. irgend bestimmt professor smokes in-front-of the door Intended: ‘A specific non-specific professor is smoking outside.’
specificity and questions of specification
179
At the same time, some skepticism is also in order. The claim that performative speech act verbs can interact in the desired way with bestimmt is not yet fully implemented. Future research should also experimentally validate the generalizations I provided for German, as the judgments are subtle, and some variation probably exists. Moreover, the interaction of specificity markers with intensional verbs such as forget or be unaware, which have negative inferences regarding the subject’s attitude towards the complement clause should be explored. Finally, I have not considered the factivity of some intensional verbs that can combine with bestimmt and I ignored intensional operators that do not involve an attitude holder such as it is possible, can, must etc. The theory needs further elaboration to be able to capture the interaction of such operators with bestimmt and gewiss.
4
Concluding Discussion
This paper consisted of two main parts. In the first part, Section 2, I introduced and motivated the erotetic theory of indefinites on three independent empirical grounds. I suggested that the erotetic theory of indefiniteness is wellsuited to capture the indefinite-interrogative affinity, it captures the connection between scopal effects and epistemic effects associated with specialized indefinite determiners and, finally, the compositional system it is couched in is suited to capture exceptional scope for indefintes because it inherently disregards scope islands. In the second part, Section 3, a case study was presented in which I showed in detail how the erotetic theory can capture the intricacies of scope taking of bestimmt-indefinites in German alongside their epistemic effects. This way, I have demonstrated how the erotetic theory of indefinites can handle scopal effects—as claimed but not technically shown in Onea (2016). In doing so, the hypothesis emerged that specific indefinites may generally associate with performative speech act verbs or intensional verbs. Showing that alongside the conceptual advantages of the theory, it can also be used to provide precise and detailed empirical predictions regarding scope and epistemic effects associated with indefinites, should be a good argument in favor of the erotetic theory of indefinites as a general framework. However, there are many open issues as well. Some I mentioned at the end of Section 3. But other issues also remain. For example, it remains to be seen whether such a theory can be applied to epistemic indefinites (maybe along the lines suggested in Onea 2016), whether further evidence can be found for the idea that specialized indefinite determiners partly raise to CP and whether further empirical support can be found for the crucial claim that indefinites raise ques-
180
onea
tions of specification in a particularly prominent way. In the same vein, studies of the diachronic development of specialized indefinite determiners will likely shed light on how exactly specialized indefinite determiners grammaticalize and whether or not these processes are compatible with the predictions of the erotetic theory of indefiniteness.
Acknowledgments Crucial ideas in this paper, especially those presented in Sections 3.1.3 and 3.3.2, are a result of insightful discussions with Ede Zimmermann, which I gratefully acknowledge. Similarly, I gratefully acknowledge that comments provided by Klaus von Heusinger, Ede Zimmermann and Malte Zimmermann on earlier work on this topic in Onea (2013) helped significantly improve the ideas presented in this paper. Furthermore, Klaus von Heusinger provided important comments on several versions of this manuscript. I am also indebted to Keir Moulton for comments on my manuscript. Finally, I am grateful to Mats Rooth and Floris Roelofsen for valuable input in early stages of this work. All remaining errors in this paper are, however, my sole responsibility and should in no way be connected to any of the mentioned scholars. Work on this paper was in part funded by the Austrian Science Fund as part of the grant i 4858 - “The erotetic and the aesthetic”, which I gratefully acknowledge.
References Abusch, Dorit. 1994. The scope of indefinites. Natural Language Semantics 2. 83–135. Aloni, Maria. 2001. Quantification under conceptual covers. Amsterdam: University of Amsterdam PhD dissertation. Aloni, Maria. 2022. Indefinites as fossils: The case of wh-based free choice. In Chiara Gianollo, Klaus von Heusinger & Maria Napoli (eds.), Determiners and quantifiers: Functions, variation, and change, 214–245. Leiden: Brill. Aloni, Maria & Angelika Port. 2010. Epistemic indefinites crosslinguistically. In Yelena Fainlein, Nicholas LaCara & Yangsook Park (eds.), Proceedings of NELS 41, 29–42. Aloni, Maria & Angelika Port. 2015. Epistemic indefinites and methods of identifications. In Luis Alonso-Ovalle & Paula Menéndez-Benito (eds.), Epistemic indefinites, 117–140. Oxford: Oxford University Press. Aloni, Maria & Robert van Rooij. 2007. Free choice items and alternatives. In Gerlof Bouma, Irene Krämer & Joost Zwarts (eds.), Cognitive foundations of interpretation. Edita KNAW.
specificity and questions of specification
181
Alonso-Ovalle, Luis & Paula Menéndez-Benito. 2010. Modal indefinites. Natural Language Semantics 18(1). 1–31. Alonso-Ovalle, Luis & Paula Menéndez-Benito. 2015. Epistemic indefinites: An overview. In Luis Alonso-Ovalle & Paula Menéndez-Benito (eds.), Epistemic indefinites. Exploring modality beyond the verbal domain, 1–27. Oxford: Oxford University Press. Beaver, David & Emiel Krahmer. 2001. A partial account of presupposition projection. Journal of Logic, Language and Information 10. 147–182. Bhat, Darbhe N.S. 2004. Pronouns. Oxford: Oxford University Press. Brasoveanu, Adrian & Donka Farkas. 2011. How indefinites choose their scope. Linguistics and Philosophy 34. 1–55. Breheny, Richard. 2003. Exceptional-scope indefinites and domain restrictions. In Matthias Weisgerber (ed.), Proceedings of the Sinn und Bedeutung 7. Arbeitspapier Nr. 114, 38–52. Konstanz: FB Sprachwissenschaft. Charlow, Simon. 2014. On the semantics of exceptional scope. New York, NY: New York University PhD dissertation. Chierchia, Gennaro. 1998. Reference to kinds across languages. Natural Language Semantics 6. 339–405. Chierchia, Gennaro. 2013. Logic in grammar. Oxford: Oxford University Press. Ciardelli, Ivano, Floris Roelofsen & Nadine Theiler. 2017. Composing alternatives. Linguistics and Philosophy 40(1). 1–36. Ciardelli, Ivano, Jeroen Groenendijk & Floris Roelofsen. 2019. Inquisitive semantics. Oxford: Oxford University Press. Cohen, Ariel & Manfred Krifka. 2014. Superlative quantifiers and meta-speech acts. Linguistics and Philosophy 37. 41–90. Condoravdi, Cleo & Sven Lauer. 2012. Imperatives: Meaning and illocutionary force. In Christopher Piñón (ed.), Proceedings of Empirical issues in syntax and semantics 9, 37–58. Csirmaz, Aniko & Anna Szabolcsi. 2012. Quantification in Hungarian. In Edward Keenan & Denis Paperno (eds.), Handbook of quantifiers in natural language, 399– 465. Dordrecht: Springer. Donnellan, Keith S. 1966. Reference and definite descriptions. Philosophical Review 77. 281–304. Ebert, Cornelia, Christian Ebert & Stephan Hinterwimmer. 2012. The semantics and pragmatics of bestimmt and gewiss. In Seda Khan, Claire Moore-Cantwell & Robert Staubs (eds.), Proceedings of the Annual Meeting of the North East Linguistic Society 40, 175–190. Ebert, Christian, Claudia Ebert & Stefan Hinterwimmer. 2013. German specificity markers: bestimmt vs. gewiss. In Christian Ebert, Cornelia Ebert & Stefan Hinterwimmer (eds.), Different kinds of specificity across languages, 31–75. Dordrecht: Springer.
182
onea
Endriss, Cornelia. 2009. Quantificational topics: A scopal treatment of exceptional wide scope phenomena. Dordrecht: Springer. Eremina, Olga. 2012. The semantics of Russian indefinite pronouns: Scope, domain widening, specificity and proportionality and their interaction. East Lansing, MA: Michigan State University PhD dissertation. Fălăuș, Anamaria. 2009. Polarity items and dependent indefinites in Romanian. Nantes: University of Nantes PhD dissertation. Farkas, Donka. 1994. Specificity and scope. In Lea Nash & George Tsoulas (eds.), Proceedings of Langues et Grammaire 1, 119–137. Paris: Université Paris 8. Farkas, Donka. 2002. Extreme non-specificity in Romanian. In Claire Beyssade (ed.), Romance Languages and Linguistic Theory 2000, 127–151. Amsterdam: Benjamins. Fodor, Janet & Ivan Sag. 1982. Speaker’s reference and quantificational indefinites. Linguistics and Philosophy 5. 355–398. Giannakidou, Anastasia & Josep Quer. 2013. Exhaustive and non-exhaustive variation with free choice and referential vagueness: Evidence from Greek, Catalan, and Spanish. Lingua 126. 120–149. Gianollo, Chiara. 2018. Indefinites between Latin and Romance. Oxford: Oxford University Press. Geurts, Bart. 2010. Specific indefinites, presupposition and scope. In Rainer Bäuerle, Uwe Reyle & Thomas Ede Zimmermann (eds.), Presuppositions and Discourse: Essays offered to Hans Kamp, 125–158. Bingley, UK: Emerald. Ginzburg, Jonathan. 1992. Questions, queries and facts: A semantics and pragmatics for interrogatives. Stanford, CA: Stanford University PhD dissertation. Groenendijk, Jeroen & Martin Stokhof. 1984. Studies on the semantics of questions and the pragmatics of answers. Amsterdam: University of Amsterdam PhD dissertation. Groenendijk, Jeroen & Martin Stokhof. 1991. Dynamic predicate logic. Linguistics and Philosophy 14. 39–100. Gutzmann, Daniel. 2013. Expressives and beyond: An introduction to varieties of useconditional meaning. In Daniel Gutzmann & Hans-Martin Gärtner (eds.), Beyond expressives: Explorations in use-conditional meaning, 1–58. Leiden: Brill. Hamblin, Charles L. 1973. Questions in Montague English. Foundations of Language 10. 41–53. Haspelmath, Martin. 1997. Indefinite pronouns. Oxford: Clarendon Press. Heim, Irene. 1982. The semantics of definite and indefinite noun phrases. Amherst, MA: University of Massachusetts PhD dissertation. Heim, Irene & Angelika Kratzer. 1998. Semantics in generative grammar. Oxford: Blackwell. Heim, Irene. 2011. Definiteness and indefiniteness. In Klaus von Heusinger, Claudia Maienborn & Paul Portner (eds.), Semantics: An international handbook of natural language meaning, vol. 2, 996–1025. Berlin: de Gruyter.
specificity and questions of specification
183
von Heusinger, Klaus. 1997. Salienz und Referenz. Der Epsilonoperator in der Semantik der Nominalphrase und anaphorischer Pronomen. Berlin: Akademie Verlag. von Heusinger, Klaus 2002. Relative specificity. In Graham Katz, Sabine Reinhard & Philip Reuter (eds.), Proceedings of Sinn und Bedeutung 6, 115–131. Osnabrück: University of Osnabrück. von Heusinger, Klaus. 2011. Specificity. In Klaus von Heusinger, Claudia Maienborn & Paul Portner (eds.), Semantics: An international handbook of natural language meaning, vol. 2, 1024–1057. Berlin: de Gruyter. Hintikka, Jaakko. 1986. The semantics of a certain. Linguistic Inquiry 17. 331–336. Ioup, Georgette. 1977. Specificity and the interpretation of quantifiers. Linguistics and Philosophy 1. 233–245. Jäger, Gerhard. 2007. Partial variables and specificity. In Uli Sauerland & Penka Stateva (eds.), Presupposition and implicature in compositional semantics, 121–162. New York: Palgrave-McMillan. Kamp, Hans. 1981/2013. A theory of truth and semantic representation. In Klaus von Heusinger & Alice ter Meulen (eds.), The Dynamics of meaning and interpretation. Selected papers of Hans Kamp, 329–369. Leiden: Brill. Reprinted from Jeroen Groenendijk, Theo M.V. Janssen and Martin Stokhof (eds.), Formal methods in the study of language, Proceedings of the Third Amsterdam Colloquium, Mathematical Center, Amsterdam, 1981, 277–322. Kamp, Hans & Ágnes Bende-Farkas. 2001. Indefinites and binding: From specificity to incorporation. Lecture Notes. University of Stuttgart. Kamp, Hans & Ágnes Bende-Farkas. 2019. Epistemic specificity from a communication theoretic perspective. Journal of Semantics 36(1). 1–51. Kaplan, David. 1978. Dthat. In Peter Cole (ed.), Syntax and semantics 9, 221–243. New York: Academic Press. Karttunen, Lauri. 1977. Syntax and semantics of questions. Linguistics and Philosophy 1. 3–44. Kellert, Olga. 2022. The evaluative meaning of the indefinite qualunque in (Old) Italian. In Chiara Gianollo, Klaus von Heusinger & Maria Napoli (eds.), Determiners and quantifiers: Functions, variation, and change, 246–284. Leiden: Brill. Kluge, Friedrich. 1899. Deutsches etymologisches Wörterbuch. Strassburg: Trübner. Kratzer, Angelika. 1998. Scope or pseudo-scope? Are there wide-scope indefinites? In Susan Rothstein (ed.), Events in grammar, 163–196. Dordrecht: Kluwer. Kratzer, Angelika & Junko Shimoyama. 2002. Indeterminate phrases. The view from Japanese. In Yukio Otsu (ed.), Proceedings of the third Tokyo conference on psycholinguistics, 1–25. Tokyo: Hituzi Syobo. Krifka, Manfred. 2001. For a structured meaning account of questions and answers. In Caroline Féry & Wolfgang Sternefeld (eds.), Audiatur vox sapientiae. A Festschrift for Arnim von Stechow, 287–319. Berlin: Akademie Verlag.
184
onea
Krifka, Manfred. 2011. Questions. In Klaus von Heusinger, Claudia Maienborn & Paul Portner (eds.), Semantics: An international handbook of natural language meaning, vo. 2, 1742–1785. Berlin: de Gruyter. Krifka, Manfred. 2014. Embedding illocutionary acts. In Thomas Roeper & Margaret Speas (eds.), Recursion, complexity in cognition, 125–155. Berlin: Springer. van Kuppevelt, Jan. 1995. Discourse structure, topicality and questioning. Journal of Linguistics 31. 109–147. Lauer, Sven 2013. Towards a dynamic pragmatics. Stanford, CA: Stanford University PhD dissertation. Onea, Edgar. 2013. Potential questions in discourse and grammar. Göttingen: University of Göttingen Habilitation thesis. Onea, Edgar. 2014. Indefinite donkeys on islands. In Todd Snider (ed.), Proceedings of SALT 23. 493–513. Onea, Edgar 2015. Why indefinites can escape scope islands. Linguistics and Philosophy 38. 237–267. Onea, Edgar. 2016. Potential questions at the semantics-pragmatics interface. Leiden: Brill. Onea, Edgar & Ljudmila Geist. 2011. Indefinite determiners and the pragmatics of referential anchoring. International Review of Pragmatics 3. 194–227. Onea, Edgar & Anna Volodina. 2011. Between specification and explanation. About a German discourse particle. International Review of Pragmatics 3. 3–32. Potts, Christopher. 2005. The logic of conventional implicatures. Oxford: Oxford University Press. Reinhart, Tanya. 1997. Quantifier scope: How labour is divided between QR and choice functions. Linguistics and Philosophy 20. 335–397. Roberts, Craige. 2012a. Information structure in discourse: Towards an integrated formal theory of pragmatics. Semantics and Pragmatics 5(6). 1–69. Roberts, Craige. 2012b. Information Structure: Towards an integrated formal theory of pragmatics. Afterword. Semantics and Pragmatics 5(7). 1–19. Ruys, E.G. 2000. Weak crossover as a scope phenomenon. Linguistic Inquiry 31(3). 513– 539. Schreiber, Peter A. 1972. Style disjuncts and the performative analysis. Linguistic Inquiry 3. 321–347. Schwarz, Bernhard. 2001. Two kinds of long-distance indefinites. Manuscript, University of Texas, Austin, TX. Schwarzschild, Roger. 2002. Singleton indefinites. Journal of Semantics 19. 289–314. Speas, Peggy & Carol L. Tenny. 2003. Configurational properties of point of view roles. In Anna Maria Di Sciullo (ed.), Asymmetry in grammar, 315–344. Amsterdam: Benjamins. Szabolcsi, Anna. 1982. Model theoretic semantics of performatives. In Ferenc Kiefer (ed.), Hungarian linguistics, 515–535. Amsterdam: Benjamins.
specificity and questions of specification
185
Szabolcsi, Anna. 2010. Quantification. Cambridge: Cambridge University Press. Truckenbrodt, Hubert. 2006. On the semantic motivation of syntactic verb movement to C in German. Theoretical Linguistics 32. 257–306. Ultan, Russell. 1978. Some general characteristics of interrogative systems. In Joseph H. Greenberg, Charles Albert Ferguson & Edith A. Moravcsik (eds.), Universals of human language, 211–248. Stanford, CA: Stanford University Press. van Valin, Robert D., Jr. 1992. A synopsis of role and reference grammar. In Robert D. van Valin, Jr. (ed.), Advances in Role and Reference Grammar, 1–164. Amsterdam: Benjamins. Winter, Yoad. 1997. Choice functions and the scopal semantics of indefinites. Linguistics and Philosophy 20. 399–467. Wisniewski, Andrzej. 1995. The posing of questions: Logical foundations of erotetic inferences. Dordrecht: Springer. Wisniewski, Andrzej. 2013. Questions, inferences, and scenarios. London: College Publications. Yeom, Jae-Il. 1998. A presuppositional analysis of specific indefinites: Common grounds as structured information states. New York: Garland. Zimmermann, Thomas Ede. 1993. On the proper treatment of opacity in certain verbs. Natural Language Semantics 1. 149–179. Zimmermann, Thomas Ede. 2006. Monotonicity in opaque verbs. Linguistics and Philosophy 29. 715–761.
chapter 6
Being Bare: A Survey of Quantifier Positions Cecilia Poletto
1
Introduction
The aim of the chapter is to provide evidence based on empirical data coming from old and modern Italian varieties that bare quantifiers (BQ s) occupy positions that are different from those of quantified nominal expressions and that this is due to the extremely impoverished internal layering of the nominal category paired to the BQ. Beghelli and Stowell (1997) in their seminal work on quantifiers already propose that different types of quantifiers can have different positions in the clause. They build on an intuition by May (1985) that the scope domain of a quantifier corresponds to its c-command domain, and distinguish five types of quantifiers showing that each type has different scopal properties and a different position. However, they do not make any real structural distinction between BQs and quantified nominal expressions. The same view is represented by other work on Romance, for instance Doetjes (1997) and Baunaz (2011), who assume that BQs are associated to a pro, which has the same categorial makeup as a full nominal expression. Differences found between the position of BQ s and that of quantified nominal expressions have been indeed noticed in the literature, but they have been attributed to the fact that BQs can be weak elements (see for instance the treatment of quantifiers like French tout / tous / rien in Cardinaletti and Starke 1999: 61). In line with Garzonio and Poletto (2017), I will argue for a reduced internal structure of BQs, which is not, as Cardinaletti and Starke propose, the pruning of the higher functional projections, but on the contrary, is deficient in its lower lexical part, an assumption which is in line with the morphological makeup of these elements. As proposed by Garzonio and Poletto (2017) I will assume that a BQ can at most contain a classifier-like noun as its lowest element and has no lexical layer at its bottom. BQs do not contain any real root category which can represent a real lexical noun. This will fall in line with the long standing proposal (first put forth by Katz and Postal 1964) that wh-pronouns have an internal ‘light’ nominal category, which I argue is akin to sortal classifiers, i.e. the structurally lowest type of classifier in languages where classifiers are
© Cecilia Poletto, 2022 | doi:10.1163/9789004473324_007
being bare: a survey of quantifier positions
187
lexically realized like Chinese (see also those light nouns like body or thing in quantifiers like somebody/anybody/nobody or something/anything/nothing), or in some cases a higher classifier corresponding to one (like in someone/anyone/no one).1 This might have important consequences on the standard idea that all functional elements are located on top of a lexical layer because they represent extended projections of lexical categories and more generally of what the internal layering of pronouns is. If BQs do not contain a lexical layer, this means that functional structure can be generated without a lexical layer at its bottom, which undermines the intuition behind the idea of extended projections that has been overtly or implicitly assumed in the syntactic literature. It might even suggest that what we call the ‘lexical category’ is paired to the instruction to tap into the lexical component outside the syntactic structure, which is most probably fixed and not contained in the actual numeration. However, the syntactic structure can also work without tapping into the lexicon at certain points, and this would be the basic property putting together all elements that are usually considered to have an internal variable like wh-items and quantifiers. In a parallel way to wh-items, I argue that all a BQ requires to be interpreted is a classifier, which represents the most minimal restrictor the quantifier c-commands. The classifier element I am referring to is the ‘sortal’ one of the type PERSON, THING, PLACE, TIME, WAY proposed by Kayne in a number of articles starting from his work on here and there in 2004 and adopted by other authors (see for instance the analysis of pronouns proposed by Leu 2014). I provide here the internal composition I propose for BQs in opposition to regular quantified nominal expressions: (1) a. [QP tutto [ClassP THING]] b. [QP tutto [DP D° …[NP N]]] If there is not even a classifier in its internal structure, the BQ cannot quantify over a void and must move to a position which allows it to c-command onto a subtree that contains at least the vP, which is the other possible domain of quantification.2
1 The element one could be located in what Zamparelli (2008) dubs PluralP and Borer (2005) ClP which distinguishes count nouns from mass nouns. 2 It might be that quantificational domains correspond to entire phases and no smaller category, but this is a claim that has to be proved on the basis on a more comprehensive set of data, therefore I leave this issue aside in the present article.
188
poletto
The article is organized as follows: in Section 2 I show why the position of BQs cannot be due to the weak status of these elements and summarize the difference between BQs which contain a classifier, like niente, and those that do not, like tutto. Section 3 is devoted to illustrating that various types of nominal expressions occur in specialized layers or circuits at different heights in the sentence structure where they display a fixed word order Subject-Indirect object-Direct object. In Section 4 I examine data from modern standard Italian which displays a different rigid word order of BQ s. In Section 5 I discard a plausible solution and in Section 6 I analyze the fact that in modern Italian the direct object BQ is higher than the other arguments as due to a problem of projection.
2
BQs Are Not Weak Pronouns
It is well known in the literature that quantifiers can have special positions which are not accessible to regular DPs. Cases of this type have been noticed in the literature, here I report the case of Icelandic studied in Svenonius (2000), a regular VO language, which still tolerates (or even requires) OV orders with the object in front of the past participle when the object is a quantified nominal expression: (2) a. Strákarnir höfðu hent miklu grjóti í bílana. the boys had thrown a lot of stones to the cars b. Strákarnir höfðu miklu grjóti hent í bílana. the boys had a lot of stones thrown to the cars ‘The boys had thrown many stones at the cars.’ (Svenonious 2000: 261) c. Strákarnir höfðu hent engu grjóti í bílana. the boys had thrown no stones to the car d. Strákarnir höfðu engu grjóti hent í bílana. the boys had no stones thrown to the cars ‘The boys had thrown no stones at the cars.’ (Svenonious 2000: 260) It is also well known since Kayne (1975) that in languages like French BQ s occupy positions that are different from those of quantified nominal expressions. French still displays a residue of OV order with BQ s, which must occur in front of the past participle unless the quantifier is focused:
being bare: a survey of quantifier positions
189
(3) a. Il n’a rien fait. he n-has nothing done ‘He hasn’t done anything.’ b. %Il n’a fait rien. he n-has done nothing c. Il a tout lu he has all read ‘He has read everything.’ d. %Il a lu tout. he has read all The fact that when focused the quantifier can also occur after the past participle led Cardinaletti and Starke (1999) to assume that facts of this type are to be analyzed on a par with other cases of weak pronouns. Notice however that the reasons why BQs cannot be focused in the vP might be different, namely the fact that the vP Focus position is lower than the position of BQ s in the low IP area (see (4) below): (4) [AspP Italian past.part. [BQ tutto/(tout)]…[French past.part. … [AspP appena/à peine…[vP[Topic] [Focus focused BQ] [VP [V°] [QP]…]]]]] Since there is no lowering movement towards a non-c-commanding position, this type of Focus is blocked because of the general constraint on movement being always to a higher position.3 In French, the past participle evacuates the vP to a very low functional position in the IP area, as shown by Cinque (1999: 46), since it raises higher than adverbs like à peine, presque and souvent, so the vP-peripheral Focus position is located lower than the past participle. Modern standard Italian does not provide us with the same empirical evidence of a pre-participial position as French does, because the past participle raises too high bypassing BQs and a number of adverbs (see Cinque 1999), as shown in (4). However, that the phenomenon is not unique to French is shown by the observation that in Italian object bare universal quantifiers occur higher 3 Actually, bare quantifiers can indeed be moved to a Focus position in the sentential left periphery in Italian, thereby showing that at least in Italian bare quantifiers can indeed bear Focus.
190
poletto
than low adverbs like bene/male ‘well/badly’ while quantified nominal expressions do not. In both cases the other order is possible if there is a focus, as shown by the focusing particle proprio, but crucially the order of BQ s and quantified nominal expressions is reversed showing that the unfocussed position of BQ s is higher than the lowest adverbs while the one of quantified nominal expressions is lower: (5) a. Ha scritto tutto bene. he.has written everything well ‘He has written everything well.’ b. Ho scritto bene (proprio) TUTTO. I.have written well (really) everything ‘I have written (really) everything well.’ c. Ha scritto bene tutto il compito. he.has written well all the homework ‘He has written all the homework well.’ d. Ha scritto tutto il compito (proprio) BENE. he.has written all the homework (really) well ‘He has written all the homework (really) well.’ A language which is more similar to modern French with respect to past participle movement is Old Italian, where the past participle does not raise higher than the vP-peripheral positions of FocusP and TopicP (see Poletto 2014). Old Italian shows that BQs cannot be treated as weak pronouns, since bare universal quantifiers like tutto ‘everything’ must occur in front of the past participle even when they are paired to a preposition, hence when they are not weak forms in a syntactic sense:4 (6) a. s’i’ mi fosse al tutto a tte gradato if I me were.1sg to.the everything to you adapted ‘if I had adapted to you in everything’ (Fiore 86)
4 All the example here are taken from the OVI database (http://artfl‑project.uchicago.edu/ content/ovi‑search‑form) and the editions of the texts correspond to those selected by the Italant project, which was the source of the OVI database.
being bare: a survey of quantifier positions
191
b. che sia per tutto detto that is for all said ‘that is said completely’ (Detto 485) c. chi ‘l tene del tutto in sé celato who it keep.3sg of.the all in himself concealed ‘who keeps it completely concealed in himself’ (Tesoretto 178) The examples in (6) show that bare universal quantifiers need to occur in front of the past participle when they are arguments as well as when they are adverbial. This is not true of quantified nominal expressions with universal quantifiers, which behave as regular definite nominal expressions: they can indeed be found in a pre-participial position, as all nominal expressions can occur in OV order in Old Italian, but the pre-participial position is not mandatory at all, as shown by Poletto (2014): tutto questo fatto, e molte altre cose (7) a. ch’egli ebbe that he had.3sg all this done and many other things ‘that he had done all this and many other things’ (Tesoro 286) b. ond’io òe perduto tutto lo mio onore where I have.1sg lost all the my honor ‘where I lost all my honor’ (Tristano 179) Summarizing, we can state that apart from the different raising span of the past participle, universal BQs are always located higher than the vP while quantified nominal expressions are not, as illustrated in (8):5 (8) past.part. [AspP [tutto/(tout)]…[vP [Topic][Focus] past.part. [VP past.part.] [QP]…]] Modern Italian Old Italian French The same pattern observed with tutto, i.e. obligatory pre-participial position with BQs and alternation between OV and VO for quantified nominal expres-
5 As for the actual position of quantified nominal expressions, there is no empirical evidence to assume that they are located in a different position than the usual argumental one, although it might be the case they also move to some very low position, if we assume an approach à la Beghelli and Stowell (1997).
192
poletto
sions (and regular DPs) is found with molto ‘much’, which according to Beghelli and Stowell (1997) belongs to a different category (group quantifiers) with respect to universal quantifiers: (9) a. Quando ha molto pugnato when has.3sg much fought ‘When he has fought a lot’ (Vizi e Virtudi 23) b. ch’egli avea molto speso in que’ servi that he had.3sg much expended in those servants ‘that he had paid a lot for those slaves’ (Seneca 61) molto dire c. non vo’ not want.1sg a.lot say ‘I do not want to say a lot’ (Fiore di Rettorica 34) d. perciò che di questo è molto detto davanti therefore that of this is.3sg a.lot said in.front.of ‘therefore a lot about this is said in what precedes’ (Rettorica 96) So, both molto and tutto must be in pre-participial position only when they are bare. The [+human] BQs tutti also obligatorily occurs in pre-participial position when it is bare:6 (10) a. Il Demonio ci avea tutti presi. the Devil us had.3sg all taken ‘The Devil took all of us.’ (Sposizioni di Vangeli 279) b. ma parve che fussero tutti vinti con lui but seemed.3sg that were.3pl all won with him ‘it seemed that they were all won with him’ (Storie contra i Pagani 135) The case of molti is more difficult to interpret since the relevant examples are not many, once all the cases of molti combined to a nominal expression are filtered out. The majority of cases of bare molti are subjects, which can indeed occur between the inflected verb and the past participle, but could be cases of subject inversion, and therefore not relevant either: 6 See Poletto (2014) for a discussion of the cases of quantified nominal expressions, which behave as regular DP s in allowing OV orders which are nevertheless not obligatory.
being bare: a survey of quantifier positions
193
(11) a. ed a molti veraci Segnori ne fossero molti mandati and to many righteous Sirs of.them were.3pl many sent ‘and many had to be sent to many valuable nobles’ (Vegezio 5) b. avegna che ne fossero molti credenti happens.3sg that of.them were.3pl many believers ‘many were believers’ (Vizi e Virtudi 83) Once all these ambiguous cases are filtered out, object bare molti must indeed occur in pre-participial position as tutto, molto and tutti: (12) a. acciò che per lo fatto di costui ne possa molti ingannare so that for the fact of him.here of.it can.3sg many cheat ‘so that through this he can cheat many’ (Vizi e Virtudi 14) b. per questa via n’ha già molti schifati e fuggiti in this way of.it’ have.3sg already many avoided and escaped ‘in this way he has already avoided many’ (Vizi e Virtudi 31) The pre-participial position of object BQs is higher than the vP as shown in structure (8) and repeated here as (13): (13) past.part. [AspP [BQ]…[vP [Topic][Focus] [v° past.part.] [VP past.part.] [QP] …]] This is attested by the fact that they always occur in front of scrambled DP s and PPs, which occur in Topic and Focus positions at the border of the vP (see Belletti 2004) as illustrated in (13) and shown by the following example: tutta in quattro parti divisa. (14) a. Vedemmo che fue parts split saw.1pl that was.3sg all in four ‘We saw that the whole was split in four parts.’ (Vizi e Virtudi 56) b. egli ci è stasera venuto un de’ suoi fratelli e he there is.3sg this.evening come one of.the her brothers and ha molto con lei favellato has.3sg much with her talked ‘This evening one of her brothers came and spoke a lot to her’ (Decameron viii–vii, 538)
194
poletto
Hence, we conclude that there is indeed a dedicated position to the left of the vP for BQs which cannot be reduced to a position for weak elements, since weak pronouns cannot be paired to a preposition while BQ s can. If we add n-words to the picture, which are another class examined by Beghelli and Stowell7 (1997), the results we obtain are not as clear-cut as the cases of tutto/tutti, molto/molti, since niente/neente/neiente can occur in various positions, as noted by Garzonio and Poletto (2018).8 They show that niente/ neente/neiente can occur before low adverbs as in (15): niente bene schifare … (15) Sì no lo potero thus not it could.3pl nothing well avoid ‘They couldn’t dodge it well at all …’ (Storia di Troia 574) Furthermore, Old Italian niente can occur after a dative or another PP, which shows that they are located in a low position in the VP: a questo fatto niente (16) a. perchè non fa because not do.3sg to this fact nothing ‘because it does not do anything to this’ (Fiore di rettorica 23) b. e non vede in lui niente perchè sia degno del and not see.3sg in him nothing because is.3sg worth of.the pane bread ‘and does not see anything in him that would make him worth of the bread’ (Esposizione del Paternostro 25) This is clearly different from French, where rien does not occupy the argumental position. As a confirmation that niente does not need to raise higher than the vP, it can also occur after the past participle even in Old Italian where the past participle does not raise outside the vP and most probably stops at the peripheral Focus or Topic position:
7 Beghelli and Stowell actually examine negative quantifiers in English. The status of n-words is still unclear in the literature, but I take them into account, since they are the counterpart of English negative quantifiers. 8 Garzonio and Poletto (2018) also consider the distribution of adverbial niente meaning ‘at all’, which I leave aside here, since it is different. Adverbial niente does not occur after the past participle or after internal arguments of the verb.
being bare: a survey of quantifier positions
195
(17) a. ch’io non t’ho tolto neente that I not from.you have.1sg removed nothing ‘that I have taken nothing from you’ (Novellino lxxii, 294) b. Dimmi, Merlino, dell’avere d’Atene fu trovato tell.me, Merlin, of.the possessions of Athens was.3sg found niente? nothing ‘Tell me, Merlin, was anything from the goods of Athens discovered?’ (Merlino 48) c. … l’altre parti della diceria, delle quali non è detto the other parts of.the message, about which not is.3sg said niente … nothing ‘… the other parts of the message, about which nothing is said …’ (Rettorica 142) neente: no lli era d. Il mercatante non mi insegnò the merchant not me taught.3sg nothing not him was.3sg neente tenuto. nothing obliged ‘The merchant taught me nothing, and nothing was due to him.’ (Novellino vii, 144) neente perduto. e. Non avea not had.3sg nothing lost ‘He lost nothing.’ (Seneca 17) nostra intenzione essere che ce ne sia f. Sì che non era that us of.it is.3sg so that not was.3sg our intention be neente renduto. nothing given.back ‘So that we did not want that anything of it would be given us back.’ (Lettera di Giachino 17) Hence, it seems that once we control for the variable of the position of the past participle, Old Italian BQs do not all behave in the same way, niente being different from both tutto/tutti and molto/molti. Garzonio and Poletto (2018) propose that in Old Italian niente can still be analyzed as containing the mor-
196
poletto
pheme —ente meaning THING and this is the reason why it does not need to move.9 This is not true of other BQs, which do not contain a lexical classifier inside their structure and therefore must move to c-command the whole vP. Hence, we find two types of BQs: those that do not contain a lexical classifier and those that do. Those that do contain the lexical classifier can stay in an argumental position, those that are completely empty of a nominal category occur outside the vP. Let us now investigate what this position outside the vP can be and the reason for this difference.
