Essays on Nominal Determination: From morphology to discourse management (Studies in Language Companion Series)

Essays on Nominal Determination

Studies in Language Companion Series (SLCS) This series has been established as a companion series to the periodical Studies in Language.

Editors Werner Abraham University of Vienna

Michael Noonan

University of Wisconsin, Milwaukee

Editorial Board Joan Bybee

Christian Lehmann

Ulrike Claudi

Robert E. Longacre

Bernard Comrie

Brian MacWhinney

University of New Mexico University of Cologne Max Planck Institute, Leipzig University of California, Santa Barbara

William Croft

University of New Mexico

Östen Dahl

University of Stockholm

Gerrit J. Dimmendaal University of Cologne

Ekkehard König

Free University of Berlin

University of Erfurt

University of Texas, Arlington Carnegie-Mellon University

Marianne Mithun

University of California, Santa Barbara

Edith Moravcsik

University of Wisconsin, Milwaukee

Masayoshi Shibatani

Rice University and Kobe University

Russell S. Tomlin

University of Oregon

Volume 99 Essays on Nominal Determination. From morphology to discourse management Edited by Henrik Høeg Müller and Alex Klinge

Essays on Nominal Determination From morphology to discourse management

Edited by

Henrik Høeg Müller Alex Klinge Copenhagen Business School

John Benjamins Publishing Company Amsterdam / Philadelphia

8

TM

The paper used in this publication meets the minimum requirements of American National Standard for Information Sciences – Permanence of Paper for Printed Library Materials, ansi z39.48-1984.

Library of Congress Cataloging-in-Publication Data Essays on nominal determination : from morphology to discourse management / edited by Henrik Hoeg Muller, Alex Klinge. p. cm. (Studies in Language Companion Series, issn 0165-7763 ; v. 99) Includes bibliographical references and index. 1. Grammar, Comparative and general--Determiners. 2. Definiteness (Linguistics) I. Müller, Henrik Høeg. II. Klinge, Alex. P299.D48E86    2008 415--dc22 isbn 978 90 272 3110 9 (Hb; alk. paper)

2007052230

© 2008 – John Benjamins B.V. No part of this book may be reproduced in any form, by print, photoprint, microfilm, or any other means, without written permission from the publisher. John Benjamins Publishing Co. · P.O. Box 36224 · 1020 me Amsterdam · The Netherlands John Benjamins North America · P.O. Box 27519 · Philadelphia pa 19118-0519 · usa

Table of contents Contributors

vii

The editors

ix

Introduction Henrik Høeg Müller and Alex Klinge

xi

Determiners and definiteness: Functional semantics and structural differentiation Peter Harder

1

Articles, definite and indefinite Michael Herslund

27

Typological correlations in nominal determination in Romance Elisabeth Stark

45

A stranger in the house: The French article de Marc Wilmet

65

Determination in endocentric and exocentric languages: With evidence primarily from Danish and Italian Iørn Korzen

79

Bare predicate nominals in Romance languages Roberto Zamparelli

101

Definiteness effect and the role of the coda in existential constructions Manuel Leonetti

131

Determination of N2 modifiers in Spanish nominal syntagmatic compounds Henrik Høeg Müller

163



Essays on Nominal Determination

Reference to individuals, person, and the variety of mapping parameters Giuseppe Longobardi

189

English th- Forms Judy B. Bernstein

213

Stating the case for þ- root and hw- root determiners Alex Klinge

233

On certain differences between noun phrases and clauses Naoki Fukui and Mihoko Zushi

265

Determination, nominalisation and conceptual processing Helle Dam-Jensen

287

The semantics and pragmatics of the possessive determiner Georges Kleiber

309

Reference, determiners and descriptive content Thorstein Fretheim and Nana Aba Appiah Amfo

337

Index

365

Contributors Nana Aba Appiah Amfo University of Ghana Judy B. Bernstein William Paterson University, USA Helle Dam-Jensen Aarhus School of Business, Denmark Thorstein Fretheim Norwegian University of Science and Technology, Norway Naoki Fukui Sophia University, Japan Peter Harder University of Copenhagen, Denmark Michael Herslund Copenhagen Business School, Denmark Georges Kleiber Université Marc Bloc, France Alex Klinge Copenhagen Business School, Denmark Iørn Korzen Copenhagen Business School, Denmark Manuel Leonetti Universidad de Alcalá, Spain Giuseppe Longobardi Università di Trieste, Italy Henrik Høeg Müller Copenhagen Business School, Denmark Elisabeth Stark University of Zurich, Switzerland

 Essays on Nominal Determination

Marc Wilmet Université Libre de Bruxelles, Belgium Roberto Zamparelli Università di Trento, Italy Mihoko Zushi Kanagawa University, Japan

The editors Henrik Høeg Müller: is Associate Professor in the Department of Spanish at the Copenhagen Business School. His research interests include nominal compounding, determination and Spanish modal auxiliaries. Alex Klinge: is Associate Professor in the Department of English at the Copenhagen Business School. His research interests include nominal compounding, nominal structure in English compared with other Germanic languages, modality and tense.

Introduction Henrik Høeg Müller and Alex Klinge This is a volume which brings together scholars of diverse theoretical persuasions who all share an interest in capturing the role that determination plays in nominals. There is a long tradition of exploring the role of determination in reference assignment and the interplay of determination with quantification. One of the primary functions of determination is to guide reference assignment, and in this way determination plays a central role in providing a link between thought, language, communication, and the world. The diversity of theoretical persuasions represented reflects the observation that different theoretical frameworks may be brought fruitfully together by a shared interest in research questions.

1. The issues Determination may be approached at all interface levels between morphology, syntax, semantics and pragmatics. Thus, linguistic forms and structures of potential interest to scholars of language span at least articles, possessives, demonstratives, quantifiers, numerals, adjectives, nouns, and the phrasal projections that they head. In addition, the role of determination is absolutely central in discourse management. Nominal determination has always been a central subject of attention in the lingustics literature. In particular, articles and demonstratives have been studied extensively. Important work has been published in German(ic) linguistics for instance by Hodler (1954) and Abraham (1997), in English linguistics by Christophersen (1939), Jespersen (1942) and Hawkins (1978). In Romance linguistics, Wilmet (1986), Renzi (1976) and Korzen (1996) constitute standard references in the field. More universallyoriented approaches are found in for instance Longobardi (1994), Kamp and Reyle (1993), Chiercia (1998), C. Lyons (1999) and Zamparelli (2000). Noteworthy collections of articles on the topic of determination have also been published, such as Auwe­ ra (1980), Vogeleer & Tasmowski (2006) and Stark, Leiss & Abraham (2007). A contribution which in many ways stands out as a landmark, however, is Abney (1987), which set in motion an entirely new direction in research which changed focus from articles and demonstratives to uncovering functional layers in nominal structure.

 Henrik Høeg Müller and Alex Klinge

Certain issues seem to recur throughout the literature such as the diachronic perspective on the development of definite articles from demonstratives through a pro­ cess of grammaticalization, the layers of functional projections, overt and covert functional categories, cross-linguistic variation, and the role of D-projection in nominals. It is widely assumed that the definite articles of Germanic and Romance languages developed from demonstrative pronouns and that at some stage in the process of grammaticalization, the articles parted ways with the demonstratives and joined up with cardinal numerals to form categories of articles overtly expressing a definite-indefinite contrast. Formal and distributional synchronic facts are however not univocal. While the so-called definite and indefinite articles seem to be in complementary distribution, certain morphological and distributional observations question the traditional assumption. If definite and indefinite articles form a category with members in complementary distribution, how do we account for data such as German der eine Arm, Old French l’un braz, and Danish den ene arm? Have the different points of origin of the definite and indefinite articles in fact been retained, i.e., are they projections of the same or of different functional categories? While the definiteness and indefiniteness are clearly to be interpreted as constituting contrasting semantic values of a ‘definiteness’ feature, the position that the indefinite article and definite article together form a formal category has been challenged. Some writers, e.g., Vangsness (1999) and Wood (2003), have argued in favour of a position where the definite and indefinite articles are in different syntactic projections and thus not in the same formal category. Another issue which has surfaced in recent years (see e.g., Olsen 1989, Bernstein, this volume, Klinge, this volume) is that the morphological perspective suggests fundamental synchronic similarities in the make-up of demonstrative pronouns and definite articles, i.e., morphological decomposition leaves us with a root d- in German der, die, das forms irrespective of whether they distribute as determiners or pronouns – they may be distinguished only through focus accent. In our quest to establish how they are different, have we overlooked what Romance and Germanic demonstrative pronouns, demonstrative determiners and definite articles still share? Similarly, the so-called indefinite article still has clear affinities with quantifiers and numerals. For instance, the indefinite article is in complementary distribution with other quantifiers and nume­ rals, and it cannot co-occur with plural N-heads, cf. *Ich habe eine viele Bücher gekauft. This is unexpected if ein is merely a marker of indefiniteness, but expected if ein is still a numeral (see C. Lyons 1999). Assigning indefinite articles, definite articles, demonstrative determiners, and other morphological material to different categories calls for a nominal structure with explicit layers of projections onto which the morphemes map. Such a structure is mani­ fested as a series of projections. So for English nominal strings such as this his final claim and a final claim, at least SpecP, DP, NumP, and NP projections might be postulated, as in the following partial representation.



Introduction 

SpecDP Spec

DP D

NumP Num

this

his

a

NP2 AP

NP1

A

N

final final

claim claim

However, cross-linguistically there is substantial variation across languages in terms of which functional categories are overtly realised and in terms of mapping between lexical material and functional categories. Thus, for instance, in the string Look at this painting, the demonstrative determiner this is associated with definiteness, but in the string I met this strange guy on the subway last night the demonstrative determiner will be interpreted as non-specific and non-definite. This duality of interpretation is not found in Romance demonstratives, so in Spanish Me encontré con ese hombre the demonstrative determiner ese can only receive a definite interpretation. In cross-linguistic perspective, several issues arise. What is the inventory of functional categories? Is the inventory universal? To what extent do we accept covert categories? And how do we identify them? Radford (2004: 141) argues that the string Ita­ lians love opera has two nominals with covert determiners, i.e., they are both DPs. That this should be so is supported by the fact that they coordinate with DPs with overt determiners, e.g., Italians love [opera] and [the finer things in life]. However, certain examples are not so clear. In a stage-level predicate constructions such as Beavers are running in the yard (from Bosque 1996: 71, and cf. Carlson 1977), such coordination is hardly as straight forward, cf.?Beavers and the grey squirrels are running in the yard. Do we then draw the conclusion that beavers is not a DP or that it is a DP with a covert D? Much effort has been invested in uncovering possible functional categories in the nominal domain (see, for instance, Zamparelli 2000 and Bernstein 2001). However, as pointed out in Stark et al. (2007: 17) “ … research for a finite list of indispensable functional categories inside nominals has not yielded a definite result shared by everyone … “ Moreover, the issue is clouded by the fact that syntactic configurations and

 Henrik Høeg Müller and Alex Klinge

morphemes which are not obviously determiners can have functions which are comparable to those of determiners (see for instance Abraham 1997 and Leiss 2007). A further issue which ranks high in the discussion of nominal determination is the semantic contribution of determiners. What do we mean by definiteness? Is definiteness a semantic primitive? Does the definite article encode uniqueness or identifiability (cf. the discussion in Abott 2004)? Is uniqueness or identifiability part of the meaning of the definite article in Many students take the train to go home for Christmas? Does the definite article only become an exponent of definiteness because it distributes into D-position in a nominal? We saw above that the demonstrative determiner this is not necessarily associated with definiteness. The contribution of determiners may also be approached in terms of their contribution to discourse management. They may be seen as processing signals which assist the hearer in identifying the referent the speaker has in mind. Prince (1979), Heim (1988), Ariel (1990), and Gundel et al. (1993) have stimulated interest in the correlation between determiners and how we cognitively access referents in discourse. A question which still remains open is whether discourse management lies at the heart of the semantics of determiners, or whether it is a function which they have acquired as a result of some other semantic content? Did the definite article arise because it facilitated discourse management? However, while identifiability is clearly part of the understanding of the way the definite article is used in discourse, examples abound which show that identifiability is not an inherent semantic feature, cf. Beware of the dog.

2. An overview of the contributions to this volume The contributions to this volume address a number of the issues raised above. They fall roughly into five main areas of concern, viz. the conceptual level of determination; the emergence and function of articles; their semantic contribution to nominal interpretation; the morphology and syntax of determiners; and, finally, the interplay and contrasts between articles, demonstratives, and possessives. Peter Harder addresses the relation between conception and coding in the way the domain of nominal determination is structured. Contrasting insights from formal and functional approaches, Harder considers the grounding function of determination in terms of a combination of arbitrary and motivated relations between form and content. Michael Herslund, Elisabeth Stark, Marc Wilmet and Iørn Korzen all address the emergence and basic functions of articles. In his contribution, Herslund takes his point of departure in the emergence of article systems and argues that the functions and values of definite and indefinite articles are different, given that definite articles were derived as pronominal heads and indefinite articles were derived as quantifiers or classifiers. Elisabeth Stark similarly takes an historical approach in that she considers the emergence from Latin of two distinct Romance determination systems coding nominal classification in terms of contoured and non-contoured referents. The two systems



Introduction 

of nominal classification are anchored in the rigid article system in the case of the central Romance languages Italian and French, whereas the more flexible article sy­ stem of Spanish and Romanian has to be supplemented by differential object marking. Marc Wilmet argues in favour of treating French de as an article in its own right. His argument draws on aspects of definition, historical facts, distribution and the paradigm of articles. Iørn Korzen draws a typological distinction between endocentric and exocentric languages. The former are languages which have relatively heavy lexical and informational weight in the centre of the predicate, i.e. in the verb, such as Danish, whereas the latter are characterized by having a relatively heavy lexical and informational weight in nominal arguments, such as Italian. This typological difference leads to the prediction that informational prominence of nominal arguments is also reflected in a more fine-grained system of determination, whereas low prominence of nominal arguments should be reflected in a less intricate determination system and a greater likelihood of noun incorporation. Korzen tests these predictions by contrasting Italian and Danish. Roberto Zamparelli, Manuel Leonetti and Henrik Høeg Müller investigate semantic implications of the presence vs. absence of determiners. Zamparelli offers an analysis of the absence of determiners in singular predicate nominals which refer to professions, roles and other relations. Zamparelli argues that while singular count nouns are normally licensed by the presence of a determiner, nouns that form bare predicates have an impoverished set of features, in particular they have no set value for gender, so they may be licensed by entering into an agreement relation with the subject of the predication. The article by Leonetti investigates the effects produced by the coda on definiteness and connects such effects to other constrains on the licensing of postverbal subjects. One central claim is that it is a clash between definiteness and Focus structures that underlies the definiteness effect. Müller contrasts two distinct manifestations of Spanish syntagmatic compounds, viz. one with the structure N1 prep. def.N2 and N1 prep. ØN2. Müller’s point of departure is that the definite article attributes to N2 either a referential reading or a prototype reading, while the zero determiner brings about an interpretation either in terms of a mass or a concept. Giuseppe Longobardi, Judy Bernstein, Alex Klinge, Naoki Fukui & Mihoko Zushi and Helle Dam discuss various interface phenomena between morphology, syntax and semantics. Longobardi’s article presents a preliminary suggestion as to how some of the crosslinguistic variation which is observable in the syntax/semantics mapping of nominal arguments, such as the well-known N-to-D raising differences between Italian and English, may be accounted for in terms of a parameter based on the presence of a Person head in the traditional D position of DPs, and whether such a Person head is strong, as in Italian, or weak, as in English. While Longobardi points out that his contribution revolves around preliminary observations and tentative conclusions, the postulation of a Person head certainly offers a highly promising solution to some of the outstanding problems. Judy Bernstein develops the idea that English words like the, they, this and existential there share an initial th- morpheme which is identified as a