3
Argumental Circuits
In the cartographic literature we often find occasional reference to sets of contiguous projections hosting the same type of nominal expressions ordered according to their thematic role so that the first element corresponds to the external thematic role, the intermediate to the recipient, and the lowest one to the internal theme/patient. A famous case are clitics, which are well known for being rigidly ordered and which always require the subject clitic to occur as the highest clitic followed by the dative and the accusative.10 This is also the case of wh-items in those Slavic languages where wh-items all move to the CP and have a fixed word order like for instance Bulgarian in (18)–(19). This effect is generally attributed to superiority (see Rudin 1988; Bošković 1997, 1998, 2002, Richards 1997, 2001, Pesetsky 2000, Grewendorf 2001, among others): (18) a. Koj kakvo pravi? who what does ‘Who was doing what?’ (Rudin 1988: 481) b. *Kakvo koj pravi? what who does ‘What is who doing?’ (Rudin 1988: 482)
9
10
Incidentally, this also shows that approaches which attribute this ordering to a prosodic condition cannot be correct, since they fail to account for the difference between French rien and Old Italian niente. Subject clitics are always the highest clitic in the string. As for objects, I leave aside the case of French le lui, which raises a problem of reordering only in the case of two third persons.
being bare: a survey of quantifier positions
197
(19) a. Kogo kak e tselunal Ivan? whom how is kissed Ivan ‘How did Ivan kiss who?’ (Bošković 1997: 234) b. *Kak kogo e tselunal Ivan? how whom is kissed Ivan Even apparent exceptions to the rigid ordering have been systematized and explained in Krapova and Cinque (2008) as effects of the internal makeup of the wh-item itself, so that properties like being marked for a [+human] feature or d-linking determine the position of the wh-item with respect to others. Krapova and Cinque show that wh-items can occur in three different layers, one for clitic doubled wh-items, one for d-linked wh-items and one for non-d-linked ones. Inside each layer wh-elements dispose along the same rigid argument order S-IO-DO the arguments have prior to wh-movement. They put forth that relativized minimality (RM) is the source of this rigid ordering, and derive it in the following way: they assume that these superiority effects are a consequence of relativized minimality so that only those wh-items which have exactly the same set of features induce RM effects, those that have a partially different feature endowment (as for instance d-linked ones in comparison to non-d-linked ones) do not. Examples which apparently violate this strict order are spurious in the sense that the features of the intervener do not perfectly match those of the moved element. Here, the possibility that also elements that have intersecting features might count as interveners is not considered, but see Rizzi (2016) for a discussion of the possible feature configurations. Furthermore, they propose that only a complete chain can count as an intervener between a moved element and its trace, which means that elements that count as interveners are only cases of nesting and no cases of crossing chains, always provided that the two elements have identical features. The two configurations are illustrated in (20a) for crossing and (20b) for nesting, the configuration banned by relativized minimality is thus (20b) but not (20a): (20) a. X………Y…….X…….Y b. X……..Y…..Y……X
If this idea is on the right track, we should find that all cases of movement of the same type of constituents, i.e. constituents that have exactly the same features,
198
poletto
must maintain the word order they display prior to movement unless they have different features. Another case of rigid ordering of the arguments has been observed in the case of floating quantifiers (FQ), which have a fixed sequence of subject, indirect object, direct object. See Cinque (1999: 116, ex. 19ff.): the examples he provides to show that the order is FQ_S-FQ_IO-FQ_DO are the following, with the unique interpretation signaled by the indexes: (21) a. Ellesi ont toutesi tousk parleé hier. leurk theyi.fem to.themk have alli.fem allk talked yesterday ‘All of them have talked to all of them yesterday.’ leurk ont tousk toutesi parlé hier. b. *Ellesi theyi.fem to.themk have allk alli.fem talked yesterday (22) a. *Ils ont bien tous compris. they have well all understood b. Ils ont tous bien compris. they have all well understood ‘They understood everything well.’ (23) a. Les fillesi lesj ont toutesi tousj lu. the girlsi themj have alli.fem allj read ‘All the girls have have read all of them.’ b. *Les fillesi lesj ont tousj toutesi lu. the girlsi themj have allj alli.fem read ai tousk toutesi montrées. (24) a. ?Je lesi leurk I themi to.themk have allk alli.fem shown ‘I have shown them all to all of them.’ b. *Je lesi leurk ai tousi toutesk montrées. I themi to.themk have alli allk.fem shown Cinque (1999) also shows that floating quantifiers and BQ s like Italian tutto occupy different positions in the low IP area, since floating quantifiers can or must cross adverbs that BQs cannot cross. He argues that floating quantifiers can occur in different layers of the low IP area, since they can occur on the right
being bare: a survey of quantifier positions
199
or on the left of lower adverbs which have a fixed position. The first FQ space is to the left of the adverb ‘completely’11 complétement in French. Cinque (1999) shows that in each circuit the order of floating quantifiers is rigidly FQ subject > FQ indirect object > FQ direct object, exactly the same found with clitics and wh-items in multiple wh-fronting languages. For instance, no floating quantifier can occur to the right of complétement in French but BQ s can (Cinque 1999: 119, ex. 37ff.): (25) a. Il a complétement tout perdu. he has completely all lost ‘He has completely lost everything.’ b. *?Il a tout complétement perdu. he has all completely lost (26) a. Elles les ont tous complétement bien refaits. they.fem them have all completely well remade ‘They have well redone all of them completely.’ b. *Elles les ont complétement tous bien refaits. they.fem them have completely all well remade Furthermore, a dative floating quantifier cannot occur after a BQ representing the direct object:12 (27) a. Je leur ai tout tous montré. I to.them have everything all showed ‘I showed them all everything.’ Cinque explicitly notices that the non-floating universal object quantifier tutto occurs in a lower position with respect to floating ones, a position he dubs completive aspect. Here is the serialization of heads he proposes (Cinque 1999: 119):13
11 12
13
The second space for FQ s is found between ‘always’ toujours and plus ‘(any)more’. The third space is between pas ‘not’ and déjà ‘already’. Notice that in French dative quantifiers occur without the preposition, to my ear this is not possible in Italian, although this difference still awaits further systematic experimental evidence. An anonymous reviewer asks for the reason for movement. Cinque (1999) does not dis-
200
poletto
(28) FQsubject FQIO FQDO complétement BQ bien As discussed above, in Old Italian bare tutto, tutti, molto, molti must occur to the left of definite DPs located in the vP left periphery. So, one might think that for each type of quantifier, i.e. universal, group quantifiers (using Beghelli and Stowell’s 1997 terminology), and n-words there is a dedicated space in which they must occur in the same rigid order S-IO-DO that has been observed for clitics, wh-items and floating quantifiers. In the next section, I will put this hypothesis to the proof showing that the same fixed sequence of arguments can be observed with different types of bare quantifiers.
4
BQs in Modern Italian
Since it has been noticed that various nominal expressions of exactly the same type occur in an argumental circuit, i.e. have a rigid ordering subject > indirect object > direct object, I will consider whether this is also true for tutto/tutti, molto/molti and n-words like niente/nessuno. BQ s clearly have a fixed word order, once focus is controlled for. Since cases in which subject, indirect object and direct object occur in postverbal position all together are rather difficult to obtain in Italian, I test them in pairs as Cinque does. If we consider the order between a postverbal subject and the direct object, we obtain the rather surprising order direct object followed by the subject:14 (29) a. Non mi ha regalato niente nessuno. not to.me has given nothing nobody ‘Nobody gave me anything.’ b. *Non mi ha regalato nessuno niente. not to.me has given nobody nothing ‘Nobody gave me anything.’
14
cuss any reason for movement. As for floating quantifiers, he just assumes that they can occupy all the position where DP s can transit, but does not state why DPs do so. As for bare quantifiers, the positions must also have to do with peculiar features of the type of nominal expression, although he does not say anything explicit about this. I abstract here from focus, which can change the basic word order. All these sentences become better if the last BQ is focused.
being bare: a survey of quantifier positions
201
(30) a. Mi hanno regalato tutto tutti. to.me they.have given all everyone ‘Everybody gave me everything.’ .
b. ??Mi hanno regalato tutti tutto. to.me they.have given everyone all ‘Everybody gave me everything.’ (31) a. Mi hanno regalato molto molti. to.me they.have given a.lot many ‘Many gave me a lot.’ b. ?*Mi hanno regalato molti molto. to.me they.have given many a.lot ‘Many gave me a lot.’ The effect is not the same for all types of BQs, it is very clear in the case of nwords, somewhat better in the case of group BQ s and rather weak in the case of universal quantifiers. Notice however, that the reason why (30b) is not really excluded could be that it is ambiguous with the case in which tutti is a floating quantifier of a preverbal subject pro.15 Also the order of direct object and indirect object is reversed with respect to what is expected on the basis of the rigid circuits of clitics, wh-items and floating quantifiers mentioned in Section 3: (32) a. Non ho regalato niente a nessuno. non I.have given nothing to nobody ‘I did not give anything to anybody.’ b. *Non ho regalato a nessuno niente. not I.have given to nobody nothing (33) a. Ho regalato tutto a tutti. I.have given everything to everyone ‘I gave everything to everybody.’
15
Once again, one has to factor out focus, which appears to be more compatible with universal BQ s than with n-words.
202
poletto
b. ??Ho regalato a tutti tutto. I.have given to everyone everything (34) a. Ho regalato molto a molti. I.have given a.lot to many ‘I gave much to many.’ b. *?Ho regalato a molti molto. I.have given to many a.lot The order between dative and postverbal subject is rather difficult to judge, since the contexts in which a postverbal subject is pragmatically allowed are restricted but there is a clear difference between the a. and the b. sentences in the examples (35)–(37): (35) a. Non lo ha regalato nessuno a nessuno. not it has given nobody to no one ‘Nobody gave it to anybody.’ b. %Non lo ha regalato a nessuno nessuno. non it has given to no one nobody (36) a. L’hanno regalato tutti a tutti. it-they.have given everything to everyone ‘Everybody gave it to everybody.’ b. %L’hanno regalato a tutti tutti. it-they.have given to everyone everything (37) a. L’hanno regalato molti a molti. it-they.have given many to many ‘Many gave it to many.’ b. %L’hanno regalato a molti molti. it-they.have given to many many The order subject-indirect object is the expected one, i.e. the subject comes before the indirect object. If we look at these examples, it is clear that the element that has a different position with respect to what is expected on the basis of the literature on clitics, wh-items and floating quantifiers is the direct object,
being bare: a survey of quantifier positions
203
which should come last in the sequence of quantifiers, but on the contrary always comes first. Since relativized minimality in the versions discussed above blocks movement of the direct object on top of a moved indirect object or subject, the only possible solution to keep Krapova and Cinque’s idea that the rigid word order of nominal expressions of the same type is due to RM effects is to argue that (nonfloating) BQs have a specialized position above the vP only for direct objects, while the subject and the indirect object can also remain in their vP-internal position. This conclusion is also supported by the fact that, as pointed out by an anonymous reviewer, object BQs and subject BQ s can be split by other elements and low adverbs, thereby showing that the subject can remain in the vP-internal position, but the direct object does not: (38) Non ha fatto niente bene nessuno. not has done nothing well no one ‘Nobody did anything well.’ (39) Non ha mangiato niente crudo nessuno. not has eaten nothing raw no one ‘Nobody ate anything raw.’ Notice that if we combine a universal and a group quantifier, or a group quantifier and an n-word, the result is the same as the one between two universal quantifiers:16 (40) a. Ho regalato tutto a molti.17 I.have given everything to many ‘I have given everything to many.’ b. *Ho regalato a molti tutto. I.have given to many everything
16 17
Once again, all these cases are possible if there is focus on the last element, just like observed for rien by Kayne (1975). Again, I abstract here from Focus, which renders the ungrammatical sequences much better. Since I am trying to establish the basic word order, Focus is a factor which can only be confusing.
204
poletto
(41) a. Ho regalato molto a tutti. I.have given a.lot to everyone ‘I have given much to everybody.’ b. *Ho regalato a tutti molto. I.have given to everyone a.lot (42) a. Non ho regalato niente a molti. not I.have given nothing to many ‘I haven’t given anything to many.’ b. *Non ho regalato a molti niente. not I.have given to many nothing Thus, we can conclude that the expectation that the sequence of non-floating BQs also conforms to the fixed order S-IO-DO observed for various argumental circuits in the literature and summarized by Cinque (1999) is not borne out for modern Italian. In particular, it seems that the odd man out is the direct object, which occurs higher than the subject and the indirect object for all types of bare quantifiers. Evidently, the question is why this is so and how to model the difference.18
5
Barest Quantifiers?
One interesting fact is that the quantifiers we have used here for the object are those that correspond to inanimates and which in many languages also have
18
An anonymous reviewer asks whether bare nominals also behave in the same way, and actually it seems that bare direct objects do behave as BQs as shown by cases like the following: (i) Non ha mangiato patate nessuno. not he.has eaten potatoes no one ‘Nobody ate potatoes.’ (ii) Hanno letto libri tutti. they.have read books everyone ‘Everybody has read books.’ However, bare nominals cannot occur in front of low adverbs like bene as BQs can: (iii) *Hanno bruciato legna bene tutti. they.have burnt wood well everyone ‘Everybody has burnt wood well.’
being bare: a survey of quantifier positions
205
an adverbial usage. Cinque (1999) already showed that tutto can be used as an adverb meaning ‘completely’ in the sense that all elements of a set have been affected by the predicate. The same is true of molto, which Cinque treats as a measure adverb akin to other measure adverbs like little, much, etc. and which he puts in the same structural space as the manner adverb bien. The element niente can also have an adverbial usage meaning ‘at all’, as observed by Garzonio and Poletto (2012, 2018) for both Old Italian and for some Italian dialects and regional colloquial varieties where sentences like the following are grammatical: (43) a. Non mi piace niente, questa storia. not to.me likes nothing, this story ‘I do not like this thing at all.’ b. Non ho dormito niente stanotte. not I.have slept nothing this.night ‘I did not sleep at all tonight.’ Here niente cannot be the object, since the verb is either unergative or a psych-verb where the structural object position is occupied by another element (questa storia, in the example above). We could thus ascribe the observation that direct object BQ s occur in front of subject and indirect object BQs to the fact that they have at their disposal an adverbial position in the low IP area, since they can be adverbs. I will first consider this hypothesis and then discard it on the basis of empirical evidence. One observation which tallies with the idea that BQ s are adverbial in nature is the fact that in grammaticalization processes the element that becomes the sentential negative marker in the case of n-words is the one corresponding to the inanimate direct object, i.e. nothing. In general, grammaticalization processes apply to those elements that are semantically weaker in their lexical endowment, and this is exactly the case of inanimate direct objects, which lack the [+human/animate] feature of agentive subjects and of recipient indirect objects. This means that the possibility to occupy an adverbial position would correspond to the barest feature endowment a BQ can have, i.e. the simple existence of an object which does not contain any bare classifier-like noun like PERSON, PLACE, WAY, TIME along the lines of what Kayne has proposed in a number of articles starting from Kayne (2004). One might object that the barest quantifiers corresponding to the inanimate direct object contains a null classifier like THING. However, since THING does not convey any other information except the existence of an object, it can evidently be suppressed more easily
206
poletto
than other classifiers which have a higher featural content and more specifications like PERSON, PLACE, WAY, TIME etc. Then, one could hypothesize that the reason why barest quantifiers are in an adverbial position has something to do with the by now standard assumption that BQ s need to c-command their domain of scope. If there is nothing under the BQ in its internal nominal projection, then the quantifier must move to a position from which it can scope over its restrictor, which it also c-commands, and this is precisely the adverbial position which scopes over the vP. The distinction between the inanimate direct object and other BQ s would thus depend on the existence of a classifier-like element like PERSON, PLACE, WAY, TIME etc. In these cases, the BQs would not need to move to an adverbial position, since the quantifier finds a domain over which it can take scope inside its nominal structure and this is given by the sortal classifier. This is the view I entertained in various recent articles (see Garzonio and Poletto 2018 among others), however this hypothesis should be tested and find independent empirical evidence. If the effects that we have noticed above concerning the special position of the direct object inanimate BQ have to do with the lack of any classifier in its internal structure, i.e. with its adverbial status, then it should not have the typical properties of nominal expressions, like being modified by adjectives and relative clauses.19 The hypothesis that BQs corresponding to the inanimate object are so devoid of internal structure that they lose their nominal status and become adverbs predicts that they should not be modified by a typical nominal modifier. I consider here three cases of typical nominal modification: a) modification by the adjectival form altro ‘other’; b) modification by an adjective; c) modification by a relative clause. All BQs can be directly modified by the element altro, ‘other’: in all cases modification by altro is grammatical:20 (44) a. Tutt’altro all-other ‘something else’
19
20
This might have consequences on our general theory of extended projections in the following sense: if barest quantifiers can exist without having an internal nominal category, but crucially still being interpreted as corresponding to the object, then we do not need a lexical root to build functional structure on top of it. Notice that for tutto truncation of the final vowel of the BQ is obligatory, it is optional for niente and impossible for molto.
being bare: a survey of quantifier positions
207
b. Nient’altro nothing-other ‘nothing else’ c. Molto altro a.lot other ‘much else’ If we consider modification by an adjective, this is possible with niente and molto, which however require the presence of the preposition di ‘of’: (45) a. Niente di interessante nothing of interesting ‘nothing interesting’ b. Molto di interessante a.lot of interesting ‘very interesting’ c. *Tutto di interessante21 all of interesting As for modification by a relative clause, once again we see a split between tutto on one side and niente, molto on the other. Tutto cannot be directly modified by a relative clause, while molto and niente can: (46) a. Mi ha detto tutto *(quello) che volevo sapere. to.me has said everything (that) that I.wanted to.know ‘He told me all I wanted to know.’ b. Non mi ha detto niente che non sapessi già. not to.me has said nothing that not I.would.know already ‘He did not tell me anything I did not know already.’ c. Mi hanno detto molto che già sapevo. to.me have said a.lot that already I.knew ‘They told me much I already knew.’ 21
Tutto can in general not be modified by an adjective except for fixed expressions like tutto di guadagnato ‘all the better’.
208
poletto
At this point one might conclude that tutto is indeed an adverb, although the data on altro attest the opposite. I think that the reason why tutto cannot be modified by an adjective or by a relative clause is that it takes a whole DP as a complement, while niente and molto can be inserted into the DP’s internal spine. So, we can conclude that there is no evidence for treating BQ s corresponding to the inanimate object as adverbial in nature. Actually, there is evidence that they at least can behave as nominal expressions in tolerating adjectives and relative clause modification. Thus, we are back to the problem as to why they occupy a special position higher than the vP.
6
A Problem of Projection
Once we have discarded the idea that direct object BQ s are adverbs in nature, we are back to the problem of understanding why they occur higher than BQ s that correspond to the subject and the indirect object. The analysis I propose for the fact that BQs corresponding to the inanimate direct object require a special position in the clause has to do with the well-known requirement that all nominal expressions require case, or are subject to a more general visibility condition which applies to all nominal expressions. Already Belletti (1999) proposed that postverbal subjects in Italian can be made visible by focus, which can be an alternative way to render a nominal expression visible when it lacks case. Although many languages mark case in the same way on all nominal expressions, there are indications that not all types of nominal expressions are made visible in the same way. One can interpret the well-known phenomenon of differential object marking (DOM) debated in the literature precisely as an instance of different visibility conditions for different types of nominal expression. As pointed out by Kennelly (1997), Zidani-Eroğlu (1997), Kelepir (2001), and more recently by Miyagawa, Wu, Koizumi (2018), in DOM languages like Turkish, the element which raises higher than the vP is not the bare indefinite one, but the definite, specific one. The same is true of the phenomenon of object case drop in Japanese, which, as Miyagawa, Wu, Koizumi (2018) notice, can drop the case morpheme more frequently when it is closer to the verb. If the object case marker is present, the object is interpreted as more topical or salient. Hence, it seems prima facie that DOM is exactly the opposite of what we see here for object BQs, since the element that raises out of the vP is the definite and topical one, not the indefinite one. Miyagawa, Wu, Koizumi (2018) derive Case as depending on a problem of projection, i.e. the fact that any syntactic object needs to be labeled. The label-
being bare: a survey of quantifier positions
209
ing algorithm looks for an asymmetry in the two elements forming the set of the new object. When it finds an X° and an XP, the asymmetry is granted and the new syntactic object is labeled with the features of the head. However, when merging two XPs or two X° there is no way for the labeling algorithm to do its job and assign a label since there is no asymmetry between the two syntactic objects that have been merged, as the two merged elements are equivalent. In Chomsky’s (2013) system the ambiguity blocking the labeling algorithm can be eliminated via two mechanisms, i.e. either by movement or by agreement. Saito (2016) proposes that Case is a third mechanism to circumvent the labeling problem, by assuming that, when case is added to a DP, it is made inert so that it cannot project, thus eliminating the ambiguity which blocks the labeling algorithm and allowing the verb as the only element which can project.22 Languages with overt morphological case like Turkish or Japanese provide us with clear evidence that definite, specific object DP s move outside the vP, while indefinites stay in their merge position in the vP. According to Miyagawa, Wu, Koizumi (2018), the reason for this is that a direct object does not raise any problem of projection because the syntactic object formed by the verb and the direct object is clearly asymmetric, the verb being a head and the object an XP. In this account the distinction between definites and indefinites remains unaccounted for and needs to be explained by an additional mechanism. Suppose however that all arguments are specifiers and that there is no complement position as recently proposed by Cinque (2020): in this case we would predict that the labeling algorithm is blocked also in the case of direct objects. This means that also for the direct object we have to use one of the three ways to block the object from projecting. This means, in turn, that either object DP s must always be case marked, or that they have to move or agree with the verb. Since in modern Italian subjects agree with the inflected verb, the relation of agreement eliminates the problem of projection. As for indirect objects, they are case marked by the preposition a and therefore become inert to project, as Saito proposes. This means that the only argument that needs to move to avoid a problem of projection is the direct object BQ, which is neither casemarked nor does it agree with the verb. As noticed above, object BQ s move indeed outside the vP, but to a very low position.23 Regular quantified expres-
22
23
In a sense, one could see case as corresponding to the second mechanism, i.e. agreement, which is not realized on the verb but on the nominal, a line of thought I will not pursue here. The position of moved object BQ s is most probably lower than the scrambling positions of languages like Japanese, where scrambling is generally located in the TP area.
210
poletto
sions are case-marked in their VP internal positions, as standardly assumed. However, since BQs do not have a fully fledged internal nominal structure they cannot be case marked as all other quantificational expressions. Hence, they are the only elements that must raise outside of the VP. Subjects, which move to SpecT and indirect objects, which are always case marked by a preposition, behave differently because they are subject to different mechanisms, as seen, and thus do not have problems of projection. So, the reason why the direct object raises is to avoid a problem of projection. This might be called a case or visibility position, but is actually just a mechanism to allow labeling. Notice that here we have a typical problem that has already been signaled in the literature, namely the problem of stopping the movement of the direct object, which attaches to another functional projection higher than the vP and thus creates a new problem of projection with the new XP. In order to stop the movement, there are two possible solutions: either the fact that this position outside the vP is a case assignment position for the direct object or the fact that the BQ taking the vP as its restrictor is equivalent to an agreement procedure. As for subject and indirect objects, as discussed, they can remain in the vP since they do not pose a problem for the labeling algorithm, although for different reasons, since the subject uses the escape mechanism of agreement while the indirect object is rendered inert by the case assigning preposition. In conclusion, the reason why we find a different word order with BQ s with respect to other types of nominal expressions like clitics, multiple wh-items and floating quantifiers is the following: BQs can indeed remain within the vP in modern Italian, their internal structure only contains a sortal classifier and they do not have features to check which force them to move like wh-items, or which pied-pipe them to higher projections like floating quantifiers. However, if they stay in their merge position, they must circumvent the problem of projection identified by Chomsky (2013). Subjects can circumvent it via agreement, indirect objects and quantified nominal expressions via overt case, while BQ s direct objects, which neither agree nor are overtly case marked in Italian, must use the third strategy, i.e. movement. This eliminates the only exception to the observation that nominal expressions of exactly the same type target specific circuits or layers of the sentence structure in a fixed word order, which is S-IO-DO. The fact that BQ s do not do so depends on their extremely poor feature endowment, which does not allow them to move for feature checking, unless they cause a problem of projection. In Italian only direct objects need to do so, since the subject and the indirect object can use other escape mechanisms. It might be interesting to find other languages in which also direct objects are case marked, since the prediction here is that the fixed word order S-IO-DO is reestablished.
being bare: a survey of quantifier positions
211
One last problem that remains to be at least briefly discussed is the difference between bare quantifiers and quantified nominal expressions and definite expressions, which do not display the same rigid ordering that bare quantifiers show. One of the reasons why this is so is that definite nominal expressions, and probably also at least specific quantified nominal expressions can be not only in a focus position, but also low topics. The possibility to use information structure related projections is the escape hatch out of the rigid word ordering that becomes evident only when we single out elements that cannot access them, so that reordering is not possible. Ideally, it should be possible to find contexts which exclude topicality, as well as focus of quantified nominal expressions and definite expressions and the ordering should be the same found with bare quantifiers. However, there is also another possibility, namely that the final interpretive position of definite DPs is higher than the one of case checking and therefore the reversal we observe with BQs direct objects remains the only indication that there is indeed a case-checking position for direct objects in languages like Italian.
7
Concluding Remarks
In this chapter I have shown that BQs behave differently from quantified nominal expressions: capitalizing on Garzonio and Poletto (2017), I have argued that they only contain a sortal classifier, and are so poor in their feature endowment that they are the only elements which display an unexpected reversal of word order so that the direct object is higher than the indirect one, contrary to what Cinque (1999) observed on the basis of floating quantifiers. I have proposed that this is due to case assignment, so that also BQ s can escape the problem of projections already identified in the literature. Subject and indirect objects use different escape mechanisms, since the subject overtly agrees with the verb while indirect objects are case marked by the preposition a. Direct objects cannot use either mechanism and therefore must move outside the vP where they can take scope and get case. This explains the observation that the word order of non-focused BQs is DO-S-IO, contrary to all expectations raised by other types of nominal expressions, which occur in circuits that display a different word order, namely S-IO-DO.
212
poletto
References Baunaz, Lena. 2011. The grammar of French quantification. Dordrecht: Springer. Beghelli, Filippo & Tim Stowell. 1997. Distributivity and negation: The syntax of each and every. In Anna Szabolcsi (ed.), Ways of scope taking, 71–107. Dordrecht: Kluwer. Belletti, Adriana. 1999. Inversion as focalization and related questions. Catalan Working Papers in Linguistics 7. 9–45. Belletti, Adriana. 2004. Aspects of the low IP area. In Luigi Rizzi (ed.), The structure of CP and IP. The cartography of syntactic structures, volume 2, 16–51. Oxford: Oxford University Press. Borer, Hagit. 2005. In name only. Structuring sense, volume 1. Oxford: Oxford University Press. Bošković, Željko. 1997. On certain violations of the Superiority Condition, AgrOP, and economy of derivation. Journal of Linguistics 33. 227–254. Bošković, Željko. 1998. Wh-phrases and wh-movement in Slavic. Paper presented at the Workshop on Comparative Slavic Morphosyntax. Indiana University, 5–7 June 1998. Downloadable at http://www.indiana.edu/~slavconf/linguistics/index.html. Bošković, Željko. 2002. On multiple wh-fronting. Linguistic Inquiry 34. 351–383. Cardinaletti, Anna & Michal Starke. 1999. The typology of structural deficiency: On the three grammatical classes. In Henk van Riemsdijk (ed.), Clitics in the languages of Europe, 145–234. Berlin: de Gruyter. Cinque, Guglielmo. 1999. Adverbs and functional heads. Oxford: Oxford University Press. Cinque, Guglielmo. 2020. The syntax of relative clauses. A unified analysis. Cambridge: Cambridge University Press. Chomsky, Noam. 2013. Problems of projection. Lingua 130. 33–49. Doetjes, Jenny. 1997. Quantifiers and selection: On the distribution of quantifying expressions in French, Dutch and English. The Hague: Leiden University PhD dissertation. Garzonio, Jacopo & Cecilia Poletto. 2012. On niente: Optional negative concord in Old Italian. Linguistische Berichte 230. 131–153. Garzonio, Jacopo & Cecilia Poletto. 2017. How bare are bare quantifiers? Some notes from diachronic and synchronic variation in Italian. Linguistic Variation 17(1). 44– 67. Garzonio, Jacopo & Cecilia Poletto. 2018. The distribution of quantifiers in Old and Modern Italian: Everything or nothing. In Ana Maria Martins & Adriana Cardoso (eds.), Word order change, 221–239. Oxford: Oxford University Press. Grewendorf, Günther. 2001. Multiple wh-fronting. Linguistic Inquiry 32. 87–122. Katz, Jerrold & Paul Postal. 1964. An integrated theory of linguistic descriptions. Cambridge, MA: MIT Press. Kayne, Richard S. 1975. French syntax. Cambridge, MA: MIT Press.
being bare: a survey of quantifier positions
213
Kayne, Richard S. 2004. Here and there. In Christian Leclère, Éric Laporte, Mireille Piot & Max Silberztein (eds.), Lexique Syntaxe, et Lexique-Grammaire / Syntax, Lexis & Lexicon Grammar: Papers in honour of Maurice Gross, 253–275. Amsterdam: Benjamins. Kelepir, Meltem. 2001. Topics in Turkish syntax: Clausal structure and scope. Cambridge, MA: MIT PhD dissertation. Kennelly, Sarah D. 1997. Nonspecific external arguments in Turkish. Dilbilim Araştırmaları [Research in Linguistics] 7. 58–75. Krapova, Iliyana & Guglielmo Cinque. 2008. On the order of wh-phrases in Bulgarian multiple wh-fronting. In Gerhild Zybatow, Luka Szucsich, Uwe Junghanns & Roland Meyer (eds.), Formal Description of Slavic Languages: The Fifth Conference, Leipzig 2003, 318–336. Frankfurt am Main: Peter Lang. Leu, Tom. 2014. The architecture of determiners. Oxford: Oxford University Press. May, Robert. 1985. Logical form. Cambridge, MA: MIT Press. Miyagawa, Shigeru, Danfeng Wu & Masatoshi Koizumi. 2018. Deriving case theory. Ms. MIT. OVI. Opera del Vocabolario Italiano data base (http://artfl‑project.uchicago.edu/conten t/ovi‑search‑form) Pesetsky, David. 2000. Phrasal movement and its kin. Cambridge, MA: MIT Press. Poletto, Cecilia. 2014. Word order in Old Italian. Oxford: Oxford University Press. Richards Norvin. 1997. What moves where when in which language? Cambridge, MA: MIT PhD dissertation. Richards, Norvin. 2001. Movement in language: Interactions and architectures. Oxford: Oxford University Press. Rizzi, Luigi. 2016. Labeling, maximality, and the head—phrase distinction. The Linguistic Review 33. 103–127. Rudin, Catherine. 1988. On multiple questions and multiple wh-fronting. Natural Language and Linguistic Theory 6. 455–501. Saito, Mamoru. 2016. (A) Case for labeling: Labeling in languages without phi-feature agreement. The Linguistic Review 33. 129–175. Svenonius, Peter. 2000. Quantifier movement in Icelandic. In Peter Svenonius (ed.), The derivation of VO and OV, 255–292. Amsterdam: Benjamins. Zamparelli, Roberto. 2008. Dei ex machina: A note on plural/mass indefinite determiners. Studia Linguistica 62(3). 301–327. Zidani-Eroğlu, Leyla. 1997. Indefinite noun phrases in Turkish. Madison, WI: University of Wisconsin-Madison PhD dissertation.
chapter 7
Indefinites as Fossils: The Case of wh-based Free Choice Maria Aloni
1
Introduction
Individual languages possess a wealth of indefinite forms that relate to each other in complex ways (Haspelmath 1997, Farkas 2002, von Heusinger 2019). English, for example, has at least four different indefinite pronouns: somebody, anybody, who, nobody. Other languages have more. These various forms typically differ in distribution and interpretation, but seem to have a common logical/semantic core. For example, consider any vs. some in English. There are contexts where they can be used without a notable meaning difference (If you hear something/anything, call me), others in which they cannot be interchanged (I didn’t meet someone/anyone, You may kiss someone/anyone, I kissed someone/#anyone). Wh-based free choice indefinites in Spanish and Italian behave like any in permissions and episodic sentences (the latter two contexts), but not under negation and, therefore, differ from any in this respect (Menéndez-Benito 2010, Aloni 2006, Chierchia 2013). Swedish, Norvegian whsom helst (Sæbø 2001), Lezgian x̂ ajit’ani, Hebrew kol and other indefinites in other languages appear to behave like Italian and Spanish free choice items. The CL-ote-series in Swahili, Latvian ar bith, French que ce soit, instead, seem to behave like English any (Haspelmath 1997). Interestingly, the German irgendseries exemplify yet another distribution/meaning pattern, resembling any in permissions, but being closer to some in episodic sentences (Kratzer and Shimoyama 2002, Aloni and Port 2010, 2015). Why so much cross-linguistic and language-internal variation in indefinite forms? What is the common core of these various indefinites? What is specific to each of them? In a seminal paper, Kratzer and Shimoyama (2002) proposed a formal analysis of indefinite meaning with the potential to give a principled answer to these questions. In this approach, the common meaning of different indefinite forms is identified in their potential to give rise to sets of propositional alternatives, which can then be bound by a variety of abstract operators. Differences among indefinite forms are then captured by assuming that different forms can associate with different such operators. This style of analysis has
© Maria Aloni, 2022 | doi:10.1163/9789004473324_008
indefinites as fossils: the case of wh-based free choice
215
reached considerable empirical success in explaining otherwise puzzling intervention effects (Kratzer 2005) and the distribution and meaning of a number of indefinite forms in different languages (e.g., Farkas 2006 for Romanian, Abrusán 2006 for Hungarian, Kim and Kaufmann 2006 for Korean, Aloni 2006 for Italian and Menéndez-Benito 2005, 2010 for Spanish). In the present article I will focus on the case of wh-based free choice indefinites and discuss Spanish and Dutch diachronic data from Aguilar-Guevara et al. (2011, 2012), which, as I shall argue, provide additional support for the abstract operators posited by analyses in Kratzer and Shimoyama’s style. Formal semantics and diachronic research are rarely combined (notable exceptions are Eckardt 2006, Deo 2015). One of the goals of this article is to show that this combination can lead to new insights and interesting questions worth future investigation. In languages with distinctive Free Choice (FC) morphology,1 a free choice inference is integrated into the conventional meaning of an indefinite form. Consider the case of Spanish plain indefinite determiner un (‘a’) vs FC indefinite determiner cualquier (‘any’). (1) Plain indefinite (Spanish un) a. Puedes traer un libro. (No traigas Guerra y Paz) can:2sg bring:inf a book not bring:imp War and Peace b. Conventional meaning: You can bring me a book c. Free choice inference: Each book is a possible option (2) Free choice determiner (Spanish cualquier) cualquier libro. (# No traigas Guerra y a. Puedes traer can:2sg bring:inf any book not bring:imp War and Paz) Peace b. Conventional meaning: You can bring me a book and each book is a possible option The free choice inference in (1) is a clear example of a conversational implicature, a defeasible pragmatic effect.2 In (2), instead, the same inference is 1 Dayal (1998), Giannakidou (2001), Jayez and Tovena (2005), Kellert (2022) and others. 2 Grice (1975), Gazdar (1979), Klinedinst (2006), Fox (2007), Aloni (2007), Franke (2011) and others.