 Henrik Høeg Müller and Alex Klinge

3rd person marker unspecified for number and gender. Contrary to a strong tradition, she argues that only the person feature, not the definiteness feature, is associated with the D heading of the functional DP projection. Klinge takes a wider perspective and discusses the pan-Germanic þ- root, with the three modern variants /θ/, /d/, /ð/, as in Icelandic þessi, German der and English the, as a morpheme contrasting in terms of fundamental ostensive processing instructions with a pan-Germanic hw- root, with the modern variants /w/ and /v/, as in Icelandic hvað, German was and English what. In their article, Fukui & Zushi point out that while much discussion of DPs since Abney has concentrated on the similarities between DPs and clauses, a number of substantial differences have been glossed over. They seek to pin down the differences and go on to formulate an analysis according to which the differences arise because nominals have a single-layered internal structure as against a dual-layered internal structure in clauses. The analysis points to a way ahead in the development of the original DP analysis proposed by Abney. Dam investigates semantic processing of three types of nominalization, viz. deverbal nouns, nominalized infinitive phrases and nominalized complementizer phrases. She argues that in these nominalizations the presence of a definite article instructs the hearer to construct an entity, whereas the nominalized complement of the article semantically denotes a situation. Dam shows how the semantic merger of entity and situation features takes place. Georges Kleibert and Thorstein Fretheim & Aba Appiah Amfo consider the semantic and pragmatic interplay and contrast between articles, demonstratives and possessive determiners. Kleibert is specifically interested in the difference between the definite article and the possessive determiner across discourse contexts where an associative anaphoric relation is established. He argues that the two items are in competition in associative discourse contexts but one central difference is that the possessive determiner introduces two entities, viz. the entity anaphorically picked out by the determiner and the determined entity. The introduction by the possessive determiner of two related entities results in individuated interpretation relative to the associative antecedent whereas the definite article has to rely on a prototypical relation with the associative antecedent. Fretheim & Aba Appiah Amfo take their point of departure in a critique of the scalar implication that a definite determiner such as the can always substitute for demonstratives such as this and that which follows from Gundel et al’s Givenness Hierarchy. The Givenness Hierarchy model predicts that the meaning of the definite article den N (that N) should be different from that of the pronoun den (that). Fretheim & Aba Appiah Amfo show that, on the contrary, in Norwegian the semantics encoded in the definite article and the pronoun is the same. In contrast to Norwegian, the NigerCongo language Akan has segmentally identical but semantically distinct definite determiners and pronouns. However, in spoken Norwegian, they show that the formal identity between the preposed definite article den and the distal demonstrative den are two distinct linguistic phenomena which cannot be conflated.



Introduction 

This volume offers many approaches to nominal determination and the correlations between internal nominal architecture and semantics, and it proposes a range of new analyses of well-known problems, but it also raises new questions and points to new areas which may prove interesting topics for future research both in functional and formal paradigms.

References Abbott, B. 2004. Definiteness and indefiniteness. In The Handbook of Pragmatics, L. R. Horn & G. Ward (eds). Oxford: Blackwell. Abney, S. P. 1987. The English Noun Phrase in its Sentential Aspect. PhD dissertation, MIT. Abraham, W. 1997. The interdependence of case, aspect and referentiality in the history of German: The case of the verbal genitive. In Parameters of Morphosyntactic Change, A. van Kemenade & N. Vincent (eds), 29–61. Cambridge: CUP. Ariel, M. 1990. Accessing Noun-phrase Antecedents. London: Routledge. Auwera, J. van der (ed.). 1980. Semantics of Determiners. London: Routledge. Bernstein, J. 2001. The DP hypothesis: Identifying clausal properties in the nominal domain. In The Handbook of Contemporary Syntactic Theory, M. Baltin & C. Collins (eds), 536–561. Oxford: Blackwell. Bosque, I. (ed.). 1996. El Sustantivo sin determinación. La ausencia de determinante en la lengua española. Madrid: Visor Libros. Carlson, G. 1977. A unified analysis of the English bare plural.  Linguistics and Philosophy 413–457. Chierchia, G. 1998. Reference to kinds across languages. Natural Language Semantics 6(4): 339–405. Christophersen, P. 1939. The Articles. A Study of their Theory and Use in English. Copenhagen: Munksgaard. Gundel, J. K., Hedberg, N. & Zacharski, R. 1993. Cognitive status and the form of referring expressions in discourse. Language 69: 274–307. Hawkins, J.A. 1978. Definiteness and Indefiniteness. A Study in Reference and Grammaticality Prediction. London: Croom Helm. Heim, I. 1988. The Semantics of Definite and Indefinite Noun Phrases. New York NY: Garland. Hodler, W. 1954. Grundzüge einer germanischen Artikellehre. Heidelberg: Carl Winter Universitätsverlag. Jespersen, O. 1942. A Modern English Grammar on Historical Principles. Part VI. Morphology. London & Copenhagen: Allen and Unwin & Munksgaard. Kamp, H. & Reyle, U. 1993. From Discourse to Logic. Dordrecht: Kluwer. Korzen, I. 1996. L’articolo italiano fra concetto ed entità, I-II [Etudes Romanes 36]. Kø­ben­havn: Museum Tuscula­num Press. Leiss, E. 2007. Covert patterns of definiteness/indefiniteness and aspectuality in Old Icelandic, Gothic and Old High German. In Nominal Determination. Typology, Context Constraints and Historical Emergence [Studies in Language Companion Series 89], E. Stark, E. Leiss & W. Abraham (eds), Amsterdam: John Benjamins.

 Henrik Høeg Müller and Alex Klinge Longobardi, G. 1994. Reference and proper names: A theory of N-movement in syntax and logical form. Linguistic Inquiry 25: 609–665. Lyons, C. 1999. Definiteness. Cambridge: CUP. Olsen, S. 1989. AGR(eement) in the German Noun Phrase. In Syntactic Phrase Structure Phenomena, C. Bhatt, E. Löbel & C. Schmidt (eds), 39–49. Amsterdam: John Benjamins. Prince, E. F. 1979. On the given/new distinction. In Papers from the 15th Regional Meeting of the Chicago Linguistic Society, P.R. Clyne, W.F. Hanks & C.L. Hofbauer (eds), 267–278. Chicago IL: Chicago Linguistic Society. Radford, A. 2004. Minimalist Syntax. Exploring the Structure of English. Cambridge: CUP. Renzi, L. 1976. Grammatica e storia dell’articolo italiano. Studi di Grammatica Italiana 5: 5–42. Stark, E., Leiss, E. & Abraham, W. (eds). 2007. Nominal Determination. Typology, Context Constraints and Historical Emergence [Studies in Language Companion Series 89]. Amsterdam: John Benjamins. Vangsnes, Ø. 1999. The Identification of Functional Architecture. PhD dissertation, University of Bergen. Vogeleer, S. & Tasmowski, L. (eds) 2006. Non-definiteness and Plurality [Linguistik Aktuell/Linguistics Today 95]. Amsterdam: John Benjamins. Wilmet, M. 1986. La determination nominale. Paris: Presses Universitaires de France. Wood, J. L. 2003. Definiteness and Number: Determiner Phrase and Number Phrase in the History of English. PhD dissertation, Arizona State University. Zamparelli, R. 2000. Layers in the Determiner Phrase. New York NY: Garland Publishing.

Determiners and definiteness Functional semantics and structural differentiation Peter Harder

University of Copenhagen Both functional and formal approaches frequently suggest that structural and semantic categories ultimately match up (possibly even universally). They arrive at this result, however, via different descriptive strategies: most functionbased approaches set up structures primarily based on semantic/functional description, subsequently looking for distributional confirmation. Formal descriptions, on the other hand, primarily focus on distributional patterns, but often aim to show that these correspond to semantic distinctions. In contrast, I try to show that the determiner category comprises heterogeneous elements whose shared function must be understood as a result of a function-based structural pattern imposed top-down (partially arbitrary, partially motivated), which carves out a specific slot in the complex noun phrase for the basic ‘grounding’ choice between definite and indefinite reference.

1. Introduction This article gives an account of determiners from the point of view of a function-based approach to linguistic structure, whose essential features are (1) that function is ontologically prior to structure in language: unless language functioned as part of human interaction, the question of its structural nuts and bolts would not arise; (2) however, structure should not be expected to match external functional needs point by point – any more than (e.g.) the structure of an education is always ideally conducive to the goal of learning. Linguistic structure is basically a way to make a complex tool serve complex ends, and, as in all human practices, shortcuts, compromises and inconsistencies are to be expected. The aim is to show how this approach can highlight phenomena that are less evident from either a purely functional or a purely structural point of view. From both these unilateral approaches, there is a logic leading to an assumption of isomorphism between structural and semantic categories. In contrast, I take my point of departure in the basic functional purpose that I see as relevant in relation to determiners, i.e.



Peter Harder

‘situational grounding’, without suggesting a direct mapping. Simplex and complex noun phrase structures constitute different ways of achieving the function of reference, and the determiner category in a complex noun phrase can only be understood in terms of an interfacing operation between semantic function and syntactic structuring, rather than as a case of the two sides matching point by point. Since my own position belongs in the functional tradition, I would like to apologize in advance for crudities in my rendering of points made in the formal generative tradition. There are numerous parallels between observations made based on formal and functionalist approaches, cf. Siewierska (1992); nevertheless, cross-references are rather scarce – not only because of conflicting beliefs but also because it is all but impossible to keep abreast of developments in both camps. I hope it may serve a purpose to try to relate the point I pursue also to formal accounts, since it runs counter to parallel and influential strands on both sides of the divide. The only assumptions about the formal generative tradition that are crucial to my account are (1) that there is a descriptive strategy in the formal tradition which aims to set up an ‘underlying’ or ‘abstract’ hierarchical syntactic structure to account for syntactic relations, and (2) that it is significant to what extent such a syntactic structure can be shown to match up with semantic facts. This assumption is based on passages such as the following in the literature on determiners:

(1) This article argues that semantically equivalent noun phrases have the same underlying phrase structure across languages, no matter whether articles exist or not (Vangsnes 2001: 249)



(2) I want to take advantage of results in the syntactic study of the internal structure of noun phrases, building in particular on work by Abney (1987), Cinque (1990), Cinque (1992) and Longobardi (1994). These works suggest that the structure of noun phrases is much more complex than the traditional picture of a single projection headed by N, with a determiner in its specifier [...] If a structure of this kind is necessary on syntactic grounds, the research project that this dissertation begins to carry out is that of putting syntactic complexity at the service of meaning (Zamparelli 2000: 2–3).

In relation to the functional tradition, my point of departure includes both the American tradition of Givon, Bybee, Hopper and Thompson and European Functional Grammar in the tradition of Simon Dik (e.g., 1989), whose central research figure in the area of noun phrase structure is Jan Rijkhoff (cf. Rijkhoff 1990, 2002). With respect to the Dik tradition, there is only one crucial assumption that my comparison depends on (explicitly confirmed by Rijkhoff, personal communication 2004), namely that in terms of the descriptive logic of his account, the category of determiners has no obvious place. This is basically due to the fact that definite and indefinite determiners belong in different ‘layers’, cf. section 5 below: there is no inherent functional reason why a



Determiners and definiteness

quantity-denoting item like one and a location-denoting item like that should belong in the same category. The extent of the parallels between morphosyntactic expression and semantic content depends on one’s descriptive strategy with respect to postulating non-overt elements. My basic assumption is that relations between elements in the clause are essentially semantic, so that we must understand syntactic patterns as reflecting patterns of collaboration and differentiation between meaning-bearing elements. This means that there is only one situation in which zero elements should be assumed to exist, namely when there is a closed paradigm such that even in the absence of a filler, meaning is conveyed (i.e., ‘no filler’ entails a particular interpretation, instead of leaving all options open, cf. Bybee 1994: 252, McGregor 2004). Basically, expressions serve the function of encoding whatever linguistic content goes into the message, and when ‘zero’ can occasionally serve that function, it should of course be recognized as an expressive option. In other cases, what you see is what you get – and this entails a much less abstract account of syntactic relations than when licences to operate with zero elements are more generous. I suggest that the generalizations that can be achieved by means of a description based on more abstract elements are not inherently wrong, but belong at a level whose object of description is not the individual syntactic string, but rather a whole set of expressions (possibly the full universal set of options) that are implicitly being compared. Although the purpose is to make some claims about determiners and noun phrase structure, the reader should thus be duly notified that these claims depend on a theoretical argument that will occupy some space especially in the first half of the article. The structure of the article is roughly as follows: Section 1 describes the relations between function and structure, focusing on the relation of ‘partial autonomy’ that aims to cut through the polarized debates on autonomy and functional motivation. Section 2 presents the top-down approach to syntactic complexity and argues that the syntactic structure of a complex expression should not be understood solely based on a strategy of ‘maximal differentiation’, but primarily on the basis of how many (collaborating) coding choices it represents.. Section 3 presents two key types of semantic functions, indexical grounding and conceptual specification, arguing that the differentiation between them at the nominal level must be seen as secondary to a logically prior differentiation at the clausal level. Section 4 relates these functions to simplex and complex noun phrases and argues for the advantages of a functional-procedural account of meaning rather than a basically referential account. Section 5 brings a number of the previous discussions together in presenting a theory of precisely what it means for a language to have a determiner category as part of its structure. Section 6 describes the implications for languages without a determiner category; and section 7 takes up the comparison with the descriptive strategies of formal accounts. Section 8 contains summary and conclusions.





Peter Harder

2. The relation between function and structure The position I defend belongs in the context of Danish functional linguistics (cf. Harder 1996, Engberg-Pedersen et al. 1996, Engberg-Pedersen et al 2005), a variety of European functionalism which has adapted some of the basic tenets of European structuralism to reflect an ‘integrative’ understanding of structure as opposed to the ‘autonomous’ understanding that characterizes both Saussure and Hjelmslev and modern generative linguistics, cf. also Newmeyer (2003). A key concept in this picture is function-based structure: the idea that linguistic structure exists as an ‘order’ or a ‘system’ imposed on elements that are fundamentally functional in nature. The basic claims of structuralism survive as ‘relative’ or ‘partial’ truths about language. Thus, linguistic structure is partially autonomous from meaning or function in the sense that you cannot derive structure from function – but without the basic linguistic functions, there would be nothing that linguistic structure could be the structure of. For each functional purpose there may be several different, more or less ‘functional’ options in terms of the way the language system allows that purpose to be served – but while it is partially arbitrary what structural options exist in a particular language, this partial arbitrariness co-exists with a bedrock of motivation in that these structural options exist only as ways of executing functional tasks. The reasoning is the same as suggested originally by Aristotle (De partibus animalium, ed. by McKeon 1941: 650, quoted from Givón 1995: 2), illustrating functionality with the properties of an axe: In order to work it has to be hard, but is can be made either of bronze or iron – no ultimate, generalizable cause can be provided for the latter choice. Typological variation exemplifies this: accusative languages and ergative languages are equally capable of encoding both one- and two-place predications (conveying, for instance, that ‘Joe broke the window’ and ‘John left’), but they impose different structural patterns on the situation. The accusative and the ergative structures exist as ways of systematizing a particular coding problem, and because structure can only be understood in relation to the function that it structures, its autonomy is only partial. A particular illustration of the implications of partial autonomy is the understanding of syntactic complexity that follows from it. Partial autonomy differs both from formal, potentially innate syntax in the generative tradition and from the purely ‘emergent’ view of grammar (cf. Hopper 1987: 141) where structural patterns derive directly from usage. Relations between clause elements should be understood in analogy with relations between elements in other types of functional wholes such as business companies. The utterance as a whole has a function to serve, and in order to succeed in doing that, each element must contribute in an appropriate manner towards the whole – just as a business only succeeds if those who constitute the company each do their job. Jobs are partially autonomous in relation to the individuals as such: in a given company, you cannot derive the job description from the employee herself (in the absence of a uniform), just as you cannot see from a particular noun phrase (in the absence of case marking) whether it serves as subject. From the founder’s perspective, it



Determiners and definiteness

is even clearer: a crowd of employees would not in themselves constitute a business company merely by being brought together – they need to assume the ‘mantle’ of structured collaboration over and above whoever they might individually be, and job descriptions can only be understood in the context of an overriding purpose (a business concept) which is being served. Conversely, the idea of the structure of a business company existing autonomously from any idea of economic life is just as incoherent as the idea of language structure existing independently of the existence of language use. Syntactic structure, in other words, is a particular kind of codified division of labour in complex utterances, assigning each expression a particular job (=function, meaning) in a larger, more complex whole. Instead of underlying structure with its disputed empirical status, we get a ‘content structure’ whose job is to describe the way meanings collaborate in the sentence. This structure resembles underlying structure in that it does not have to mirror the expression side point by point (demonstratives are definite but have no separate expression indicating definiteness). But unlike underlying abstract syntax, content syntax is accessible to empirical inspection in the same way that expressions are, by virtue of the ability of native speakers to recognize certain types of expressions as conveying certain types of meaning rather than others. Nonisomorphism between syntax on the expression side and content side can be exemplified in the area of determiners in Danish, cf. (3): (3) a. b.