216
aloni
no longer cancellable: adding the continuation ‘Don’t bring War and Peace’, which would contradict the inference, results in oddity.3 FC indefinites typically require a licensor to be felicitous. For example, Spanish cualquier is ungrammatical in episodic sentences (# Juan trajo cualquier libro ‘John brought cualquier book’). One of the challenges for a semantic theory of FC indefinites is to arrive at an account of their FC meaning, which also explains their restricted distribution. Menéndez-Benito (2005, 2010) and Aloni (2006) proposed analyses of Spanish and Italian FC indefinites, which meet this challenge. In these accounts, couched in Kratzer and Shimoyama’s alternative semantics, FC items necessarily associate with two covert operators, exh and [∀], with the latter quantifying over the sets of alternatives introduced by the indefinite form. One of the questions arising for these approaches, however, was how to justify the necessary association of FC indefinites with these abstract operators. In this article I will review two diachronic corpus studies on wh-based FC items from AguilarGuevara et al. (2011, 2012) and, in view of these data, I will conjecture that, at least in the studied cases, the association with exh and [∀] is inherited from earlier universal-like uses of the wh-based forms. The next section summarizes Menéndez-Benito (2005, 2010) and Aloni (2006)’s account of FC indefinites. Section 3 reviews the results of the Spanish and Dutch diachronic studies from Aguilar-Guevara et al. (2011, 2012). The first diachronic study, which investigates Spanish cualquier(a) from 1200s to 1900s, shows an already established indefinite, which is only marginally developing. This motivated the second study on Dutch FC item wie dan ook (‘who then also’), which instead clearly shows a wh-based FC indefinite “in status nascendi”. Section 4 proposes a semantic analysis of the development phases of the Spanish and Dutch indefinite forms in the style of Eckardt (2006). Kratzer and Shimoyama’s style abstract operators, which can be argued to be inherited from early uses of the indefinite wh-morphology, will play a crucial role in the explanation of the diachronic change these constructions went through before they reached their current meaning and distribution.
3 An attractive idea, already present in Grice’s seminal work, is that specialized indefinite forms may have emerged as result of historical processes of conventionalisation of originally pragmatic inferences: “[…] it may not be impossible for what starts life, so to speak, as a conversational implicature to become conventionalized […].” (Grice 1975: 58).
indefinites as fossils: the case of wh-based free choice
2
217
Indefinites in Alternative Semantics
Alternative-based analyses of indefinites identify the common meaning of different indefinite forms in their potential to generate sets of propositional alternatives (Kratzer and Shimoyama 2002, Kratzer 2005), just like questions do (Hamblin 1973, Karttunen 1977, Groenendijk and Stokhof 1984). Sentences containing the indefinites forms in (3a) induce the set of alternatives represented in (3b): (3) a. Somebody/anybody/nobody/who fell
b. d1 fell d2 fell d3 fell …
Such a set of propositions can be bound by a variety of operators with different quantificational force. Examples of such operators are defined in (4) (where W is the logical space, i.e. the set of all possible words, and A ⊆ Pow(W) is a set of propositions).4 (4) a. b. c. d.
[∃](A) = ⋃(A) [∀](A) = ⋂(A) [Neg](A) = W ∖ ⋃(A) [Q](A) = A
The core idea is that different indefinite forms have emerged as an indication of necessary association with different matching operators as illustrated in (5) (see Onea 2022 for a different but related view): (5) a. b. c. d. e.
[∃] (somebody fell) [∀] (anybody fell) [Neg] (nobody fell) [Q] (who fell) …
f. d1 fell d2 fell d3 fell …
4 Technically the framework of Alternative Semantics assumed here faces a number of wellknown problems when it comes to meaning composition (see, e.g., Shan 2004). Various solutions have been proposed, including the inquisitive one defended by Ciardelli et al. (2017). The analysis presented in this paper can be easily reformulated in an inquisitive semantics setting by adding there the relevant abstract operators. Given that the technicalities of meaning composition are not the main concern of this short contribution, I will use the somehow old-fashioned alternative semantics formulations, which are easier to parse for a non technically-oriented reader.
218
aloni
In (5), the expressions in parenthesis denote one and the same set of propositional alternatives, illustrated in (5f). The various propositional operators in (5a) to (5e) quantify over this set. It is easy to see that the analysis of FC any proposed in (5b) is too simpleminded as it would predict a universal interpretation for a sentence like (6) with no explanation of its ungrammaticality: (6) #Anyone fell. a. [∀] (anyone fell) d1 fell d2 fell d3 fell … b. [∀](anyone fell) = [∀] ({that d1 fell, that d2 fell,…}) = {that everyone fell} Additionally, the truth conditions that would be predicted for the grammatical sentence (7) have been argued by Menéndez-Benito (2005) to be too weak: (7) Anyone can fall. a. [∀](anyone can fall) ◇ d1 fall ◇ d2 fall ◇ d3 fall … b. [∀](anyone can fall) = [∀] ({that d1 can fall, that d2 can fall,…}) = {that anyone can fall} According to Menéndez-Benito, the meaning representation in (7b) fails to capture the unrestricted freedom of choice expressed by the modal sentence. Consider the following scenario (Menéndez-Benito 2005: 60–63): (8) One of the rules of the card game Canasta is: when a player has two cards that match the top card of the discard pile, she has two options: (i) she can take all the cards in the discard pile or (ii) she can take no card from the discard pile (but take the top card of the regular pile instead). In this scenario, (9) is judged false. An analysis along the lines of (7), however, would predict (9) to be true. (9) In Canasta, you can take any of the cards from the discard pile when you have two cards that match its top card. To fix these problems, Menéndez-Benito (2005) assumed that the interpretation of universal free choice items involves the application of an exclusiveness operator, Excl, which transforms Hamblin alternatives into sets of mutually exclusive propositions. Applying [∀] immediately after Excl as in (10) yields a contradiction which explains the ungrammaticality of (6). In (11) the modal
219
indefinites as fossils: the case of wh-based free choice
operator ‘intervenes’ which avoids the contradiction and delivers the desired universal free choice meaning. (10) #Anyone/Cualquiera fell. alt: only d1 fell only d2 fell … a. [∀](Excl(cualquiera fell)) b. [∀] ({that only d1 fell, that only d2 fell,…}) = ⊥ (11) Anyone/Cualquiera may fall. alt: ◇ only d1 fell a. [∀](◇(Excl(cualquiera fall))) b. [∀] ({that ◇ only d1 fell, that ◇ only d2 fell,…}) ≠ ⊥
◇ only d2
fell …
Aloni (2006) extended Menéndez-Benito’s analysis employing rather than Excl a more general operation of exhaustification, exh, assumed to play a role also in the semantics of free relatives and wh-interrogatives:
(12) a. Free relative: John ate [DP what Bill cooked] b. Wh-interrogative: John knows [Q what Bill cooked]
Building on Cooper (1983) and Jacobson (1995), Aloni assumed that free relatives and wh-interrogatives are born with the same meaning, a predicative meaning, but type-shift differently: free relatives type-shift into an entitydenoting expression, wh-interrogatives into a proposition-denoting one. (13) what Bill cooked a. ( John ate) [DP what Bill cooked] b. ( John knows) [Q what Bill cooked]
type: ⟨e, ⟨s, t⟩⟩ type: e type: ⟨s, t⟩
The common meaning of (13a) and (13b) is an exhaustive property of type ⟨e, ⟨s, t⟩⟩ denoting the set of pairs ⟨x, v⟩ where x is the maximal collection of things that Bill cooked in v:
(14) what Bill cooked type: ⟨e, ⟨s, t⟩⟩ a. exh[what, λx. Bill cooked x] b. {λxλv. x is the maximal collection of things that Bill cooked in v} In the case of wh-interrogatives, this property will type-shift into a question/ proposition denotation via the shift⟨s,t⟩ rule, which as output yields a partition of the logical space:5 5 If exh[αe , β⟨e,⟨s,t⟩⟩ ] of type ⟨e, ⟨s, t⟩⟩ denotes the property of exhaustively satisfying β with
220
aloni
(15) ( John knows) [Q what Bill cooked] a. shift⟨s,t⟩ (exh[what, λx. Bill cooked x]) type: ⟨s, t⟩ b. {Bill cooked nothing, Bill cooked only d1 , Bill cooked only d2 , Bill cooked only d1 and d2 , … } In the case of free relatives, the same property can type-shift into a DP denotation via the shifte rule.
(16) ( John ate) [DP what Bill cooked] a. shifte (exh[what, λx. Bill cooked x]) b. {the maximal collection of things that Bill cooked in w0 }
type: e
Via point-wise functional application this denotation combines with the denotation of the rest of the sentence to yield the singleton set containing the proposition that John ate the things that Bill cooked. Eventually this set will be bound by [Q], the operator wh-pronouns necessarily associate with. Here, [Q] is the identity function as defined in (4). Therefore, as illustrated in (17), this analysis explains the definite reading that the free relative obtains in this sentence. (17) a. John ate [DP what Bill cooked] b. [Q] ( ate(j)(shifte (exh[what, λx. Bill cooked x]))) c. {that John ate the things that Bill cooked in w0 }
In Aloni’s (2006) proposal (Italian) wh-based FC indefinites trigger the application of exh, just like wh-words do, but rather than necessarily associating with [Q] as plain wh-pronouns do, they associate with a universal propositional quantifier [∀]: (18) a. Plain wh-pronoun: [Q]…exh[who, λxψ(x)]… b. Wh-based FC item: [∀]…exh[fci, λxψ(x)]…
On this analysis, wh-based FC items are correctly predicted to be ungrammatical in episodic sentences (and under necessity modals), and FC inferences under possibility modals are readily derived as semantic entailments as in respect to domain α, shifte (exh[αe , β⟨e,⟨s,t⟩⟩ ]) of type e denotes the maximal set of individuals from domain α which satisfies β, and shift⟨s,t⟩ (exh[αe , β⟨e,⟨s,t⟩⟩ ]) of type ⟨s, t⟩ denotes the partition determined by the question ‘which individuals from domain α are β’. See Aloni (2006) for definitions. Notice that to obtain partitions, Aloni assumes domains of plural individuals including ∅.
indefinites as fossils: the case of wh-based free choice
221
Menéndez-Benito (2005, 2010), but improving on Menéndez-Benito also subtrigging (which in Section 3 will be called UFC uses) is easily explained (as well as universal readings in comparative clauses, such as Mary is taller than anyone else, see Aloni and Roelofsen 2014): (19) #Anyone fell. a. [∀](shift⟨s,t⟩ (exh[anyone, fell])) b. nobody fell only d1 fell only d2 fell … (20) Anyone can fall. a. [∀](◇(shift⟨s,t⟩ (exh[anyone, fall]))) b. ◇ nobody fell ◇ only d1 fall ◇ only d2 fall … (21) Anyone who tried to jump fell. a. [∀](↓shifte (exh[anyone, who tried to jump]) fell) b. d1 fell d2 fell In the first two structures the value produced by exhaustification undergoes the shift(s,t) rule yielding the partition represented in (19b). In (19a), each alternative in this partition is stated to be true resulting in a contradiction. This explains why universal FC items are out in plain episodic sentences. In (20) the elements of the partition are further expanded by the modal operator. Universal quantification in this case does no longer result in a contradiction. Finally, in the subtrigged case (21), exhaustification crucially occurs inside the DP. Therefore, the value it produces undergoes the Shifte rule yielding as output in w the sum of people who tried to jump in w. To avoid trivial quantification, ↓ applies to this sum to produce a set of singular individuals.6 The VP denotation ⟦fell⟧w,g applies to the latter set producing the set of Hamblin alternatives represented in (21b). Since this set occurs in the scope of a universal operator, the sentence obtains the desired interpretation: everyone who tried to jump fell. One of the questions arising for Aloni’s (2006) approach, however, was how to justify the necessary association of FC items with exh and [∀]. In the next section we summarize the results of two diachronic studies conducted on whbased free choice items presented in Aguilar-Guevara et al. (2011) and then conjecture that at least for the studied cases, while exh plausibly comes from 6 The operation ↓ maps plural individuals back into their atomic elements. (i) Illustration: a. ⟦α⟧w,g = {a + b} a singleton set of plural entities b. ⟦ ↓ α⟧w,g = {a, b} a multi-membered set of atomic alternatives
222
aloni
wh-morphology, the emergence of [∀] in (18b) is triggered by earlier universallike uses of the wh-based forms. 3
Diachronic Studies
In this section we summarize the results of two corpus based diachronic studies presented in Aguilar-Guevara et al. (2011, 2012) investigating the development of Spanish cualquier(a) and Dutch wie dan ook. These constructions, in addition to the free choice meaning, share the property of being compounds and containing an interrogative word meaning ‘who’ or ‘which’ within their constituents. Section 4 will then present a semantic analysis of the (conjectured) phases of development of these two items building on Aloni’s (2006) account, presented in the previous section. In total, Aguilar-Guevara et al. (2011, 2012) conducted synchronic and diachronic corpus studies of free choice (FC) and epistemic indefinite (EI)7 forms from six different languages.8 In the synchronic research they studied the following items: English FC any, Czech FC kterýkoli, Italian FC (uno) qualunque, Spanish FC cualquier(a), Dutch FC wh dan ook and German EI irgend-series. In the diachronic research they studied Spanish cualquier(a), Dutch wie dan ook and German irgend-series. This section, which largely overlaps with AguilarGuevara et al. (2011), summarizes the key findings of the Spanish and Dutch diachronic portion of this research.9 I will first briefly describe the methodology adopted and report on the relevant synchronic results.
7 Jayez and Tovena (2006), Alonso-Ovalle and Menéndez-Benito (2010), Zamparelli (2007), Aloni and Port (2010, 2015), Gianollo (2018) and others. 8 The corpus studies reported in this section were conducted as part of a NWO-funded project that ran from 2008 to 2012 at the University of Amsterdam and are joint work with Ana Aguilar-Guevara (Spanish), MA (Italian), Angelika Port (German), Radek Šimík (Czech), Stephanie Solt (English) and Machteld de Vos (Dutch). Tikitu de Jager, Hedde Zeijlstra and Katrin Schulz were also involved in various phases of this project. All annotated data are available at https://osf.io/z2j9e/. See Aguilar-Guevara et al. (2012) for full documentation. 9 For an analysis of the German diachronic data, see Port and Aloni (2021). The German data are not included in the present discussion because irgend-indefinites are better classified as epistemic indefinites than free choice indefinites. In Alternative Semantics they are indeed analysed as necessarily associating with the existential propositional quantifier [∃] and not the universal [∀] (Kratzer and Shimoyama 2002).
indefinites as fossils: the case of wh-based free choice
223
3.1 Methodology In these corpus studies coders annotated randomly selected occurrences of indefinite forms according to a number of categories. The starting point for the identification of the relevant categories was Haspelmath’s (1997) typological survey. Haspelmath identified 9 main functions for indefinite forms organized in an implicational map. Aguilar-Guevara et al. (2011, 2012) assumed the following extended version of Haspelmath’s map motivated by a more detailed classification for Negative Polarity and Free Choice Items. The newly introduced functions are in boldface in the following illustrations: (22) An extended version of Haspelmath’s map
(23) Functions on the map Abbr Label a. SK specific known b. SU specific unknown
Example Somebody called. Guess who? I heard something, but I couldn’t tell what. c. IR irrealis You must try somewhere else. d. Q question Did anybody tell you anything about it? e. CA conditional antec. If you see anybody, tell me immediately. f. CO comparative John is taller than anybody. g. DN direct negation John didn’t see anybody. h. AM anti-morphic I don’t think that anybody knows the answer. i. AA anti-additive The bank avoided taking any decision. j. FC free choice You may kiss anybody. k. UFC universal free choice John kissed any woman with red hair. l. GEN generic Any dog has four legs. A function in this framework can be identified with a pair consisting of a syntactic context and a semantic interpretation. In order for an indefinite to qualify for a function, it must (i) be grammatical in the syntactic context the function specifies; and (ii) have the semantics that the function specifies. For
224
aloni
example, any does not qualify for the specific functions SK and SU because it is ungrammatical in episodic sentences, marked as # in (24), while somebody does not qualify for the comparative function, CO, or the free choice function, FC, because it does not have the universal meaning these functions specify, marked as * in (25): (24) a. Somebody/#anybody called. Guess who? b. I heard something/#anybody, but I couldn’t tell what.
[SK] [SU]
(25) a. You may kiss anybody/*somebody. ‘For every individual x it holds that you may kiss x.’ b. John is taller than anybody/*somebody. ‘For every individual x it holds that John is taller than x.’
[FC] [CO]
During annotation the functions in (23) were identified with logico-semantic interpretations and a number of diagnostic tests organized in a decision tree were used to assign to each instance of an indefinite a function on the map.10 In ambiguous cases, such as (26), if the context did not disambiguate the intended reading, the instances were annotated with both possible functions. To keep the randomly chosen occurrences stable the readings were counted as 0.5. (26) If she can solve any problem, she’ll get a prize. a. (‘existential’) If there is any problem she can solve, … b. (‘universal’) If she can solve every problem, …
[CA] [FC]
While these tests proved useful for most cases, there were examples for which the decision tree was inconclusive. For some of these cases Aguilar-Guevara et al. (2011, 2012) introduced new off-map functions, such as the indiscriminacy function, IND, which was added to cover examples like (27) from Horn (2005) (see also Kellert 2022).
10
An assessment of the methodology by measuring inter-annotator agreement with the kappa coefficient has been carried out in January 2011. Five annotators coded 100 randomly chosen examples from the British National Corpus. Each example contained one marked occurrence of some (20 examples) or any (80 examples). The average kappa score obtained was poor in general (kappa: 0.52), but it improved significantly (kappa: 0.69) when internal distinctions within the specificity area and the negative area were disregarded. For details see Aloni et al. (2012).
indefinites as fossils: the case of wh-based free choice
225
Indiscriminacy (27) I do not want to go to bed with just anyone anymore. I have to be attracted to them sexually. (English) For the diachronic studies more off-map functions were added to label uses which were not strictly indefinite, like no matter, adposition, and free relative uses, for which we give here an illustration in Czech, English and Italian: No matter jsme v kterékoli zemi, všude nacházíme slušné (28) Ať už let already be:1pl in any country everywhere find:1pl polite lidi. (Czech) people ‘No matter in which country you are, you can find polite people everywhere.’ Adposition (29) You may choose an apple, any apple.
(English)
Free Relative (30) Ha aiutato chi è caduto. have:3sng helped who be:3sng fallen ‘She helped who fell.’
(Italian)
The synchronic studies attested the following distributions for the Dutch and Spanish indefinites on the extended Haspelmath’s map: (31) Spanish cualquier(a)
226
aloni
(32) Dutch wh dan ook
In what follows we summarize the main results of the diachronic research on these two items. 3.2 Results of Diachronic Studies 3.2.1 Spanish cualquiera Cualquiera (pronoun), or cualquier (determiner), translated to English as whatever, whichever, whoever or any, and composed of cual (‘which/who’) plus quier(a) (‘want:3.pres.subj’), has been claimed to have emerged in Spanish as result of a grammaticalization process through which free relative clauses were reanalyzed as indefinite noun phrases (see Company Company and Pozas-Loyo 2009). (33) Hypothesized grammaticalization process for ‘cualquiera’ a. Free relative clause Haga en él cual castigo quiera. do on him which punishment want:3.pres.subj b. Phrasal compound Haga en él cual quiera castigo. do on him which want:3.pres.subj punishment c. Indefinite Haga en él cualquier(a) castigo. do on him whichever punishment Aguilar-Guevara et al. (2011) could not witness this process, which presumably occurred in early stages of the history of Spanish. Figure 1 shows the number of occurrences of the construction they found in El Corpus del Español in the four periods selected for their study. As the graph illustrates, recurrent occurrences of cualquiera, as a word, are found already in the first documentations of Spanish, which date back to the thirteenth century. Furthermore, the pres-
indefinites as fossils: the case of wh-based free choice
figure 1
Number of occurrences of cualquiera per million of words adapted from aguilar-guevara et al. 2011
figure 2
Functions covered by cualquiera in 1200s, 1500s, 1700s and 1900s adapted from aguilar-guevara et al. 2011
227
ence of cualquiera doubles between the 1200s and the 1500s, reaching a similar proportion to that documented for the 1900s. Aguilar-Guevara et al. took the latter fact as an indicator that the use of the construction is consolidated at least since the sixteenth century. The distribution of the functions on the extended Haspelmath’s map that cualquiera covered throughout these periods is given in Figure 2. As the graph illustrates, the use of the construction remained pretty stable throughout the centuries with the FC function as clearly the most dominant since the first period.
228
aloni
Interestingly, the UFC function displays a remarkable decrease starting in the 1500s, when the use of the construction consolidated. Aguilar-Guevara et al. tentatively attributed this to the fact that cualquiera, as part of its grammaticalization, occurs less and less frequently accompanied by post-nominal modifiers such as restrictive relative clauses and prepositional adjuncts, which typically serve as licensor of free choice items in subtrigging UFC uses (e.g., John kissed any woman #(with red hair)). I will return to this issue in the analysis section. The last important observation is that two more off-map functions, namely IND and no matter, appear in the 1500s and gain presence by the 1900s. The late emergence of the no matter function will turn particularly interesting in light of the development of the Dutch indefinite wie dan ook. As said, given the early grammaticalization of cualquiera, and stable distribution of its functions, Aguilar-Guevara et al. (2011) could not really attest much of the process this compound went through in order to behave as it does nowadays. This motivated the study of wie dan ook, an indefinite comparable to cualquiera in meaning and (partly) in form, but that emerged in Dutch more recently and that even in these days appears to be ‘less’ grammaticalized than its Spanish counterpart. 3.2.2 Dutch wie dan ook The Dutch diachronic study consisted of the analysis of occurrences of wie dan ook (‘who then also’) in written Dutch historical corpora (CD-ROM Middelnederlands (270 texts before 1300), DBNL (4458 text from 1170–2010)) (de Vos 2010). The first occurrence found was from 1777. The period of this indefinite’s existence was therefore divided into four phases, each covering 55 years of the indefinite’s evolution. The outcome, illustrated in Figure 3, shows that wie dan ook went through a four-staged process of grammaticalization. Aguilar-Guevara et al. (2011) describe the four stages as follows: 3.2.2.1 Stage i The first phase in the grammaticalization of wie dan ook as an indefinite is formed by three forms of the no matter-function. Characteristic of types of no matter constructions is that the wh dan ook is not part of the main clause yet: they all consist of either a wh-clause and a main clause, or a wh-clause within a main clause, as illustrated in (34): (34) a. Wie dan ook naar het feest komt; ik zal blij zijn. ‘Whoever comes to the party; I will be happy.’
indefinites as fossils: the case of wh-based free choice
figure 3
229
Four stages in the grammaticalization of wie dan ook adapted from aguilar-guevara et al. 2011
b. [Wie dan ook naar het feest komt]i ; hiji zal blij zijn. ‘[Whoever comes to the party]i ; hei will be happy.’ c. Jan, (of ) wie dan ook hij mag zijn, zal blij zijn. ‘Jan, (or) whoever he may be, will be happy.’ These forms occur around the same time. Together, they seem particularly frequent in the first phase, forming a significant majority of the total amount of occurrences here, with this relative amount decreasing in the three phases that follow (see the brown bars in Figure 3). 3.2.2.2 Stage ii In the following stage in the development of wie dan ook as an indefinite, no matter constructions are shortened to so-called adpositions, thus getting one step closer to becoming a grammaticalized indefinite. Adpositions have the following form: […, [wie dan ook], … ]. They are shortenings of the no matterfunction, formed by the ellipsis of the predicate. Although they do not form a separate wh-clause next to or within a main clause anymore, they are still not part of the actual sentence and therefore no real indefinites: they merely modify the noun they are placed after, conveying an ignorance inference as in (35) or an indifference meaning, as in (36):
230
aloni
(35) Jan, wie dan ook, is blij. Jan who then also is happy ‘Jan, whoever, is happy.’ (36) Als er iemand, wie dan ook, naar het feest komt, zal ik if there someone who then also to the party comes will I blij zijn. happy be ‘If someone, whoever/anyone, comes to the party, I will be happy.’ As the orange bars in Figure 3 show, this adpositional modification with a wie dan ook is particularly frequent in the second phase in the development of this indefinite. 3.2.2.3 Stage iii The third phase, the free relative-stage, shows a further integration of the wie dan ook-clause into the sentence, though still not a full integration either. The Free Relative (FR) function, the most frequent use in this phase, forms another spinoff of the no matter construction. However, whereas no matter constructions still form combinations of wh-clauses (wie dan ook + predicate) and a main clause, the FR-function is more integrated than that, with the wie dan ook + predicate not forming a separate clause, but an actual part of the main clause, typically the subject. Examples of the FR-function have the following form: [[wie dan ook + predicate](,) VP], as illustrated in (37): (37) Wie dan ook naar het feest komt(,) zal blij zijn. who then also to the party comes will happy be ‘Whoever comes to the party(,) will be happy.’ However, these subjects consisting of wie dan ook + predicate are often followed by a comma, thereby perhaps indicating that they are still seen as slightly standing outside of the actual sentence. Yet omitting the part starting with wie dan ook would give an incomplete thus ungrammatical sentence. This is a specific feature of the FR-function dominating the third phase; both the no matterclauses and the adpositions can still be left out, of course sometimes causing a change in meaning of the sentence, but never with an incomplete sentence as a result.
indefinites as fossils: the case of wh-based free choice
231
3.2.2.4 Stage iv In this last stage of the grammaticalization of wie dan ook, the word group has finally become an indefinite. Examples of this kind form integrated parts of the sentence, with a plain wie dan ook, without any kind of predicate modifying it, being either subject or object: […[wie dan ook]…]. (38) Je mag wie dan ook uitnodigen voor het feest. you may who then also invite for the party ‘You may invite anyone to the party.’ Indefinite uses of wie dan ook exist from 1833 onwards, and their number increases in every phase, finally forming a vast majority of the occurrences in the fourth phase, as the graph in Figure 3 illustrates. Overall, Aguilar-Guevara et al. (2011) concluded that the process of grammaticalization of wie dan ook as an indefinite roughly followed four stages, starting off as a no matter construction in a separate wh-clause, slowly evolving into an adpositional modifier on its own, while also turning into a part of the main clause with predicate, eventually yielding to the true and plain indefinite wie dan ook as part of a sentence. The initial hypothesis of Aguilar-Guevara et al. (2011) was that FC indefinites emerged as the result of a process of conventionalization of an originally pragmatic inference. The envisaged ‘conventionalization’ was in fact quite difficult to test because conversational implicatures are by definition not overtly expressed. The testing would have to consist in checking for a raising frequency of a conversational implicature of sentences with unmarked indefinites, then a development of a morpheme which captures the implicature and then its grammaticalization. Alternatively, the morpheme that had already been used in a plain indefinite use would change its function—the implicature would be built in. The latter is not what was found for Dutch (or conjectured for Spanish). Yet, the observed development of wie dan ook is consistent with the former, with the various appositive constructions within wie dan ook started out its life as forms which express the original implicature and later get grammaticalized. More precisely, the grammaticalization path that was found for wie dan ook could be interpreted as a path from a conversational implicature, via a conventional implicature in the sense of Potts (2005)11 to a conventional meaning (i.e. core / at-issue semantics).
11
According to Potts (2005), adpositives express conventional implicatures, i.e. not at-issue meanings.
232
aloni
(39) a. Jij mag iemand uitnodigen. (plain indefinite + conversational implicature) ‘You may invite someone’ b. Jij mag iemand, wie dan ook (hij mag zijn), uitnodigen. (plain indefinite + conventional implicature) ‘You may invite someone, whoever (he may be)’ c. Jij mag wie dan ook uitnodigen. (new FC indefinite) ‘You may invite anyone’
4
The Emergence of wh-based FC: Towards an Analysis
From the research summarized in the previous section we can conclude that in the development of Spanish cualquiera and Dutch wie dan ook the same constructions were involved though not in the same order: Spanish: free relative (conjectured) > free choice indefinite > no matter uses Dutch: no matter uses > adposition > free relative > free choice indefinite
In this section, extending Aloni (2006), I propose a semantic analysis of these different phases where differences in meaning and distribution among free choice indefinites, free relatives, adpositions and no matter constructions are derived by different combinations of a small number of independently motivated semantic operations. In view of this analysis, the development of Spanish and Dutch free choice can be explained in terms of changes that affect these different operators and their possible combinations. A further consequence concerns the nature of the conventionalization of the free choice implicature that the Dutch and Spanish cases appear to illustrate. In an influential approach, Chierchia (2013) proposed to treat the free choice inference triggered by FC indefinites as a type of scalar implicature resulting from the application of a grammatical version of an operation of exhaustification. The diachronic analysis presented here suggests a different explanation: the free choice inference triggered by these indefinites is not the result of exhaustification but rather follows from the application of an operator of universal quantification [∀] inherited from the constructions the newly developed free choice items derive from: (universally read) free relatives in the case of Spanish and no matter constructions in the case of Dutch. Let us have a closer look.
indefinites as fossils: the case of wh-based free choice
233
4.1 The Emergence of Spanish Free Choice Grosu and Landman (1998) observed that free relatives are ambiguous between a definite and a universal reading as illustrated in (40): (40) We will veto three-quarters of whatever proposals you make. a. Of the proposals: three-quarters won’t make it. b. For each proposal: three-quarters of it will be vetoed.