Content: Expression: Content: Expression

def (red (house)) det røde hus def (house) hus-et

(3a) represents the form used in Danish with adjectival premodification (where higher scope on the content side is mirrored by linear precedence on the expression side). (3b) departs from this mirroring pattern because a solitary definite article is expressed by suffixation. There exists no single overall mirroring pattern between (expression-) syntactic and semantic (= content-syntactic) relations. (Cf. Longobardi 2001: 587f for a discussion of the relation between various types of postposed articles and the D position in abstract syntax). It follows from what I have said above that in emphasizing this non-isomorphism I am also concerned to establish a difference with the parts of the functional tradition. Bybee, Perkins & Pagliuca (1994: 1) see language-internal systems as purely epiphenomenal, seeing linguistic categories as “universal phenomena with language-specific manifestations”; and Langacker (2004) sums up the field of nominal structure and grounding without invoking distributional classes such as determiners, basing his analysis on a semantic analysis of the relevant items (or “fillers”, cf. below), including zero. In contrast, it is a central claim of the analysis below that in analysing a (category of a) language, it is necessary to specify exactly in what way it structures the semantic task(s) that it enables the speaker to perform, and that we do not get the whole picture by looking simply at the semantic content of the relevant items in a universal perspective.





Peter Harder

3. Syntactic complexity in a top-down perspective: Minimal vs. maximal differentiation Syntax can be approached either bottom-up (from word to utterance) or top-down (from utterance to word). Although the two approaches can be designed so as to match up automatically, the top-down approach based on functional differentiation, cf. above, can be used to raise some questions that a bottom-up account will tend to sidestep. The bottom-up, compositional perspective must take for granted the existence of an inventory of lexical ‘items’ with properties that are already defined before the syntactic process can begin: without items, there is nothing to combine. Some of these properties are purely ‘semantic’ in the sense that they indicate aspects of the meaning potential of the item; in English, for instance ‘male’ in the case of the noun bull. Other properties are ‘syntactic’ in the sense that they specify the combinatorial potential of the form; that bull is a ‘noun’ implies, among other things, that the word in English can take the head position after premodifying adjectives (leaving aside the semantic implications of nounhood). Such ‘syntactic’ properties foreshadow the existence of certain types of larger syntactic constituents – which is why lexical properties have occupied an increasingly dominant role in generative syntactic description. From the top-down perspective the cards are not automatically stacked in the same way. In itself, the message content corresponding in English to I do not know provides no royal road to advance predictions of what lexical properties the ultimate constituents will need to have. As pointed out by Hjelmslev (1943/1966: 47), different languages will codify it with a greatly varying number of constituents with greatly varying lexical-distributional properties. Few languages will have any equivalent of English do; in Finnish the negative element patterns like an auxiliary verb, while in written French it will split into two items ne and pas; spoken English dunno could be regarded as an unanalysable holophrase. That it is the top-down perspective which allows these perspectives is due to the fact that it is more basic than the bottom-up approach: it mirrors the evolutionary scenario (individual words arrived on the scene after holophrastic signals, cf. Deacon 1997). This is also why it does not presuppose an already existing, codified division of labour between different types of items (such as nouns, etc) – these only emerge as a result of the process of differentiation. One of the advantages of taking this perspective is that we do not have to assume ‘maximal differentiation’ as the natural format of description. If we take the bottom-up perspective, on the other hand, beginning with the inventory of items, we have to devise a theory that will afford a precise place for the actual items with their distributional properties. In doing so, the obvious goal is to provide a descriptive format that captures the full potential of syntactic differentiation in the language – just as the factory’s description of a Swiss army knife has to include the full panoply of implements in order to be adequate. An item-based language description will be adequate if it can offer a place to all occurrences of these items – but such a description is not guaranteed to be optimal for the description of any particular whole utterance. We can describe



Determiners and definiteness

the utterance stop! with the full positional template in Diderichsen’s topological description of Danish (cf. Diderichsen 1946), or a fully extended generative clause tree, but we will miss the sense in which imperatives are structurally similar to interjections like out or shit. As will be obvious, there is an alternative perspective, that of describing only how much structural differentiation is actually used. The latter perspective is likely to offer a better angle on the relationship between function and structure. With the analogy of a Swiss army knife, it would not be very functional if a Swiss army knife always had to be used with everything hanging out: as pointed out, e.g., by Mackenzie (1998), holophrases are optimal from a processing point of view. From that point of view, speakers are likely to follow a Zipf-style maxim: ‘do not use more structure than strictly necessary!’ The same point applies even more clearly for structures set up so as to be crosslinguistically valid. A positional template covering many languages is useful for showing the differences between them – but for any single language, the template is overcomplex as a description of positional relations inside the language (cf. Dik & Kahrel (1992), where a positional template is set up that can offer slots for a range of different languages, but no language differentiates between all of the slots available). Part of an adequate structural description of linguistic utterances in a given language is a description of exactly how much structure is used in order to encode them. In the case of determiners, a functional-structural account must therefore describe where and how they come in, viewed in the perspective of the downward differentiation of whole utterance meaning in specific languages, including how the kind of differentiation that they reflect compares with less complex coding patterns. Like all properties of language systems, this top-down/bottom-up duality reflects in abstract form some properties of the basic situation of the speaker engaged in actual language use. Two sets of processes are presupposed when the human speaker encodes a complex message: first, a differentiation process whereby aspects of a whole intended message become factored out into encodable fragments available to the speaker. In that process, each differentiated item presupposes the existence of the larger, whole utterance meaning to which they are designed to contribute. Secondly, the whole utterance presupposes a compositional process in which each element is inserted in a freely compiled, meaningful semantic whole. In that process, the creation of the larger whole presupposes an inventory of differentiated items that need to be brought into a relevantly meaningful relationship. The same double perspective on task management crops up in other situations where a problem cannot be solved in one fell swoop. A marine archaeologist who wants to recover a sunken ship and cannot extract it as a whole must first take it to pieces before she can put it together to display it as a whole.





Peter Harder

4. Indexical grounding and conceptual content: Two levels of functional differentiation Above, following Deacon 1997, I mentioned that the evolutionary background for the rise of human languages was the existence of communicative systems based on holophrastic signals. Characteristic of these systems is that all such calls are inherently indexically ‘grounded’ in the immediate context: it does not make sense to ask a lion that is roaring, ‘do you mean now?’ Although this is a primitive type of signal, having them clearly makes functional sense, and human societies also make use of them, cf. the siren of an ambulance. Purely symbolic-conceptual meaning (such as is associated with the noun ambulance as opposed to the siren) is by definition independent of direct situational links and hence cannot have their immediacy. For the same reason it may be argued that there is an inherent link between the rise of conceptual sign meaning and syntax (cf. Harder 1996, 1999a; Deacon 2003): if purely symbolic-conceptual meaning is to make functional sense, it has to be possible to use accompanying signals that provide the necessary indexical grounding in the context of communication. Thus conceived, grounding can be seen as a sub-function which arises when a full utterance meaning becomes differentiated into subparts, one of which is symbolicconceptual. From this top-down perspective, grounding is basically an utterance-level phenomenon: it is the symbolic-conceptual content of the whole utterance that needs to be grounded. But differentiation of utterance meaning does not stop here. The existence of the clause format illustrates that the semantic division of labour can go much further. In order to understand the precise role of the determiner category, two syntactic levels are necessary: clausal grounding and nominal grounding. There are links between them: in the standard subject-predicate format, the grounding of the subject plays an important role in grounding the clausal message – but it is syntactically important to understand that there are two different levels involved. The clearest exemplification of other types of grounding than the subject-predicate pattern is the existence of so-called ‘thetic’ judgements (cf. Kuroda 1972) as opposed to the standard ‘categorical’ type that correspond to the subject-predicate format. Prototype thetic judgements are presentational statements like there is nothing to be afraid of, where no entity is singled out as being the subject/topic of conversation. In such clauses, grounding operates only at the clausal level (signalled by there and present tense). This is a more elementary form of statement than the categorical type. The conceptual differentiation that leads from one to the other has been analysed by Strawson (1959:  202), who describes the simplest type of non-categorical clauses as ‘featureplacing’ sentences: a feature specification + an indexical expression anchoring it in the situational ground (there is food here; it is cold). Strawson (1959: 212) points out that this basic form of statement provides an analysis of what a subject term involves, semantically and metaphysically speaking: a subject term contains within it the idea of a fact of the feature-placing kind, since a particular object can be reduced to a location for a property. A tree is a manifestation of treehood in a particular place, as it were. As



Determiners and definiteness

Strawson makes clear, the (superimposed) idea of a particular entity adds extra properties over and above the feature-placing content, which are very convenient for practical purposes. Since we conceive of a particular entity as something that can be reidentified at a later time, we can impose an extra layer of order and stability on the ontological space we occupy by populating it with particulars: instead of merely a shimmering process of feature manifestations we get a universe that includes familiar objects to which we can relate and adapt. Coding something as a nominal expression means attributing entity status to it, as opposed to having just features and locations. The result of this extra level of complication is that we have two points at which grounding can occur: we can ground something at the nominal level, as an entity description (food) by appending a specialized grounding expression to it (that food) as well as at the whole utterance meaning (there is food); and a grounded entity description can then in turn be part of a grounded utterance meaning (that food is delicious). The priority of the top-down perspective is reflected in the fact that it is only at the clausal level that grounding is obligatory (unattached symbolic content does not constitute a message), while separate grounding at the clausal level is an additional subtlety. Typologically, this is also manifested in the fact that not all languages have this two-level grounding pattern (at least in the European form); the Salishan languages are the most well-known case where it is not clear that nouns are distinct from other predicates (cf. Fortescue 2004). In these languages, the non-distinctness of nominal predicates goes with a clause structure in which top-down, utterance-level patterns, imposing a pragmatically motivated structural pattern on the clause, cut across all lower-level distinctions.

5. The structure of nominal expressions With the differentiation that brings about a nominal level of analysis with separate expressions for grounding, the stage is set for a structural analysis of noun phrases based on these functional foundations. In relation to most of the literature on nominal meaning, however, there is a difference when it comes to what exactly nominal meaning is supposed to be. Much of the discussion of the relation between syntax and semantics, especially in the nominal domain, has been conducted under the assumption that ‘semantics’ was by definition a matter of logical, referential-semantic properties. In a functional approach, however, continuing the business analogy, meanings are like job descriptions: they specify what language users should do in order to construct the right utterance meanings (cf. Davies and Isard 1972 for a computation-based fleshing out of this idea). Applied to nominal meaning, this idea entails that referential properties are not semantically basic, but rather a consequence of what happens as a result of functional meanings ‘doing their jobs’, i.e. when they work successfully in context. The complex noun phrases that in the philosophical literature are known as definite descriptions





Peter Harder

illustrate the division of labour described above, between two sets of linguistic instructions, one associated with the conceptual construct and one associated with the way it should be related to the world of discourse. The properties of the referent arise only as the end product of carrying out the relevant procedures. Thus existence-and-uniqueness (the classical Russellian entailments of definite descriptions, cf. Russell 1905 and also Hawkins 1978) arise only as prerequisites for carrying out the procedure that is encoded by markers of definiteness: the ‘grounding’ instruction to perform an act of identification between the nominal expression and the relevant available referent. Reference, therefore, is the pragmatic end point of the process, an aspect of the situational pragmatics of interpretation, not the encoded meaning. In a rough-andready paraphrase, a key example of the literature on reference can be glossed as follows (cf. Harder 1976, 1992): (4) Encoded meaning of The present King of France: ‘to-be-identified (reigning now (King of France))’. If coded meaning is viewed as instructional, the expression is semantically unproblematic; there is no problem in understanding the full encoded complex instruction – but the attempt to follow it up ’pragmatically’ in order to achieve actual reference, as instructed by the code, sends the hearer on a wild goose chase. It is not the language that is deceptive, it is the speaker. To take another example, the fact that reference is not criterial for encoded meaning means that ‘genericity’ is typically a matter of the nature of the intended situational referent rather than of the encoded instructions. Thus the statements the lion/a lion is a mammal, lions are mammals are semantically different, but in all cases the interpretation process can up with a judgement applying to the whole ‘genus’. Conversely, every evening at six o’clock a heron flies over the chalet (cf. Lyons 1977: 188) has the same semantically encoded referential instructions associated with a heron whether it is interpreted as referring to a specific heron or not. In describing the semantics of nominal expressions based on definite descriptions, however, we are getting ahead of ourselves in terms of the top-down approach. From a differentiation perspective, the logical first step is to look for ways in which reference to an object in the world of discourse can be accomplished without further syntactic sub-differentiation. Two simplex coding strategies for reference offer themselves: (5) a. The situational-deictic strategy, reminiscent of pointing: that; you! b. The naming strategy, sticking a coding label directly on an individual (‘I christen thee Jane Smith’) There is an interesting relationship between these communicative-functional strategies and logical properties of expressions. The social act of sticking a label on someone leads to the logical property of rigid designation, because if a name ‘sticks’, it goes with the named entity regardless of where she is. Similarly, the situational-deictic strategy, identifying the referent via its role in the interaction, is associated with the category of



Determiners and definiteness

personal pronouns (cf. Bhat 2004), and leads to variability of denotation, viewed from the point of view of the entity. The label Peter Harder goes around with me – while the situational-deictic role (you) goes with the utterance event, getting redefined for each new utterance. But both types of nominal expression have constant semantic functions associated with them: (1) you identifies X as being addressed (2) Jane Smith identifies X as the bearer of that name. The structure of complex nominal expressions must be understood in that perspective. The rise of common nouns can be understood as a differentiation from proper names – a path that is familiar in the ontogenetic perspective. In the beginning, Mummy, with a capital M, is a rigid designator (the one and only), and only later analogy leads to flexible designation (my mummy and your mummy). The development can be understood as a special case of the rise of context-independent symbolic-conceptual meanings, and accordingly it shares the basic constraint discussed in section 4, that in order to be used for identifying reference it becomes dependent on a grounding mechanism. In terms of structural differentiation, this is the functional rationale for the differentiation of nominal expressions into a determiner slot (D for short) and a ‘nouny’ slot (N for short). D is the functional-structural slot for the grounding operation and N for the symbolic-conceptual specification. Both have a function in relation to the purpose of identification, but via different functional strategies: While D is the home of subtypes of situational-deictic pointing type operations, N provides generalizable descriptions. Either slot may provide the criterial difference: THIS tree, not THAT tree; the BIRD, not the BEARD. But the special job of establishing identifying reference so as to ground the whole statement in a situationally available referent is linked specifically with the D slot. This is suggested by several facts. First of all, the D slot is the locus of those expressions which signal that the whole expression is to bring about identifying reference (e.g. by means of the rather than a/some). The content of N may enable but cannot signal definite reference in itself. A description that fits only one referent but is preceded by the indefinite article, such as a man who found the North Pole first, remains an indefinite description, not an identifying reference to Robert Peary. The same thing is reflected in the fact that certain expressions can either stand alone or be accompanied by a N expression, depending on whether a conceptual specification is necessary:

(6) Look at that (car)!

The whole expression can achieve its purpose without the N expression but not without the D expression. Hence, an endocentric interpretation of that tree must have the D expression that as its head (cf. Harder 1976). In the paradigm case of identifying reference, the structural division of labour in complex nominal expressions is therefore functionally clear-cut.





Peter Harder

6. Structural patterns and semantic relations: What does it mean for a language to have a determiner category? Although in the introduction it was stressed that all ways of structuring the content side are necessarily arbitrary in the sense that there are always many different ways of ‘cutting the pie’, it may appear as if the argument in sections 3–4 in fact suggested that there was a well-defined functional-semantic rationale for English nominal structure, sufficient to justify all structural slots. The use of only definite determiners as examples has underpinned this somewhat simplistic picture. However, in the case of the determiner slot, it now needs to be stressed that a significant fact about it is that determiners do not have a shared, overriding positive semantic content. Deriving the category ‘determiner’ directly from positive functional content is impossible. A natural conclusion would then be to say that determiners must be a purely formal aspect of linguistic structure. Syntactic structure, working according to its own inherent non-functional logic, requires an item to make the structure complete, the precise content of which is beside the point. According to the rationale of the account I am suggesting, however, that would be equally wrong. But how can one justify a midway point between these two clear-cut options? The centrepiece of the argument is that a function-based (and usage-based) view of structure can only be possible if we allow structure to impose a measure of order and systematicity on language that goes beyond the totality of functional motivations impinging upon the linguistic code. A ‘structure’ that always transparently reflected all online, emergent functional pressures would not really correspond to its name. With the business analogy, an employee with a given function may be recruited to serve other functions, but if there is no ‘job description’ over and above what each person does from moment to moment, the company really does not have any corporate structure (extreme ‘adhocracy’ in the terms of Mintzberg, 1979). From the point of view argued here, this would not be a functionally ideal way of organizing a functional system. The key justification for the existence of structure in functional systems is the need to impose a manageable order on the way functions are served. Although it is optimal from the individual’s point of view if there is always a bus exactly when you need to take one, this would not be an optimal demand to live up to for whoever had to organize the bus service. This logic, I suggest, applies also to language. In order to encode complex messages, speakers and listeners need an overall level of simplicity that overrides pressures to include a large number of individual functional considerations. Constraints on constituent order (cf. Hawkins 1994) are one example. Restrictions on the variability of morphological coding patterns, including the complexity of layering, are another example. The existence of obligatory constituents (cf. Harder 1999b) must be understood as a reflection of this: If there is a pervasive and well-entrenched coding pattern, it may be functionally optimal to use it even in cases where its functional motivation is weak or absent, simply in order to be able to stick to the same pattern in as many cases as



Determiners and definiteness

possible. As a child learns to lock her bicycle as a safety measure, she may gradually find it easier to just do it always, rather than speculate about whether it is strictly necessary in each particular case. Similarly, in a well-functioning transport system, you will sometimes see an empty bus. I suggest that the rise of a well-defined determiner slot that incorporates a variety of semantically different types of expressions must be understood as the result of a functional bias towards a certain degree of uniformity. In terms of Hawkins (1994), such a paradigm at the left edge of a constituent can assist the function of efficient recognition of nominal constituents (regardless of content), thus easing overall processing. But the co-presence of a variety of meanings in the same slot is also fully compatible with something that is both structural in the classical sense of European, Saussurean structuralism, and functional in the sense of providing a semantically motivated slot. In terms of this picture, the determiner slot is functionally motivated not in terms of positive content, but in terms of the paradigmatic choice associated with it – the choice between identifying and non-identifying reference. From the American, predominantly syntagmatic perspective, this is not an obvious kind of functional motivation, since syntagmatically speaking, motivation emerges from the positive contribution of each semantic element. But if processing of entity-denoting expressions depends crucially on their status in terms of how they relate to the existing inventory of relevant items (cf., e.g., Givón 2001: 476 for a grid of choices), it may be practical enough so that languages sometimes reserve a slot for elements relevant to this kind of decision rather than for elements with the same positive semantic content. The development whereby such a slot arises can be characterized by the standard features of grammaticalization (on the development of the article system in French, cf. Schøsler 2001: 96). In the stage before a determiner system arises, we have a situation that can be exemplified by Latin. In it, elements signalling positive grounding can stand alone as simplex nominal expressions, or be combined with a noun providing a conceptual specification; finally, they can also be left out when the context makes them unnecessary, and a noun on its own can denote a situationally grounded entity: HOC est corpus filii (‘This is the body of the son’, simplex); In HOC SIGNO vinces (‘In this sign you will conquer’, combination of indexical and conceptual information); SIGNUM dare ‘(‘to give Ø/a/the sign’, bare conceptual noun, may be inferentially grounded, cf. section 6). This system can then change via a route in which a demonstrative, in this case ille, gradually becomes more and more frequent as a marker of definite reference. The crucial development is when the absence of any element indicating definite reference comes to indicate that definite reference is not a possible interpretation (the logical change from ‘no marking of X’ to ‘marking of no X’). This situation can be described as a case where the slot becomes partially autonomous of the filler, and an unfilled slot thus acquires a meaning of its own. In the system that arises, the functional role of a determiner is dependent on its relationship with a nominal head – because it represents a differentiation of the unified ‘nominal referential function’ into two designated sub-slots: one for the choice of ‘grounding’ (definite vs. indefinite) and one for symbolic-conceptual specification.





Peter Harder

Structurally speaking, determiners are therefore not identical to ‘free’ pronouns, except if you decide to describe free pronouns in terms of ‘maximal differentiation’ – in which case they are clearly more analogous to determiners than to nominal heads. But ‘maximal differentiation’, as we have seen, is only one possible approach – the other is to view structure top-down and postulate no more differentiation (hence, no more structure) than strictly necessary. From that point of view, free pronouns are not ‘determiner phrases’, but simplex, undifferentiated expressions, making do with one of the two basic referential strategies discussed in section 4 above. The duality of ‘slot’ and ‘filler’ which is crucial to the understanding of functional differentiation means that fillers as such need not be definable in terms of category membership – as long as it can be seen when they fill a particular slot. A case in point is the numeral one, which can occupy both the D slot and a premodifying slot (although not at the same time). In Danish the corresponding numeral is marked by a ‘definiteness’ suffix when it occurs after the definite article: (7) Alt afhang af én ting Everything depended on one thing (8) Alt afhang af den en-e ting Everything depended on that ’one-definite’ thing In terms of the division of labour described above, the element ‘one’ can either occupy the D slot, in which case it signals ‘indefiniteness’ by default (it stands in lieu of the indefinite article, as it were, since singular count nouns cannot take zero) – or it can take the post-determiner slot where it signals a property of the to-be-identified referent. In languages like English with a determiner paradigm, the absence of definite determiner means that the NP is indefinite rather than ‘unspecified for definiteness’. This contrasts with for instance ‘cardinality’ where the absence of a number in front of a plural noun means that any figure is possible. The D slot can thus be specifically devoted to the paradigmatic choice between definiteness and indefiniteness, even while fillers of the slot may have a variety of other semantic contributions (typically quantity, but also possession in languages where possessives are also determiners). ’Articles’ as a special class of items arise when some fillers become specialized for the D slot, especially as minimal indicators of the content that is bound up with the slot itself, i.e. (in)definiteness. The development of definite articles out of demonstratives is also a process of historical differentiation, whereby definiteness (= ‘to be identified’) emerges as a separate element, after previously existing only as part of a larger semantic whole (‘to be identified as proximal/distal’). The proximal/distal element can similarly be coded separately, by being coded via the ‘here/there’-contrast, as in varieties of English this here/that there (cf, also Longobardi 2003). In Danish, where the proximal/ distal contrast in demonstrative ‘fillers’ has been lost, this is the only way to encode it (den her/ den dér).



Determiners and definiteness

The issue of what constitutes the head is very complex and will not be taken up in detail here (but cf., Langacker 2004: 86–87). However, the special group of articles as a filler class is clearly dependent in relation to the nominal head, synchronically as well as diachronically: articles arise only as items specialized for that particular slot, and they function only as means of grounding the symbolic-conceptual meaning expressed in the N slot of a complex nominal expression. From a functional point of view, the rise of articles is analogous to the case of an organism evolving by Darwin-finch type speciation to occupy a particular environmental niche, differing from its predecessor species precisely by its adaptation to that particular niche. Whatever one favours as an overriding generalization when it comes to head status, the balance of power between the determiner and the head noun is very different in an obvious case of an endocentric demonstrative such as (3), look at that (car)! and in the case of articles, including zero.

7. Languages without determiners and the relationship between coded and inferential information An analysis that attaches great importance to semantic content elements and is sceptical about underlying forms must have an account of how to distinguish between semantic content that is coded and semantic content that is pragmatically inferred. A key role in this account is the role of semantic dependence relations (cf. Langacker 1991, Harder 1996). As a basic example, a verbal predicate is semantically dependent on argument terms, because verbal meaning is incomplete (as pointed out already by Diogenes Laertius): an event cannot take place without participants. This creates two possible situations in languages: either this dependency relation means that a participant has to be coded (the verb promote requires an object, for instance), or (conversely) the verb is all you need to encode, because the dependency relation forces a reading where necessary participants are included. For example, in languages like Chinese it is very difficult to set up obligatory participants, and hence equally difficult to operate with valency in the classical sense. Everything that is given by the situation or inferrable can basically be left out. This second type of situation has some claim to being basic: no more coding than strictly necessary. One might imagine an ideal type of language in which only new and non-inferrable information is coded. Strawson’s feature-placing statements could work without coding the place if it is situationally given, making only the feature explicit, as in Fire! or Timber!. Less basic examples are utterances without deictic tense and without encoded distinctions between interrogative and declarative sentence types, and without subjects (‘pro-drop’); in such cases more inferential work has to be done in order to bring about the canonical relations with the ground than in languages like English and Danish, where all these things typically have to be encoded. The dependency relation triggers the inferential activity via the basic functional assumption that utterances have to be meaningful, which includes being grounded in





Peter Harder

some way (cf. the discussion in section 4). This mechanism is really the same whether the information has to be coded in the language or not – but language structure is not the same in the two cases. Having obligatory subjects is a structural feature of a small exotic group of European languages, while the mechanism of verbs being understood in relation to the participants in the verbal event is universal. Another way of saying the same thing is that linguistic context is a special case of context in general: If you can find the information in the message itself, you need not look to the context to fill out the missing specifications – but in either case, the necessary information has to be provided. The normal situation is that grounding is to some extent encoded, since alternative grounding options may need to be differentiated – but the task of grounding by definition always goes beyond coding, since the code is necessarily general and the context in which you need to situate the message is specific. Information that is not explicitly coded can be inferred from context via the familiar mechanism of Gricean implicature. But the linguistic context can also provide the background against which a non-explicit piece of information can be inferred. I use the term syntagmatic implicature about the mechanism whereby item meanings become enriched in order to fill out the linguistic slot in which they are placed. The rationale that triggers the inference is the same, but there is a point in recognizing an intra-linguistic variety because it stresses the fact that pragmatic mechanisms do not remain politely outside the clause boundary, but are ubiquitous. In the collocation my pregnant cousin there is no need to look outside the linguistic expression to trigger an inference that the cousin is female. Compared to the strategy of postulating underlying forms to account for non-explicit content, this descriptive strategy works on the assumption that languages differ in how much information is contained in what is coded, leaving the rest to pragmatic work dependent on the context and the interlocutors. The question of ‘more or less coding’ thus involves a question of at what point you are left to your contextual knowledge – not of whether the code does it explicitly or implicitly. The code is almost by definition explicit in what it does for you, as long as zero elements are accounted for in terms of their dependence on precisely definable relations between structural items that are explicitly present. Applied to the analysis of languages without a determiner system, absence of obligatory definiteness marking means that the interpretation works by linking the descriptive content with the discourse universe directly, rather than via an encoded linking instruction. An example is Finnish (cf. Chesterman 1991: 144): (9) (23) UKKO oli tuvassa. ‘An OLD MAN was in the cottage’ […] if this sentence is taken out of context, the definiteness of the subject is in fact ambiguous. A previous mention of an old man, or a situationally known one, would make the noun definite. The translation given of (23) is simply the first that comes to mind (…)



Determiners and definiteness 

In other words, speakers of Finnish must have an interpretation strategy associated with ‘bare NPs’ that is attuned to previous mention or situational knownness, and thus differs from the strategies of English speakers. Such strategies in Finnish have to do more work than the analogous contextualization strategies for English speakers, because in English you get more help from the code (on this point). In this section, I have discussed the inferential accomplishment of the grounding function. It would also be possible to discuss inferential accomplishment of the categorizing function associated with the head noun, but that would take us beyond the scope of this paper. However, in discussing the status of free-standing pronouns in relation to the determiner category, the same logic that was outlined above applies: if elements are not coded overtly, it is necessary to be precise about the way in which context and coding collaborate in making them available to the receiver. Some freestanding pronouns such as any can only be understood in relation to a category specification that must be either coded or directly retrievable from context, e.g., if the speaker is pointing to a bowl full of tangerines and asks do you want any? Other pronouns (that was mentioned as an example) do not similarly depend on an implicit category specification. No uniform structural pattern can capture such differences. Universals of language, in this picture, are mostly universals of relevant functions: ‘reference to objects’, ‘encoding of symbolic-conceptual meaning’, ‘expressing who does what to whom’ are universal, but neither as formal or functional properties of the code itself: None of the linguistically encoded ways of achieving those functions are universal (cf. also Croft 2001). Only if pragmatic properties are smuggled into the code via criteria that are actually pragmatic rather than code-based (e.g. if the category of ‘nouns’ were to be defined as words primarily used to refer to objects) will this appear to be a universal of the code itself, rather than of the code-in-communicative-use.