(definite) (universal)
Aloni (2006) captures the difference between (40a) and (40b) by assuming that the latter is further bound by a propositional universal quantifier: (41) We will veto three-quarters of whatever proposals you make. a. FR: [Q](P(shifte (exh[whatever, S]))) b. UFR: [∀][Q](P(↓shifte (exh[whatever, S])))
(definite) (universal)
As we saw in Section 2, (41a), which denotes a singleton set of propositions, immediately characterizes the definite reading of the sentence. To characterize the universal reading, we need to further apply [∀] (and ↓ to avoid vacuous quantification). We can then conjecture that the emergence of [∀] in association with Spanish cualquier(a) was triggered by early universally read free relative uses of the wh-form, as illustrated in (42). (42) Conjectured development of Spanish cualquier(a) a. FR: [Q](P(shifte (exh[item, S]))) (free relative) (universal free relative b. UFR: [∀][Q](P(↓shifte (exh[item, S]))) (UFR)) c. UFC: [∀](P(↓shifte (exh[item, S]))) (UFC indefinite) (FC indefinite) d. FC: [∀](◇(shift⟨s,t⟩ (exh[item, S]))) In agreement with hypothesis (33) concerning the grammaticalisation of cualquiera, we conjecture that the wh-form cual (‘what’) originally employed in free relative constructions (42a) started combining with a propositional quantifier [∀] to generate universal reading of free relative uses as in (42b). These then developed into subtrigged UFC uses where a wh-based indefinite directly associates with [∀] as illustrated in (42c). Both UFR and UFC uses employ [∀] in combination with an individual shifted notion of exhaustification shifte (exh[α, β]). Only in a later phase full-fledged FC uses emerged where [∀] combines with a propositional shifted exhaustification, shift⟨s,t⟩ (exh[α, β]), which we can assume was already present in the language to
234
aloni
generate ordinary wh-interrogative uses. In yet a later phase no matter uses emerged as well but as the Spanish diachronic data demonstrate these were not instrumental to the development of the free choice indefinite. Although Aguilar-Guevara et al. (2011) did not witness the emergence of cualquier(a) in the corpus study, the high frequency of UFC uses in phase one of their dataset (see the orange bar in Figure 2) would have a natural explanation if cualquier(a) were indeed born with the UFC function as conjectured in (42). Dutch wie dan ook, instead, followed a different development pattern with no matter as its first use. As explained below, no matter constructions require the application of [∀] giving rise to a different explanation of what triggered the emergence of [∀] in the development of Dutch free choice. 4.2 The Emergence of Dutch Free Choice Recall that Aguilar-Guevara et al. (2011) have encountered four uses of wie dan ook in their diachronic study: 1. No matter
wie dan ook + predicate, [main clause]
(43) Wie dan ook naar het feest komt; ik zal blij zijn. ‘Whoever comes to the party; I will be happy.’ 2. Adposition
…, [wie dan ook], …
(44) Als er iemandi , wie dan ooki , naar het feest komt, zal ik blij zijn. ‘If someone, whoever/anyone, comes to the party, I will be happy.’ 3. Free relative
[wie dan ook + predicate] (,) VP
(45) Wie dan ook naar het feest komt(,) zal blij zijn. ‘Whoever comes to the party(,) will be happy.’ 4. FC indefinite
… [wie dan ook] …
(46) Je mag wie dan ook uitnodigen voor het feest. ‘You may invite anyone to the party.’ The free relative and FC indefinite examples will be analysed as in Aloni (2006). Henceforth I will write exha [α, β] to indicate shifta (exha [α, β]).
indefinites as fossils: the case of wh-based free choice
235
(47) a. Wie dan ook naar het feest komt(,) zal blij zijn. b. [Q](happy(exhe [wie dan ook, λx.x comes to party])) (definite) c. [∀][Q](happy(↓ exhe [wie dan ook, λx.x comes to party])) (universal) ‘Whoever comes to the party(,) will be happy.’ (48) a. Je mag wie dan ook uitnodigen voor het feest. b. [∀](◇(exh⟨s,t⟩ [wie dan ook, λx.you invite x to the party])) ‘You may invite anyone to the party.’ In what follows I extend Aloni (2006) with an explicit analysis of no matterconstructions and adpositions. 4.2.1 No Matter I propose to analyze no matter-constructions as unconditionals building on Rawlins (2008): (49) a. Wie dan ook naar het feest komt; Jan zal blij zijn. b. [∀](exh⟨s,t⟩ [wie dan ook, λx.x comes to party](λi □i ϕ)) ‘Whoever comes to the party; Jan will be happy.’ The wie dan ook clause acts here as a restrictor of the domain of quantification of the modal in the main clause as in Rawlins (2008). Thus no matterconstructions seem to require a modal (‘zal’ or similar) in the main clause and this appears to be confirmed by the corpus data. In the present formalization, sentence (49) asserts that all of the following propositions must be true: If nobody comes, Jan will be happy;12 If only a comes, Jan will be happy; if only b comes, Jan will be happy, and so on. This is enough to capture the indifference flavour of the sentence (‘It doesn’t matter who comes to the party, Jan will be happy in any case’). Furthermore, since the wie dan ook clause denotes a partition of the logical space, the main clause ‘Jan will be happy’ is entailed by (49b) as it should be. Note that in (49), the wie dan ook clause cannot be interpreted on its own and must be integrated in the unconditional construction. Indeed wie dan ook naar het feest komt (‘wie dan ook comes to the party’) is ungrammatical in isolation. This is captured in this analysis because [∀](exh⟨s,t⟩ [wie dan ook, λx.x comes to party]) is a contradiction. 12
This is different from Rawlins (2008), who does not have the ‘nobody’ alternative and therefore does not produce partitions of the logical space.
236
aloni
To be interpretable on its own the wie dan ook clause would need the presence of an operator which intervenes between [∀] and exh, for example a possibility modal like mag (‘may’). This is precisely what appears to happen in the following stage of the development of the Dutch item, the adposition phase (Stage ii), or at least so we conjecture. In adpositions, the wie dan ookclause gets interpreted independently (possibly on a different level) of the main clause. We might conjecture that frequent cases with explicit mag (‘may’) inside the wie dan ook-clause in the late no matter phase triggered the transition to the next adposition phase where wie dan ook-clauses with (implicit) mag (‘may’) get an independent interpretation. 4.2.2 Adpositions Aguilar-Guevara et al. (2011) found two variants of ‘adpositions’ in their data: 1. Clause adposition:13 …, wie dan ook P, … (50) Jan, wie dan ook hij mag zijn, is blij. ‘Jan, whoever he may be, is happy.’ 2. Plain adposition: …, wie dan ook, … (51) Jan, wie dan ook, is blij. ‘Jan, whoever, is happy.’ In the corpus they encountered and labeled two possible interpretations for plain adpositions: 1. Indifference: (52) Jij mag iemand, wie dan ook, uitnodigen. ‘You may invite someone, anyone.’ 2. Ignorance: (53) Jan, wie dan ook, is blij. ‘Jan, whoever (Jan may be), is happy.’
13
Clause adpositions were annotated as no matter (type UN3) in Aguilar-Guevara et al. (2011), see (34c).
indefinites as fossils: the case of wh-based free choice
237
In clause adpositions and other no matter-constructions we have two propositions, which must somehow be integrated: one expressed by the main clause, and the other by the wie dan ook-clause. (54) a. main clause: ϕ b. wie dan ook clause: wie dan ook P In ordinary no matter-constructions these two propositions are integrated in a conditional structure, as we saw above. In clause adpositions, instead, the wie dan ook clause is interpreted on its own: building on Dayal (2004) (and somewhat simplifying) we assume that it is connected to the main clause by wide scope conjunction (see also Potts 2005): (55) a. No matter (unconditional): wie dan ook P ⇒ ϕ b. Clause adpositions: ϕ and wie dan ook P
We further assume that wie dan ook adpositions always contribute a proposition. If the predicate P is not overtly given, as in the case of plain adpositions, it must be pragmatically supplied or syntactically reconstructed. We assume the following possible resolutions/reconstructions for P in these cases: (56) a. P is resolved to the predicate of the main clause ϕ b. P is reconstructed as x mag zijn (‘x may be’) As we will see the first resolution is only possible if ϕ is a FC licensing context, and will produce indifference meanings. The second resolution is only possible if ϕ provides a proper antecedent for anaphor x and produces ignorance meanings. Let’s have a closer look. As we said, building on Dayal, we assume that in adpositions the wie dan ook clause is connected to the main clause by wide scope conjunction:14 (57) ϕ and wie dan ook P Such construction requires a FC licensing operator in the wie dan ook clause, otherwise the second conjunct, analyzed as (58), would be contradictory.
14
Or maybe disjunction since many examples from the corpus have explicit of (‘or’), but there are other possible explanations for this, e.g., of as adposition marker.
238
(58) [∀] Op (exh⟨s,t⟩ [wie dan ook, λx.ψ])
aloni
For the cases of plain adjunction, if ϕ, the main clause, contains such a FC licensor, we resolve Op and ψ in (58) to elements of ϕ and obtain an indifference reading. Example (59) illustrates such a resolution: (59) a. Jij mag iemand, wie dan ook, uitnodigen voor het feest. b. [∃]◇([iemand, λx.you invite x to the party]) ∧ [∀]◇(exh⟨s,t⟩ [wie dan ook, λx.you invite x to the party]) c. Predicted meaning: ‘You can invite someone and you can invite anyone’ If ϕ does not contain such a licensor, we reconstruct P as ‘x mag zijn’, and we obtain an ignorance reading: (60) a. Jan, wie dan ook (hij mag zijn), is blij. b. happy(j) ∧ [∀]◇(exh⟨s,t⟩ [wie dan ook, λx.x = j]) c. Predicted meaning: ‘Jan is happy and Jan might be anyone.’ To summarize, I propose the following analyses for the different uses of wie dan ook that were found in the diachronic corpus of Aguilar-Guevara et al. (2011): 1. No matter (61) a. Wie dan ook naar het feest komt; ik zal blij zijn. b. [∀](exh⟨s,t⟩ [wie dan ook, λx.x comes to party](λi □i ϕ)) ‘Whoever comes to the party; I will be happy.’ 2. Adposition (62) Ignorance a. Jan, wie dan ook ( Jan mag zijn), is blij. b. happy(j) ∧ [∀]◇(exh⟨s,t⟩ [wie dan ook, λx.x = j]) ‘Jan, whoever he might be, is happy.’ (63) Indifference a. Jij mag iemand, wie dan ook (hij is), uitnodigen voor het feest. b. [∃]◇([iemand, λx.you invite x to the party]) ∧ [∀]◇(exh⟨s,t⟩ [wie dan ook, λx.you invite x to the party]) ‘You may invite somebody, anybody, to the party.’
indefinites as fossils: the case of wh-based free choice
239
3. Free Relative (64) a. Wie dan ook naar het feest komt zal blij zijn. b. [Q](happy(exhe [wie dan ook, λx.x comes to party])) (definite) c. [∀][Q](happy(↓ exhe [wie dan ook, λx.x comes to party])) (universal) ‘Whoever comes to the party will be happy.’ 4. Indefinite (65) a. Je mag wie dan ook uitnodigen voor het feest. b. [∀](◇(exh⟨s,t⟩ [wie dan ook, λx.you invite x to the party])) ‘You may invite anyone to the party.’ Again this provides a detailed analysis of the different phases of development of wie dan ook using different combinations of a small number of semantic operations. In the first no matter phase, the universal quantifier [∀] is still part of a conditional structure, while in the adposition phase, [∀] operates on an independent (adposed) proposition. We observe that in the adposition phase we have already all semantic ingredients of the later indefinite phase, but, in phase 2, wie dan ook still contributes an independent proposition, not yet integrated in the main clause. We conjecture that the free relative phase facilitates the full integration we observe in phase 4 possibly triggering the shift from a ‘clausal’ to a nominal use of the wh-compound. 4.3 Comparison Comparing the Dutch data with the Spanish ones, we observe that the developments of cualquiera and wie dan ook appear to constitute evidence against unidirectionality in the acquisition of new functions: while the Dutch item was born with the no matter function, the Spanish item starts its development from a free relative into a plain indefinite and only later allows the no matter function to emerge. Dutch: no matter > adposition > free relative > indefinite Spanish: free relative > indefinite > no matter
Our semantic analysis does not make any prediction with respect to the directionality of development of these three uses but our conjecture concerning the emergence of [∀], namely that it is triggered by earlier universallike uses of the wh-based form, does put constraints on possible development
240
aloni
patterns. For example, if our conjecture is correct, the following developments are ruled out: # indefinite > no matter > free relatives # indefinite > free relatives > no matter
The following configuration instead is compatible with our conjecture but might still be implausible since a free relative phase might be required to trigger the shift from a clausal to a nominal use of the wh-form: ?? no matter > indefinite > free relatives
A very recent corpus-based diachronic research on Italian fc indefinite qual si sia showed that the emergence of this item between the 14th and the 17th century followed a path similar to that of Dutch wie dan ook (Degano 2019 and Degano and Aloni 2021):15 Italian: embedded wh-clause > no matter > adposition > indefinite
The Italian data provide additional evidence that the adposition phase might be needed to trigger the separation of [∀] from the conditional structure operative in the no matter phase. No free relative examples were found in this study, but the data were too few (a total of 55 examples) to draw any definitive conclusion concerning the role of these constructions. In view of the evidence from Italian we might want to modify our conjecture about the development of Spanish free choice. The fact that two of the indefinites we could observe in ‘status nascendi’ developed from a no matter use suggests the following new conjecture for Spanish with a no matter phase preceding the indefinite phase and then disappear to emerge again in a much later phase: Spanish (new): no matter/adposition, free relative > indefinite > no matter
If this is right,16 we could assume that [∀] is always inherited from a conditional construction with no need to posit universally read free relative uses. 15 16
See Napoli (2013), Gianollo (2018) and Kellert (2022) for other diachronic studies of Italian indefinites. Data from Company Company (2016) seem to confirm this hypothesis, see her example (12c) on page 523.
indefinites as fossils: the case of wh-based free choice
241
This hypothesis would still leave open the possibility that a free relative phase is needed for the grammaticalisation of the indefinite form but then only for morpho-syntactic reasons.
5
Conclusion
I defended an account of wh-based free choice items where free choice inferences are derived as semantic entailments in an alternative based semantics. I further presented an analysis of the different phases of development of Spanish cualquiera and Dutch wie dan ook using different combinations of a relatively small number of semantic operations and, inspired by diachronic findings, I conjectured (contra, e.g., Chierchia 2013) that the universal-like flavour of these items derives from the application of an operation of universal (propositional) quantification which the item inherits from its source constructions: universally interpreted free relatives in the case of Spanish and no matter constructions in the case of Dutch (and Italian). To further refine this conjecture and test its predictions we would need to study the developments of many more wh-based FC indefinites, but this must be left to future work.
Acknowledgements The data presented in this article were collected as part of the NWO-funded project ‘Indefinites and Beyond’. I am grateful to all the colleagues and students who collaborated to that project, in particular Ana Aguilar-Guevara and Machteld de Vos who collected and annotated the Spanish and Dutch data. I would also like to thank Chiara Gianollo and Klaus von Heusinger for their guidance and their infinite patience, and Marco Degano for the graphs and the storage of the data at https://osf.io/z2j9e/.
References Abrusán, Marta. 2006. Even and free choice any in Hungarian. In Proceedings of Sinn und Bedeutung 11, 1–15. Barcelona: Universitat Pompeu Fabra. Aguilar-Guevara, Ana, Maria Aloni, Tikitu de Jager, Angelika Port, Radek Šimík, Stephanie Solt & Machteld de Vos. 2012. A corpus of indefinite uses annotated with semantic functions: Documentation. Tech. rep. ILLC, University of Amsterdam.
242
aloni
Aguilar-Guevara, Ana, Maria Aloni, Angelika Port, Radek Šimík, Machteld de Vos & Hedde Zeijlstra. 2011. Semantics and pragmatics of indefinites: Methodology for a synchronic and diachronic corpus study. In Stefanie Dipper & Heike Zinsmeister (eds.), Proceedings of the DGfS Workshop “Beyond Semantics: Corpus-based investigations of pragmatic and discourse phenomena”, 1–16. Ruhr-Universität Bochum: Bochumer Linguistische Arbeitsberichte. Aloni, Maria. 2006. Free choice and exhaustification: An account of subtrigging effects. In Proceedings of Sinn und Bedeutung 11, 16–30. Barcelona: Universitat Pompeu Fabra. Aloni, Maria. 2007. Expressing ignorance or indifference. Modal implicatures in BiOT. In Balder ten Cate & Henk Zeevat (eds.), Logic, language, and computation. 6th International Tbilisi Symposium on Logic, Language, and Computation. Batumi, Georgia, September 12–16, 2005, Revised Selected Papers, 1–20. Berlin: Springer. Aloni, Maria & Angelika Port. 2010. Epistemic indefinites crosslinguistically. In Proceedings of NELS 40, 29–43. Aloni, Maria & Angelika Port. 2015. Epistemic indefinites and methods of identifications. In Luis Alonso-Ovalle & Paula Menéndez-Benito (eds.), Epistemic indefinites, 117–140. Oxford: Oxford University Press. Aloni, Maria & Floris Roelofsen. 2014. Indefinites in comparatives. Natural Language Semantics 22(2). 145–167. Aloni, Maria, Andreas van Cranenburgh, Raquel Fernández & Marta Sznajder. 2012. Building a corpus of indefinite uses annotated with fine-grained semantic functions. In Proceedings of LREC, 1511–1515. Alonso-Ovalle, Luis & Paula Menéndez-Benito. 2010. Modal indefinites. Natural Language Semantics 18. 1–31. Chierchia, Gennaro. 2013. Logic in Grammar. Polarity, Free Choice, and Intervention. Oxford: Oxford University Press. Ciardelli, Ivano, Floris Roelofsen & Nadine Theiler. 2017. Composing alternatives. Linguistics and Philosophy 40(1). 1–36. Company Company, Conceptión. 2016. Gramaticalización y cambio sintáctico. In Javier Gutiérrez-Rexach (ed.), Enciclopedia de la lingüística hispánica, vol. 2, 515–526. London: Routledge. Company Company, Concepción & Julia Pozas-Loyo. 2009. Los indefinidos compuestos y los pronombres genérico-impersonales omne y uno. In Concepción Company Company (ed.), Sintaxis histórica de le lengua española (Segunda parte: La frase nominal), 1073–1219. México City: Fondo de Cultura Económica-Universidad Nacional Autónoma de México. Cooper, Robin. 1983. Quantification and syntactic theory. Dordrecht: Reidel. Dayal, Veneeta. 1998. Any as inherently modal. Linguistics and Philosophy 21. 433– 476.
indefinites as fossils: the case of wh-based free choice
243
Dayal, Veneeta. 2004. The universal force of free choice any. In Johan Rooryck (ed.), The Linguistic Variation Yearbook 4, 5–40. Amsterdam: Benjamins. Degano, Marco. 2019. Meaning through time: A diachronic and semantic study of Italian Free Choice. Amsterdam: University of Amsterdam MA thesis. Degano, Marco & Maria Aloni. 2021. Indefinites and free choice: When the past matters. Natural Language and Linguistic Theory. https://doi.org/10.1007/s11049‑021‑09518‑x Deo, Ashwini. 2015. Diachronic semantics. Annual Review of Linguistics 1. 179–197. Eckardt, Regine. 2006. Meaning change in grammaticalization. An enquiry into semantic reanalysis. Oxford: Oxford University Press. Farkas, Donka F. 2002. Specificity distinctions. Journal of Semantics 19. 213–243. Farkas, Donka F. 2006. Free Choice in Romanian. In Betty J. Birner & Gregory Ward (eds.), Drawing the boundaries of meaning: Neo-Gricean studies in pragmatics and semantics in honor of Laurence R. Horn, 71–94. Amsterdam: Benjamins. Fox, Danny. 2007. Free Choice and the theory of scalar implicatures. In Uli Sauerland & Penka Stateva (eds.), Presupposition and implicature in compositional semantics, 71–120. London: Palgrave MacMillan. Franke, Michael. 2011. Quantity implicatures, exhaustive interpretation, and rational conversation. Semantics and Pragmatics 4(1). 1–82. Gazdar, Gerald. 1979. Pragmatics: Implicature, presupposition, and Logical Form. New York: Academic Press. Giannakidou, Anastasia. 2001. The meaning of free choice. Linguistics and Philosophy 24. 659–735. Gianollo, Chiara. 2018. Indefinites between Latin and Romance. Oxford: Oxford University Press. Grice, Herbert Paul. 1975. Logic & conversation. In Peter Cole & Jerry Morgan (eds.), Syntax and Semantics 3: Speech acts, 41–58. New York: Academic Press. Groenendijk, Jeroen & Martin Stokhof. 1984. Studies on the semantics of questions and the pragmatics of answers. Amsterdam: University of Amsterdam dissertation. Grosu, Alexander & Fred Landman. 1998. Strange relatives of the third kind. Natural Language Semantics 6. 125–170. Hamblin, Charles L. 1973. Questions in Montague English. Foundations of Language 10. 41–53. Haspelmath, Martin. 1997. Indefinite pronouns. Oxford: Oxford University Press. von Heusinger, Klaus. 2019. Indefiniteness and specificity. In Jeanette Gundel & Barbara Abbott (eds.), The Oxford Handbook of Reference, 146–167. Oxford: Oxford University Press. Horn, Laurence R. 2005. Airport ’68 revisited: Toward a unified indefinite any. In Greg N. Carlson & Francis Jeffry Pelletier (eds.), The Partee effect, 179–205. Stanford: CSLI Press. Jacobson, Pauline. 1995. On the quantificational force of English free relatives. In
244
aloni
Emmon Bach, Eloise Jelinek, Angelika Kratzer & Barbara Partee (eds.), Quantification in natural language, 451–486. Dordrecht: Kluwer. Jayez, Jacques & Lucia M. Tovena. 2005. Free choiceness and individuation. Linguistics and Philosophy 28. 1–71. Jayez, Jacques & Lucia M. Tovena. 2006. Epistemic determiners. Journal of Semantics 23. 217–250. Karttunen, Lauri. 1977. Syntax and semantics of questions. Linguistics and Philosophy 1. 3–44. Kellert, Olga. 2022. The evaluative meaning of the indefinite qualunque in (Old) Italian. In Chiara Gianollo, Klaus von Heusinger & Maria Napoli (eds.), Determiners and quantifiers: Functions, variation, and change, 246–284. Leiden: Brill. Kim, Min-Joo & Stefan Kaufmann. 2006. Domain restriction in freedom of choice: A view from Korean indet+na items. In Proceedings of Sinn und Bedeutung 11, 375–389. Barcelona: Universitat Pompeu Fabra. Klinedinst, Nathan. 2006. Plurality and possibility. Los Angeles: University of California, Los Angeles PhD dissertation. Kratzer, Angelika. 2005. Indefinites and the operators they depend on: From Japanese to Salish. In Greg N. Carlson & Francis Jeffry Pelletier (eds.), Reference and quantification: The Partee Effect, 113–142. Stanford: CSLI Press. Kratzer, Angelika & Junko Shimoyama. 2002. Indeterminate pronouns: The view from Japanese. In Yukio Otsu (ed.), Proceedings of the Third Tokyo Conference on Psycholinguistics, 1–25. Tokyo: Hituzi Syobo. Menéndez-Benito, Paula. 2005. The grammar of choice. Amherst, MA: University of Massachusetts PhD dissertation. Menéndez-Benito, Paula. 2010. On universal Free Choice items. Natural Language Semantics 18. 33–64. Napoli, Maria. 2013. When the indefinite article implies uniqueness: A case study from Old Italian. Folia Linguistica 47(1). 183–236. Onea, Edgar. 2022. Specificity and Questions of Specification. In Chiara Gianollo, Klaus von Heusinger & Maria Napoli (eds.), Determiners and quantifiers: Functions, variation, and change, 130–185. Leiden: Brill. Port, Angelika & Maria Aloni. 2021. The diachronic development of German irgendindefinites. Tech. rep. ILLC, University of Amsterdam. Potts, Christopher. 2005. The logic of conventional implicatures. Oxford: Oxford University Press. Rawlins, Kyle. 2008. Unifying unconditionals and if-unconditionals. In Proceedings of Semantics and Linguistic Theory (SALT 18), 583–600. Sæbø, Kjell Johan. 2001. The semantics of Scandinavian free choice items. Linguistics and Philosophy 24. 737–788. Shan, Chung-chieh. 2004. Binding alongside Hamblin alternatives calls for variable-
indefinites as fossils: the case of wh-based free choice
245
free semantics. In Proceedings of Semantics and Linguistic Theory (SALT 14), 289– 304. de Vos, Machteld. 2010. WH DAN OOK: The synchronic and diachronic study of the grammaticalization of a Dutch indefinite. Amsterdam: University of Amsterdam BA thesis. Zamparelli, Roberto. 2007. On singular existential quantifiers in Italian. In Ileana Comorovski & Klaus von Heusinger (eds.), Existence: Semantics and syntax, 293– 328. Berlin: Springer.
chapter 8
The Evaluative Meaning of the Indefinite qualunque in (Old) Italian Olga Kellert
1
Introduction
This article deals with the synchronic and diachronic distribution of the evaluative meaning of the Italian indefinite qualunque ‘ordinary’.1 (1) Era un giorno qualunque, un giovedì, anonimo, piuttosto triste. was a day any a Thursday anonymous rather sad ‘It was an ordinary day, a Thursday, anonymous, rather sad.’ (CORIS, MON2008_10, Venuto al mondo2 Evaluative Meaning (Eval): I, the speaker, find the day unremarkable/ordinary. This indefinite has a Free Choice Interpretation (FCI) under modal verbs in Italian (see Aguilar-Guevara et al. 2010, Aloni 2022, Chierchia 2006, Stark 2006, among others): (2) Puoi scegliere qualunque libro. can.2sg choose any book ‘You can choose any book.’ Conventional meaning: You can choose a book and Free Choice Interpretation (FCI): each book is a possible option.
1 Qualunque is derived from Latin qualiscumque composed from qualis ‘of which kind’ + cumque ‘ever’ (qualiscumque > qual[is]-umqua[m] > Old and Modern Italian qualunque. See Meyer-Lübke (1899: 57), Stark (2006: 98) for further references on the etymological status). 2 The primary sources of the corpus examples in this article are listed in the reference lists provided by the corpus builders referred to in the bibliography section.
© Olga Kellert, 2022 | doi:10.1163/9789004473324_009
the evaluative meaning of the indefinite qualunque
247
The sentence in (2) expresses ‘I inform my addressee that all the books are permitted possibilities for her’. According to Aguilar-Guevara et al. (2010) and Aloni (2022), the sentence in (2) has a universal inference which is often paraphrased as ‘For every book x you can take x.’ Other Romance languages have similar indefinites with similar interpretations that depend on postnominal and prenominal position, as discussed in Alonso-Ovalle and Menéndez-Benito (2018), Rivero (2011), Fălăuş (2015), and Vlachou (2012). FCIs are usually not felicitous in episodic contexts that refer to past events unless they are licensed by relative clauses that have a modal interpretation (socalled ‘subtrigging’ cases). In the following example the relative clause modification triggers a modal interpretation of ‘anything possible’ (see Quer 2000 on subtrigging in Spanish): (3) Ha mangiato qualunque cosa le venisse messa sotto il have.3sg eaten any thing her came put under the naso. nose ‘She ate anything that was put in front of her.’ The FCI is usually reserved for qualunque that acts as a prenominal determiner as in (2) or (3). However, postnominal qualunque can also have a free choice interpretation (FCI) if it is embedded under modal verbs: (4) Puoi scegliere un numero qualunque tra quelli della lista. can.2sg choose a number any among those of.the list ‘You can choose any number from the list.’ Conventional meaning: You can choose a number from the list and Free Choice Interpretation (FCI): each number of the list is a possible option. The indefinite qualunque can also occur between the indefinite article and the noun (e.g., un qualunque libro), although this configuration is significantly less frequent than un libro qualunque, as will be shown in Section 2.1 on frequency of distribution. In this case, the interpretation is the same as with postnominal qualunque. Aloni and Port (2013) show that indefinites such as Italian qualunque in (6) have an indiscriminative interpretation under negation, as has been already observed for English any under negation by Horn (2000) in (5) (see Section 4):
248
kellert
(5) I don’t want to go to bed with just anyone anymore. I have to be attracted to them sexually. (6) Gianni non ha letto un libro qualunque. Gianni neg has read a book any ‘Gianni didn’t read just any book.’ The Italian indefinite qualunque is also possible in episodic contexts that refer to past events without negation and without subtrigging as shown in (7). In this case, qualunque must be used with an indefinite article and cannot act as a determiner (as observed by Aloni and Port 2013). In episodic contexts as in (7), the most common interpretation of qualunque is the Random Choice Interpretation (RCI), as was already shown for Spanish by AlonsoOvalle and Menéndez-Benito (2011). The authors have demonstrated that RCI usually occurs with verbs that imply some decision-making process by the agent, so-called agentive or volitional verbs such as buy in (7): (7) a. Gianni ha comprato un libro qualunque. Gianni bought.3sg a book any b. *Gianni ha comprato qualunque libro. book Gianni bought.3sg any Gianni bought a book (existential inference) RCI: ‘Gianni bought the book he chose randomly.’ One way to explain the ungrammaticality in (7b) is to assume that prenominal qualunque must express a universal statement about all books or all things. This interpretation contradicts the use of existential determiner un and the episodic tense that refers to an event in which one book was bought. This explanation has been suggested in the literature by many authors, such as Chierchia (2006, 2013). The common practice in the literature on FCI s is to distinguish existential FCIs from universal FCIs. Alonso-Ovalle and Menéndez-Benito (2011) note for Spanish free choice indefinites such as cualquiera that in certain episodic contexts with volitional verbs such as (8), cualquiera ‘any’ can have the evaluative interpretation of ‘unremarkable’. A similar observation can be made for Italian, as shown in (8). This interpretation can be made salient in contexts with a salient speaker that does not correspond to the agent Gianni:
the evaluative meaning of the indefinite qualunque
249
(8) Gianni ha comprato un libro qualunque. Sarebbe stato meglio se avesse comprato un libro migliore come la Bibbia. ‘Gianni bought an unremarkable book. It would have been better if he had bought a better book, such as the Bible.’ Gianni bought a book (existential inference) Eval: Gianni bought a book, that I, the speaker, find unremarkable. Another crucial difference between RCI and Eval is that the latter can occur with demonstrative noun phrases (see Section 2.1 on the frequency of distribution of demonstratives and definite determiners): (9) questa ragazza qualunque this girl any ‘This ordinary girl’ (Eval only, # RCI, # FCI) There is still a debate in the literature as to whether RCI and Eval represent a case of polysemy, as suggested by Alonso-Ovalle and Menéndez-Benito (2011) or whether both readings are semantically related. In this article, I will argue that Eval is semantically related to RCI and that the evaluative reading is the result of judging the random choice of doing something (e.g., buying a book) as not being the best choice in a decision system, which relies on rational agents that make distinctive and rational choices. In this article, I will focus on the evaluative meaning of postnominal qualunque (Eval) and relate this interpretation to the other interpretations. The syntactic configurations with their respective interpretations are summarized in the following schema. The most important characteristics of the qualunque that modifies a noun and co-occurs with an indefinite article, as in un N qualunque (henceforth “the nominal modifier qualunque”), is that it is ambiguous and its interpretation mainly depends on the syntactic and semantic context; that is, modal verbs trigger FCI, negation triggers the indiscriminative interpretation, and an episodic context with a volitional verb triggers RCI. Copular verbs in present or past tense trigger Eval: Interpretation of modifier Linguistic context qualunque FCI ‘any’ Modal contexts Indiscriminative ‘just any’ Negation RCI ‘random’ Volitional/agentive verbs in episodic contexts
250 Eval ‘unremarkable’
kellert
Copular verbs, demonstrative noun phrases, episodic contexts with a salient speaker’s judgment
Despite the existent synchronic and diachronic studies of qualunque that focus on its free choice character as represented in (2), as discussed in the literature by Aloni and Port (2013), Chierchia (2006, 2013), Stark (2006), and Becker (2014), among others, there are no synchronic or diachronic studies focusing on the evaluative interpretation of qualunque. In this chapter, I will try to fill the gap with respect to qualunque and describe its distribution, as well as its syntactic and semantic properties in synchrony and diachrony. My main focus will be answering the question of how Italian qualunque developed an evaluative meaning component and how the evaluative meaning is related to other interpretations such as FCI, RCI, and the indiscriminative interpretation. I will support the idea that the evaluative meaning is connected to the modifier postnominal position of qualunque. The modifier function is the result of a grammaticalization process whereby qualunque lexicalizes into an indefinite from a relative clause (for lexicalization from relative clauses into FCIs, see Company-Company and Pozas-Loyo 2009: 1084, Haspelmath 1997: 134, among others). I adopt the definition of ‘average, common, standard’ used in Alonso-Ovalle and Royer (2020: 57). The noun modifier with the meaning ‘average’ expresses that the majority of individuals denoted by the noun in UN N qualunque does not stand out with respect to some property (e.g., as a man or as a student) and only a smaller portion of individuals are ranked high (see Sections 3 and 4 for details). I argue that the ranking of individuals can be derived from Free Choice under certain semantic and pragmatic conditions, namely when the modifier qualunque is used under negation with the meaning ‘not just any’. In this configuration, the modifier qualunque is focused and contrasted to some specific alternative. This specific alternative is usually defined by the context as special with respect to an average. The contrastive focus construction with the global meaning ‘not just any, but special, extraordinary’ leads to the reinterpretation of ‘just any’ as ‘not special’ or ‘not extraordinary’. This contextual meaning has been lexicalized and is now used in positive contexts in Modern Italian such as in è un uomo qualunque ‘he is an ordinary man’. The chapter is structured as follows: Section 2 provides a syntactic description of postnominal qualunque in Modern Italian, and Section 3 gives a definition of the evaluative meaning in Modern Italian. Section 4 explains how Free Choice and the evaluative meaning (Eval) are connected. It also introduces a hypothesis on how the semantic change from Free Choice to Eval took place. This diachronic hypothesis is tested in Section 5. Section 6 presents an anal-
the evaluative meaning of the indefinite qualunque
251
ysis of the structure of postnominal qualunque from Old to Modern Italian, which will explain some data from Section2. A summary will be presented in Section7.
2
Modifier qualunque in Modern Italian
In order to describe the modifier qualunque in Modern Italian, I used the corpus CORIS of 20th-century written Italian. It is tagged for lemma and part of speech. Moreover, I used native speaker judgments to test some hypotheses about the properties of modern qualunque. This section is organized as follows. Section 2.1 describes the syntactic distribution of qualunque in CORIS. Section 2.2 investigates postnominal qualunque under negation in CORIS, and Section 2.3 analyzes its syntactic category by using data from CORIS and speaker judgments. 2.1 Syntactic Distribution This section focuses on the distribution of the postnominal modifier qualunque, as the prenominal position with indefinite nouns is very rare. The corpus data show that there is a significant correlation between the string [un/a Noun qualunque]3 (henceforth UN N Qln) and copular verbs, that is [‘be’ N Qln]. From all occurrences of [Verb N Qln] (173 occ. in total) copular verbs represent nearly 60% of the pattern (i.e., 100 occ. of [‘be’ UN N Qln] out of 173 occ. of [Verb UN N Qln]).4 In all examples of the type [‘be’ UN N Qln], Qln has a predicative function, as illustrated in (10) and (11). Moreover, 89 % of copular verbs are in the present or past tense indicative as in (11), without any overt modal verb, modal verbs being typical for FCI contexts: dietro un successo. (10) esempio perfetto della casualità che sta example perfect of-the randomness that be.3sg behind a success Fu un film qualunque. was.3sg a film any ‘perfect example of the randomness behind a success. It was an ordinary/unremarkable movie.’ (CORIS, MON2001_04, Repubblica)
3 I use square brackets [] to represent the string. 4 I used this search algorithm: [pos=“V.*”] “un|una” [pos=“NN”] “qualunque”.