8. Some tentative remarks about the relationship between functional semantics and formal generative linguistics In the introduction, I placed my own account as an alternative to both extant functional and formal theories. Above in various places I have tried to argue that accounts based on abstract syntactic categories provide descriptions that are more complex than the structure of the expressions that are being described. I have also suggested an alternative rationale, under which such abstractions may be useful. Now that I have presented my own story, I would like to briefly round off that discussion and specify where the main differences are. Originally, cf. Chomsky (1957), the difference between generative linguistics and description based on functional-semantic foundations was very clear: generative theory was about formal distribution and not semantics, because semantics was deemed to be too unsystematic to be amenable to formal description; the same basic assumption remained until at least 1977 (cf. Chomsky 1977: 30–31). When semantics was



Peter Harder

taken up within the generative tradition, however, a considerable area of shared interest between functional and generative linguists emerged: how to capture semantic generalizations while linking them up with distributional generalizations. As examples of convergent tendencies can be mentioned the notion of layering, cf. Siewierska (1992) including also the parallels between NP and clause structure, cf. Langacker (1991: 143, 194), Abney (1987), Hengeveld (1989), Rijkhoff (1990, 2002), Zamparelli (2000), Bernstein (2001) etc. As part of this development, the theories also came to reflect an awareness of hierarchical relations between higher, ‘functional’ elements and lower, lexical/ conceptual elements as well as the links between position and scope. It is remarkable also that a point is made both on the functionalist side (cf. Langacker 2004: 86; 94) and on the formal-generative side (cf. Longobardi 2001: 589) of the connection between grounding/determination and the ability to serve as a full nominal expression. In the generative literature, Longobardi’s (2003) general theory of relations between semantic and syntactic properties of nominal expressions reflects interests which are also crucial in a functionalist perspective. One major problem, which is also shared (cf. Harder 1996: 231), is the problem in distinguishing clearly between semantic and distributional generalizations, corresponding to two senses of the term ‘underlying’ (1) = ‘reflecting abstract distributional generalization’, and (2) = ‘reflecting semantic relations’ (cf. also Matthews 1993). This is a foundational issue with both an ontological and a methodological side: the ontological question of the relation between meaning and structure gives rise to the methodological issue of what criteria apply in assigning syntactic as opposed to semantic properties to linguistic expressions. The following three quotations illustrate the point. The quotation from Chomsky below was originally pointed out to me by Helge Dyvik: (10) (1) (2)

John i [VP seems [S ti to [VP be sad ]]] seems (sad (John))

In theories of the sort I will be considering here, (1) is the S-structure representation of the sentence, while (2) – understood as indicating that sad is predicated of John and seems of the proposition sad (John) is reminiscent of notations of familiar logics. The null hypothesis, within the theories considered here, is that (1) is also the LF-representation. While it would be simple enough to design an algorithm to convert (1) into (2) or something like it, empirical evidence would be required to support any such move…I will tentatively assume that representations such as (1) are indeed appropriate for LF. (Chomsky 1981: 35)

Chomsky is arguing that a description justified on distributional, syntactic grounds should be assumed (for reasons of descriptive economy) also to be valid as a description of logical, semantic relations. This view clearly contrasts with a basic functional assumption, namely that providing an intuitively satisfactory description of the semantic relations is a point in itself – abstract syntactic structure cannot take over the burden, simply because meaning is meaning and needs to be described as such. As far



Determiners and definiteness 

as I can see, the same possibility of choosing whether allocate a problem to syntax or to semantics is also present in the following quotation from Longobardi (2003: 253): (11) [these conclusions] suggest that a very close and abstract mapping exists between syntax and semantics [… ] On the other hand, the evidence suggests that the semantic differences, at least in this case, need not be stated as a primitive semantic parameter […], but can well be reduced to differences in the abstract morphosyntax of languages. The question is, what are the criteria for assigning syntactic as opposed to semantic properties to expressions? In the two cases above, the issue is that the semantics need not be specified if a syntactic representation could ‘handle’ the issue. In 12 below, what I see as a semantic difference is used as a direct argument for the presence of a syntactic property: (12) (1)

a. [John] smiled b. [Every dog] barked c. Mary is [a person]

(…) 1 c belongs to a syntactic category different from (a) or (b), since it lacks the topmost DP layer which is present in (a) and (b). Unlike (1)a and (1)b, (1)c denotes a property. Zamparelli (2000: 3–4).

From a functional-semantic point of view, the fact that a person denotes a property is basically a semantic fact, not a syntactic fact. There is indeed a relation between the property-denoting function and the absence of a determinative layer, since the determiner is roughly speaking superfluous when only the property rather than an instantiation is involved (cf. also Langacker 2004). It is also true that the fact that a person ends up with this semantic job is due to its (content-)syntactic position, i.e. the encoded relation between a person and the rest of the clause. However, it seems doubtful to interpret the syntactic dimension as reflecting the absence of a phrase-internal top layer. If we look at the way encoded meanings interact functionally in the clause, we can explain the fact that ‘a person’ ends up denoting a property rather than a referent as reflecting the fact that a person fills the predicate (or ‘subject complement’) slot after the copula verb – which (by syntagmatic implicature, cf. above) adds an ascriptive function to its inherent semantic potential (roughly, = “indefinite, singular member of the ‘person’ category”). The expression side, i.e. the linear position in the predicative slot after a form of the verb to be, correlates with a content-side relationship involving property ascription. If we put the same expression in a different syntactic slot, i.e. after I just killed ___,,the same semantic potential ends up denoting an indefinite specific referent. In predicative position, the meaning of the indefinite article is not absent, merely duplicated by information elsewhere in the clause (cf. the ‘empty bus’ principle, discussed in section 6).



Peter Harder

The fact that the article is superfluous in predicative function is in some cases reflected in its absence; in English sometimes after as, and in other languages also in positions corresponding to c above. But presence/absence of an indefinite article in such cases does not invariably correlate specifically with property ascription vs. specific indefinite reference. In Danish property-ascribing NPs, for instance, there is no indefinite article if the noun denotes a profession (han er Ø lærer = he is a teacher), but there is one if it is a term of abuse (han er et fjols = he is a fool). In either case, the semantic content needs to be specified independently, whether it can be correlated with (expression) syntax or not. An attractive feature of the descriptive practice of assigning underlying syntactic form is that on the basis of such forms predictions can be made that ‘account for’ or ‘explain’ the presence of both semantic and ‘surface’ (=’expression-’) syntactic properties that would otherwise stand unexplained. In a generative grammar where one of the purposes is to uncover relations between syntactic and semantic elements, this has the consequence that from a beginning where semantics was assumed to have no systematic relationship with syntax, generative linguists increasingly aim to provide descriptions that establish a point-by-point correspondence between syntax and semantics, setting up underlying elements where surface ones are absent to amend gaps in the correspondence. Although from a functional point of view such a strategy overburdens the description of individual sentences in individual languages, it may be illuminating in other ways: if we view abstract syntactic descriptions as involving metacategories designed to bring different syntactic systems into motivated relations with one another, their status is not open to the same objections. Abstractions may be very economical when we see them as spanning a large body of cross-linguistic generalizations (cf. also Thrane 1999: 21). Also, positing abstract relations between structures that are superficially different has the heuristic advantage of forcing the question of what motivates the choice between alternatives that would otherwise simply have lived apart – enriching also functionalist agenda. One example where generative-style structural apparatus is brought into service as a means of showing how closely related languages can solve problems of expressive options differently is Klinge (2001: 170). The problem is past participles used as noun modifiers and it is demonstrated how Danish, German and English go separate ways towards achieving the same semantic potential for reasons associated with the differences in the expression systems, illustrating the practical ‘muddling-through’ fashion that I see as characteristic of function-based structure. The difference of perspective when it comes to abstract, cross-linguistic syntax can be illustrated with reference to the impressive general theory of noun phrase structure and semantics found in Longobardi (2005). Space does not permit attempts to do justice to the subtlety of the theory, only to illustrate where a functional approach would be different. Longobardi formulates a very simple and general principle (applying to all classes of nominal expressions, including proper names and pronouns) linking up the semantic function of reference with the D position in the syntax of noun phrases. As



Determiners and definiteness 

part of this general theory, he provides arguments concerning the issue of where determinerless proper names belong: in the D position or the N position? In terms of the overall descriptive strategy, convincing arguments are advanced that determinerless proper names belong in the D position, undergoing the operation N-to-D raising. A striking example from Italian of the way this rule works is provided by the following data (additional details have been left out): (12) a. L’ antica Roma fu la città più importante del Mediterraneo Gloss: The ancient Rome was the most important city of the Mediterranean b. *Antica Roma fu la città più importante del Mediterraneo Gloss: Ancient Rome was the most important cuity opf the Mediterranean c. Roma antica fu la città più importante del Mediterraneo Gloss: Rome ancient was the most important city of the Mediterranean The main features of the analysis is that the b form is ungrammatical because if there is no article, Italian requires the proper name to raise to the determiner position (compare example c), thus “acquiring object reference”. From a functional point of view, the generalizations achieved by Longobardi would also be significant and interesting, bringing out, e.g., the functional affinity between (inherently definite) determiners and (inherently definite) proper names that motivates the position of the proper name Roma in the c example. The maximal differentiation that is inherently bound up with a tree that always specifies both a D and an N position, even if we have only a monolithic Roma, is also useful when distributional regularities are compared across languages, enabling Longobardi to provide a number of simple and elegant comparisons. But in functional terms, this does not warrant a description in terms of which reference is bound up with an abstract structural D-position, also when the whole nominal expression is he or Joe.

9. Summary and conclusions I have tried to outline a picture of the nature of determiners viewed in terms of a function-based structure, with the structure of the semantic content side as the most important dimension. The approach to the syntactic organization of clause meaning is basically top-down, viewed as a matter of differentiation rather than composition: investigating how types of meaning in a whole utterance can be factored out into syntactic sub-slots. A fundamental type of differentiation that is relevant for understanding determiners in this perspective is the differentiation into grounding and symbolicconceptual meaning, which in an evolutionary perspective are successors to holophrastic calls with fused indexical and descriptive content (analogous to the siren of an ambulance). Such a differentiation may operate both at clausal level (there’s food) and at the argument level (that car); and the two levels are linked in that (in the Standard



Peter Harder

Average European pattern) the subject term in a categorical judgment adds situational grounding to the whole utterance (the house is on fire). The existence of the determiner slot, however, does not reflect a structure that emerges directly from function – rather, it illustrates the partial arbitrariness that is endemic to all functional systems. The slot hosts not only grounding elements such as demonstratives but also elements that provide no link to the indexical ground (indefinites like a, one and some). This is why, from a purely functional view, the fillers of the determiner paradigm cannot motivate the category solely via an ‘emergent’ semantic rationale. Descriptions of noun phrase structure based on the positive semantic contribution of different types of elements thus would not predict any determiner category. Structural slots, like job descriptions in a business company, impose a structure which is partially autonomous from the inherent properties of the elements themselves, and also from the overall semantics of the utterance meaning: structure cannot be derived from function (alone). The determiner slot is reserved not for a particular type of positive meaning, but for a choice: specifying whether or not the symbolicconceptual part of the nominal expression is to be read as denoting a referent that needs to be identified. Differentiation creates structure and enables flexibility of coding – but there is a cost-benefit mechanism operating to keep structural complexity down to what is optimal for achieving the relevant communicative purpose. No overall matching between semantics jobs and syntactic complexity is therefore possible. The split between a determiner slot and a nominal, descriptive slot is only optimal if there is not a simplex referential strategy available, e.g., via a proper name or a pronoun, which would then fill out the whole argument slot. This avenue of approach contrasts with the ‘maximal differentiation strategy’ where all elements are placed with respect to the most complex distributional pattern available – which predicts a determiner slot for all nominal structures. By the same logic, it contrasts with a descriptive practice which posits structural elements standing for meaning that is not encoded but arises via inference, i.e. by interpreting coded instructions with reference to the situational context of utterance. A very simple form of language might leave all grounding to inference rather than code any explicit grounding instructions; with respect specifically to nominal expressions, many languages have no obligatory slot for grounding but use only grounding instructions when they are individually motivated in the actual situation. Such languages, including Latin and Finnish, depend on inferential strategies that are more elaborate, and coding strategies that are less elaborate (on this particular point) than languages with determiner systems. In understanding determiners, it is thus essential to see them both in terms of their functional contribution and their structural slot. The slot is (partially) autonomous in relation to their semantic contribution, as exemplified by the fact that one only becomes an indefinite determiner in virtue of being seen as occupying the D slot (by syntagmatic implicature) – there is no contradiction between the meaning ‘cardinality 1’ and definiteness. Similarly, the D slot itself is dependent on the existence of an



Determiners and definiteness 

N slot in relation to which its syntagmatic role is defined; and the relative importance of the two elements depends on what fillers go into the slots. This does not mean that it is inherently wrong to set up a super-category subsuming both the category of freestanding pronouns and the category of determiners, or to set up cross-linguistic patterns that constitute the most economical formats of comparison between syntactic systems. But such generalizations should not be seen as expressing the syntactic structure of simple clauses in individual languages.

References: Abbott, B. 2004. Definiteness and indefiniteness. In The Handbook of Pragmatics, L.R. Horn & Gregory Ward (eds). Oxford: Blackwell. Abney, S.P. 1987. The English Noun Phrase in its Sentential Aspect. MIT. Anderson, J.M. 2004. On the grammatical status of names. Language 80(3): 435–474. Bakker, D. & Siewierska, A. 2004. Towards a speaker model of Functional Grammar. In Mackenzie & Gómez-Gonzalez (eds), 325–365. Baltin, M. & Collins, C. (eds.) 2001. The Handbook of Contemporary Syntactic Theory. Oxford: Blackwell. Bernstein, J. 2001. The DP hypothesis: Identifying clausal properties in the nominal domain. In Baltin & Collins (eds), 536–561. Bhat, D.N.S. 2004. Pronouns. Oxford: OUP. Bybee, J. 1994. The Grammaticalization of zero. In Perspectives on Grammaticalization, W. Pagliuca (ed.). Amsterdam: John Benjamins. Bybee, J., Perkins, R. & Pagliuca, W. 1994. The Evolution of Grammar. Tense, Aspect, and Modality in the Languages of the World. Chicago IL: University of Chicago Press Chesterman, A.1991.On Definiteness. Cambridge: CUP. Chomsky, N. 1957. Syntactic Structures. The Hague: Mouton. Chomsky, N.. 1977. Essays on Form and Interpretation. Amsterdam: North-Holland. Chomsky, N. 1981. Lectures on Government and Binding. Dordrecht: Foris. Coene, M. & D’hulst, Y. (eds). 2003. From NP to DP. Vol 1. The Syntax and Semantics of Noun Phrases. Amsterdam: John Benjamins Croft, W.A. 2001. Radical Construction Grammar. Syntactic Theory in Typological Perspective. Oxford: OUP. Deacon, T. 1997. The Symbolic Species. The Co-evolution of Language and the Human Brain. Harmondsworth: Penguin. Deacon, T. 2003. Universal Grammar and semiotic constraints. In Language Evolution, M.H Christiansen & S. Kirby (eds), 11–139. Oxford: OUP. Diderichsen, P. 1946. [1968]. Elementær dansk grammatik. København: Gyldendal. Dik, S.C. 1989. The Theory of Functional Gramma. Vol r 1: The Structure of the Clause. Dordrecht: Foris. Dik, S.C. & Kahrel, P. 1992. ProfGlot: A multi-lingual natural language processor, Working Papers in Functional Grammar 45. Diogenes Laertius. 1965. Lives of Eminent Philosophers. With an English translation by R.D. Hicks. Cambridge MA: Harvard University Press.



Peter Harder Engberg-Pedersen, E., Fortescue, M., Harder, P., Heltoft, L. & Falster Jakobsen, L. (eds.). 1996. Content, Expression and Structure: Studies in Danish Functional Grammar. Amsterdam: John Benjamins. Engberg-Pedersen, E., Fortescue, M., Harder, P., Heltoft, L., Herslund, M. & Falster Jakobsen, L. 2005. Dansk funktionel lingvistik. København: Københavns Universitet & Roskilde Universitetscenter.. Giorgi, A. & Longobardi, G. 1991. The Syntax of Noun Phrases. Cambridge: CUP. Harder, P. 1976. En strukturel og funktionel beskrivelse af bestemthed i moderne engelsk. MA thesis, Univer­sity of Copenhagen. Harder, P. 1990. The semantics and pragmatics of reference. In Pragmatics and its Manifestations in Language [Copenhagen Studies in Language 13], L. Lundquist & L. Schack Rasmussen (eds), 41–78. Copenhagen: Copenhagen Business School. Harder, P. 1992. Semantic content and linguistic structure in functional grammar. On the semantics of ‘nounhood’. In Layered Structure and Reference in a Functional Perspective, M. Fortescue, P. Harder & L. Kristoffersen (eds), 303–327. Amsterdam: John Benjamins. Harder, P. 1996. Functional Semantics. A Theory of Meaning, Structure and Tense in English [Trends in Linguistics. Studies and Monographs 87]. Berlin: Mouton de Gruyter. Harder, P. 1999a. Function, cognition and layered clause structure. In Cognitive Semantics. Meaning and Cognition, J. Allwood & P. Gärdenfors (eds), 37–66. Amsterdam: John Benjamins, Harder, P. 1999b. Partial autonomy. Ontology and methodology in cognitive linguistics. In Cognitive Linguistics: Foundations, Scope and Methodology, T. Janssen & G. Redeker (eds), 195–222. Berlin: Mouton de Gruyter. Hawkins, J.A. 1978. Definiteness and Indefiniteness. A Study in Reference and Grammaticality Prediction. London: Croom Helm. Hawkins, J.A.1994. A Performance Theory of Order and Constituency. Cambridge: CUP. Hjelmslev, L. 1966 [1943]. Omkring sprogteoriens grundlæggelse. København: Københavns Universitet. (English translation, 1953. Prolegomena to a Theory of Language. Bloomington IN: Indiana University). Hopper, P.J. 1987. Emergent grammar. In Proceedings of the Thirteenth Annual Meeting of the Berkeley Linguistics Society, J. Aske, Beery, N., Michaelis, L. & Filip, H. (eds), 139–157. Berkeley CA: Berkeley Linguistics Society. Klinge, A. 2001. Notes on the structure of the English determiner phrase in a comparative perspective. In Iconicity and Structure, E. Engberg-Pedersen & P. Harder (eds). Copenhagen: Copenhagen University. Kuroda, S-Y. 1972. The categorical and the thetic judgement. Foundations of Language 9: 153–185. Langacker, R.W. 1991. Foundations of Cognitive Grammar, Vol II. Stanford CA: Stanford University Press. Langacker, R.W. 2004. Remarks on nominal grounding. Functions of Language 11(1): 77–113. Longobardi, G. 2001. The structure of DPs: Some principles, parameters, and problems. In Baltin & Collins (eds), 562–604. Longobardi, G. 2003. Determinerless nouns. A parametric mapping theory. In M. Coene & Y. D’hulst (eds), 239–254. Longobardi, G. 2005. Toward a unified grammar of reference. Zeitschrift für Sprachwissenschaft 24, 1: 5–44.



Determiners and definiteness  Mackenzie, J.L. 1998.The basis of syntax in the holophrase. In Functional Grammar and Verbal Interaction, M. Hannay & A.M. Bolkestein, 267–295. Amsterdam: John Benjamins. Mackenzie, J.L & Gómez-Gonzáles, M.A. (eds). 2004. A New Architecture for Functional Grammar. Berlin: Mouton de Gruyter. Matthews, P. 1993. Grammatical Theory in the United States from Bloomfield to Chomsky. Cambridge: CUP. Mintzberg, H. 1979. Structuring of Organizations. Englewood Cliffs NJ: Prentice-Hall. Newmeyer, F. 2003. Grammar is grammar and usage is usage. Language 79(4): 682–707. Rijkhoff, J. 1990. Explaining word order in the noun phrase, Linguistics 28: 5–42. Rijkhoff, J. 2002. The Noun Phrase [Oxford Studies in Typology and Linguistic Theory]. Oxford: OUP. Schøsler, L. 2001. Reanalysing structure. The modern French definite article, its predecessors and development. Acta Linguistica Hafniensia 33: 91–108. Strawson, P.F. 1959. Individuals. An Essay in Descriptive Metaphysics. London: Methuen. Thrane, T. 1999. Forklaringstyper i lingvistikken. Funktionelle Fodnoter juni 1999, 1–26. English Department, University of Copenhagen. Zamparelli, R. 2000. Layers in the Determiner Phrase. New York NY: Garland.

Articles, definite and indefinite Michael Herslund

Copenhagen Business School In the Indo-European languages where a definite article exists, it is historically derived from a demonstrative pronoun. The hypothesis of this paper is that the origin of the definite article is the creation of noun phrase structure by the subordination of a noun to a demonstrative pronoun. This process is described for the Romance languages and Danish. In languages where an indefinite article exists, it is historically derived from the numeral ‘one’. This origin of the indefinite article points out two possible directions for its further evolution: it can continue as a quantifier or it can become a classifier. The quantifier road is illustrated by the plural of the indefinite article in Old French and Spanish, the classifier use by the two indefinite articles of Modern French. Definite and indefinite articles thus have different functions and values: definite articles are pronominal heads, indefinite articles are quantifiers or classifiers

1. Introduction Many languages have articles, but there is a marked asymmetry between so-called definite and indefinite articles. The first aspect of this asymmetry is the fact that languages may have a definite but not an indefinite article, e.g. Ancient Greek. The second is that the two kinds of articles, where they both exist, perform very different duties. What they have in common, though, is the fact that in some languages they apparently enter into a paradigmatic relation so that they – often together with demonstrative and possessive determiners – occupy the same position or slot in the noun phrase, the determiner slot: (1)

English: a the my this that

N



Michael Herslund

It is this paradigmatic relation (mutual exclusiveness) that accounts for the common denomination as ‘articles’.

2. The definite article In presumably all Indo-European languages where there is a definite article, it is historically derived from a demonstrative pronoun by processes of semantic bleaching and syntactic cliticisation.

Definites as (cliticised) pronouns

2.1

The hypothesis put forward in this paper is that the origin of the definite article is to be sought in the subordination of a noun (and its possible dependents) or any other kind of word to a demonstrative pronoun, cf. Herslund (2002, 2003b). The origin of the definite article can be illustrated by Ancient Greek, where one of the uses of the article is to turn something else into a noun, as in the well-known example, the name of Christ: (2) ho christos ‘the anointed’ In this phrase, the definite article – which historically is a demonstrative pronoun, its singular forms ho, he:, to stemming directly from the Indo-European pronoun series *so, sa, tot – together with a participle creates (of the verb chrio: ‘anoint’) a noun phrase. So, historically speaking, the emergence of the definite article is the result of the combining of a pronoun with a noun or something else, thereby creating a noun phrase structure, which can be represented as the following dependency structure with examples from German and French: (3)

Det

N

Article

Noun/Adjective/Participle

das

Buch/Schöne/erlebte

le

livre/beau/vécu

‘the

book/beatiful/experienced’



Articles, definite and indefinite 

I shall not go into the details of the arguments in favour of this analysis, which is at variance with the traditional analysis, which sees the noun as the head of the noun phrase and the article as subordinated to it. I just want to point out a couple of facts which support it. The proposed structure, with the determiner as the head, is of course a variant of the now widely favoured DP-analysis of contemporary versions of GB-grammars. But the pronominal origin and nature of the article has been recognised for a long time. So for instance by the 18th Century French grammarian Nicolas Beauzée, cf. the following quotation from Wilmet (1986): (4) “Beauzée identifie l’article et le pronom: Avez-vous lu la grammaire nouvelle? – Non, je la lirai bientôt” (Wilmet 1986: 31) ‘Have you read the new grammar? – No, but I shall read it soon’

The same correspondence can be observed in Danish: (4’) Har De læst den nye grammatik? – Nej, men jeg læser den snart. ’Have you read the new grammar? – No, but I shall read it soon’ The identification of the definite article with the 3rd person personal pronoun is of course obvious, but both synchronically and diachronically one has in these cases to assume a demonstrative pronoun as the origin of both, as is also well known. What I want to argue is that also synchronically the definite article is a reduced and cliticised version of a demonstrative pronoun. In certain environments the (clitic) article thus alternates with its non-clitic, demonstrative counterpart, which in such contexts, however, has no demonstrative, i.e. deictic, value. This appears in French from “gapping” structures like the following: (5) la maison de mes parents et celle ___ de mes grand-parents ‘the house of my parents and that ___ of my grand-parents’ Whereas the article can cliticise to the following noun in the first conjunct, there is no noun in the second, and the full (non-clitic) form of the underlying demonstrative pronoun reappears. Such conditions are different in different Romance languages. In Spanish for instance, the article is retained in the second conjunct: (6) la casa de mis padres y la ___ de mis abuelos ‘the house of my parents and that ___ of my grand-parents’ This indicates a much more pronoun-like status of the Spanish article, a status that is confirmed by other features of Spanish syntax. As antecedents of relative clauses,



Michael Herslund

French and Italian must use the non-cliticised (demonstrative) version of the article, but here again Spanish uses the short form: (7) El que no quiere participar en los trabajos … ‘The one who doesn’t want to participate in the work …’ The pronoun-like status of the Spanish definite article is further corroborated by the use of a definite subject noun phrase together with a first or second person plural verbal inflection, but without any other pronoun such as would be required in most other languages: (8) Los que te quisimos para siempre te odiaremos (F.C. Barcelona supporters to Portuguese player Figo after his change from Barcelona to Real Madrid) ‘We who loved you will forever hate you’ los madrileños gustamos de la noche (Conde 41) ‘we Madrilenes like the night’ no creo que seamos tan distintos los pueblos de Europa (Conde 88) ‘I don’t think that we the peoples of Europe are so different’

The Romance languages thus exhibit an interestingly graded distribution in this respect: (9) Spanish: a. la casa roja y la ___ amarilla b. la casa de Jorge y la ___ de Ana (10) French: a. la maison rouge et la ___ jaune b. la maison de Georges et celle ___ d’Anne (11) Italian: a. la casa rossa e la/quella ___ gialla b. la casa di Giorgio e quella ___ di Anna ‘the red house and the yellow (one)’ ‘the house of George and that of Ann’ As it appears from the examples (9a) – (11a), both Spanish, French and Italian can cliticise the pronoun to a following adjective, but Italian actually prefers the non-cliticised demonstrative already at this stage, cf. Renzi (1988: 422 f.). And whereas Spanish, as seen above, sticks firmly to the clitic form even before a prepositional complement, both French and Italian have only the non-clitic demonstrative in this case, cf. (9b) – (11b). Historically, French once was in line with Spanish insofar as Old French, where examples like the following are common, also had the clitic form in such cases: (12) a. le cheval noir et le ___ blanc ’the black horse and the white (one)’



Articles, definite and indefinite

b. le cheval Lancelot et le ___ Gauvain ‘the horse of Lancelot and the (one) of Gawain’ The alternation between (clitic) article and non-clitic demonstrative in the Romance languages can then be represented as the following hierarchy of possibilities: (13)

___ N

___ Adj

___ PP

Spanish

Art

Art

Art

French

Art

Art

Dem

Italian

Art

Dem

Dem

In Danish the facts are different, but they point in the same direction. The demonstrative origin of the definite article is well known, but in Old Norse it became enclitic, following the noun. This is the origin of the Modern Danish definite inflection. But as soon as the noun is modified by a preposed adjective, the full form of the article, which is identical with the demonstrative, reappears anteposed to the noun group, where the adjective then takes a definite ending, cf. (14): (14)

bog ’book hus ’house

bog-en book-def hus-et house-def

den rød-e bog the red-def book’ det rød-e hus the red-def house’

Gapping structures in Danish show a similar distribution to the ones in (9a) through (11a), cf. (15) (the b-examples with a genitive are more complicated and will be dealt with in 2.2 below): (15) det røde hus og det gule ___ ’the red house and the yellow(one)’ The cliticisation and reduction of the demonstrative in cases like these – which could be supplemented by similar data from other languages − is accompanied by a semantic bleaching so that the former demonstrative pronoun in its clitic version as an article just states the identifiable existence of whatever the subordinate noun denotes – whether referentially or generically. This means that the definite article “points outwards” to the surrounding text or discourse and anchors the noun phrase in that environment.

2.2

Definite articles and other heads

So far we have only considered noun phrases with pronouns – becoming clitic articles – as their heads. But in fact other things can occur in the head position or determiner slot instead of or alternating with this pronoun. This is most noticeably the case with genitive phrases. In this respect it is instructive to compare Danish and the Romance





Michael Herslund

language Romanian. Like English, Danish has a preposed genitive with the ending -s. This ending is actually a clitic element, but the important fact in the present context is that the genitive phrase occupies the same determiner slot as the definite article and thus makes the entire noun phrase definite, whether the genitive phrase is itself definite (16a) or indefinite (16b), cf. Herslund (2001a, b): (16) a. b.

bogen-s rød-e omslag ’book-def-gen red-def cover’ en bog-s rød-e omslag ’a book-gen red-def cover’

That the genitive phrase actually makes the entire noun phrase definite, is seen from the fact that an adjective must occur in the definite form. The structure of the phrase will accordingly be like (16’): Det

(16’)

Gen

Adj

N

en

N

bog-en

-s

rød-e

omslag

bog

-s

rød-e

omslag

The head status of the genitive phrase is further demonstrated by gapping structures like the ones of (9b) – (11b), where in the second conjunct only the head remains, cf. (15’) and (15): (15’) Georgs hus og Annas ___ ’George’s house and Anna’s’



Articles, definite and indefinite 

Like Danish, Romanian has an inflectional, postposed definite article, and its genitive is as in Danish, details apart, inflectional (or clitic), but like the article it is postposed. This goes smoothly as long as the core noun of the genitive phrase itself is definite: (17) istori-a poporu-lui român history-def people-def-gen Romanian ‘the history of the Romanian people’ And the dependency structure of (17) is (17’): (17’)

Det

Det

Gen

N

N

Adj

istori-

-a

poporu-

-lui

român

Under the ‘determiner-as-head’ analysis, the genitive determiner phrase thus governs a cataphoric definite article. This explains one of the more special features of Romanian grammar, viz. a double system of definite articles, which is in fact the well-known alternation between the definite article and its demonstrative counterpart (cf. 1.1) in a new disguise. But when the core noun of the dependent phrase is itself indefinite, one has to introduce a non-inflectional definite article – actually a reduced demonstrative − agreeing with the lexical core noun, as in (18): (18) a. b.

o istorie a poporu-lui român ‘a history def people-def-gen Romanian’ un creion al prietenu-lui român ‘a pencil def friend-def-gen Romanian’



Michael Herslund

The dependency structure of these phrases is accordingly (18’): (18’)

Det

Det

N

Gen

N

Q

Adj

o

istorie

a

poporu-

-lui

român

un

creion

al

prietenu-

-lui

român

The topmost determiner – the genitive phrase – has to govern a headed phrase, i.e. a definite article. Since the governed phrase has no definite article (but an indefinite one here symbolised by Q for quantifier; see further section 3. below), a definite article has to be introduced – in much the same way as the preposed Danish genitive induces the definite inflection of an adjective (but no definite article). This indicates that the genitive phrase occupies the topmost determiner position as a definite article, and that that fact has to “materialise” in the shape of a non-inflectional definite article. At the same time it shows that the (anteposed) indefinite article is not in itself a head, i.e. it is not an article − since it cannot be governed by a genitive phrase in the way the definite article can – a sure sign that definite and indefinite articles are not on an equal footing. And the assumed paradigmatic relation between the two kinds of articles is, in some languages at least, an illusion. The two observations – the pronominal nature, the head status and ensuing textual function of the definite article, and the “non-head status” of the indefinite article − make a natural transition to the discussion of the indefinite article.