252
kellert
(11) Era un giorno qualunque, un giovedì, anonimo, piuttosto triste. was a day any a Thursday anonymous rather sad ‘It was an ordinary day, a Thursday, anonymous, rather sad.’ (CORIS, MON2008_10, Venuto al mondo) All examples with the copula in present or past tense like (10) and (12) have an evaluative interpretation that can be paraphrased as ‘ordinary, unremarkable, standard, common, normal’. This result is not surprising given that qualunque in [‘be’ UN N Qln] denotes a description or a predicate that applies to the subject of the copular clause. The subject is described as being ‘common, standard, unremarkable’. The corpus contains many appositive clauses following [‘be’ UN N Qln] that express the similarity of the subject of the copular clause to many other individuals of the same comparison class. The following corpus example asserts that the story under discussion is similar to many other stories of this kind: (12) Come si vede è una storia qualunque, come tante. as refl see be.3sg a story any like many ‘As one can see, it is an ordinary story like many.’ The corpus also contains some examples of copular verbs with [UN N Qln] in the subjunctive, such as those in (13) and (14). (13) Come se quella fosse stata una camera qualunque del as if that be.sbjv.3sg been a room any of-the nostro viaggio. our trip ‘As if this (room) had been any room of our trip.’ il cinema come fosse un (14) Io sono un attore che fa be.sbjv.3sg a I be.1sg an actor who make.3sg the movies as lavoro qualunque. job any ‘I’m an actor that makes movies as if it were any kind of job.’ All examples found in the subjunctive (7.51%) have a counterfactual interpretation; that is, they express that some event did not take place in the actual world or that some state does not apply to the actual world. In these cases, [Un N Qln] has a Free Choice Interpretation. The example in (13) expresses that the room under discussion could have been any other room of the trip: room
the evaluative meaning of the indefinite qualunque
253
A, room B, or room C, but in fact it was a different or some special room. The example in (14) expresses that the speaker does his job as if it were any kind of job: job A, job B, or job C, but in fact it is not. To summarize so far, when the copula refers to some present or past state or event in the actual world, [UN N Qln] has an evaluative interpretation or expresses similarity among individuals of the same comparison class. When, instead, the subjunctive copula has a non-episodic interpretation (i.e., it triggers different world situations, as in (13) and (14)), [UN N Qln] has a free choice any interpretation. The latter description is a well-known characterization of the distribution of free choice items in general: free choice items appear only in non-episodic contexts that do not trigger existential presupposition of some actual state or event (see Giannakidou 1997, Quer 2000, among many others). Note that the pattern [Verb Qln N], where qualunque appears in prenominal position, does not show the same tendency to appear with copular verbs in present or past tense that has been observed above with postnominal qualunque. Only 20 (= 1.96%) of the total 1,018 occurrences have a copular verb. The fact that [Qln N] was found more often than [N Qln] (1018 = 85,5 % vs. 173 = 14,5%) shows that qualunque is more often used in the prenominal position than postnominally. As the evaluative interpretation is only possible with postnominal qualunque, the frequency distribution of qualunque shows that the evaluative interpretation is a rare interpretation in contrast to the free choice interpretation. What about other verbs that are not copular verbs and appear with [UN N Qln] (i.e., 41% of Verb UN N Qln)? These verbs are various transitive and intransitive verbs where [UN N Qln] has the function of either an object as in (15) or a subject as in (16). Their common feature is that the majority of them appear in non-episodic contexts and trigger a free choice interpretation of [UN N Qln], which can be represented by different alternatives (shown in parentheses): (15) Cercherei un lavoro qualunque. look.cond.1sg a job any ‘I would look for any kind of job.’ (It can be this work or some other) un nome qualunque. (16) Basta be.enough.3sg a name any ‘Any name is enough.’ (It can be this name or some other) There are also few cases of [UN N Qln] with an RCI. In this case, [UN N Qln] appears in object position with verbs that imply some decision-making process by the speaker or agent with respect to a choice of alternatives denoted by
254
kellert
the object. The following example states that the speaker/agent will randomly choose some girl and dance with her: (17) Ma lo sai che faccio? Io prendo una ragazza qualunque I take.1sg a girl any e ballo con lei sotto i tuoi occhi. ‘You know what I’ll do? I’ll take some random girl and dance with her in front of you.’ Copular verbs do not represent the same kind of verbs illustrated in (15), as they usually do not imply any decision-making process by some agent that can make a choice with respect to some alternatives expressed by [UN N Qln].5 UN N Qln is thus not interpreted as RCI under copular verbs (see also Alonso-Ovalle and Menéndez-Benito 2011, 2018). The generalization on the distribution of UN N qualunque can be summarized as follows on the basis of the corpus data from CORIS. In non-episodic contexts, it has an FC interpretation. With a present tense or past tense copula, it has an evaluative interpretation. In contexts with verbs that enable some decision-making process by the agent, UN N qualunque has an RC interpretation. 2.2 Postnominal qualunque under Negation Let us now look into the pattern [copula UN N Qln] in more detail. In 41 % of these cases the copular verb is under negation (i.e., 41 occ. with [Neg copula UN N Qln] out of 100 occ. of [copula UN N Qln]): (18) Ed io non sono un passeggero qualunque. and I neg be.1sg a passenger any Sono invece il direttore generale della banca Huddleston & Bradford di Westminster. ‘And I’m not just any passenger. I’m the general manager of the Huddleston & Bradford bank in Westminster.’ (CORIS, NARRATTrRomanzi, La grande rapina al treno) The fact that 41% these cases are in the context of negation is significant because usually the configuration [Neg copula UN N Adjective] only occurs at
5 If the copular predicative noun implies some intentionality like the noun choice, RCI is also possible under copular verbs, e.g., è una scelta qualunque ‘it’s a random choice’.
the evaluative meaning of the indefinite qualunque
255
the rate of 19% in the context of [copula UN N Adjective]. We must thus conclude that negation and predicative UN N qualunque correlate for some reason, which needs to be explained. Recall from examples in Section 1, such as (6), that postnominal qualunque under negation has the indiscriminative interpretation.6 The appositive prepositional phrase con una donna straordinaria in (19) expresses the contrast of the preceding prepositional phrase con una donna qualunque. It represents a specific alternative that makes the predication true: (19) non con una donna qualunque, con una donna straordinaria neg with a woman any with a woman extraordinary ‘not with any kind of woman, with an extraordinary woman’ It is reasonable to assume that (19) expresses a denial of free choice and that the result of this denial is the discriminative choice of a specific alternative. Aloni (2022) argues that negation in examples like (19) operates not on the propositional level, but on the level of pragmatic meaning and more specifically of implicatures, since Aloni (2022) treats free choice as an implicature.7 The crucial observation is that qualunque under negation as in (19) can be also interpreted as ‘ordinary’, that is, ‘not with an ordinary woman, but with an extraordinary woman’. It is thus possible to assume that the indiscriminative interpretation and the evaluative interpretation (Eval) are related. This conclusion is desirable, as it implies a unified analysis of different interpretations of qualunque. As will emerge from the diachronic investigation, Eval appeared much later than the indiscriminative interpretation. It is thus reasonable to conclude that Eval was derived from the indiscriminative interpretation and not the reverse. In summary, the majority of examples with the pattern [Verb UN N Qln] contain present or past tense copular verbs. The string [Copula UN N Qln] appears significantly more often under negation than the string [Copula UN N ADJ]. This observation leads us to the conclusion that the negation must be the core trigger of the evaluative interpretation of postnominal qualunque. As we saw, the negation is not propositional, but operates on the level of the free choice implicature. Given that negation of qualunque is negation of free
6 Prenominal qualunque is less used under negation in the corpus than postnominal qualunque. The explanation for the difference in frequency is focalization (see Section 4). 7 Note that it is still under debate what status the Free Choice Interpretation has, whether it is a presupposition (see von Fintel 2000), part of an assertion (Alonso-Ovalle and MenéndezBenito 2018) or an implicature (Aloni 2022).
256
kellert
choice, which can be reanalyzed semantically as ‘not ordinary’ (see Section 4), it is highly probable that the ‘ordinary’ meaning of qualunque in affirmative contexts is an expansion of a meaning evolution that starts out with negation of free choice at some diachronic stage (i.e., having first the interpretation of ‘not ordinary’) and then expands to the affirmative contexts without negation that are observed in synchrony. The steps of this process of semantic change are summarized in (20): (20) Neg V UN N qualunque with interpretation as ‘not just any/not ordinary’ (Diachronic Step 1: restriction to negation) > lexicalization of the ‘ordinary’ meaning of UN N qualunque (Diachronic Step 2: ‘ordinary’ is part of the lexical meaning) > copula UN N qualunque with ‘ordinary’ interpretation (Diachronic Step 3: use of UN N qualunque ‘ordinary’ in affirmative contexts) This hypothesis will be tested in Section 5. 2.3 Syntactic Category of qualunque This section investigates the syntactic status of postnominal qualunque, that is, whether it behaves similarly to the prenominal determiner qualunque or whether it should be instead analyzed as an adjective or adverb. Until now, it was implicitly assumed that postnominal qualunque should be analyzed as an adjective. However, it is important to investigate this aspect in more detail. As will be shown in this section, postnominal qualunque has a hybrid nature. It shows properties of an adjective (e.g., it can be coordinated with an adjective and it has the meaning of an adjective ‘ordinary, common, normal, simple’), but, at the same time, it cannot be analyzed as a prototypical adjective, due to the impossibility of modification with degree adverbs such as molto ‘very’ and the absence of agreement. Postnominal qualunque cannot occur as a predicate on its own (see (21)), nor does it allow any (degree) modification (see (22)). However, both processes (predication and degree modification) are acceptable with semantically comparable adjectives such as ‘ordinary, common, unremarkable, standard’: (21)
Maria è (molto) ordinaria/ (*molto) qualunque.8 any Maria be.3sg very ordinary very ‘Maria is very ordinary/*very qualunque.’
8 Note that the predicative function of qualunque is possible under FCI interpretation in an adpositive clause:
the evaluative meaning of the indefinite qualunque
257
(22) Maria è una ragazza molto ordinaria/ * qualunque.9 Maria be.3sg a girl very ordinary any ‘Maria is a very ordinary/*qualunque girl.’ Moreover, postnominal qualunque remains invariable and does not agree with the noun, in contrast to adjectives with a similar meaning: (23) Maria è una ragazza qualunque/ordinaria. Maria e Lucia Maria be.3sg a girl any/ordinary Maria and Lucia sono due ragazze qualunque/ordinarie. any/ordinary be.3pl two girls ‘Maria is an ordinary girl. Maria and Lucia are two ordinary girls.’ (24) Gianni è un ragazzo qualunque/ordinario. Gianni e Marco any/ordinary Gianni and Marco Gianni be.3sg a boy sono due ragazzi qualunque/ordinari. be.3pl two boys any/ordinary ‘Gianni is an ordinary boy. Gianni and Marco are two ordinary boys.’ Another piece of evidence showing that postnominal qualunque is not a simple adjective comes from the indefiniteness of a noun modified by qualunque. Postnominal qualunque modifying indefinite nouns [UN N Qln] appears significantly more often than modifying definite nouns [IL N Qln], as in la donna qualunque ‘the ordinary woman’ (506 occ. of UN N Qln vs. 13 occ. of IL N Qln, i.e., 97.5% vs. 2.5%). This frequency of distribution tells us something about the syntactic and semantic status of postnominal qualunque. Postnominal qualunque is not a simple adjective, as its use with a definite article or a demonstrative, as shown in (25) and (26), is significantly less frequent than with indefinite articles. This is not the case with postnominal adjectives.
(i)
Dammi un oggetto qualunque esso sia. ‘Give me an object whatever this object might be.’ 9 Kellert (2021) has analyzed an online survey in which 135 native speakers of Italian gave their judgments on: i. molto/abbastanza/tanto qualunque ‘very ordinary’, and on ii. più/meno qualunque di ‘more ordinary than’. Almost 20 % accepted degree modification in i but only 5 % accepted it in ii. This shows that qualunque is still in the phase of a linguistic change towards an evaluative adjective, as it does not allow all linguistic operations allowed by degree adjectives.
258
kellert
(25) questa ragazza ordinaria this girl ordinary ‘This ordinary girl’ (26) questa/la donna qualunque (infrequent) vs. una donna qualunque (frequent) One possible explanation for the distribution of postnominal qualunque is that indefinite nominal phrases are usually interpreted as referentially vague and anti-specific. Referential vagueness expresses uncertainty about the reference of the indefinite noun (see Giannakidou and Quer 2013, among others). Free Choice Items or indefinites with the Random Choice Interpretation (RCI) have also been described in the literature as being referentially vague and antispecific (see Fălăuş 2015, Alonso-Ovalle and Menéndez-Benito 2011, among others). As the evaluative interpretation of qualunque ‘unremarkable’ is still correlated with indefinite noun phrases, as the frequency distribution of [UN N Qln] has shown, it is reasonable to conclude that the evaluative interpretation of qualunque is very much related to FCI or RCI, as these interpretations are also correlated with indefinite noun phrases (see Section 1). Therefore, Eval must be somehow derived from FCI or RCI in Italian (see Section 2.2 for the same conclusion). Otherwise, it is difficult to explain why postnominal qualunque with the Eval interpretation remains statistically rare with definite nominal phrases or demonstratives. The data from (27) to (29) suggest that postnominal qualunque is not a simple adjective. However, postnominal qualunque can be coordinated with certain adjectives. This observation suggests that qualunque does have an adjectival property, as coordination usually requires similarity of the syntactic categories that are conjoined: (27) Forse era una cosa stupida e qualunque, forse un’ maybe be.3sg a thing stupid and any maybe a assurda speranza. absurd hope ‘Maybe it was a stupid and ordinary thing, maybe an absurd hope.’ (CORIS, NARRATTrRomanzi, Il deserto dei Tartari) (28) avventura che anima la gente semplice e qualunque adventure that revive.3sg the people simple and any ‘An adventure that revives simple and ordinary people’ (CORIS, NARRATTrRomanzi)
the evaluative meaning of the indefinite qualunque
259
(29) La sua idea della vita è in un certo senso, e detto con rispetto, un’ idea an idea “qualunque e normale”. any and normal ‘His idea about life is in some sense, and I say it with respect, a simple and normal idea.’ (CORIS, STAMPAPeriodici) The semantic class of adjectives with which qualunque can be coordinated, can be described as follows. The modifier qualunque can only be conjoined by e ‘and’ if the adjective with which it is conjoined is compatible with the meaning ‘ordinary’ or ‘normal’; otherwise, native speakers do not accept the coordination: (30) # un uomo straordinario e qualunque a man extraordinary and any ‘An extraordinary and ordinary man’ The coordination of adjectives with some positive meaning that is not interpreted as ‘ordinary or common’ (e.g., ‘beautiful’) must be expressed by contrastive coordination with ma ‘but’: (31) Gianni è un uomo bello ma qualunque. Bello ma Gianni be.3sg a man beautiful but any beautiful but insignificante. insignificant ‘Gianni is a beautiful but ordinary man. He is beautiful but insignificant.’ Given the lack of agreement, the unacceptability of degree modification and predication of qualunque, we must infer that the ‘ordinary’ meaning is not yet fully lexicalized. Note, however, that similar indefinites in other Romance languages show a more advanced stage of lexicalization of the evaluative meaning, as Vlachou (2012) has shown for French très quelconque ‘very ordinary’ and as Kellert (in preparation) has shown for Argentinian Spanish cualquiera, as in la peli es (re) cualquiera ‘the movie is (quite) ordinary/unremarkable’. On the other hand, coordination with adjectives is possible and thus reveals that postnominal qualunque has some kind of adjectival nature. It is a hybrid category between a determiner and an adjective and is in the process of lexicalizing the evaluative meaning.
260
kellert
table 8.1
Distribution of postnominal and prenominal qualunque in CORIS
Postnominal modifier Qln Determiner Qln episodic contexts modal contexts copula [indic. pres./past] negation Verb coordination with adjectives degree modification with molto agreement with the noun indefinite nominal phrase demonstrative nominal phrase
+ + + + + – (frequent) + (infrequent) – +(frequent) +(infrequent)
– + – + – – – – –
2.4 Summary Table 8.1 summarizes the most important properties identified in this section. The differences between the determiner qualunque and postnominal qualunque are highlighted in gray. In the following section, I will define the evaluative meaning, which will account for some properties mentioned in this table.
3
Definition of Eval
In order to define the evaluative meaning of qualunque, I assume Alonso-Ovalle and Royer’s (2020) definition of ‘average’ as applied to the nominal modifier komon in Chuj, which is a Mayan language spoken in Guatemala and in Mexico. In (32) komon can convey that Xun does not stand out as a student. A ranking of students is invoked with respect to how good their grades are. (32) Context: Xun is a student with average grades. [ Komon estudyante ] waj Xun. komon student clf Xun ≈ ‘Xun is an average/unexceptional student.’ Komon with the ‘common’ meaning ranks its argument in a contextually determined scale of (types of) individuals. It conveys (i) that the satisfier of the NP (Xun) is like most other individuals (denoted by the NP in (32)) and (ii) that it
the evaluative meaning of the indefinite qualunque
261
is ranked around the middle of the scale (hence the satisfier of the NP is ‘average’ relative to the relevant scale). See Alonso-Ovalle and Royer (2020: 58) and their Appendix for a formal representation of (i) and (ii). The ‘average’-meaning of komon is predicted to be deviant with nouns that describe entities (or portions of matter) that are hard to rank with respect to each other or with nouns that cannot be ranked (since they describe singletons). This prediction is borne out empirically for komon (see Alonso-Ovalle and Royer 2020: 58): (33) a. ?Komon tumin jun k’en tik. komon money one clf dem Intended: ?‘This is average money.’ b. #Ix-w-il k’en komon uj. pfv-a1s-see clf komon moon Intended: #‘I looked at the average moon.’ Postnominal qualunque on the meaning ‘average, standard, normal’ is like komon, as it expresses that the referent denoted by questa ragazza ‘this girl’ in (34) does not stand out compared to other individuals in the NP extension. The ‘average’ meaning entails similarity or equivalence between the NP satisfier and other individuals of the same comparison class with respect to a property associated with a girl such as her behavior or some other property. This is why this use is often accompanied by appositive clauses that express similarity among individuals, such as UN N Qln come tante ‘like many other’ (see Section 2.1). Like komon, postnominal qualunque is also deviant with nouns that are hard to rank, as in (35): (34) questa ragazza qualunque this girl any ‘This normal girl’ (35) #questa seconda guerra mondiale qualunque this second war worldwide any ‘This normal second world war’ The meaning ‘average’ expresses that most individuals ranked near the middle range are equal with respect to some property (e.g., with respect to a girl’s behavior) and are thus hard to discriminate. Hence, the ‘average’ meaning of UN N Qln is related to indiscriminacy of individuals of the NP extension. Note
262
kellert
that Free Choice Items such as English whatever and any have been described in the literature with respect to indiscriminacy as well, as they trigger alternatives that are hard to discriminate or that the speaker or some agent does not care to discriminate (see Horn 2000, Vlachou 2012, among others, and Section 4). There is thus a semantic relation between Free Choice, Indiscriminacy and ‘average’, as will be discussed in Section 4. It is possible to assume that the negative meaning of postnominal qualunque such as ‘unremarkable’ or ‘nothing special’ is a pejorative version of the meaning ‘average’.10 Note that pejoration of the meaning ‘common/standard’ is a very well-known process in many languages, as in Latin vulgaris ‘common’ >> French vulgaire ‘vulgar, obscene’, etc. (see Borkowska and Kleparski 2007: 43). The question that still needs to be answered is how the ‘average’ meaning came into existence and how it is related to indefinites with Free Choice meaning. This question is answered in the next section, which relates the evaluative meaning with negation of FCI. As will be shown in this section, it is the denial of indiscriminacy in context of Free Choice Items under negation (e.g., not just any or non…. qualunque) that triggers the meaning ‘not average’. In Italian the positive meaning ‘average’ lexicalizes on qualunque in positive contexts such as è un uomo qualunque ‘he is an average/ordinary man’.
4
From ‘Not Just Any’ to ‘Average’
As will be argued in this section, it is the meaning ‘not just any’ that triggers the meaning ‘average’ or ‘not outstanding’ of qualunque in diachrony. Let us first look into the definition of ‘not just any’. The ‘not just any’ interpretation of FCIs was first described by Horn (2000). The speaker in (5), repeated here in (36), says that she wants to distinguish or discriminate among the possible candidates she wants to go to bed with. In other words, the speaker wants to be selective. According to Horn, not negates the indiscriminacy, thus not just any has an anti-indiscriminative interpretation (see Section 1): (36) I don’t want to go to bed with just anyone anymore. I have to be attracted to them sexually.
10
Probably the reason why ‘average’ can be qualified as ‘bad’ under certain circumstances is because humans in general prefer distinguished entities/individuals over ones that are not distinguished. A company usually prefers to hire someone with distinguished qualities than with normal/average qualities.
the evaluative meaning of the indefinite qualunque
263
Any triggers focus alternatives {candidate A, candidate B, candidate C, …} (see Horn 2000, Chierchia 2006, Fălăuş 2015, among others). The function of just is to interpret these alternatives on a scale of goodness or some other pragmatically determined value and to downgrade them, that is, to put these alternatives at the bottom of the scale of goodness, as suggested by Lee (1987) and by Coppock and Beaver (2011) in order to define the scalar function of ‘just’ and ‘merely’. This is why every candidate denoted by just anyone in (36) has a low value of goodness or, to put it differently, is a bad candidate. The speaker thus wants a good candidate (e.g., sexually attractive) to go to bed with. Jayez and Tovena (2005) show that the indiscriminative meaning of the free choice indefinite n’importe quoi (lit. ‘it is not important what’) in French occurs not only under modal verbs like a usual free choice item, but also in episodic contexts under agentive verbs: (37) Il a acheté n’importe quoi. he has bought not important what. ‘He bought something randomly.’ Italian Un N Qln is not restricted to episodic contexts and agentive verbs. Actually, the most frequent use of Un N Qln is together with a copular verb, as Section 2.1 has shown.11 There is no agent or choice of an agent involved in copular sentences. I thus need an analysis of ‘not just any’ independent of episodic contexts and agentive verbs. Fălăuş (2013) offers a more general analysis of ‘not just any’ without restriction to agentive verbs. According to her, qualunque in (38) has the denotation of an existential quantifier like some in English (see Fălăuş 2013 for a formal representation). Moreover, qualunque triggers alternatives {a, b, c …}; for example, qualunque libro triggers alternative books {Don Quijote, War and Peace, …}. Postnominal qualunque is focused, which triggers a domain restriction on the quantification of ‘some’. The sentence with a focused un libro qualunque is interpreted as Gianni did not read a book in the initially considered domain of quantification (say D), but he read a book in some other, special domain (say D′). The book in the special domain is singled out pragmatically. The spe-
11
Actually, in French as well, n’ importe quoi can appear under copular verbs and have an evaluative meaning there, as shown in (i). However, it is different from qualunque, as it has undergone nominalization in these contexts. I leave the distinction for future research. (ii) C’ est (du) n’ importe quoi. ‘It’s nonsense.’
264
kellert
cial domain can contain a difficult book, for instance. The domain restriction is thus established pragmatically. The contribution of the focus on qualunque is to restrict the domain of quantification of qualunque by individuals in D minus the special domain D′. (38) Gianni non ha letto un libro qualunque. Gianni neg has read a book any ‘Gianni didn’t read just any book.’ The negation and focus interact in (38). Negation creates a complement of the NP extension based on the ranked focus alternatives. If qualunque were not focused in (38), negation would operate on every domain alternative and qualunque would be interpreted as a Negative Polarity Item similar to English any under negation. One would get a different interpretation, namely the plain negation ‘Gianni did not read any book’. However, such interaction with plain negation is expressed by negative words such as niente/nessuno ‘anything/any’ in Italian: (39) Gianni non ha letto niente / nessun libro. Gianni neg has read nothing no book ‘Gianni did not ready anything / any book.’ This proposal explains how focus and free choice alternatives interact. However, it does not explain why the focused qualunque as well as other Romance FCIs have developed the evaluative meaning of ‘ordinary, average, common’. There must be some semantic or pragmatic relation between the meaning ‘not just any’ and ‘common, average, ordinary’. I argue that the meaning ‘not just any’ and the meaning ‘average’ converge under the condition that the alternatives of the NP extension in non UN NP qualunque are interpreted as ‘equal’ or ‘not distinctive’ for the speaker. The relevant alternatives are those contained in the (D–D′ domain). The focus on qualunque restricts the domain to non-distinctive alternatives and leaves out the distinctive one. This distinctive alternative is usually established by the context as exceptional or special:12
12
Distinctive qualities are usually interpreted as exceptional. For instance, in English, with distinction means an exceptionally good grade.
the evaluative meaning of the indefinite qualunque
265
(40) Non è un uomo qualunque, ma eccezionale. neg be.3sg a man any but exceptional ‘He is not just any man, he is exceptional.’ The example in (40) denies that the person in question can be described as every possible type of man, say {type A, type B, type C, …} or more concretely {a man that seldom practices sports, drives an average car, speaks one language, … } and asserts that he is a distinctive type of man who has the predicate of being exceptional, for instance, an athlete who drives a Porsche and speaks five languages. The assertion logically implies that all alternatives of qualunque in (40) have the predicate of being not exceptional or ordinary. This prediction is borne out empirically. In the corpus CORIS, the majority of distinctive alternatives (i.e., those in D′) are clearly interpreted on a scale of specialness/exceptionality.13 (41) Non è un leader qualunque, è l’ anima dell’ ala neg be.3sg a leader any be.3sg the soul of-the wing legalista. legalist ‘He is not just any leader, he is the soul of the legalist wing’ (42) Solo che Judith non è una donna qualunque, è sua be.3sg his only that Judith neg be.3sg a woman any sorella. sister ‘Only, Judith is not just any woman, it’s his sister.’ The meaning ‘average’ of Free Choice Items such as qualunque expresses a relation between non-distinctive alternatives and a distinctive one by ranking the majority of the alternatives around the middle of a scale of specialness and ranking only one alternative or a smaller portion of alternatives at the top. This is why one still finds a significant part of postnominal qualunque in copular sentences under negation in Modern Italian, which can be described as a corrective negation in line with Repp (2009) (see Section 2.2), which is the trigger for the lexicalization of the ‘average’ meaning of qualunque in positive copular clauses such as è un uomo qualunque ‘he is an ordinary/not special man’ (see further Section 5 on diachrony).
13
I used the following search string: “non” “è” “un|una” [pos=“NN”] “qualunque”.
266
kellert
The analysis makes the correct prediction that Eval should be possible in any language that allows focalization of FCIs. This is indeed true for English ‘just any’. The only difference with Italian is that English ‘just any’ has more restrictive licensing conditions: it only appears under negation. The fact that other languages allow the ‘ordinary, common, normal’ meaning of FCIs under certain conditions, such as the presence of negation, is another argument that the ‘ordinary, common, normal’ meaning does not represent lexical ambiguity with the FCI ‘any’, as has been suggested by Alonso-Ovalle and Menéndez-Benito (2011, 2018) for cualquiera in European Spanish. As the ‘ordinary’ meaning is a derived inference, obtained by ranking the majority of free choice alternatives around the middle of a scale, as argued in Section 3, the question arises whether the ‘ordinary’ meaning is an implicature and if so, whether one can cancel this implicature and thus cancel out the ‘ordinary’ meaning of postnominal qualunque.14 As has been argued by Grice (1989), conversational implicatures are usually cancellable because they highly depend on contextual factors. Conventional implicatures usually resist cancellation because they convey a side comment by the speaker and do not contribute to the relevant meaning of the sentence (Potts 2005, Simons et al. 2010). Let us see whether it is possible to cancel out the ‘ordinary’ meaning. The following example was presented to several native speakers to have them assess its acceptability: (43) Gianni è un uomo qualunque, ma non è banale né Gianni be.3sg a man any but neg be.3sg simple neg ordinario. ordinary ‘Gianni is a man qualunque, but he is neither simple nor ordinary.’ Most native speakers accept this sentence, while some do not. Those who do not accept it find that it triggers a contradiction; that is, the first sentence expresses that Gianni is a simple or ordinary man, whereas the second conjoined sentence expresses the opposite, resulting in a contradiction. Other native speakers assume that qualunque can mean ‘normal’ or ‘like many other people’, which does not necessarily mean ‘ordinary’. In this case, the example in (43) expresses that Gianni is like many other men, but not necessarily ordinary. So far, it seems that there is no clear consensus among native speakers with
14
I thank an anonymous reviewer for pointing this out to me.
the evaluative meaning of the indefinite qualunque
267
respect to cancellation of the ‘ordinary’ meaning. It is very probable that the different judgments reflect the degree of lexicalization of the ‘ordinary’ meaning. Those who have lexicalized the meaning are likely to interpret (43) as a contradiction. Cancellation will be studied in detail in the future. The next question is why prenominal qualunque did not lexicalize the meaning ‘average’ under negation (see Section 2.2). It is because it is a common property of Romance languages to express focalization of quantifiers or adjectives by placing them in postnominal position (see Bernstein 2001, Cinque 2010: ch. 6, Gianollo 2018: ch. 3 among others). As prenominal qualunque is usually not focused, the mechanism for deriving Eval from ‘not just any’ does not operate there. This is why Eval correlates with postnominal and not prenominal qualunque. The question now is whether it is possible to find diachronic evidence that the use of postnominal qualunque under negation was attested before its use in affirmative copular clauses. This is borne out empirically, as the next section will show.
5
Diachrony of qualunque
In this section, the diachrony of qualunque will be investigated. The data will show that qualunque with free choice interpretation and qualunque with random choice interpretation appear before qualunque with an evaluative interpretation. 5.1 Corpus Description, Methodology In order to investigate the diachronic distribution of qualunque, the MIDIA corpus (an acronym for Morfologia dell’Italiano in Diacronia) has been chosen. It is built to be a representative and balanced diachronic corpus of written Italian texts from 1200 to 1945, fully annotated with the indication of the lemma and the part of speech (PoS), as well as literary genre: expository prose, literary prose, juridical prose, personal prose, scientific prose, poetry, or spoken language mimesis. The MIDIA corpus contains 25 texts of similar size for each time period15 and genre. It has in total more than 7.5 million tokens or word
15
First period: 1200–1375, formation of Tuscan-centered Old Italian; Second period: 1376– 1532, consolidation of Italian outside Tuscany; Third period: 1533–1691, standardization of Italian in the late Renaissance, Mannerist, and Baroque periods; Fourth period: 1692–1840, the birth of Modern Italian: the age of Arcadia, the Enlightenment, and Romanticism; Fifth period: 1841–1947, the language of the political unification of Italy.
268
kellert
forms. The corpus search focused on qualunque as lemma in the whole corpus. The total corpus provided a total of 3,438 occurrences of the word. From this number all cases of postnominal qualunque were extracted for the linguistic analysis.16 5.2 Distribution of qualunque in MIDIA As Table 8.2 shows, postnominal qualunque occurs only 120 times out of 3,438 occurrences in total (i.e., only 3,5%). Prenominal qualunque occurs more often in this corpus, as Table 8.3 shows (i.e., 60%). The rest of the tokens represent other functions than nominal modifiers (e.g., qualunque di ‘any of’) (i.e., 35 %). Let us compare this distribution of prenominal and postnominal qualunque in MIDIA to that of modern Italian in CORIS as discussed in Section 2. As the numbers in (44) show, the percentage of postnominal qualunque increases by 11% in the Modern Italian corpus CORIS: (44) Prenominal and Postnominal Qln in MIDIA and CORIS Prenominal Qln Postnominal Qln MIDIA 96.5% 3.5% CORIS 85.5% 14.5% Moreover, Table 8.2 shows that the frequency in the last two periods of postnominal qualunque is higher than in the first three periods. Given that the general number of qualunque does not rise in the last two periods (see Table 8.3), one can conclude that there is a slight increase of the use of postnominal qualunque indeed. The data considered show that bare nouns modified by qualunque (e.g., vizio qualunque ‘vice whichever (it is)’, in filosofia qualunque ‘in philosophy whichever it is’) appear earlier (already in the 13th c.) than indefinite nouns (after the 17th c.). Bare nouns modified by qualunque are no longer used after the 17th century. This is expected, as Old Italian had bare indefinites (and bare kind-referring nouns), which are no longer possible in Modern Italian (see Renzi 2010: 331). As for the distribution over text genre of postnominal qualunque, no real change can be observed because the same distribution over text genres can be observed with respect to qualunque in general. Law texts and scientific texts 16
Part of the methodology and analysis of this article is the result of my work on quantifiers in Old Italian in the project “Quantification in Old Italian” funded by the German Research Foundation (DFG) (PI s: Guido Mensching and Cecilia Poletto). I’m very thankful to Marika Francia, who assisted in the data extraction and annotation.