3. The indefinite article In the Indo-European languages where there exists an indefinite article, it is historically derived from the numeral ‘one’ by similar processes of semantic bleaching and syntactic cliticisation as the definite article. But contrary to the definite the indefinite article “points inwards” by determining the semantic/referential value of a noun (or



Articles, definite and indefinite 

whatever constitutes the lexical core of the noun phrase). It is thus common for languages with indefinite articles not to use them when the intended reading of a noun is non-specific or purely intensional, cf. e.g. the case of Old French as described by Carlier (2001) and Herslund (2002), Old Italian as described by Stark (2002), or Bulgarian as described by Guenchéva (1994). In such languages, it is indeed doubtful whether the numeral has reached the stage of a proper, i.e. rather automatically inserted, indefinite article. The origin of the indefinite article as a quantifier points out two possible directions for its further evolution in a given language: it can either continue as a kind of neutral quantifier or it can become a classifier as well. The quantifier road is illustrated by the plural of the indefinite article in Old French and Spanish (cf. Herslund 2003a), the classifier use is illustrated by Modern French where the indefinite article has two shapes: the traditional indefinite article un, and the so-called partitive article du (cf. Herslund 1998, 2003c).

3.1

Indefinites as quantifiers

Many languages wouldn’t dream of inflecting their indefinite article meaning ‘one’ in the plural. Indeed, what could the meaning be of such a rather contradictory plural of ‘one’? Many languages use the bare noun for indefinite plural, or props like indefinite pronouns meaning ‘some’ or ‘any’ when more explicit quantification is needed. But among the Romance languages both Old French and Spanish, together with Catalan and Portuguese, have plural forms of the indefinite article. They are special uses however, not employed automatically. The normal plural of Spanish una casa ‘a house’ is not unas casas, but the bare plural noun casas ‘houses’. What is then the use and meaning of unas casas? Both in Spanish and in Old French the plural of the indefinite article is used to denote collectives, i.e. structured collections of items. This is quite understandable because such collections are units with internal plurality – exactly the meaning of the two morphemes ‘un’ and ‘plural’ constituting the apparently contradictory plural of ‘one’, Old French uns, unes, Spanish unos, unas. In dealing with these forms one should notice that, semantically, pluralisation is homogenisation, cf. Langacker (1991: 77), Herslund (1998, 2003a, c). A moment’s reflection will bear out this apparently strange statement. What is in fact a homogeneous entity if not an entity whose division yields other entities of the same kind, viz. a lump of butter, some water, etc.? The same is true of plural sets, the division of which also yields entities of the same kind, viz. a collection of typewriters or whatever – so plurals are semantically homogeneous. The end meaning of our indefinite plural is thus inevitably a plural unit. This accounts for the different readings associated with the indefinite plural: a set, a series, a



Michael Herslund

pair or lexicalised as a plurale tantum. The following selection of examples will serve to illustrate this: (19) Set: Old French: Spanish:

uns echés uns degrez unes armes uns dens unas tierras unos dientes unos versos

‘set of chessmen’ ‘staircase’ ‘armour’ ‘teeth, denture’ ‘estate’ ‘teeth, denture’ ‘stanza’

Series: Old French: Spanish:

uns cops unos golpes unos minutos unas semanas

‘series of blows’ ‘series of blows, knocks, knocking’ ‘a couple of minutes’ ‘a couple of weeks’

Pair: Old French: Spanish:

uns esperons unes botes uns soliers uns ganz unos ojos unos pechos unas cortinas unos labios unas medias

‘pair of spurs’ ‘pair of boots’ ‘pair of shoes’ ‘pair of gloves’ ‘pair of eyes’ ‘bosom’ ‘curtain’ ‘pair of lips’ ‘pair of stockings’

Pluralia tantum: Old French: Spanish:

unes forces unes letres unas tenazas unas antiparras

‘a pair of scissors’ ‘a letter’ ‘tongs, pincers’ ‘spectacles’

The last case is of course the extreme, and most cases of pluralia tantum seem to denote pairs. The notion of ‘pair’ is of course also the most salient version of a plural unit, and in some languages it grammaticalises as a dual. The unit interpretation can be substantiated by authentic examples of its use from Spanish, where either the context or the structure of the sentence – or both – confirm the collective interpretation of the indefinite plural: (20) Desde la oscuridad del cuarto (…) unos ojos escrutaban a las dos mujeres (Conjura 27) ‘From the dark of the chamber a pair of eyes watched the two women’



Articles, definite and indefinite 

We learn from the context – and we are told by the very use of unos − that a pair of eyes, i.e. only one person, is lurking in the dark – and not several, or even, say, two one-eyed persons. In other cases the inherent ambiguity of a plural – collective or distributive reading − is lifted by the use of unos pointing to the collective interpretation: (21) Eran como unos niños que pierden a su madre de vista (Conjura 79) ‘They were like (a group of) children who loose sight of their (common) mother’ Dos mujeres (…) llevaban largas capas y unas capuchas amplias cubrían su cabeza (ib. 204) ‘Two women (…) they wore long capes and wide hoods covered their head(s)’ The crucial feature in the interpretation here is the use of the singular in the object (su madre, su cabeza), which underlines the unit interpretation of the subject (unosS → singular object). The same interpretation is also clearly relevant in cases like the following: (22) Del campo llegaban, a caballo, unos cazadores de altanería (Conde 116) ‘From the field arrived, mounted, a group of falconers’ Le asaltaron, cuando podía dormir, unos sueños minuciosos y terribles (Conde 220) ‘He was assaulted, when he could sleep, by (series of) detailed and terrible dreams’ Unos golpes sonaron en la puerta (Conjura 58) ‘A series of knocks sounded on the door’ In these examples, the subjects, whatever they do, do it together! The distributive reading, on the other hand, where plural subjects do not appear as a group, can be emphasised by the use of the indefinite pronoun algún ‘some’: (23) Era una zona más montañosa en la que, como perdidos, aparecían algunos viñedos (Conjura 83) ‘It was a more mountainous region, where some vineyards appeared, as lost (here and there)’ algunos vendedores voceaban sus mercancías (Conjura 252) ‘some salesmen were crying out their goods’ The plural in the object mercancías of the last example is indicative of the distributive reading thereby matching the plural of the subject (algunosS → plural object).



Michael Herslund

3.2

Indefinites as classifiers

If a language does not make a distinction between countable (units, heterogeneous) and non-countable (mass, homogeneous) nouns – as is the case in Chinese according to the analysis of Lyons (1977: 460 ff.) – it needs a device to signal whether an indefinite noun is to be understood as denoting a unit or a mass. Such a language is Modern French, and such a device is classifiers. The system of the Modern French indefinite articles has all the features of a classifier system, cf. Allan (1977) and Herslund (1998, 2003c). A French noun such as cheval ‘horse’ can, when introduced into a discourse, be presented either as a unit (heterogeneous), or as a mass (homogeneous): (24) a. un cheval b. du cheval The unit interpretation is obvious; the mass reading is illustrated in (25): (25)

Cela, c’est du cheval! ‘That’s a pretty horse, that!’ Ce soir, on mange du cheval. ‘Tonight we have horse meat for dinner’ Il y a en elle du cheval. ‘She looks rather horse-like’ Il y a dans sa famille de la roulotte et du cheval. ‘There is some cart and horse (i.e. gypsy or circus connotations) in his family’

The system of the French indefinite article-classifier is set out in (26): un cheval

(26)

une eau Singular

Heterogeneous

du cheval de l’eau

Plural

des chevaux

Homogeneous

des eaux

One crucial feature of this system is the existence of an indefinite article in the plural too, a fact which should be related to this other fact that all French nouns have a plural. And that the plural indefinite article is morphologically the same as the homogeneous singular (de + definite article) falls out from the observation above that the plural is in itself a mass or homogeneous term, cf. Langacker (1991: 77). That is also why the distinction in the singular is neutralised in the plural. The Modern French



Articles, definite and indefinite 

system is thus radically different from the system in Old French, where there was no “partitive article”. Many grammars and grammarians pretend that the “partitive article” only combines with uncountables, cf. e.g. Kleiber (1994), but the following selection of examples shows that also countables freely combine with this article and hence receive a homogeneous interpretation – a fact highlighted also by the necessity of translating certain of the examples into an English plural: (27) “Tonnerre de Dieu! cria Weiss, ils amènent du canon!”   (...) les Bavarois étaient en train de mettre en position une pièce, au coin de la place de l’Église (Zola Débâcle 265) ‘Thunder of God! Weiss shouted, they bring along artillery (…) the Bavarians were putting a canon in position at the corner of the church square’ C’est plus que de la collaboration, c’est de l’alliance militaire (Pottecher Pétain 382) ‘That is more than collaboration, that’s military alliance’ un grand patriote de 50 ans (...) qui réussit à faire de la victoire avec de la défaite (Express 17.10.86, 66) ‘a great patriot of 50 years (…) who succeeded in making victory out of defeat’ “Il doit y avoir du canard en quantité!” disait le substitut à M. Grandmaison en observant les terrains d’alentour (Simenon Port 46) ‘There must be lots of ducks! the deputy said to M. Grandmaison while observing the terrain around them’ Eux qui venaient de casser allégrement du prisonnier révolté (Pennac Marchande 75) ‘Those who had just joyfully massacred revolting prisoners’ De la clientèle disséminée, qui butine par-ci par-là, ignorante de ce qui se joue (Pennac Bonheur 272) ’Scattered clientele, which are gleaning here and there ignorant of what is at stake’ The following examples show the same nouns with the two articles: (28) A chaque inspiration, c’était comme s’il respirait du feu (Villiers Santiago 104) ‘With every breath it was as if he breathed fire’ Où est la différence entre un feu de chez moi et celui-ci? (Vercors Silence 38) ‘Where is the difference between a fire from my home and this one’ Vanderputte disait que le motif représentait la gloire et que l’objet avait de la valeur (Gary Vestiaire 56) ‘Vanderputte said that the motive represented the glory and that the object had value’



Michael Herslund

les bibelots qui se trouvaient dans l’appartement avaient une grande valeur (Gary Vestiaire 57) ‘the bibelots that were in the flat had a great value’ − Tu as du remords? (Dard Pain 196) ‘− Do you feel remorse?’ Je te supplie simplement de ne pas laisser ta vie s’enliser dans un remords sans cause (Gallo France 285) ‘I only beg you not to let your life be sucked down into a remorse without reason’

je compose de la musique (Vercors Silence 35) ‘I compose music’ C’est une musique inhumaine (Vercors Silence 44) ‘It’s an inhuman music’

The functioning of the system is most clearly seen, however, when the same noun is presented in the same stretch of discourse with the two articles: (29) il me semble qu’on peut reconnaître une certaine manière d’attaquer la première syllabe avec une aspiration, ou plutôt de l’aspiration (Cornulier Etudes 51) ‘I think one can recognise a certain way of attacking the first syllable with an aspiration or rather (some) aspiration’ Ou alors, c’était d’un bien étrange bétail qu’il s’agissait! Du bétail qui allait régulièrement au cinéma (Merlino Jargonautes 116) ‘Or else it was a very strange cattle it was a question of! Cattle that went to the movies’ − Il y a du boeuf gros sel et, avant cela, vous pourriez prendre du pâté de campagne... Sa voix sonnait-elle autrement que d’habitude quand, à la porte de la cuisine elle lança: – Un pâté et un boeuf gros sel! (Simenon Marie 85) ‘− There is beef with coarse salt and before that you can have a country pie … Did her voice have a different sound than usually when she shouted at the kitchen door: − A pie and a beef with coarse salt!’ Veux-tu un café? (...) un militaire apporta un plateau avec du café (Arnothy Ami 177) ‘Do you want a coffee? (…) a soldier brought a tray with coffee’

4. Conclusions If the indefinite article is not an article in the same sense as the definite article, i.e. a pronominal head of a noun phrase, but rather a quantifier or classifier, one might



Articles, definite and indefinite 

expect some important differences to follow. There are indeed such differences, two of which will be briefly touched upon here. The first major difference between definite and indefinite articles is the very obvious fact that even in languages where the definite article has become postposed to the noun and reduced to an inflectional ending, the indefinite article still precedes the noun like other numerals and quantifiers in general: Noun bog carte kniga ‘book’

(30) Danish: Romanian: Bulgarian:

Definite noun bog-en carte-a kniga-ta ‘the book’

Indefinite noun en bog o carte edna kniga ‘a book’

The other difference is that if the two “articles” do not occupy the same slot in the noun phrase one might expect them to combine. That is also the case in languages where the indefinite article and the numeral ‘one’ are formally identical. Just as the definite article as seen in section 2. above sometimes alternates with its non-clitic (demonstrative) counterpart, the indefinite article too alternates with its non-clitic counterpart, the numeral ‘one’ when following the definite article, cf. the following examples from Old French, Danish and German: (31) Old French: Danish: German:

l’un braz den ene arm der eine Arm the one arm ‘one (of the) arm(s)’

What examples like these show is that the indefinite article in these languages has not totally severed its ties with the numeral: in combinations with the (clitic) definite article it cannot cliticise itself – it is stressed in Danish and German! – and it retains the definite inflection in Danish and German which is characteristic of adjectives following a definite article: (31’)

Det

N

Q den der

en-e ein-e

arm Arm



Michael Herslund

Only in English have the ties between article and numeral been cut – there is no longer any resemblance between the clitic article a and the numeral one – so that these two may actually combine in expressions like ‘just a one’ thus validating the paradigm of (1), but only for English. Definite and indefinite articles have different functions and values. That the two kinds of determiners have been classified together as ‘articles’ introducing known and new nouns, respectively, is understandable, but these values follow directly from their different semantic contents and they do not warrant the assumption of a syntactic class of definite and indefinite articles.

References Allan, K. 1977. Classifiers. Language 53: 285–311. Carlier, A. 2001. La genèse de l’article un. Langue française 130: 65–88. Guenchéva, Z. 1994. Document: edin « un » et l’indétermination en bulgare. L’indéfini. Faits de langues 4: 113–120. Herslund, M. 1998. Le français, langue à classificateurs? In La ligne claire. De la linguistique à la grammaire. Mélanges offerts à Marc Wilmet à l’occasion de son 60e anniversaire, A. Englebert, M. Pierrard, L. Rosier & D. van Raemdonck (eds), 65–73. Louvain-la-Neuve: Duculot. Herslund, M. 2001a. The Danish -s genitive: From affix to clitic. Acta Linguistica Hafniensia 33: 7–18. Herslund, M. 2001b. Affixe ou clitique? Le cas du génitif danois en -s. In Clitiques et cliticisation. Actes du colloque de Bordeaux, octobre 1998, C. Muller, P. de Carvalho, L. Labrune, F. Lambert & K. Ploog (eds), 147–157. Paris: Honoré Champion. Herslund, M. 2002. Incorporation and transitivity in Romance. In Complex Predicates and Incorporation. A Functional Perspective [Travaux du Cercle Linguistique de Copenhague XXXII], O. Nedergaard Thomsen & M. Herslund (eds), 175–206. København: Reitzel. Herslund, M. 2003a. Le pluriel de l’article indéfini en ancien français. In La cognition dans le temps. Etudes cognitives dans le champ historique des langues et des textes, P. Blumenthal & J.-E. Tyvaert (eds), 75–84. Tübingen: Max Niemeyer. Herslund, M. 2003b. Article et pronom. Réflexions sur le syntagme nominal. In La syntaxe raisonnée. Mélanges de linguistique générale et française offerts à Annie Boone à l’occasion de son 60e anniversaire, P. Hadermann, A. van Slijcke & M. Berré (eds), 105–116. Bruxelles: Duculot. Herslund, M. 2003c. Articles et classificateurs. Cahiers Ferdinand de Saussure 56: 21–33. Kleiber, G. 1994. Nominales. Essais de sémantique référentielle. Paris: Armand Colin. Langacker, R.W. 1991. Foundations of Cognitive Grammar. Volume II. Descriptive Application. Stanford CA: Stanford University Press. Lyons, J. 1977. Semantics I-II. Cambridge: CUP. Renzi, L (ed.). 1988. Grande grammatica italiana di consultazione. Bologna: Il Mulino. Stark, E. 2002. Indefinitness and specificity in Old Italian. Journal of Semantics 19: 315–332. Wilmet, M. 1986. La détermination nominale. Paris: PUF.