269
the evaluative meaning of the indefinite qualunque table 8.2
Distribution of postnominal qualunque in MIDIA
Expository Law Private Poems Prose Scientific Theatre Total 1200–375 1376–532 1533–691 1692–840 1841–1947 Total
table 8.3
0 0 0 11 8 19
3 1 1 12 11 28
0 0 0 0 3 3
0 0 0 0 1 1
0 0 0 3 4 7
0 0 1 31 28 60
0 0 0 0 2 2
3 1 2 57 57 120
Distribution of prenominal qualunque in MIDIA
Expository Law Private Poems Prose Scientific Theatre Total 1200–1375 1376–1532 1533–1691 1692–1840 1841–1947 Total
15 22 24 60 52 173
420 419 253 185 92 1369
0 25 6 20 32 83
4 5 4 9 6 28
24 29 18 26 16 113
28 54 80 55 38 255
2 6 9 8 4 29
493 560 394 363 240 2050
are the most frequent text genres in which qualunque appears, as shown in Tables 8.2 and 8.3. The increase of postnominal qualunque might suggest that new interpretations appeared in connection with the postnominal position of qualunque in the last two periods such as the evaluative meaning reported in Section 1. I will look at the interpretation of postnominal qualunque in detail in Section 5.3. 5.3 Interpretation of Postnominal qualunque From all occurrences of postnominal qualunque (120 in total), there are 90 occurrences with a modal meaning (with FCI or RCI interpretation), illustrated by the following examples: (45) I metafisici che credono che la questione del libero the metaphysicians that believe.3pl that the question of-the free arbitrio possa continuare ad avere un senso qualunque will can.sbjv.3sg continue to have a sense any
270
kellert
all’ infuori di ogni implicazione teologica rassomigliano a to-the out of every implication theological resemble.3pl to quegli … those FCI: ‘Metaphysicians who believe that the question of free will can continue to have any meaning outside of the realm of any theological implication resemble those …’ (MIDIA, Scienza positiva, , 1901, Libero arbitrio ed imputabilità morale) (46) Un villaggio qualunque di Lombardia potrebbe offrire un a village any of Lombardy can.cond.3sg offer a soggiorno meno sgradevole di quella … stay less unpleasant of that FCI: ‘Any village in Lombardy could offer a less unpleasant stay than that …’ (MIDIA, Fosca, 1869, xi.2) alle Alpi un apostolo qualunque della (47) egli vuole of-the he want.3sg to-the Alps a apostle any pronuncia e della frase fiorentina pronunciation and of-the sentence Florentine ‘… he wants any apostle of the Florentine pronunciation and sentence structure … for the Alps’ (MIDIA, Proemio all’Archivio Glottologico Italiano, 1872) There are some ambiguous cases between the FCI, the indiscriminative (Ind) and Eval interpretation in the 19th century in the following examples. All ambiguous cases are represented by indefinite NP s: (48) Crede lei che un giovanotto —un giovanotto qualunque— a young-man any possa non farsi più nessuno scrupolo, nessun rimorso …? FCI: ‘Do you believe that a young man—any young man—could have no scruple, no remorse …?’ Ind: ‘Do you believe that a young man—no matter which—could have no scruple, no remorse …?’ Eval: ‘Do you believe that an ordinary young man could have no scruple, no remorse …?’ (MIDIA, Pensaci Giacomino!, 1916, 268) I found only nine occurrences with a clear evaluative interpretation. Most examples with Eval refer to a male individual, and all examples are distributed over the text genre ‘theater’ or ‘prose’:
the evaluative meaning of the indefinite qualunque
271
(49) Se il vostro mastice non serve a nulla, voi siete un imbroglione you be.2pl a cheater qualunque. any ‘If your cement does not function at all, you are (just) an ordinary cheater.’ (MIDIA, La giara, 1917, 260) (50) Mi credevo un uomo nella vita, un uomo qualunque, me believe.1sg a man in-the life a man any che vivesse così alla giornata, una scioperata vita in fondo, benché piena di curiosi pensieri vagabondi … ‘I believed myself to be a man in life, an ordinary man, who lived from day to day, an idle life, though full of curious wandering thoughts …’ (MIDIA, Uno, Nessuno e Centomila, 1924–1926, iii, 14) The first cases of postnominal qualunque under negation appear very late (19th c.). The negation scopes over the implicature of Free Choice, and functions, as described in Section 4, as corrective negation. The result of this corrective negation is the assertion of a special alternative (‘an ultimate effect’), as expressed in the statement in (51) (see Section 2.2 for qualunque under negation in Modern Italian): (51) L’arte è arte in quanto produce, non un effetto qualunque, neg an effect any ma un effetto definitivo. ‘Art is art if it produces, not just any effect, but an ultimate one.’ (MIDIA, Del romanzo storico, 1830) Due to low numbers of postnominal qualunque in informal text genres such as theatre, prose, and private texts, it is impossible to test the development of different functions found in Modern Italian and discussed in Sections 1 and 2 in our diachronic corpus (MIDIA), despite the apparent representativeness of this corpus (on which see Section 5.1). There are two possible explanations for this. One is that the corpus does not provide enough diachronic data from relevant text genres that matter for the observation of the development of the evaluative interpretation of postnominal qualunque. The second explanation is that the evaluative function might be a very recent diachronic innovation that is not covered by our diachronic data at all. The second explanation can be corroborated by the following observation. In the early and mid-1940s, the NP uomo qualunque starts to be used in popu-
272
kellert
lar works of children’s literature and satirical and political journals, as well as in radio programs and politics. All examples refer to a ‘common man’ or to ‘a man that is indifferent with respect to political ideologies, especially in politics’ (see Cocco 2014: 396). Note that some of the examples have a definite article instead of an indefinite article, which might be interpreted as a step towards the development of qualunque as an adjective (see Section 2.3 for discussion): (52) Pinocchio uomo qualunque Pinocchio man any literally: ‘Pinocchio an ordinary man’ (Cortelazzo and Zolli 1996: 1010) (53) Uomo qualunque man any ‘ordinary man’ (nickname for the radio announcer Elio Nissim on Radio Londra) (Cortelazzo and Zolli 1996: 1010) (54) fronte dell’ uomo qualunque front of-the man any ‘front of the indifferent man’ (political movement) (Cortelazzo and Zolli 1996: 1010) In order to test the two explanations, the DiaCORIS written corpus has been included in the research. It contains data from the 19th to the 20th century and is the diachronic counterpart of the synchronic corpus CORIS investigated in Section 2. The advantage of this corpus is that the majority of the data represents literary genres that can potentially contain the speaker’s evaluation and can therefore potentially contain the evaluative meaning of qualunque. 5.4 UN N qualunque in DiaCORIS The DiaCORIS corpus contains 334 cases of UN N qualunque, which have been classified with respect to the evaluative, free choice, random choice, or ambiguous interpretation, that is, [+/- Eval], [+/- FCI/RCI]. FCI/RCI is the most common interpretation, as shown in Table 8.4. This observation is expected, given that FCI is a more general meaning than Eval, as discussed in Section 2 (see Table 8.4). All occurrences found with Eval in DiaCORIS belong to subcorpora such as (everyday) press, essays, and literary texts. No single occurrence was found in academic or scientific texts. Eval is thus restricted for text genres as we have already seen in MIDIA, which is expected given that speaker’s evaluation is usually not appropriate in academic or scientific texts. This observation brings us back to our question as to why there is not enough data with Eval
the evaluative meaning of the indefinite qualunque table 8.4
273
Semantic interpretation of [UN N Qln] in DiaCORIS
Total
FCI/RCI Eval
Ambiguous
100% 54.79% 24.55% 20.66 % Total (334 occ.) (183 occ.) (82 occ.) (69 occ.)
in MIDIA as discussed in Section 5.3. As DiaCORIS contains more data of the literary genre than MIDIA, it is not surprising that the evaluative meaning of qualunque is more represented in DiaCORIS than it is in MIDIA (23% of Eval in DiaCORIS and 7% of Eval in MIDIA). As for negation, the corpus contains 18 occurrences of negation out of all 334 occurrences (i.e., 20% of all verbs).17 The first occurrences were observed between the late 1860s and early 1870s. All cases show a contrast between negation of Free Choice and a special alternative, similarly to what we saw in Section 5.3 for the data from MIDIA and in Section 2 for the data from CORIS. I also find UN N qualunque appearing in non-negative contexts such as infinitive constructions that express ‘hidden modality’, as in (55) (see Bhatt 2006 for hidden modality): (55) A: Che cosa intendete di fare? B: […] Ritornare, giustificare con un pretesto qualunque la mia come.back justify with an excuse any the my rinuncia alla licenza. rejection to-the leave ‘A: What do you want to do? B: (I want to) Come back and justify with any (possible) excuse my giving up the right to a leave of absence.’ (DiaCORIS, Fosca, 1869) After 1870, UN N qualunque is found appearing with perfective verbs. In (56) the sentence asserts that some person S went to visit someone (Mr. Giggi) with
17
A reviewer correctly hinted at the observation that the occurrence of negation is lower at this diachronic stage than in Modern Italian as shown in Section 2. The difference in frequency can be explained by the fact that 41 % of negation in CORIS was identified in copular contexts and not with all types of verbs in Modern Italian. It seems that at this diachronic stage, UN N qualunque was not used as much in a predicative position under copular verbs as in Modern Italian.
274
kellert
some excuse. Moreover, qualunque conveys the meaning that S has chosen his excuse randomly without any good motivation or careful judgment: (56) Context: Some person (S) wants to talk to Mr. Giggi, who is not very sociable and does not like to be disturbed. S does not have a good reason to disturb Mr. Giggi, so S finds some random excuse to visit him: Con un pretesto qualunque andò a trovarlo. with an excuse any go.3sg to visit-him ‘He[S] went to visit him [Mr.Giggi] with some random excuse.’ (DiaCORIS, Mastro Titta, 1891) Due to the random and careless choice of the excuse, the example in (56) conveys the meaning that there is no good motivation for the excuse. The excuse can be interpreted in this context as a bad excuse. This evaluative meaning is a contextual meaning, as it highly depends on the speaker of the utterance in (56) evaluating the random excuse according to common standards of choosing and filtering explanations. A good explanation for visiting someone like a friend is usually not random, but highly selective and very well motivated; for instance, one good reason to visit a friend is that we missed him/her. Random human acts are thus biased towards bad behavior in a context of a human evaluator. The crucial aspect of this analysis is that the evaluative interpretation of qualunque is a pragmatic side effect of its interpretation as RCI. The evaluative interpretation is not a case of polysemy of RCI as assumed by Alonso-Ovalle and Menéndez-Benito (2011, 2018) and by Alonso-Ovalle and Royer (2020). In summary, postnominal qualunque (UN N Qln) appears later than prenominal qualunque (i.e., qualunque N). UN N qualunque was observed under negation, which triggered a denial of unextraordinary alternatives. UN N qualunque also occurred under perfective verbs with RCI before it was observed under copular verbs in present or past tense with the evaluative meaning ‘ordinary/unremarkable’. We can thus summarize the two diachronic steps as follows. First, qualunque with Eval was observed under negation, and it was subsequently observed in affirmative contexts without negation: (57) Diachronic Step 1: UN N Qln under negation and episodic contexts with RCI Diachronic Step 2: UN N Qln with Eval under copular verbs This diachronic path confirms our idea that Eval is derived from RCI. In the next section I will provide a syntactic analysis explaining why postnominal qualunque can be licensed in positive contexts by looking into the
the evaluative meaning of the indefinite qualunque
275
diachronic origin of qualunque. The syntactic analysis presented will also explain other properties identified in Section 2.
6
Diachronic Origin of qualunque
In Section 2, I showed that postnominal qualunque acts like a nominal modifier similar to adjectives and that it can appear in episodic or copular contexts. One possibility for explaining the diachronic emergence of the nominal modifier function of qualunque and its occurrence in positive contexts is to assume that qualunque represents a subtrigging case. The notion of subtrigging was introduced above when discussing the example in (3), and it was shown that subtrigging licenses postnominal qualunque in the absence of an overt modal. Before I show how subtrigging works in positive contexts without an overt modal, let us demonstrate the phenomenon of subtrigging by means of an example with a modal verb, since qualunque was first used with modal verbs and it is only at a subsequent stage that it started to be used in non-modal contexts (see Section 5). I assume that qualunque introduced a (reduced) relative clause (RRC) at some diachronic stage before it grammaticalized into a nominal modifier. At this stage, qualunque is interpreted as a predicate of a subjunctive copula (see also Cinque 2010 for the predicative analysis of postnominal adjectives and Rivero 2011 for the predicative nature of postnominal cualquiera in Spanish). The assumed structure is represented in (58): (58) Puoi sposare un uomo [RRC qualunque egli sia]. ‘You can marry a man, whatever property he has/whatever identity he has.’ Assertion: You can marry a man. Free Choice Inference: Every alternative (property/identity) is a possible option. Note that qualunque was also used as a pronoun in Old Italian similar to chiunque ‘whoever’ (see Stark 2006: 155). It is thus possible to interpret qualunque in (58) as a free choice between individual alternatives (i.e., ‘You can marry this man or some other man’). Is there any diachronic evidence for the Reduced Relative Clause analysis of postnominal qualunque? One indirect piece of evidence for this analysis is related to the fact that many FCIs in the Romance languages represent
276
kellert
cases of grammaticalized relative clauses, as with Italian qualsiasi, qualsivoglia (lit. ‘which (ever) it might be’, ‘which (ever) one want’) (see Haspelmath 1997: 38–39, Stark 2006: 97–98, among many others). There are numerous cases of qualunque inside a relative clause with a copular verb in Old Italian corpus OVI: (59) Io vi chiamo a testimonio che questo popolo (e nominalo, I you.pl call.1sg to testimony that this people and name-it qualunque ello si sia) è oltraggioso … any he refl be.sbjv.3sg be.3sg offensive ‘I call you to testify that this people (and name it, whatever it may be) is offensive …’ (OVI, Tito Livio, 1350, a59) qualunque ella si (60) Adunque sofferiamo questa onta she refl thus suffer.1pl that disgrace any sia be.sbjv.3sg ‘Thus I suffer that disgrace (of) whatever (kind) it may be …’ (OVI, Tito Livio, 1350, b259) è (61) niuno affetto, o vero accidente qualunque egli sia, no-one affect or real accident any he be.sbjv.3sg be.3sg tanto universale e tanto comune a tutte le cose, quanto much universal and much common to all the things as.much amore love ‘No passion, or accident of reality, whatever it is, is as universal and as common to all things, as love’ (MIDIA, Varchi, Due Lezioni, 1543) I will now discuss the syntax in more detail. The syntax of the transparent relative clause structure with qualunque and qualsiasi is represented as follows. I assume that, at the beginning of the grammaticalization, qualunque is analyzed as a predicate of a reduced free relative clause represented as a complementizer clause (CP). Qualunque is the predicate of a subjunctive copula inside a predicate phrase (PrP) embedded into the relative CP. At this stage, qualunque is a wh-element with a wh-feature [+wh] that motivates its movement to Spec,CP (see Kellert 2015 on wh-features):
the evaluative meaning of the indefinite qualunque
277
(62) [TP Puoi [VP sposare [DP un [NP uomo [CP qualunquej [+wh] [PrP pro (sia) tj ] ]]]]] ‘You can marry any (kind of) man (whatever property he has).’
TP Puoi
VP
sposare
DP un uomo
NP CP
qualunquej
PrP pro (sia) tj
Another indefinite item of Italian, qualsiasi (lit. ‘whichever it might be’, qualsia-si), reflects the analysis in (62) even better, as it realizes the copula (sia) morphosyntactically: (63) [TP Puoi [VP sposare [DP un [NP uomo [CP qualj [PrP prok siasik tj]]]] ‘You can marry any (kind of) man (whatever property he has).’ In Modern Italian the relative clausal structure (CP) disappears and qualunque (as well as qualsiasi) is no longer perceived as clausal. These indefinites have lost the wh-feature and have been reanalyzed as the exponents of some functional category FP within the nominal phrase. However, qualsiasi and qualunque have preserved their predicative status from their diachronic origin; that is, postnominal indefinites are syntactically closer to adjectives than to determiners: (64) [TP Puoi [VP sposare [DP un [NP uomo [FP qualsiasij [PrP tj]]]]]] ‘You can marry any (kind of) man (whatever property he has).’ The prenominal qualunque took a different path, as discussed for Catalan FCI qualsevol in Kellert and Enríque-Arias (submitted) and other Romance FCI s (see Francia and Kellert submitted and Kellert in preparation). Indefinites like
278
kellert
qualunque start out as relative clauses as in (65) and are reanalyzed as determiners in (66): (65) [DP Ø [NP accidente [CP qualunquej [PrP ello siak tj]]]] accident any he be.sbjv.3sg. ‘any accident’ (66) [DP qualunque [NP accidente]] ‘any accident’ The element qualunque originated as a Relative Clause and then lexicalized into a single word; then, depending on the prenominal or postnominal position, this new single-word category was either reanalyzed as a lexical category (i.e., as a postnominal modifier) or as a grammatical category (i.e., as a determiner). Only the latter development can be defined as a process of grammaticalization. (67) qualunque as a reduced relative clause → lexicalized word → adjective (like) → determiner How does this analysis capture the licensing of FCI s such as qualunque/qualsiasi in positive contexts in Italian such as in (7), repeated here in (68)? (68) Gianni ha comprato un libro qualsiasi/qualunque. Gianni have.3sg bought a book any ‘Gianni bought some (random) book.’ I assume in line with Quer (2000) and Rivero (2011) that postnominal qualunque/qualsiasi represent lexicalized cases of subtrigging due to their relative clause nature. Recall from Section 1 that subtrigging can license FCI s in positive contexts. Romance indefinites have lexicalized the modality component in their lexical semantics; that is, the subjunctive copula (e.g., in qualsiasi) or some other subjunctive verb (e.g., in qualsivoglia ‘which one might want’) as well as the generalized18 adverb -unque ‘ever’ (e.g., in qualunque ‘which ever’) have become part of the indefinite compound. No longer is an external overt modal needed to license the postnominal indefinite (see Rivero 2011).
18
Quer (2000) and references therein discuss generalized adverbs such as usually that license FCI s.
the evaluative meaning of the indefinite qualunque
279
The indefinite compound licenses itself via the lexicalized modal subjunctive -siasi inside the indefinite qualsiasi or the generalized adverb -unque inside qualunque. Prenominal qualunque has a different status. It is a determiner that needs to be licensed by an external modal verb, similar to negative words that need to be licensed by negation, as argued by Zeijlstra (2011) and many others. I formalize the difference between the two elements as follows. The determiner qualunque has an uninterpretable feature that needs to be licensed by an external modal verb or a modal operator (see, e.g., (69)), whereas the nominal modifier in (70) already has an interpretable feature due to the modal verb inside qualunque/qualsiasi. This is why the postnominal indefinite does not need an overt modal to be licensed:19 (69) Modal verb [interpretable modal] … Determiner qualunque/ qualsiasi [uninterpretable modal] (70) Perfective verb … un N qualunque/qualsiasi [interpretable modal] This analysis explains some properties identified in Section 2 for postnominal qualunque, such as the lack of degree modification and the lack of agreement, which I discuss more closely below. It is a common assumption in the literature that postnominal adjectives in Italian have the syntactic structure of a relative clause (see Cinque 2010). I have shown that postnominal qualunque originates in a reduced relative clause. This explains why postnominal qualunque behaves similarly to adjectives in some respects, such as with respect to coordination (see Section 2). Degree modification is not possible under the relative clause analysis of qualunque because qualunque is analyzed as a predicate of a (reduced) relative clause rather than as a (degree) adjective. The lack of degree modification of postnominal qualunque is thus expected under our analysis: (71) Gianni è un uomo (*molto) qualunque. Gianni be.3sg a man (very) any ‘Gianni is an ordinary man.’ 19
Rivero (2011) assumes that the lexical modal verb is interpreted as a modal operator that creates a modal context and licenses cualquiera ‘whatever/whichever’ on the logical form (LF): (ii) LF: [TP [Modal quiera ] [TP Juan compró [NP [NumP una ] revista cual [Modal quiera ]]] ‘Juan bought some random journal.’ I make a slightly different assumption in (70) by saying that qualunque already has an interpretable feature. It does not need movement at LF and checking at all.
280
kellert
The lack of agreement is explained by the fact that the wh-element qualunque is a functional element that is invariable in Old Italian. As this element continues to be a functional element in Modern Italian, as shown in the tree representation in (62), although the syntax of postnominal qualunque is different from that in Old Italian, it is thus expected that this element will not acquire plural or gender morphology in Modern Italian. However, as Kellert (in preparation) shows, there are Romance language varieties such as Argentinian Spanish in which the evaluative meaning of the indefinite has triggered a much stronger reanalysis into a lexical category than the reanalysis in Modern Italian of qualunque, which has led to degree modification and plural agreement of indefinites with evaluative meaning.
7
Conclusion
This chapter started with an analysis of the evaluative meaning of Modern Italian qualunque, which was interpreted as ‘average’ on its neutral meaning and as ‘unremarkable, nothing special, just ordinary’ on its evaluative meaning. The ‘average’ meaning was understood as a ranking of individuals or entities denoted by the noun in N qualunque (e.g., studiante qualunque) on a scale of specialness. The individuals (e.g., students) are positioned in the middle of this scale; that is, they are neither interpreted as especially good nor especially bad (but as average). In certain pragmatic contexts, this interpretation of ‘average’ can shift towards a more negative meaning as ‘nothing special’. It was argued that this shift towards a more negative meaning was first diachronically triggered by the interplay of negation and focus, which yielded the meaning ‘not just any’. The negation is interpreted as a denial of free choice alternatives, which creates a contrastive reading between indiscriminative alternatives and some distinctive alternative established by the context as exceptional: un uomo qualunque, è (72) Gianni non è un uomo Gianni neg be.3sg a man any be.3sg a man straordinario. extraordinary ‘Gianni is not just any man. He is an extraordinary man.’ The use of evaluative qualunque in positive contexts was explained diachronically by the lexicalization of the meaning of qualunque as ‘not outstanding’ emerging in negative contexts. In order to explain why the postnominal and
the evaluative meaning of the indefinite qualunque
281
not the prenominal qualunque acquired the evaluative meaning, the grammaticalization of postnominal qualunque was argued to play an important role in the meaning change. Qualunque as well as qualsiasi started out as relative clauses and lexicalized into postnominal modifiers with the possibility of being focused and negated similar to postnominal adjectives, which triggered the evaluative meaning. Another result from grammaticalization is that postnominal qualunque/qualsiasi encoded the modal component (-unque/-siasi) in the lexical word. The effect of this is that postnominal qualunque started to be used in affirmative contexts without any overt modal verb or modal operator in the sentence. This process results in the self-licensing of the indefinite in non-modal contexts. The diachronic origin of Modern Italian qualunque (i.e., its relative clause origin) also explains other properties of Modern Italian qualunque such as the lack of degree modification and the lack of agreement.
Reference Corpora CORIS DiaCORIS MIDIA OVI
Corpus di Italiano Scritto. http://corpora.dslo.unibo.it/coris_eng. html. Corpus Diacronico di Italiano Scritto. http://corpora.dslo.unibo.it/ DiaCORIS/. Morfologia dell’Italiano in Diacronia. http://www.corpusmidia.uni to.it. Opera del Vocabolario Italiano. https://artfl‑project.uchicago.edu/ content/ovi.
References Aguilar-Guevara, Ana, Maria Aloni, Angelika Port, Radek Simík, Machteld de Vos & Hedde Zeijlstra. 2010. Indefinites as fossils: A synchronic and diachronic corpus study. http://citeseerx.ist.psu.edu/viewdoc/citations;jsessionid=CD15B18B5FE8867E65AB CC403713633B?doi=10.1.1.186.9553 (18 March, 2021). Aloni, Maria. 2022. Indefinites as fossils: The case of wh-based free choice. In Chiara Gianollo, Klaus von Heusinger & Maria Napoli (eds.), Determiners and quantifiers: Functions, variation, and change, 214–245. Leiden: Brill. Aloni, Maria & Angelika Port. 2013. Epistemic indefinites cross-linguistically. In Yelena Fainleib, Nicholas LaCara & Yangsook Park (eds.), Proceedings of NELS 41, 29–43. Alonso-Ovalle, Luis & Paula Menéndez-Benito. 2011. Expressing indifference: Spanish
282
kellert
un NP cualquiera. In Neil Ashton, Anca Chereches & David Lutz (eds.), Proceedings of SALT 21, 333–352. Alonso-Ovalle, Luis & Paula Menéndez-Benito. 2018. Projecting possibilities in the nominal domain: Spanish uno cualquiera. Journal of Semantics 35. 1–41. Alonso-Ovalle, Luis & Justin Royer. 2020. Random-choice modality: The view from Chuj (Mayan). Proceedings of Sinn und Bedeutung 24, 48–65. Becker, Martin. 2014. Welten in Sprache: Zur Entwicklung der Kategorie „Modus“ in romanischen Sprachen. Berlin: de Gruyter. Bernstein, Julie. 2001. Focusing the ‘right’ way in Romance determiner phrases. Probus 13. 1–29. Bhatt, Rajesh. 2006. Covert modality in non-finite contexts. Berlin: de Gruyter. Borkowska, Paulina & Grzegorz A. Kleparski. 2007. It befalls words to fall down: Pejoration as a type of semantic change. Studia Anglica Resoviensia 4. 33–50. Cinque, Guglielmo. 2010. The syntax of adjectives: A comparative study. Cambridge, MA: MIT Press. Chierchia, Gennaro. 2006. Broaden your views: Implicatures of domain widening and the “logicality” of language. Linguistic Inquiry 37. 535–590. Chierchia, Gennaro. 2013. Logic in grammar: Polarity, free choice, and intervention. Oxford: Oxford University Press. Cocco, Maurizio. 2014. Il qualunquismo storico: Le idee, l’organizzazione di partito, il personale politico. Cagliari: Università degli Studi di Cagliari PhD dissertation. Company Company, Concepción & Julia Pozas Loyo. 2009. Los indefinidos compuestos y los pronombres genéricos-impersonales omne y uno. In Concepción Company Company (ed.), Sintaxis histórica de la lengua española. Segunda parte: La frase nominal, vol. 2, 1073–1222. México: Fondo de Cultura Económica y Universidad Nacional Autónoma de México. Coppock, Elizabeth & David Beaver. 2011. Sole sisters. In Neil Ashton, Anca Chereches & David Lutz (eds.), Proceedings of SALT 21, 197–217. Cortelazzo, Manlio & Paolo Zolli. 1996. Dizionario etimologico della lingua italiana. Bologna: Zanichelli. Fălăuş, Anamaria. 2013. Broaden your views, but try to stay focused: A missing piece in the polarity system. In Ivano Caponigro & Carlo Cecchetto (eds.), From grammar to meaning: The spontaneous logicality of language, 81–107. Cambridge: Cambridge University Press. Fălăuş, Anamaria. 2015. Romanian epistemic indefinites. In Luis Alonso-Ovalle & Paula Menéndez-Benito (eds.), Epistemic indefinites: Exploring modality beyond the verbal domain, 60–81. Oxford: Oxford University Press. von Fintel, Kai. 2000. Whatever. In Brendan Jackson & Tanya Matthews (eds.), Proceedings of SALT 10, 27–40. Francia, Marika & Olga Kellert. Submitted. Argentinian Spanish qualunque and Italian
the evaluative meaning of the indefinite qualunque
283
qualunque. In Olga Kellert, Malte Rosemeyer & Sebastian Lauschus (eds.), Romance indefinites: From semantics to pragmatics. Berlin: Language Science Press. Giannakidou, Anastasia. 1997. The landscape of polarity items. Groningen: Rijksuniversiteit PhD dissertation. Giannakidou, Anastasia & Josep Quer. 2013. Exhaustive and non-exhaustive variation with anti-specific indefinites: Free choice versus referential vagueness. Lingua 126. 120–149. Gianollo, Chiara. 2018. Indefinites between Latin and Romance. Oxford: Oxford University Press. Grice, Paul. 1989. Studies in the way of words. Cambridge, MA: Harvard University Press. Haspelmath, Martin. 1997. Indefinite pronouns. Oxford: Oxford University Press. Horn, Laurence R. 2000. Any and -ever: Free choice and free relatives. In Adam Zachary Wyner (ed.), Proceedings of IATL 15. 71–111. Haifa: University of Haifa. Jayez, Jacques & Lucia M. Tovena. 2005. Free-choiceness and non-individuation. Linguistics and Philosophy 28. 1–71. Kellert, Olga. 2015. Interrogative und Exklamative: Syntax und Semantik von multiplen wh-Elementen im Französischen und Italienischen. Berlin: de Gruyter. Kellert, Olga. 2021. Variation and change of unspecific indefinites: A case study of language contact in Río de la Plata area (talk given at LMU, 7.1.2021). Kellert, Olga. In preparation. Non-modal meanings of Free Choice Items. Kellert, Olga & Andrés Enríque-Arias. Submitted. On the diachrony of Catalan indefinite qualsevol. In Olga Kellert, Malte Rosemeyer & Sebastian Lauschus (eds.), Romance indefinites: From semantics to pragmatics. Berlin: Language Science Press. Lee, David. 1987. The semantics of just. Journal of Pragmatics 11. 377–398. Meyer-Lübke, Wilhelm. 1899. Grammatik der romanischen Sprachen: Romanische Syntax. Leipzig: Reisland. Potts, Christopher. 2005. The logic of conventional implicatures. Oxford: Oxford University Press. Quer, Josep. 2000. Licensing free choice items in hostile environments: The role of aspect and mood. SKY Journal of Linguistics 13. 251–268. Renzi, Lorenzo. 2010. L’articolo. In Giampaolo Salvi & Lorenzo Renzi (eds.), Grammatica dell’italiano antico, 297–347. Bologna: Il Mulino. Repp, Sophie. 2009 Negation in gapping. Oxford: Oxford University Press. Rivero, Maria. 2011. Cualquiera posnominal: Un desconocido cualquiera. Cuadernos de la Asociación de Lingüística y Filología de la América Latina 3. 60–80. Simons, Mandy, Judith Tonhauser, Craige Roberts & David Beaver. 2010. What projects and why. In Nan Li & David Lutz (eds.), Proceedings of SALT 20, 309–327. Stark, Elisabeth. 2006. Indefinitheit und Textkohärenz: Entstehung und semantische Strukturierung indefiniter Nominaldetermination im Altitalienischen. Berlin: de Gruyter.
284
kellert
Vlachou, Evangelia. 2012. Delimiting the class of free choice items in a comparative perspective: Evidence from the database of French and Greek free choice items. Lingua 122. 1523–1568. Zeijlstra, Hedde. 2011. On the syntactically complex status of negative indefinites. Journal of Comparative Germanic Linguistics 14. 111–138.
chapter 9
Bare and Indefinite Nominal Predicates in the History of German Svetlana Petrova
1
Subject and Goals
In present day standard German (PDG), indefinite noun phrases used as complements of copula verbs like sein ‘be’ and werden ‘become’, also called predicative indefinite noun phrases, give rise to variation regarding the presence of the determiner that is illustrated in (1)–(2): (1) Peter ist (*ein) Arzt / (*ein) Italiener / (*ein) Katholik. Peter is a physician an Italian a catholic ‘Peter is a physician / an Italian / a catholic.’ (2) Peter ist *(ein) Freund / *(ein) Genie / *(ein) Optimist. Peter is a friend a genius an optimistic person ‘Peter is a friend / a genius / an optimistic person.’ It is assumed that the variation in (1) and (2) is semantically driven, in that the determiner is missing in those cases in which the predicate expresses affiliation to an institutionalized social role, while it is required in all remaining contexts, in which the subject referent is characterized as a representative of some class of individuals (see Engel 2004: 316, Eisenberg 2006: 463, Duden 92016: 331, Berman 2009, Hallab 2011 and Geist 2014). Behaghel (1923) provides minimal pairs like those in (3a–b), in which predicates involving the same lexical expression receive a different interpretation, depending on the presence or the absence of the determiner. While the bare variants are most naturally interpreted as denoting a regular activity, e.g. a profession, or the actual origin of the subject referent, the cases displaying a determiner are understood as expressing a prototypical property that the subject referent shares with other representatives of the denoted class (physical appearance, behaviour, capacities, etc.):
© Svetlana Petrova, 2022 | doi:10.1163/9789004473324_010
286
petrova
(3) a. Er ist Athlet / Preuße / Künstler. he is athlete Prussian artist ‘He is a professional athlete / Prussian by birth/ a professional artist.’ b. Er ist ein Athlet / ein Preuße / ein Künstler. he is an athlete a Prussian a creative person ‘He has an athletic appearance / behaves like a Prussian clerk / is creative.’ Summarizing these observations, Geist (2014) proposes a classification of indefinite noun phrases that accounts for the mapping between the meaning and form of nominal predicates in standard PDG in the way given in (4a–b). She introduces the basic distinction between role nouns (Class A nouns) and class nouns (Class B nouns), which correlate with two different types of formal expressions. Class A nouns appear as bare predicates (henceforth BNP s), while class B nouns display the indefinite article (hence assigned the label INP s1): (4) a. Class A (role nouns) BNPs professions (Lehrer ‘teacher’), nationalities (Italiener/in ‘Italian’), hobbies (Alpinist ‘alpinist’), functions (Oberarzt ‘senior physician’), convictions (Pazifist ‘pacifist’), occupations (Student ‘student’), religious denominations (Katholik ‘catholic’) b. Class B (class nouns) INPs sub-kinds of humans (Mann ‘man’), properties (Riese ‘giant’), epitheta (Held ‘hero’), swear words (Idiot ‘idiot’), etc. A similar distribution is also found in other languages displaying an indefinite determiner, e.g. modern Dutch (de Swart, Winter and Zwarts 2005) or the Romance languages (Zamparelli 2008). These observations suggest that the drop of the determiner with Class A predicates might follow from a more abstract, universal property. But at the same time, as Duden (92016: 331) also remarks, the PDG picture presented in (4a–b) is subject to manifold variation. Berman (2009) presents data from contemporary newspapers, showing that Class A nouns may take a determiner without any change in meaning. Even more significant is the picture in some contemporary dialects, in which Class
1 I stick to the abbreviations used in Geist (2014). Of course, in both cases, the noun phrases are indefinite in semantic terms, independently of the question if the determiner is present or not.
bare and indefinite nominal predicates
287
A predicates regularly appear with the determiner. In the first place, as shown in (5a–b) and (6a–b), this is the case in Bavarian (see Glaser 1996 and Kolmer 1999 as well as the references cited therein) and Alemannic (Weber 1987), i.e. in the Upper German dialects, which are located in the South of the German speaking area. These dialects are known as particularly ‘article-friendly’, in that they display a determiner in contexts in which the modern standard language rejects it, e.g. with mass and abstract nouns (see Glaser 1996 and Kolmer 1999). However, the use of the article in predicates involving role nouns cannot be considered a phenomenon restricted to these Southern varieties only. HennMemmesheimer (1986) provides examples from Palatine, a Western Central variety closest to Upper German. In addition, the example in (7) taken from a famous popular song written in the urban variety of Cologne shows that the use of the determiner with Class A nouns is also found in Ripuarian, representing the northernmost extension of the Western Central German dialectal area: (5) a. I bin a Nadaren gwes. I am a sewer been ‘I was [= used to work as] a sewer.’ (Kolmer 1999: 67) b. Si isch e Sängeri. she is a singer ‘She is a singer [by profession].’ (Kolmer 1999: 67) (6) a. Sai Grossfatta was a Ästaraicha. his grandfather was an Austrian ‘His grandfather was an Austrian.’ (Glaser 1996: 151) b. Er isch en Schwyzer. he is a Swiss ‘He is Swiss.’ (Weber 1987: 207) (7) Ich ben ne Franzus, ich kom mem Napoleon. I am a Frenchman I came with.the Napoleon ‘I am a Frenchman, I came with Napoleon.’ (Unser Stammbaum, a song by Bläck Fööss) Finally, the dialectal situation is mirrored by diachronic evidence presented in Behaghel (1923), who shows that nouns denoting crafts, occupations, or ethnicity/provenance may take a determiner in the historical stages of German, as demonstrated in (8a–b):
288
petrova
(8) a. Kain war ein Accherman. Cain was a peasant ‘Cain was a peasant.’ (Genes. 1219; cit. in Behaghel 1923: 87) b. Der selbe bapst was von der geburt ein Römer. the same pope was from the birth a Roman ‘The same pope was Roman-born.’ (Stretl. Chr. 2, 8; cit. in Behaghel 1923: 87) This data raises the question of what governs the distribution of BNP s and INPs in the historical stages of German and which changes lead to the establishment of the contemporary situation, including the mismatch between the standard language and the dialects mentioned above. It challenges previous scenarios according to which the presence of BNP s in PDG results from incomplete grammaticalization of the indefinite determiner. According to such scenarios, the indefinite determiner evolves from the numeral ‘one’ and spreads over to indefinite noun phrases in a gradual fashion, attaching to referentialspecific expressions first and to non-referential expressions, incl. predicative ones, only later, towards the end of the grammaticalization process (see Givón 1981 and Heine 1997). Following this line of argumentation, it might be suggested that in a more fine-grained model of stepwise grammaticalization, Class A nouns constitute a semantic sub-type of expressions which have not received the indefinite determiner yet (see Szczepaniak 22011: 81). However, given the data in (8a–b), it cannot be assumed that the situation in the PDG standard is a straightforward continuation of the diachronic process taking place in the dialects, as they seem to have undergone further steps in the grammaticalization process of the indefinite determiner already during Middle High German times. The present paper will pursue a different scenario. It will argue that the principles underlying the use of the determiner in the domain of nominal predication remain constant over time, while the changes result from (extra-linguistic) factors determining the classification of states and activities as social roles on the one hand, and from the existence of empty determiners in the system of the respective variety on the other. The paper is structured as follows. Section 2 describes the properties of the dataset and the basic facts regarding the distribution of BNP s and INP s in the historical stages of German. It shows that nominal predicates referring to institutionalized functions and activities that are typical for the respective historical period reject the determiner over the entire history of German, suggesting that the affiliation to social roles is a diachronically constant property
bare and indefinite nominal predicates
289
underlying the semantics of BNPs. For those nouns displaying variation in the use of the determiner, this variation depends on additional conditions, such as coordination/disjunction, or the referential status of the subject expression. Section 3 integrates these finding into Longobardi’s (1994, 2005) account on the availability of a semantically governed category of empty determiners [D e], including the option of N-movement to the empty D-head. Section 4 summarizes the findings.
2
Role Nouns in the Diachronic Stages of German
2.1 Corpus and Methods of Search For the purpose of the present investigation, I compiled a database containing predicative indefinite noun phrases denoting crafts, occupations, ethnicity, religious denominations, etc., i.e. lexical equivalents of the PDG Class A nouns given in (4a). In addition, I collected nouns denoting affiliations to institutionalized functions typical of the respective historical period. The aim is to examine whether the indefinite determiner is used with these nouns in the historical examples, given that in standard PDG the article is missing in this domain. The historical periods that are relevant for the investigation are Early New High German (ENHG, 1350–1650) and Middle High German (MHG, 1050– 1350). The earliest attested period of German, Old High German (OHG, 750– 1050), is inconclusive because an overt form of the indefinite determiner is rare in this period as a whole, and because it is practically non-existent in predicative noun phrases involving all kinds of nouns (see Behaghel 1923: 87, Oubouzar 2000, Schrodt 2004: 26 and Petrova 2015). The data from ENHG was retrieved from the Bonner Frühneuhochdeutschkorpus (FnhdC), an electronically available corpus containing material from 40 texts from the various dialectal regions and sub-periods. The MHG data was collected by searching two electronic resources, namely Referenzkorpus Mittelhochdeutsch (REM) and the text archive of the Middle High German dictionary (MWB Online). I adopt the spelling and the references to the text sources of the MHG examples used by these two electronic resources.2 I examined the dialectal distribution of the examples included in the database, to make sure that the presence of INPs is not due to texts from Upper 2 I use the abbreviations of MWB Online to refer to the text sources of the examples taken from there, see http://www.mhdwb‑online.de/quellenverzeichnis.php (last accessed on May 7th, 2021).
290
petrova
table 9.1
Dialectal distribution of BNP s and INP s in MHG and ENHG in the database
MHG
BNP INP Total
ENHG
Upper German
Central German
Total
Upper German
Central German
Total
78 41 119
84 29 113
162 70 232
9 20 29
3 1 4
12 21 33
German only, which is known for its strong preference for indefinite articles with Class A nouns even today (see Glaser 1996, Kolmer 1999 and the argumentation in section 1). Table 9.1 shows the dialectal distribution of BNP s and INP s contained in the database (in absolute numbers). The numbers in Table 9.1 show that the presence of the determiner with Class A nouns cannot be regarded as an Upper German phenomenon. Indeed, the number of INPs is higher in Upper German than in Central German, but this can be due to the fact that Upper German examples overwhelm as a whole, especially in the ENHG part of the database. Nevertheless, Upper German displays relative high numbers of BNPs as well. Note also that the standard statistics tests reveal a p-value that is not significant at p < .05 in none of the periods, suggesting that there is no statistically significant correlation between the dialectal area and the type of nominal expression in the database. This suggests that the distribution of BNPs and INPs follows from some underlying, language-internal principles in both periods and dialectal zones. Finally, I searched for individual role nouns in Middle Low German (MLG, 13th to 16th century) by way of example, using Referenzkorpus Mittelniederdeutsch / Niederrheinisch (REN). The examples in (9a–c) suggest that Class A nouns similarly appear as INPs in the Low German varieties attested at roughly the same period of time: (9) a. Wel he oc en copman wesen wants he also a merchant be-inf ‘if he also wants to be a merchant’ (Brem._StR_1303,04) b. Vnde myn Uader is eyn ackerman and my father is a peasant ‘and my father is a peasant’ (Buxteh._Ev. 1480)
bare and indefinite nominal predicates
291
c. De en iodinne was who-fem a Jew-fem was ‘who was Jewish’ (Brem._Sächs._Wchr) In the remaining part of the paper, I will discuss the MHG and the ENHG data. The situation in MLG will be left aside for further research. 2.2 Early New High German I will start with ENHG, the period during which fundamental processes leading to the consolidation of a national standard variety take place (see Hartweg and Wegera 2005). It is necessary to remark in advance that the indefinite determiner is already firmly established in this period, which is visible from the fact that indefinite noun phrases regularly occur with a determiner, both as arguments (10) and as predicates (11): (10) Eß ist eyn baum also geheissen. it is a tree so called ‘There is a tree that is called that name.’ (FnhdCX > 243) (11) Cytrus ist eyn baum. citrus is a tree ‘Citrus is a tree.’ (FnhdCX > 243) The indefinite determiner also appears in nominal predicates involving typical representatives of Class B nouns, such as sub-kinds of humans, properties, epitheta and swear words. All of the items that were tested and found in predicative function in FnhdC display the indefinite determiner in the way shown in (12a–d): (12) a. Daz jch ein man bin vn̅ nit ein weib that I a man am and not a woman ‘that I am a man and not a woman’ (FnhdCX > 223) b. Nemrot was eyn resse. Nemrot was a giant ‘Nemrot was a giant.’ (FnhdC > 253) c. du bist ein Held. you are a hero ‘You are a hero.’ (FnhdC > 147)
292
petrova
table 9.2
BNP INP Total
Distribution of BNP s and INP s in ENHG
n
%
12 21 33
36.4 63.6 100
d. ob er ein tor waͤ r if he a fool was-pst.sbjv ‘if he were a fool’ (FnhdCX > 223) With the equivalents of the PDG Class A nouns, there is variation, in that both BNPs as well as INPs can be found. The distribution of the different types of nominal predicates in ENHG is represented in Table 9.2. The overall number of conclusive examples obtained from FnhdC is small (n=33), which is due to the restrictedness of the dataset, but nevertheless, some principles governing the distribution of BNPs and INP s can be determined. Basically, three groups of examples can be identified. In the first one, there is a strong preference for the bare variant, as shown in (13a–c). The nouns involved in this group are Burgermeyster ‘mayor’, Rentmeister ‘treasurer’ and Schreiber ‘clerk’, as well as Abt ‘abbot’ and Bischoff ‘bishop’. These nouns denote functions in public administration or positions in the hierarchy of the church. In addition, these positions share the property of uniqueness, in that they typically exist only once at a time in the respective domain. In the database, nominal predicates referring to such types of positions occur as BNP s in 9 out of 10 examples, the single exception being the example in (14), which involves the noun Zölner ‘tax collector’. Note, however, that the particular example refers to events taking place at the period of time described in the New Testament, and not to events settled in the context of the historical period of the respective scribe. In addition, the property of uniqueness does not apply to this type of appointment: (13) a. de Burgermeyster gewest was who mayor been was ‘who was a mayor’ (FnhdCX > 151)
bare and indefinite nominal predicates
293
b. do he Rentmeister was when he treasurer was ‘when he was a treasurer’ (FnhdCX > 151) c. Herr Ferdinand Stenglin ist Stadtschreiber. Mr Ferdinand Stenglin is clerk ‘Mr Ferdinand Stenglin is a clerk.’ (FnhdCX > 127) (14) Mattheum der ein Zoͤlner […] gewesen ist Matthew-acc.sg who a tax collector been is ‘Matthew who was a tax collector’ (FnhdC > 135) In the second group of examples, the determiner is present in all of the attested instances (a total of 10 cases in the database), some of which are given in (15a– b). This group involves nouns like Bawr ‘peasant’, Poet ‘poet’, Schneider ‘tailor’ (incl. compounds) or Zimmermann ‘carpenter’, i.e. nouns that denote crafts and occupations. (15) a. das er ein Bawr […] war that he a peasant was ‘that he was a peasant’ (FnhdC > 255) b. sein Vater war ein Schneider his father was a tailor ‘His father was a tailor’ (FnhdC > 237) In the third group of examples, there is variation regarding the use of the determiner with one and the same noun, even within the same text source, as (16a–b) and (17a–b) suggest. This group involves expressions like Maister/Lerer ‘teacher’ and Bot ‘messenger’, i.e. expressions denoting appointments. Note, however, that in contrast to the first group, or to contemporary uses of role nouns such as ‘teacher’ referring to a socially codified role, the nouns discussed in the present examples denote appointments settled in the private sphere, not in the public domain. This suggests that the historical setting in which these nouns are used plays a role as an extra-linguistic factor shedding light on the relation between the meaning and the form of a linguistic expression. More precisely, it can be seen that in this group, INP s make up 10 out of 13 instances, which means that the respective nouns behave more frequently like those denoting crafts and occupations, and less frequently like those denoting institutionalized functions or activities.
294
petrova
(16) a. Dieser ist maister gewesen des manns Pitagore. this.one is teacher been the-gen man-gen Pitagoras ‘This one was a teacher of Pythagoras.’ (FnhdC > 223) b. Vnd ein maister Archillej gewesen and a teacher Achilles-gen been ‘and [was] a teacher of Achilles’ (FnhdC > 223) (17) a. die stat oder dz land dennen derselb bot ist the city or the land rel-dat.pl the.same.one messenger is ‘the city or the land of which he is a messenger’ (FnhdC > 213) b. ich bin ain bot Priamus. I am a messenger Priamus ‘I am a messenger of Priamus.’ (FnhdC > 221) Let us summarize these findings. Although there is variation regarding the use of the indefinite determiner in nominal predicates containing equivalents of Class A nouns in PDG, the drop of the determiner in ENHG applies in particular in those cases in which the noun denotes affiliation to an institutionalized position in some administrative domain. In this respect, BNP s in ENHG share the distribution of those in PDG. But unlike PDG, the uniqueness of the type of position or social role is another relevant property applying to BNP s in ENHG. This conclusion is supported by observations made visible by the minimal pair in (18a–b): (18) a. Context: So weit ist es kommen/ daß auch dem Eydschwure nicht mehr zu trauen ist = ‘It has become as bad as there is no trust in oath.’ Allein der Himmel wird Richter seyn wie hoch dieser Meineyd only the Heaven will judge be-inf how hard this perjury soll bestrafft werden. should punished become-inf ‘Only heaven will be judge of how hard this perjury should be punished.’ (Christian Weise, Jugendlust, Leipzig 1684; Upper Saxonian) b. Context: Wir brauchen einen gerechten Keyser der dem Reich nichts entziehe sondern das entzogene wieder herzubringe = ‘We need a fairminded emperor who will not take away anything from the public but who would give back what has been taken away.’
bare and indefinite nominal predicates table 9.3
BNP INP Total
295
Distribution of BNP s vs INP s in MHG
n
%
162 70 232
69.8 30.2 100
der der ein Richter und nicht ein Fuͤ hrer sey who a judge and not a leader is-prs.sbjv the-gen.pl offentlichen Landraͤ uber public thieves-gen ‘who would act as a judge and not as a leader of thieves of public property’ (Sigmund von Birken, Spiegel, Nürnberg 1668; East Franconian) While in the a-example, the bare noun describes someone who is a judge in its literal meaning, referring to the single judge at doomsday, the indefinite noun in the b-example describes an emperor who should behave as a prosecutor of the thieves of public goods, i.e. who acts as a judge, without being one in the literal sense of this notion. 2.3 Middle High German 2.3.1 Basic Distributional Facts Let us consider the use of the indefinite determiner in nominal predicates in the MHG period. Luckily, the database obtained for this period is considerably larger than that for ENHG, comprising a total of 232 examples. The distribution of the various types of NPs is represented in Table 9.3. Let us look at the factors responsible for the distribution of these types of nominal predicates in MHG. As observed for ENHG above, there are groups of nouns which reject the indefinite determiner in all of the instances found in the attestation. First, nominal predicates describing one of the court appointments typical for the period, like trucheze ‘steward’, marschalc ‘marshal/Master of the Horse’, cemerere ‘treasurer’, skence ‘cupbearer’, never occur with the indefinite determiner (a total 26 instances). Representative examples are given in (19a–b): (19) a. Sindolt der was scenke. Sindhold he was cupbearer ‘Sindhold, he was cupbearer.’ (MWB Online, NibB 11,3)
296
petrova
b. Hûnolt was kamerære. Hûnold was treasurer ‘Hunold was a treasurer.’ (MWB Online, NibB 11, 4) Second, as also seen for ENHG, nouns describing higher positions in the structure of the church, such as abbet ‘abbot’, bischof ‘bishop’, babes ‘pope’, strictly reject the article (a total of 73 instances). An example is given in (20): (20) Dô bat der keiser Fôcas den herren, der dô bâbest was then asked the emperor Focas the man who then pope was ‘Then Emperor Focas asked the man who was pope at that time’ (MWB Online, Eracl 2376) Again, the special feature applying to these nouns is that they express affiliation to an institutionalized social role. In addition, like shown for ENHG above, this position usually exists only once at a time in the respective administrative domain. In sum, denotation of a unique social role reveals to be the relevant factor triggering the omission of the indefinite determiner in MHG as well. As for all remaining types of nouns, i.e. those denoting crafts, occupations, religious denominations, or national origin, two additional factors can be determined, which govern the use of the determiner in the predicate. In the following two sub-sections, the role of these factors for the selection of BNP s (a total of 63 instances) and INPs (a total of 70 instances) will be investigated. 2.3.2 Single vs Conjoint/Disjoint NPs First, it can be observed that coordination and disjunction of nominal predicates strongly favour the drop of the determiner. Consider the minimal pairs in (21a–b) and (22a–b): (21) a. kain waſ ein accherman Cain was a peasant ‘Cain was a peasant’ (MWB Online, Gen 611) b. Der altere wart jagire und accherman the elder became hunter and peasant ‘The elder one became a hunter and a peasant’ (MWB Online, Gen 1072)
bare and indefinite nominal predicates
297
(22) a. Er was kurtoys, sîn vater was ein Franzoys he was well-mannered his father was a Frenchman ‘He was well-mannered, his father was a Frenchman’ (MWB Online, Parz 46, 22) b. ist Kingrûn Franzoys od Bertûn is Kingrûn Frenchman or Breton ‘Is Kingrûn a Frenchman or a Breton’ (MWB Online, Parz 195, 28) This effect is expected because it is known from the literature that coordination and disjunction lead to article omission across the board, i.e. with all types of singular indefinite noun phrases, irrespectively of their semantic class, animacy status, or grammatical functions. An example from modern German is given in (23), see also Bisle-Müller (1991).3 The same applies for MHG as well (see Paul 2007: 380 and Desportes 2000). The examples in (24a–b) show that conjoint indefinite NPs with class nouns in argument position also show up without an article in MHG: (23) Normalerweise will ein Amerikaner Haus und Auto kaufen und normally wants an American house and car buy-inf and für die Ausbildung der Kinder sorgen. for the education the-gen.pl children care-inf ‘Normally, an American wants to buy a house and a car and to care for the education of his children.’ (Christoph Dieckmann, Oh! Great! Wonderful! Anfänger in Amerika; google books [09.03.2018]) (24) a. Ir mochte ravber noch diep. mer genemen noch geſchaden her could robber nor thief more take.away nor damage ‘No robber or thief might take away from her anything else or do damage to her’ (14_1- bair-PUV-G > M319-G1) b. Suaz mir wolf oder dîep genam what me wolf or thief took.away ‘whatever a wolf or a thief took away from me’ (MWB Online, Gen 1472)
3 For the same effect of coordination on definites, see Heycock and Zamparelli (2003).
298
petrova
table 9.4
BNP INP
Single vs. conjoint/disjoint role nouns in MHG
Single NP
Conjoint/disjoint NP s
38 67
25 3
Fisher Exact p < .05
In view of these observations, it comes as no surprise that for the database including predicative role nouns in MHG, we obtain a statistically significant result on the effect of coordination and disjunction on the drop of the determiner, as indicated in Table 9.4. To conclude, nouns referring to crafts, occupation, nationalities etc. significantly more often occur as BNPs if they build the predicate of the clause in coordination or disjunction with another NP. This property, however, applies to all indefinite nouns in the period, i.e. to argumental ones as well as to nonargumental ones. In the next section, I will focus on non-conjoint nominal predicates to retrieve an additional feature that correlates with the omission of the determiner in MHG. 2.3.3 Referential Properties of the Subject Referent Another factor relevant for the distribution of BNP s and INP s in MHG is the referential status of the subject expression. Consider the minimal pairs in (25a–b), (26a–b) and (27a–b): (25) a. Asclêpîus ein Arzât was. Asclepius a physician was ‘Asclepius was a physician.’ (MWB Online, RvEBarl 10083) b. dâ sol nieman Arzât wesen. there should nobody physician be-inf ‘There, nobody should be a physician.’ (MWB Online, SM:St 3: 2, 7) (26) a. Aber ſebaſtianus der waz eín richts. but Sebastianus he was a judge ‘But Sebastianus was a judge.’ (14_1-rhfrhess-PU-G > M407-G1)
bare and indefinite nominal predicates
299
b. Wer Richter sin muge whoever judge be-inf may-prs.sbjv ‘whoever might be a judge’ (MWB Online, Sachsenspiegel 1, lv) (27) a. ſancti antonij der was ein muͦ nich St. Antonius he was a monk ‘St Antony, he was a monk’ (13_2-omd-PV-G > M408-G1) b. Er ſcol uon rechte imer muͦ nich ſin ſwer hi nicht he should by justice forever monk be-inf whoever here not ſlet daz ſwert. hits the sword ‘Whoever does not fight with the sword here truly deserves to become a monk forever.’ (12_2-bair-V_Rol-X > M205P-N1) Note that in the a-examples, in which the nominal predicate takes a determiner, the subject expression displays properties of a discourse referent in the sense of Karttunen (1976), i.e., the subject denotes a single individual present in the linguistic discourse. By contrast, in the b-examples, in which the nominal predicate is a bare noun, the subject is represented by a non-referring expression. In (25b), this is the negative quantifier nieman ‘nobody’. In (26b), the subject is a w-pronoun introducing a free relative clause. Finally, in (27b), the formal subject er is the correlate of a free relative clause which quantifies over individuals. It can be observed that in general, negative and quantified subject expressions trigger bare predicates, as shown in (28) and (29). Such expressions either reject the presence of an individual, or they refer to the members of a nonnearly defined set of individuals: (28) mag mít rehte keín gebure rihter geſín may with justice no peasant judge be-inf ‘It is not without good reason that no peasant may become a judge’ (13_2alem-PU-G > M339-G1) (29) Eîn îegelích man der rihter íſt every man who judge is ‘whoever is a judge’ (13_2-alem-PU-G > M339-G1) In addition, there is a series of clauses with bare predicates the subjects of which are realized as ordinary lexical expression. Consider (30)–(33) where the subject expressions are referential personal pronouns:
300
petrova
(30) vn̄ ſol er andreſt rihter werden and should he for.a.second.time judge become-inf ‘If he becomes a judge for a second time’ (13_2-alem-PU-G > M339-G1) (31) er ne ſî Rihtare he neg be-prs.sbjv judge ‘unless he is a judge’ (12_2-bairalem-PV-G > M061B-G1) (32) Bin ich richtere daz Ich sie stritē hiezze. am I judge that I them fight-inf command-pst.sbjv ‘Am I a judge such that I should let them fight.’ (14_1-thurhess-V_Herb2-X > M541H2-N) daz waiz wol arzet. (33) Wer hat dir geſaget daz ich ſi who has you told that I be-prs.sbjv physician this knows well uon himel min trehtin daz ich arzat niene bin. from heaven my lord that I physician never-NEG am ‘Who told you that I was a physician. Lord in Heavens knows that I am none.’ (12_2-bair-V_Kchr2-X > M121y2-N) Note, however, that these cases involve conditionals (30), hypothetical clauses (31), rhetorical questions (32), and declaratives whose truth is rejected (33). In other words, these are cases in which the subject expression is in the scope of a semantic (conditional, interrogative, or negative) operator. There is a parallel between these contexts and those involving non-referential subjects. Recall from the literature (e.g. Giannakidou 2011 for an overview) that these contexts license negative polarity items (NPIs) cross-linguistically, i.e. referentially deficient elements which are “unable to introduce discourse referents on their own” (Giannakidou 2011: 1662). I assume that, similarly to NPI s, lexical expressions in the respective operator-bound contexts lose their ability to introduce discourse referents, because the proposition that the sentences convey does not apply in any possible world. Let us test the effect of the referentiality of the subject expression on the choice of BNPs vs. INPs in the MHD data. The relevant figures are presented in Table 9.5. It can be observed that non-referentiality of the subject referent triggers BNPs in a statistically significant way in the MHG database. But it is questionable if this condition is restricted to nouns in predicative use. It is known from the literature that indefinites in MHG often lack a determiner if used nonreferentially, e.g. generically or in the scope of negation (Desportes 2000, Paul
bare and indefinite nominal predicates table 9.5
301
Effect of subject referentiality on BNP s vs. INPs in MHG
Referential subjects
Non-referential subjects
17 64
46 6
BNP INP Fisher Exact p < .05
2007: 378–382 and Jäger 2007), irrespectively of the semantic class or the grammatical function of the noun phrase.4 Some examples are given in (34) and (35): (34) vndᵉ wart nie gaſt geminnet baz and became never guest loved better ‘Never was a guest loved more’ (13_1-obd-V-G > M342-G1) (35) den wîp ze liebe ie gewan whom lady to love ever received ‘Whom a lady ever found as a lover’ (MWB Online, Iw 57) It is highly probable, consequently, that in MHG, the observed effect of nonreferentiality and non-factivity on article drop is not restricted to predicates but applies to indefinites in general. 2.4 Interim Conclusion The diachronic corpus study presented above reveals two basic results. First, it turns out that in both ENHG and MHG, denotations of institutionalized types of activities systematically reject the indefinite determiner when used as predicates. In this respect, BNPs in historical German resemble those in standard PDG, in that in all cases, they assign to the subject referent the “property of being involved in some well-established activities or states” (Geist 2014: 96). Consequently, denoting social roles is a constant property associated with BNPs throughout the entire history of German. There is, however, a significant difference, in that social roles denoted by BNP s in historical German are restricted to only one position at a time present in the respective domain of reference. In this sense, historical representatives of the class of social roles display the property of uniqueness. This condition is obviously weakened in 4 Similar factors governing the distribution of the zero article are also hold in Medieval French, see Becker (2013).
302
petrova
the course of time. What is important in PDG is that institutionalization of the respective position or activity takes place, independently of the number of individuals associated with this position. But this is a change in the conception, and therefore in the inventory of social roles, that is basically non-linguistic in nature. While the inventory of states and activities perceived of as social roles changes on extra-linguistic, socio-cultural grounds, the drop of the determiner is a diachronically constant feature associated with the predictive use of the denotations of these social roles. Second, in those cases in which there is variation regarding the use of the determiner, e.g. with predicates denoting crafts and occupations, we identified two factors supporting the drop of the determiner. These are i. coordination and disjunction and ii. non-referentiality of the subject expression. There are, however, reasons allowing us to assume that these factors probably apply to indefinites in these periods in general, not only to nouns in predicative use. The follow-up question is why role nouns constantly occur as BNP s and how the difference between standard PDG and the recent dialects using INP s instead should be explained.
3
Theoretical Explanation
3.1 Previous Accounts The previous formal literature on the variation between BNP s and INP s has pursued an explanation according to which there is a relation between the lack of a determiner and the semantic deficiency of the phrase. According to Zamparelli (2008), the basic property of BNPs is gender deficiency. More precisely, he argues that role nouns are deficient for gender but receive gender specification via agreement with the subject NP, which is in a close relation with the nominal predicate in the argument structure of the copula verb. In English, which has lost the gender distinction, BNPs are obsolete because there is no feature to transmit via Agree any longer. Note that this approach is unable to capture the situation in the modern German dialects presented in the introduction, whose predicates expressing social roles may take a determiner although the gender specification is preserved. In addition, Geist (2014) puts forward a substantial counter-argument on the basis of PDG data presented in (36). Note that although German has gender specification, the subject does not necessarily transmit its gender feature to the nominal predicate, which is visible from the fact that a mismatch between the overt gender features of the subject and the nominal predicate may occur:
bare and indefinite nominal predicates
303
(36) Antoinette war unschuldiges Opfer Antoinette-fem was innocent-neut victim-neut ‘Antoinette was an innocent victim’ An alternative approach is put forward by de Swart, Winter and Zwarts (2005), claiming that the semantic deficiency of nouns expressing social roles results from the structural deficiency of the type of syntactic phrase that they represent. In short, there is a basic difference between BNP s and INP s in terms of phrase structure, in that INPs are considered DP s with a fully-fledged, complex internal structure of the type given in (37), while BNP s represent a less complex type of phrase, a single NP, lacking the projections NumP and DP, respectively. (37) [DP D [NumP Num [NP N]]] This approach is based on Dutch data, more precisely on some languagespecific facts regarding the possibility of adjectival modification of class and role nouns, as well as on the presence or absence of overt inflectional marking on modifiers of the respective classes. Note that BNP s do not allow modification, except for some event-related (internal) adjectives specifying a sub-class of professions etc., e.g. managing director, artistic director, etc. Apart from this highly restricted class of modifiers, the broad class of evaluative adjectives like good, successful etc. automatically requires the indefinite determiner if applied to a role noun. Consider, however, the inflectional behaviour of modifying adjectives in the following Dutch data taken from de Swart, Winter and Zwarts (2005: 452): (38) a. Jan is werkloos visser. Jan is unemployed fisherman ‘Jan is an unemployed fisherman.’ b. Jan is *klein visser. Jan is short fisherman ‘Jan is a short fisherman.’ c. Jan is een klein-e visser. Jan is a short-INFL fisherman d. Jan is *werkloz-e visser. Jan is unemployed-INFL fisherman
304
petrova
Assume that the expression werkloos visser ‘(an) unemployed fisherman’ is used to denote a sub-kind of fishermen, thus possible without an article in Dutch, see (38a). The evaluative adjective klein ‘short’ in (38b) does not support a similar sub-kind interpretation but rather enforces a class-noun interpretation of the head noun, requiring the indefinite determiner, as shown in (38c). In this case, the adjective additionally displays an inflectional feature, the morpheme schwa (-e). Crucially, this inflectional feature is missing on the internal adjective werkloos in (38a), and it leads to ungrammaticality if added, as shown in (38d). Applying the complex DP-structure in (37), de Swart, Winter and Zwarts (2005) interpret these facts in the following manner. They assume that the indefinite determiner is located in the DP-layer, while the inflectional feature is situated in NumP (de Swart, Winter and Zwarts 2005: 453). The ban of inflection on adjectives supporting a sub-kind interpretation, like in werkloos in (38d), is taken to suggest that the respective adjective is not located in NumP but rather in the NP, together with the noun. From this, they conclude that BNP s have an impoverished structure, lacking a DP and a NumP, with the property of number deficiency being their basic characteristics. For the most part, the facts explained for Dutch hold for PDG as well. Crucially, modification of BPNs is possible only if the adjective is part of the denotation of a sub-kind of activity, while evaluative adjectives require the determiner in the way shown for Dutch above. In a minimal pair like (39a–b), there is a clear difference in meaning between the two sentences. The modified BNP in (39a) means that someone is a member (or a follower) of the Free Democratic Party in Germany, while (39b) says that a democratic person enjoys individual freedom (in contrast to being in prison, etc.): (39) a. Er ist freier Demokrat. he is free democrat ‘He is a member / a follower of the Free Democratic Party.’ b. Er ist ein freier Demokrat. he is a free democrat ‘He is a democrat enjoying personal freedom.’ However, as also remarked by Geist (2014: 91) passim, there is a crucial difference between the Dutch and the German examples regarding the presence of inflection on internal modifiers of BNPs. Note that such modifiers display obligatory adjectival inflection in German, as also visible from the ungrammaticality of the non-inflected variants in (40). Consequently, the lack of inflection
bare and indefinite nominal predicates
305
on event-related modifiers in BNPs is a specific property of Dutch, rather than a universal property supporting the conclusion that BNP s lack NumP crosslinguistically. (40) Peter ist {*leitend / leitend-er Oberarzt} / {*technisch / Peter is managing managing-INFL director technical / technisch-er Assistent}. technical-INFL assistant ‘Peter is a senior physician / a technical assistant.’ Geist (2014) puts forward another argument suggesting that BNP s are number deficient. She evokes the well-known fact that, in PDG, role nouns are indifferent regarding the number specification of the corresponding subject expression, also accounted for by Duden (92016) and Berman (2009). Note that while INPs necessarily agree in number with their subjects (41a), singular BNP are compatible with both singular and plural subjects (41b): (41) a. Beide Brüder wurden *ein Held / Helden. both brothers-PL became a hero-SG/ heroes-PL ‘Both brothers became heroes.’ (Geist 2014: 89) b. Beide Brüder wurden Ingenieur / Ingenieure. both brothers-PL became engineer-SG / engineers-PL ‘Both brothers became engineers.’ (Geist 2014: 98) Crucially, both de Swart, Winter and Zwarts (2005) and Geist (2014) explain the lack of NumP as the reason for the lack of a determiner in BNP s, assuming that the latter is the spell-out of a singular Num°. This, however, cannot account for role nouns taking a determiner in the modern German dialects. In the following section, I will adopt the idea that BNPs are number deficient, but will provide an alternative explanation for the drop of the determiner in nominal predicates referring to social roles. 3.2 Alternative Proposal Geist (2014) accounts for a crucial property of BNP s in PDG according to which they do not describe individual, episodic events but rather general dispositions, e.g. properties qualifying someone to be engaged in an institutionalized kind of activity. In other words, predicates referring to social roles, e.g. Lehrer sein ‘to be a teacher’ do not describe what a person does in a particular situation but denote that a person is authorized to hold an institutionally defined type of
306
petrova
position, e.g. that of a teacher. By virtue of this property, Geist (2014) establishes a parallel between BNPs and so-called generic, or characterizing sentences, which also lack an episodic interpretation. E.g., the sentence Mary handled the mail from Antarctica is true in its interpretation that Mary was in charge of the mail from Antarctica even if such mail in fact never arrived (Krifka et al. 1995: 9). Adopting the similarity between generic sentences and predicates denoting social roles, Geist (2014) concludes that BNPs in PDG denote well-established activities, or event kinds ek that are ascribed to human individuals. While I agree that there is a principled similarity between predicates ascribing social roles to individuals and sentences generalizing over prototypical situations, I reject the event kind interpretation of BNP s and propose an alternative analysis according to which role nouns, like proper names, are rigid designators in the sense of Kripke (1980). By definition, rigid designators, e.g. proper names, denote the same entity in any possible world in which this entity exists. I would like to apply this notion to role nouns as rigid designators in the sense that they establish reference to a unique function, property or activity that is subject to well-known institutionalized, social or legal regulations. Role nouns are labels used to refer to entities in the extra-linguistic world that are fixed in terms of regulations, approvals and legal procedures, and they are identifiable as such by the members of that community. The rigidity of denotation of nouns referring to social roles is especially evident in the context of the MHG and ENHG attestation, as the respective types of activities correspond to single and therefore maximally identifiable entities within the domain of reference. In addition, I want to assume that social roles share another property of proper names and other non-mass singular nouns with a restricted domain of denotation that is studied in Longobardi (1994, 1996, 2005), namely that they are raisable to the empty D-head, provided that this empty category is present in the determiner system of the language. I want to apply Longobardi’s (1994, 2005) theory on N-movement to [e D], which enables us to account for the bare vs. indefinite distinction in the diachronic data, but also for the establishment of the modern German dialectal situation in a uniform, straightforward manner. Following seminal work by Abney (1987), Longobardi (1994, 2005) argues in favour of an analysis of definite descriptions like the book which is in line with the structure in (42), i.e. the noun book is the head of an NP that is selected as a complement of a higher projection DP whose head hosts the determiner the. This analysis extends to indefinite DPs with an overt indefinite determiner like a(n) in D. In addition, Longobardi (1994, 2005) observes a series of contexts in which a (definite) determiner can be missing. In Italian, this applies to some well-defined uses of DPs as non-arguments (e.g., DP s used as vocatives, pred-
bare and indefinite nominal predicates
307
icatives as well as exclamatives), but also to some argumental DP s sharing the property of number deficiency (singular mass nouns, plural count nouns not necessarily implying plurality of referents, and existential indefinites). Longobardi assumes that such nominal expressions have the same underlying structure as in (42), but that, crucially, they involve an empty determiner [D e], as shown in (43), which is subject to lexical and semantic restrictions: (42) [DP [D Det] [NP [N N ]]] (43) [DP [D e] [NP [N N ]]] Let us apply this structure to the data discussed for historical German. Recall that in MHG, nouns denoting crafts, occupations, origin, etc., represent regular class nouns which take the determiner unless used in coordination or disjunction, or under conditions rejecting the referential status of the subject expression. For the sake of explicitness, I will repeat the minimal pair in (25a– b) as (44a–b) below. Recall the observation that in MHG, the omission of the determiner under coordination and disjunction, or if the subject expression is non-referential, is not restricted to nominal predicates but also applies for DP s as arguments. Assuming that the bare form of the noun results from licensing an empty determiner under the condition of number deficiency, the structure of these nouns in (44a–b) can be represented along the lines of (45a–b) and (46a–b), respectively: (44) a. Asclêpîus ein arzât was. Asclepius a physician was ‘Asclepius was a physician.’ (MWB Online, RvEBarl 10083) b. Dâ sol nieman arzât wesen. there should nobody physician be-inf ‘There, nobody should be a physician.’ (MWB Online, SM:St 3: 2, 7) (45) a. [DP [D Det] [NP [N N ]]] b. [DP [D ein] [NP [N arzât ]]] (46) a. [DP [D e] [NP [N N ]]] b. [DP [D e] [NP [N arzât ]]] Let us turn to nouns denoting social roles. We observed that they appear as BNPs throughout the history of German. Let us adopt for role nouns in BNP s
308
petrova
the status of a special class of proper names uniquely referring to well-known institutionalized states and activities typical of the respective period. According to Longobardi (1994 and 2005) proper names in Italian are subject to Nmovement to [D e ], as represented in (47a–b). Extending this type of movement to proper names denoting social roles, the structural representation of a case like (19a) = (48) would be that in (49): (47) a. [DP [D e] [NP [N N ]]] b. [DP [D Ni ] [NP [N ti ]]] (48) Sindolt der was scenke. Sindhold, he was cupbearer ‘Sindhold, he was a cupbearer.’ (MWB Online, NibB 11,3) (49) [DP [D scencei ] [NP [N ti ]]] The crucial question is whether there is evidence for N-movement to [D e] in the respective historical periods of German. Let us test this by adopting Longobardi’s (1994 and 2005) diagnostics, namely the difference in the position of proper names relative to modifiers and possessives, depending on the presence or absence of an overt determiner. The empirical facts relevant for Italian are as follows: In the presence of a determiner (il) possessives or modifying adjectives precede the proper name, as in (50a–b) and (51a–b), while in the absence of a determiner, this order is obligatorily reversed, i.e. the noun precedes the possessive or the modifier, see (52) and (53). Longobardi (1994) accounts for this fact by claiming that in the cases in which the determiner is missing, the proper name leaves its underlying position in N° and raises to the empty Dhead, as shown in (47a–b) above: (50) a. Il mio Gianni ha finalmente telefonato. the my Gianni has finally called ‘My Gianni has finally called.’ b. *mio Gianni ha finalmente telefonato (51) a. È venuto il vecchio Cameresi. is arrived the old Cameresi ‘The old Cameresi has arrived.’ b. *È venuto vecchio Cameresi
bare and indefinite nominal predicates
309
(52) Gianni mio ha finalmente telefonato. Gianni my has finally called ‘My Gianni has finally called.’ (53) È venuto Cameresi vecchio. is arrived Cameresi old ‘The old Cameresi has arrived.’ Crisma (1997, 1999) has argued that N-movement to [D e] takes place in Old English (OE), accounting for the lack of the determiner in one lexically restricted case that she examines, namely when the noun god ‘god’, sharing the status of a proper name, is modified by the adjective ælmihtig ‘almighty’. Crisma (1997) finds out that a putative order in which the determiner co-occurs with the post-nominal adjective, i.e. the pattern se god ælmihtig ‘the God almighty’, is not attested in her database. Instead, only the post-nominal bare variant god ælmihtig is found (see also Crisma 1999). In the reverse case, when the proper name follows the adjective, the determiner may appear, leading to patterns like se ælmihtig god or ælmihtig god.5 This distribution of pre- and postnominally modified proper names with and without a determiner supports a N-to-D movement analysis for OE. Let us look at historical German. For OHG, I searched all proper names modified by attributive adjectives found in Referenzkorpus Altdeutsch (REA),6 and determined the position of the adjective relative to the noun in the presence or absence of a determiner. As (54a–d) shows, all four putatively possible patterns are found in the data: 5 In addition, Crisma (1999) observes that the pattern DET–ADJ–N (se ælmihtig god) is strongly preferred in DP s used as arguments, while the bare variant of this order (ælmihtig god) is typical for non-arguments, such as vocatives and nominal predicates. According to her, this sheds light onto a potential structural DP/NP-mismatch between nominal expressions in these two syntactic domains. However, as the example in (54b) suggest, the bare ADJ–N variant is possible in argumental DP s in OHG as well, see also Petrova (2019) for a detailed corpus examination on the argumental/non-argumental distribution of various ADJ–N orders in OHG and MHG. 6 REA involves two relevant annotations, namely ordinary adjectives applied to proper names as well as adjectives as parts of proper names. Both options were tested. The queries in i. and ii. yield proper names with adjectives in the order ADJ–N, while those in iii. and iv. yield the orders N–ADJ. In brackets, I provide the total of hits that each of the queries produced. Note that for each of the queries, various instances had to be removed from the database on manual search, because the respective examples were inconclusive. i. pos=“ADJ” & pos=“NE” & lang=“goh” & #1.#2 & #2_=_#3 (74 hits) ii. pos=“ADJE” & lang=“goh” & #1_=_#2 (93 hits) iii. pos=“ADJN” & pos=“NE” & lang=“goh” & #2.#1 & #2_=_#3 (52 hits) iv. pos=“ADJNE” & lang=“goh” & #2.#1 (2 hits).
310
petrova
(54) a. Dhes chiuuarin Iesuses the-gen.sg true-gen.sg.weak Jesus-gen.sg ‘of the true Jesus’ (DDD-AD-Isidor_1.1 > I_DeFide_6)
DET–ADJ–N
b. Quam Almahtic Got. ADJ–N came almighty God ‘The almighty God came.’ (DDD-AD-Monsee_1.1 > MF_5_FH.XLI) c. Dâr piutit der Satanasz altist heizzan there offers the Satan oldest hot-acc.sg.strong Lauc. DET–N–ADJ blaze ‘There, the oldest Satan offers raging flames.’ (DDD-AD-Kleinere_Althochdeutsche_Denkmäler_1.1> M_Muspilli) Hártmuatan joh Wérinbrahtan d. Krist hálte Christ save-prs.sbjv Hartmut-acc.sg and Werinbraht-acc.sg gúatan. N–ADJ good-acc.sg.weak ‘Christ save Hartmut and the good Werinbracht.’ (DDD-AD-Otfrid_1.1 > O_Otfr.Hartm) But the distribution of these patterns, presented in Table 9.6, shows that the critical pattern in (54c), namely the one in which both the determiner and the noun precede the adjective, is attested only once. Thus, I conclude that this instance is an exception. Moreover, this example is found in the poetic text of Muspilli, allowing us to assume that this pattern is due to metrical demands. Upon the rest of the examples, it can be concluded that the proper name and the determiner are in complementary distribution in the position preceding the modifying adjective, which is consistent with the N-to-D hypothesis. The same situation holds for MHG as well. I searched the prose texts included in Referenzkorpus Mittelhochdeutsch (REM), removing instances in which Latin loan material occurs in the examples. Again, all four putative orders can be identified in the data, see (55a–d):7
7 The query in i. yields the order ADJ–N, the one in ii. is used to find N–ADJ orders. Examples containing Latin material were excluded. i. pos=“ADJA” & pos=“NE” & #1.#2 & meta::genre=“P” (1901 hits). ii. pos=“ADJN” & pos=“NE” & #2.#1 & meta::genre=“P” (4 hits, 2 invalid).
bare and indefinite nominal predicates table 9.6
ADJ–N N–ADJ
311
The presence of a determiner in modified proper names in REA
+DET
–DET
62 1
46 47
(55) a. Der heligo chriſt vvar geboren cebetlehē. DET–ADJ–N the holy Christ was born to.Bethlehem ‘The Holy Christ was born in Bethlehem.’ (11–12_1-obd-PV-X > M035-N1) b. Eya Gutter Ihc yeah good Jesus ‘Yeah, good Jesus’ (14_1-rhfrhess-PV-X > M407y-N0) ſynai heiligen c. An dᵉme by the-dat.sg Sinai holy-dat.sg.weak ‘by the holy Sinai’ (12_2-wmd-PVU-X > M188y-N1) d. Biſchoffes Otten ſelgen bishop-gen.sg Otto-gen.sg blessed-gen.sg.weak ‘of the blessed Bishop Otto’ (14_1-ofrk-PU-G > M356-G1)
ADJ–N
DET–N–ADJ
N–ADJ
But again, the quantitative distribution of these patterns, represented in Table 9.7, is rather telling. The figures in Table 9.7 reveal that post-nominal modification is practically non-existent in MHG prose texts, given that the total of conclusive examples involving the N–ADJ order amounts to 2 examples.8 Only one of these displays the critical order, in which both the determiner and the noun precede the adjective (i.e. DET–N–ADJ in (55c)).
8 If the restriction on prose texts is removed, the number of total hits for the N–ADJ order increases to 90 instances. The respective query is pos=“ADJN” & pos=“NE” & #2.#1. The number of critical patterns of the type DET–N–ADJ increases by only one additional example, though.
312 table 9.7
ADJ–N N–ADJ
petrova The presence of a determiner in modified proper names in REM
+DET
–DET
319 1
15 1
In sum, proper names preceding attributive adjectives do not involve a determiner in OHG and MHG. This is in line with an analysis according to which in these cases, the proper noun undergoes movement to an empty D-head. Let us sketch the later development of bare nouns towards standard PDG and the modern German dialects. First, as already mentioned, it is assumed that, due to extra-linguistic factors, the inventory of nouns referring to institutionalized states and activities is enriched by further representatives, e.g. professions, nationalities, etc. This explains their systematic use as BNP s in standard PDG, in other words, nouns which enter the class of denotations of social roles also acquire their ability of undergoing N-to-D movement, as is typical for proper names. But in some modern dialects, role nouns still take a determiner. A crucial fact regarding these dialects is, however, that they display a determiner with nouns that are number deficient or existentially bound, as outlined by Glaser (1996) for Bavarian. Note that in contrast to the standard language, the indefinite determiner is present with singular abstract (56), singular mass (57) as even plural count nouns in existential use (58): (56) Gebts doch a rua. give modP a silence ‘Be quiet.’ (Glaser 1996: 151) (57) I brauchad a kloageld firn automatn. I need-pst.sbjv a change for-the machine ‘I need some change for the vending machine.’ (Glaser 1996: 151) (58) Mägsd a kirsch. like-you a cherry-PL ‘Do you like some cherries.’ (Glaser 1996: 160)
bare and indefinite nominal predicates
313
Glaser (1996) points out that the expansion of the indefinite determiner in Bavarian is known from earlier dialectal data reaching back to the MHG period, and that the respective forms are not ubiquitous, but highly productive. The assumption is that this dialect loses the property of licensing [D e], which results in article use across the board, incl. the ban on N-to-D movement for role nouns in nominal predicates.
4
Conclusions
Starting from the observation that standard PDG gives rise to a semantically driven distribution of BNPs and INPs which (at first glance) is not existent in some of the modern dialects or in the historical stages of the language, the question was asked as to what governs the use of the indefinite determiner in the attested diachronic stages of German and how the establishment of the modern German situation can be explained. The corpus study of MHG and ENHG showed that the basic factor correlating with BNPs, namely the affiliation of the subject referent to a unique socially established, institutionalized kind of state or activity is consistently present throughout the entire attestation. But the inventory of such activities changes on extra-linguistic grounds, leading to differences regarding the use of the determiner over time. In addition, factors favouring the omission of the determiner, such as coordination/disjunction or non-referentiality of the subject referent, were identified, which however do not only apply for indefinites as predicates but also for indefinites as arguments. Adopting to the data a diachronically constant DP–NP model, involving the additional option of licensing lexically governed empty determiners (59), the two types of bare nouns identified in the historical corpus received two different structural representations, given in (59b–c): (59) a. [DP [D D] [NP [N N ]]] b. [DP [D e] [NP [N N ]]] c. [DP [D Ni ] [NP [N ti ]]] It was argued that singular count nouns in coordination / disjunction as well as in non-referential environments are number deficient, occupying the Nposition in an DP–NP model involving an empty determiner (59b). Bare nouns in predicates denoting social role are different. Adopting the assumption that such nouns act as proper names, and assuming that proper names undergo Nmovement to (empty) D (Longobardi 1994), it was argued that these nouns are
314
petrova
subject to this type of movement in OHG, as represented in (59c). On the basis of corpus evidence, it was shown that N-to-D movement is present in OHG and MHG, because the attested distribution of determiners with modified proper names in these periods correspond to that in languages in which N-to-D movement takes place. The later development is straightforwardly representable along the lines of this model. Assuming an extra-linguistically triggered change in the conceptualization of activities as social roles over time, we observe an increase in the inventory of nouns undergoing N-movement to D towards standard PDG, with ENHG being the transitional period proper. In other words, BNP s with role nouns retain their structural representation given in (59c). In addition, some original class nouns were re-analysed as role nouns, thus sharing the property of N-movement to D and the structure in (59c). The development takes another path in those dialects which do not display BNPs, such as Bavarian. The fact that these dialects lose bare nouns in almost all environments indicates that they lose the option of licensing [D e] for nouns that are number deficient or existentially bound. In other words, the option in (59b) becomes obsolete in these varieties. The loss of [D e], in turn, leads to the loss of N-to-D movement, leading to the presence of a determiner in predicates denoting social roles.
Electronic Resources MWB Online REA 1.1 REM REN FnhdC
http://www.mhdwb‑online.de/index.html (last accessed on May 7th, 2021) https://korpling.org/annis3/ddd/ (last accessed on May 7th, 2021) https://www.linguistics.rub.de/annis//annis3/REM/ (last accessed on May 7th, 2021) https://corpora.uni‑hamburg.de/annis/ren (last accessed on May 7th, 2021) https://korpora.zim.uni‑duisburg‑essen.de/annis/ (last accessed on May 7th, 2021)
References Abney, Steven Paul. 1987. The English noun phrase in its sentential aspect. Cambridge, MA: MIT PhD dissertation.
bare and indefinite nominal predicates
315
Behaghel, Otto. 1923. Deutsche Syntax. Eine geschichtliche Darstellung. Band I. Die Wortformen. A. Nomen. Pronomen. Heidelberg: Winter. Becker, Martin. 2013. Competing zones: The zero article in Old French and its loss in the history of the French language. In Sofiana Chiriacescu (ed.), Proceedings of the vi Nereus International Workshop “Theoretical implications at the syntax/semantics interface in Romance”, Arbeitspapier 127, 1–17. Fachbereich Sprachwissenschaft, Universität Konstanz. Berman, Judith. 2009. The predicate as a source of grammatical variation. In Andreas Dufter, Jürg Fleischer & Guido Seiler (eds.), Describing and modeling variation in grammar, 99–116. Berlin: de Gruyter. Bisle-Müller, Hansjörg 1991. Artikelwörter im Deutschen. Semantische und pragmatische Aspekte ihrer Verwendung. Tübingen: Niemeyer. Crisma, Paola. 1997. L’articolo nella prosa inglese antica e la teoria degli articoli nulli. Padova: Università di Padova PhD dissertation. Crisma, Paola. 1999. Nominals without the article in the Germanic languages. Rivista di Grammatica Generativa 24. 105–125. Desportes, Yvon. 2000. Artikel im Mittelhochdeutschen. Lässt sich Paul Valentins Modell des Artikelsystems im heutigen Deutsch auf das Mittelhochdeutsche übertragen? In Yvon Desportes (ed.), Zur Geschichte der Nominalgruppe im älteren Deutsch. Festschrift für Paul Valentin, 213–253. Heidelberg: Winter. de Swart, Henriette, Yoad Winter & Joost Zwarts. 2005. Bare predicate nominals in Dutch. In Emar Maier, Corien Bary & Janneke Huitink (eds.) Proceedings of Sinn und Bedeutung 9, 446–460. Nijmegen: NCS. Duden, die Grammatik, 9th edition. 2016. Edited by Angelika Wöllstein et al. Mannheim: Dudenverlag. Eisenberg, Peter. 2006. Grundriß der deutschen Grammatik, 2: Der Satz. Stuttgart: Metzler. Engel, Ulrich. 2004. Deutsche Grammatik. München: Iudicium. Geist, Ljudmila. 2014. Bare predicate nominals in German. In Ludmila Veselovská & Markéta Janebová (eds.), Complex visibles out there. Proceedings of the Olomouc Linguistic Colloquium 2014: Language use and linguistic structure, 83–106. Olomouc: Palacký University. Giannakidou, Anastasia. 2011. Negative and positive polarity items. In Klaus von Heusinger, Claudia Maienborn & Paul Portner (eds.), Semantics, 1660–1712. Berlin: de Gruyter. Givón, Talmy. 1981. On the development of the numeral ‘one’ as an indefinite marker. Folia Linguistica Historica 2. 35–53. Glaser, Elvira. 1996. Morphologie und Funktion des unbestimmten Artikels im Bairischen. In Hans-Werner Eroms & Hermann Scheuringer (eds.), Proceedings of Schriften zur Literatur und Sprache in Oberösterreich 5. Sprache an Donau, Inn und Enns.
316
petrova
Vorträge des Symposions zu Sprache und Kultur des altbairischen Raumes 1994, 149– 169. Hallab, Amina Christine. 2011. Referentialität, Prädikation und die Struktur der Nominalphrase: Kontrastierung artikelloser nominaler Prädikate und Indefinita in Kopulasätzen im Deutschen. Hamburg: Kovač. Hartweg, Frédéric & Klaus-Peter Wegera. 22005. Frühneuhochdeutsch. Eine Einführung in die deutsche Sprache des Spätmittelalters und der frühen Neuzeit. Tübingen: Niemeyer. Heine, Bernd. 1997. Cognitive foundations of grammar. Oxford: Oxford University Press. Henn-Memmesheimer, Beate. 1986. Nonstandardmuster. Ihre Beschreibung in der Syntax und das Problem ihrer Arealität. Tübingen: Niemeyer. Heycock, Caroline & Roberto Zamparelli. 2003. Coordinated bare definites. Linguistic Inquiry 34(3). 443–469. Jäger, Agnes. 2007. No changes: On the history of German indefinite determiners in the scope of negation. In Elisabeth Stark, Elisabeth Leiss & Werner Abraham (eds.), Nominal determination. Typology, context constraints, and historical emergence, 141– 170. Amsterdam: Benjamins. Karttunen, Lauri. 1976. Discourse referents. In James David McCawley (ed.), Notes from the Linguistic Underground, 363–385. New York: Academic Press. Kolmer, Agnes. 1999. Zur MASS/COUNT-Distinktion im Bairischen: Artikel und Quantifizierung. Cologne: University of Cologne MA thesis. Krifka, Manfred, Francis Jeffry Pelletier, Gregory Norman Carlson, Alice Ter Meulen, Gennaro Chierchia & Godehard Link. 1995. Genericity: Introduction. In Gregory Norman Carlson & Francis Jeffry Pelletier (eds.), The Generic Book, 1–124. Chicago: University of Chicago Press. Kripke, Saul. 1980. Naming and necessity. Oxford: Blackwell. Longobardi, Giuseppe. 1994. Reference and proper names: A theory of N-movement in syntax and logical form. Linguistic Inquiry 25(4). 609–665. Longobardi, Giuseppe. 1996. The syntax of N-raising: a minimalist theory. OTS Working Papers 96(5). 1–55. Longobardi, Giuseppe. 2005. Towards a unified grammar of reference. Zeitschrift für Sprachwissenschaft 24. 5–44. Oubouzar, Erika. 2000. Zur Entwicklung von ein in der Nominalgruppe des Althochdeutschen. In Yvon Desportes (ed.), Zur Geschichte der Nominalphrase im älteren Deutsch. Festschrift für Paul Valentin, 255–268. Heidelberg: Winter. Paul, Hermann. 2007. Mittelhochdeutsche Grammatik. 25th edition. Tübingen: Niemeyer. Petrova, Svetlana. 2015. Synchronic variation and diachronic change in the expression of indefinite reference: Evidence from historical German. Zeitschrift für Sprachwissenschaft 34(2). 213–246.
bare and indefinite nominal predicates
317
Petrova, Svetlana. 2019. Proper nouns and the structure of the DP in Old and Middle High German. In Natascha Pomino (ed.) Proceedings of the viii Nereus International Workshop “Morphosyntactic and semantic aspects of the DP in Romance and beyond”, 27–40. University of Wuppertal. Arbeitspapier. Fachbereich Sprachwissenschaft, Universität Konstanz. Schrodt, Richard. 2004. Althochdeutsche Grammatik ii: Syntax. Tübingen: Niemeyer. Szczepaniak, Renata. 22011. Grammatikalisierung im Deutschen: Eine Einführung. Tübingen: Narr. Weber, Albert. 31987. Zürichdeutsche Grammatik. Ein Wegweiser zur guten Mundart. Zürich: Rohr. Zamparelli, Roberto. 2008. Bare predicate nominals in Romance languages. In Henrik Høeg Müller & Alex Klinge, Essays on nominal determination: From morphology to discourse management, 101–130. Amsterdam: Benjamins.
Index of Subjects accessibility 15, 119–121 additive 55, 74–75, 77, 84 adposition 225, 229, 230, 232, 234, 236– 240 aggregate 19, 20, 55, 58, 76, 84 alternative 21, 22, 135, 136, 141, 214, 216, 217, 218, 221, 250, 253, 263, 264– 265 Alternative Semantics 11, 131, 133, 135–136, 141–142, 146, 165, 216, 217, 222, 241 ambiguity 209, 224, 249 anaphora 14, 15, 112–113, 146, 175 antecedent 96–103, 109–115, 120–121, 137 atomic 58, 67, 73, 77 bare nouns 3, 7, 9, 17–18, 21, 29–31, 35, 37–38, 44, 60–63, 65, 69–72, 77–81, 312–314 bare quantifiers 12–13, 186–211 bestimmt 16, 132, 148, 155–167, 169–179 bleaching 17 bounded / unbounded 55–56, 58, 67, 68, 73, 76–79, 80, 85 Cartographic approach 103–108, 196 case 45, 208–210 Catalan 64, 277 certain 15–16, 130, 131, 133, 146–147 choice function 134, 138, 140 Chuj 260–261 class nouns 18, 286, 297, 307, 314 classification 8, 17, 19, 20, 55–59, 67–75, 78– 79, 81–82, 84–86, 286 classifier 13, 37, 67–69, 79, 80, 186–187, 196, 205–206, 211 coherence 14, 15, 110, 112 collective 19, 69–70, 74, 75, 76 competition 6, 7, 23, 31, 32, 42, 120, 122 conceptual cover 135, 157, 161, 163, 165 conventionalization 22, 216, 231, 232 count 8, 20, 37, 39, 44, 45, 58, 60, 68, 71, 74, 76–77, 81, 84, 307, 312–313 cualquier(a) 21, 215, 216, 222, 225, 226, 233, 275 cumulative 58
D 5, 37 empty 4, 36, 39, 45, 62, 80, 288–289, 306–308, 312–313 N-to-D movement 10, 44, 289, 306, 308– 309, 312–314 Danish 58 DP hypothesis 2–3, 5 definite article 5, 6, 7, 8, 37, 40, 61, 64, 79, 80 definiteness 123 weak 6 degree 256–257, 279 demonstrative 43, 249 directionality 17, 21, 239 discourse 13–14, 16, 152–153, 299–300 discrete 55, 69, 74, 76, 77, 84 dislocation 103–108, 115–117 divisive 37, 39, 40, 58, 77 DOM (Differential Object Marking) 72–73, 133, 208 Dutch 21, 46, 216, 222–228, 231–236, 239– 241, 286, 303–305 embedding 132, 158, 168 empty category 64, 80, 104–108, 118–119, 289, 306 English Early Middle 42 Irish 31 Jamaican 31 Middle 31, 32, 41 Modern (Present Day) 7, 29–30, 32, 34, 42, 45, 86, 214, 222–226, 302 Old 20, 31, 41, 43, 45, 46–49, 309 erotetic 11, 15, 131–133, 151–154, 167, 179 evaluative 12, 22, 246–281, 303–304 exhaustification 21, 219, 221, 232, 233 expletive 96 extended projection 3–4, 13, 187 feature 55, 58, 77, 81, 197, 205, 209, 210, 276, 279, 302, 304 focus 12, 106, 189–190, 201, 203, 208, 211, 250, 263–264, 267 Franco-Provençal 82 free choice 22, 214–241, 246–250, 253
320 French Modern (Present Day) 13, 21, 29–30, 38, 44, 45, 57, 60–61, 63, 64–68, 69, 71, 74– 83, 188–189, 198–199, 214, 259, 263 Old 19, 21, 43, 44, 45, 69, 301 Gascon 64 gender 56, 75, 79, 302 genericity 7–8, 20–21, 29–49, 56, 63, 223, 306 German Early New High 18, 289–295, 301, 306, 313 Middle High 18, 288–291, 295–301, 306– 307, 309–314 Middle Low 290–291 Modern (Present Day) 10, 18, 20, 34, 41, 78–79, 130–185, 214, 222, 285–286, 288–289, 294, 297, 301–302, 304–306, 312–314 Old High 289, 309, 312, 314 Upper 10, 287, 290 gewiss 12, 16, 131–132, 148, 151, 155–169, 176– 179 grammaticalization 8, 13, 17–22, 32, 55, 56, 57, 59, 67, 69, 74, 82, 83–84, 85, 86, 146, 164, 205, 226, 228–231, 250, 276, 278, 288 Hungarian 139, 148 Icelandic 188 identifiability 157–158, 166 implicational map 223 implicature 22, 255, 266 conventional 161, 162, 176, 231, 232 conversational 215, 216, 231, 232 incorporation 3 indefinite 217, 223–226, 228–232, 239, 240, 247 article 5–6, 8, 9, 16–17, 30, 56, 61, 69, 78, 81, 83, 85, 130, 286, 290 determiner 9, 10, 16, 60–61, 67–68, 76– 77, 151, 154, 286, 288–289, 291, 294–296, 301, 303–304, 306, 312–313 dependent 139 epistemic 133, 135 free choice 11, 12, 21, 135, 215–217, 232– 234, 240, 241
index of subjects negative 11 pronoun 130, 139–140, 178, 214 specialized 10, 12, 21, 130–133, 139, 141– 142, 146–154, 176–179, 216 specific 56, 57, 63, 70–74, 80, 132, 134, 179, 223, 224 wh-based 21, 139–141, 216, 232, 233, 241 indiscriminacy 224, 225, 228, 247, 248, 255, 261–263 individuated type 33, 34 information structure 15, 95–101, 211 interrogative 219, 222, 234 pronouns see wh-pronouns interrogative-indefinite affinity 11–12, 16, 139–141, 142, 152, 179, 300 intervention effect 197, 215 Italian Modern 13, 14, 44, 49, 58, 60–62, 63, 64– 68, 69–83, 96–103, 109–114, 189–190, 200–208, 210, 214–216, 220, 222, 225, 240, 241, 251–260, 306, 308 Old 13, 22, 190–196, 200 kind 20, 30, 35, 62, 63, 70, 80 labeling 208–210 Latin 43, 74–75, 79, 82, 146–147, 154, 246, 310 mass 8, 9, 20, 55–59, 67–68, 70, 71, 74–77, 79, 80, 83, 84, 287, 307, 312 modal 218–221, 235, 236, 249, 273, 275, 278– 279 molto 192–193, 201–204, 207–208 N-to-D movement see D negation 12, 63, 71, 82, 137, 158–160, 168–169, 214, 223, 249, 250, 254–256, 264, 265, 271, 273, 274, 300 negative polarity item 300 n-word 194–195, 200 niente 194–196, 200–208 no matter 21–22, 225, 228–241 nominal predicates 9, 18, 286, 288, 291–292, 294–296, 298, 305, 307, 309, 313 null subject 14–15, 94–95, 96–103, 115–123 number 7–8, 17, 19, 20–21, 38, 44–45, 56–58, 59, 61, 65–67, 74–76, 78, 81–85, 304– 305, 307, 312–314
321
index of subjects Occitan 19, 64, 69, 82 operator 216–220, 232, 236, 237 abstract 21, 214, 215, 216, 217 intensional 136, 151, 158–166, 169–177 optionality 62
relativized minimality 197, 203 rigid designator 306 role nouns 18, 286–287, 289–290, 293, 298, 302–303, 305–307, 312–314 Romanian 61, 64, 72, 79, 133, 154
partitive 8–9, 19–20, 38, 44, 57, 60–61, 64, 67–68, 71–74, 76, 78–83 personal pronoun 15, 94, 122, 289 Piedmontese 8, 19, 20, 64, 82–86 plural 8–9, 19, 29–49, 55–86, 307, 312 Portuguese Brazilian 63–64, 96 presupposition 144, 150, 177, 253 prominence 15, 110, 121
Sardinian 64 scalar implicature 232 scope 62, 63, 71–73, 300 exceptional 138, 141–142, 152, 163 intermediate 173 Seinsart 8, 55, 58, 68, 78, 81 semantic type 4–6, 144 Skolem function 134, 137–139, 161 Spanish Argentinian 259, 280 Modern 14, 58, 60–62, 63, 64–68, 69– 82, 106, 107, 114, 214–216, 222, 225–234, 239–241 Old 21, 22, 68, 226–228 specificity 6, 16, 29, 56, 57, 63, 70–74, 80, 130–179, 224 speech act operators 132, 155, 166, 174–178 subtrigging 221, 228, 247, 275
qualsiasi 276–279 qual si sia 240 qualunque 12, 22, 246–281 quantifier 56, 61, 65, 67, 69, 72, 73, 74, 79, 84, 85, 86, 112–114, 138, 142, 158–162, 186–211 floating 198–202, 210 Generalized Quantifier 4, 5 QP 4, 5, 187 negative 299 propositional quantifier 220, 222, 233, 239 quantifier positions 13, 186–211 quantifier raising 4, 138, 142, 143 question 11, 14, 16, 131–154, 166–173, 176–179, 217, 219, 220, 223 potential 152, 153 reanalysis 21, 277–278 relative clause 250 free 21, 219, 220, 225, 226, 230–241, 299 reduced 22, 275–278, 279
topic 14–15, 96–103, 115–117, 118–119, 211 tutto 84–85, 190–193, 199, 201–208 unconditional 235, 237 uniqueness 6, 10, 18, 292, 294, 301 weak definite see definite wh-based indefinite see indefinite wh-pronoun 139–141, 186, 196–197, 220 wie dan ook 21, 216, 222, 228–241