Articles, definite and indefinite 

Quotations from the following texts in French and Spanish: Christine Arnothy, L’ ami de la famille. Livre de poche. B. de Cornulier & F. Dell, éds. Etudes de phonologie française. Editions du C.N.R.S. Frédéric Dard, Le pain des fossoyeurs. Presses pocket. Max Gallo, France. Livre de poche. Romain Gary, Le grand vestiaire. Folio. J. Merlino, Les jargonautes. Stock. Daniel Pennac, Au bonheur des ogres. Folio. Daniel Pennac, La petite marchande de prose. Folio. Frédéric Pottecher, Le procès Pétain. J.C. Lattès. Georges Simenon, Le port des brumes. Presses pocket. Georges Simenon, Marie qui louche. Presses de la Cité. Vercors, Le silence de la mer. Livre de poche. Gérard de Villiers, SAS. L’ ordre règne à Santiago. Plon. Emile Zola, La débâcle. Livre de poche. L’ express. Néstor Luján, ¿Decidnos Quién mató al conde? Plaza & Janés. José Calvo Poyato, Conjura en Madrid. Plaza & Janés.

Typological correlations in nominal determination in Romance Elisabeth Stark

University of Zurich

This paper discusses divergences and significant typological correlations found in the family of Romance languages, specifically French, Italian and Spanish. It proposes to reinterpret the complex system of indefinite nominal determination in two central Romance languages, viz. French and Italian, which both feature an indefinite article and a partitive article, as a device of nominal classification in a broad sense, marking the conceptually important distinction between a single, contoured referent and a non-contoured substance. It is argued that this classification system arose when nominal declension in Latin, which differentiated these two referentially highly relevant cognitive concepts via overt gender and number affixes, was partially or completely lost. In contrast to modern central Romance languages, which require obligatory (indefinite) determination in almost every argument position, modern peripheral Romance languages like Romanian or Spanish, possessing a simpler and more flexible system of determination, developed a system of differential object marking in order to unambiguously indicate contoured and highly individualized referents in direct object position.

1. The problem: Different systems of indefinite nominal determiners in Romance noun phrases Despite some well-known and fruitful generalizations and hypotheses assuming homogenous semantic and syntactic systems of nominal determination for all Romance languages (e.g. Chierchia 1998, Longobardi 2001), the data in (1) demonstrates that there is considerable variation: (1) a.

Sp.: Has visto *(un) águila? Fr.: As-tu vu *(un) aigle? It.: Hai visto *(un) aquila? Rom.: Ai văzut (un) vultur? (Did you see an eagle?)



Elisabeth Stark

b. c. d. e. f.

Sp.: Compro pan. Fr.: J’achète *(du) pain. It.: Compro (del) pane. Rom.: Cumpăr (nişte) pîine. (I buy (some) bread). Sp.: Me falta agua. Fr.: Il me faut *(de l’)eau. It.: Mi occorre (dell’)acqua. Rom: Îmi trebuie (nişte) apă. (I need (some) water). Sp.: Demostró paciencia en esta situación. Fr.: Elle montra *(de la) patience dans cette situation. It.: Dimostrò (*della) pazienza in questa situazione. Rom.: Demonstră răbdare în această situaţie. (She showed patience in this situation) Sp.: Veo (a unos) estudiantes en el edificio. Fr.: Je vois *(des) étudiants dans le bâtiment. It.: Vedo (degli) studenti nell’edificio. Rom.: Văd (nişte) studenţi în clădire. (I see (some) students in the building) Sp.: Salen estudiantes del edificio. Fr.: Il sorte *(des) étudiants du bâtiment. It.: Escono?(degli) studenti dall‘edificio. Rom.: Ies studenţi din clădire. ((Some) students leave the building)

In Spanish, French, Italian and Romanian, indefinite nominals with existential reading show quite heterogeneous characteristics in argument position. Table 1 presents an overview over the three most frequent and grammaticalized indefinite determiners used with indefinite nominals in argument position:1 Bare noun phrases in argument position occur in Spanish, Italian and Romanian under restricted grammatical conditions: in fact, only abstract nouns can appear freely in bare noun phrases in argument position even in the singular (cf. 1d); bare plurals surface postverbally in subject and object position independently of the lexical category of the noun (normally with non-specific interpretation of the nominal, cf. 1e and 1f). Bare singulars are also possible with “mass-denoting nouns” in postverbal subject and object position in Spanish, Italian and Romanian (see 1b and 1c), again with nonspecific interpretation of the nominal. Conversely, the only Romance language which almost never permits bare noun phrases in argument position is French. 1. Including ‘zero’ as a possible null determiner for the sake of a similar underlying syntactic structure (cf. Longobardi this volume), let us tentatively suppose, then, that mass/plurals, unlike singulars, can be introduced by an empty determiner.



Typological correlations in nominal determination in Romance 

Table 1.  Distribution of indefinite nominal determiners in four Romance languages French

Italian

‘Zero’: abstract / “mass-denoting”, rarely: “entity-denoting“ nouns (nonspecific): singular.

‘Zero’: rarely with abstract nouns (only in more or less idiomatic expressions).

‘Zero’: abstract / “mass-denoting” (non-speci­ fic): singular.

‘Zero’: abstract / “mass-denoting”, “entity-denoting” (non-speci­ fic): singular.

Plural noun phrases (mostly postverbally).

No bare plurals.

Plural noun phrases (mostly postverbally).

Plural noun phrases (mostly postverbally).

Uno: singular countable noun phrases.

Un: singular countable noun phrases.

Uno: singular countable noun phrases.

Un: singular count­ able noun phrases.

No ‘partitive article’.

Du: abstract / “mass-denoting” in noncountable singular noun phrases.

Del: “mass-denoting” in non-countable sing­ ular noun phrases.

No ‘partitive article’.

Spanish

Romanian

As for overt indefinite nominal determination, every Romance language possesses a so-called indefinite article derived from the Latin numeral unus, ‘one’, which accompanies singular count noun phrases. 1a demonstrates that it seems possible for some Romanian speakers to accept even an “entity-denoting noun” like vultur, ‘eagle’, with a non-specific reading without un. Only two out of the four Romance languages discussed here, namely French and Italian, have a further indefinite determiner, the socalled ‘partitive article’, derived from the composition of Latin de and the definite article. It marks indefinite non-countable singular noun phrases2, usually with “mass-denoting nouns”, in pre- and postverbal subject and object position. In addition, it is obligatory with abstract nouns in French and optional in Italian (see examples in 1b, 1c and 1d). In describing the facts in these terms, following Löbel (1993: 192ff.) and with reference to Gil’s (1987) typology, I assume a fundamental difference between the lexical categories “mass-denoting”, “entity-denoting” and “abstract noun” (N), which derive from characteristics of the potential (extra-linguistic) referents (additivity, divisibility and so on) and which are based on denotational properties of the head noun, and the countability or non-countability of entire noun phrases. This last opposition is a grammatical category or a syntactic feature depending on the internal syntactic structure of the noun 2. I will not discuss the whole functional range and semantic properties of its morphological plural here, which seems to be the normal indefinite plural article and which is fully grammaticalized in French and optional in Italian.



Elisabeth Stark

phrase and it is characterized by the possibility of forming a morphological plural and/or to show compatibility with certain indefinite determiners (French/Italian: uno vs. del – NumP- or DP-level, see below). This assumption is justified by the fact that virtually any noun in Romance languages (like in any language with a grammaticalized countability distinction in this sense) can in principle appear in any kind of noun phrase: (2) a. b.

Sp.: Has comido águila? Fr.: As-tu mangé de l’aigle? It.: Hai mangiato (dell’) aquila? Rom.: Ai mîncat (nişte) vultur? (Did you eat (some) eagle?) Sp.: Compro un pan. Fr.: J’achète un pain. It.: Compro un pane. Rom.: Cumpăr un pîine. (I buy one (a certain amount/piece of) bread)

Even if these examples seem semantically marked,3 due to prototypical affinities between “mass-denoting nouns” (like engl. bread) and non-countability, and between “entity-denoting nouns” (like engl. eagle) and countability (as already discussed for English in Allan 1980), they are grammatically well-formed and their ‘mass’ or ‘count’ readings derive exclusively from the prenominal indefinite determiners (‘zero’, ‘partitive’ or indefinite article).

2. An explanation proposal and its problems In order to explain the striking differences between French and the other Romance languages concerning the possibility of permitting bare plurals or bare (‘mass’) singulars in argument position, the following correlation has often been observed (cf. e.g. Schroten 2001): the loss of overt morphological number marking in nouns correlates with the necessity of number marking via determiners in spoken French.4 Le trait pertinent qui distingue […] l’espagnol du français est la présence du nombre dans la prononciation du nom (Schroten 2001: 196; similarly Wanner 2001: 1699).

3. Cf. Behrens 1995: 47-50, Corbett 2000: 86f.; see also the sortal interpretation or “Artenplural” mentioned by Krifka 1991: 414f. for “mass-denoting nouns” in countable plural NPs and the unique meaning of the morphological plural in languages with grammaticalized countability: it is always understood as additive, “diskrete Gesamtheiten von Objekten derselben Art” (Link 1991: 418). 4. Cf. also Delfitto/Schroten (1991: 157): “...and bare nouns cannot be interpreted since there is no number affix which can be raised to the D-position at LF”.



Typological correlations in nominal determination in Romance 

However, even if this correlation explains the degree of how obligatory explicit nominal determination in argument position is in French, it is not precise enough to explain the considerably different behaviour of the other Romance languages in this respect, as they exhibit morphological number marking also in their spoken varieties. Italian seems to behave in a particularly striking way when compared to Spanish or Romanian: Italian has a quite restricted distribution of bare NPs (recall 1a to 1f) despite overt morphological number marking, and like French but unlike Spanish or Romanian, it possesses a ‘partitive article’. Consequently, it may be inferred that there may be more to an adequate explanation than just the problem of overt morphological number marking in Romance. These facts give rise to the following questions: From Latin to Romance: 1. From a diachronic perspective how can the loss of nominal inflection and the rise of obligatory nominal determination be explained in conjunction with the ‘countability distinction’? 2. Why are there different systems of overt indefinite determiners (singular) in central Romance (French, Italian) vs. peripheral Romance languages (Romanian, Spanish)? 3. Can we find a relation between the two major syntactic innovations in Romance languages as compared to Latin, i.e. the introduction of obligatory and explicit (indefinite) nominal determination in argument position and the phenomenon of “Differential Object Marking” (DOM) (cf. Bossong 1997)? The remainder of this article will propose some tentative answers to these questions. Sections 3 and 4 will deal with Latin and Romance nominal morphology and its possible implications for the distribution of bare noun phrases in argument position (questions 1 and 2). Section 5 will present and discuss again well-known correlations between the existence of some special indefinite determiners and the DOM-phenomenon in Romance, and then develop a new interpretation of the function as well as the potential origin of DOM in Romance languages (question 3). I will try to show that both the Romance systems of indefinite determiners and the phenomenon of “Differential Object Marking” can be considered as devices of nominal classification in a broad sense, replacing the ancient Latin nominal inflection that was partially or completely lost.

3. A look at Latin and Romance nominal inflection 3.1

Latin

Classical Latin possessed a complex declension system divided into 5 classes, which required obligatory and overt marking of the morpho-grammatical categories case, of which there are 5, gender, of which there are 3, and number, of which there are 2. Even



Elisabeth Stark

though clear-cut correspondences between gender, declension class and ‘semantic’ or lexical noun class cannot be assumed (in contrast to the situation suggested for ProtoIndo-European in Ralli 2002), there was some ‘classification potential’ for nouns sharing the same lexical root but differing in gender and/or number: (3) a. b. c.

caseus, ‘one single (piece of) cheese’, olea, ‘olive’ / ‘olive tree’ caseum, ‘cheese as a substance’, oleum, ‘oil’ acinus/acinum, ‘berry’, acina, ‘grape’ frumentum, ‘wheat’, frumenta, ‘corn’5

(3a) shows lexical roots with masculine and feminine gender, resulting in “entity-denoting nouns”, whereas the nouns from the same root in (3b) with neuter gender are “mass-denoting nouns”. In addition, (3c) shows the well-known ‘collective’ semantics of the Latin neuter plural ending in –a (cf. Schön 1971, Windisch 1973). Although theses oppositions are not systematic, the Latin neuter and especially the Latin neuter plural in –a – both unambiguously marked in spoken and written varieties – can be re-interpreted as a partly generalised ‘classification system’ denoting mainly the opposition between ‘single, contoured object’ (e.g. one piece of cheese, one olive, one berry) and ‘non-contoured substance’ (e.g. cheese, oil) or ‘collective’ (grape). That this important semantic opposition is as much related to gender as to number is shown by the fact that, unlike the plural in modern Indo-European languages, including the Romance languages, the Latin plural is neither automatically interpreted as additive (cf. Link 1991) nor restricted to “entity-denoting nouns”: (4) a. frigora caloresque, ‘an intense heat and cold’: plural indicating intensification b. acquae, ‘waters’, cerae, ‘wax tablets’: different appearances of a substance6

Although the Latin plural can have a sortal reading, bare plurals of abstract or “mass denoting-nouns” are not automatically re-categorized as for instance in modern Romance languages (compare Fr. huile, ‘oil’, des huiles, ‘different sorts of oil’). Virtually any Latin noun can be pluralized, and in fact frequent occurrences of plurals of “mass-

5. Cf. in detail Hofmann/Szantyr ((1997) [1965]: 7-10), Meisterfeld (1998: 56ff.) and for late Latin analogical neuter plurals following the same pattern cf. Morani (2000: 228). 6. Cf. Kühner/Stegmann (31955: 69, 73), Hofmann/Szantyr (1997) [1965]: 18, 21).



Typological correlations in nominal determination in Romance

denoting nouns” or abstract nouns, as in (4) above, are attested.7 This fact, together with the absence of compatibility restrictions for (optional) indefinite determiners with nouns (Lat. quidam or aliquis combine freely with abstract, “mass-denoting” and “entity-denoting nouns”) indicates that Latin had no grammaticalized “countability distinction” at the level of noun phrases (cf. Löbel 1993). If Latin inflectional affixes, which mark declension class, gender and number indicate oppositions between ‘contoured single object’, ‘substance’ and ‘collective’, they can be considered as classification devices in the following sense: they are part of the universal dimension of nominal apprehension, which is a central universal operation of establishing reference: First of all, so it seems, one has to be able to express that something is a thing [= dimension of APPREHENSION, E.S.]. Only then can it be named: The dimension of NAMING […] Following that, it can be referenced: The dimension of DETERMINATION. (Seiler 1986: 9) APPREHENSION is the universal operational dimension with corresponding subdimensions which explicate the grasping and representation of concepts corresponding to objects or things by means of language. (Seiler 1986: 145)

Consequently, apprehension concerns chiefly the classification of the denotation of the noun phrase as “an undifferentiated concept or as an individual” (Lehmann 1991: 206, see also Meisterfeld 2000: 328).

3.2

French and Italian

The main morphological changes in the nominal system from Latin to Romance are well-known and comprise the complete loss of morphological case, a reduction of gender (especially the loss of the neuter gender), contrasted with a solid formal preservation of number, but with now only the ‘additive plural meaning’ left. In addition, Modern Standard French shows the complete loss of the declension classes (already in Old French, cf. Delfitto/Schroten 1991: 180f.). Gender and number are usually marked (in the phonetic code) only by prenominal determiners: (5) un ami/une amie – des ami(e)s [œnami/ynami – dezami] ‘a male friend’/‘a female friend’ ‘male or female friends’8 7. Cf. Iturrioz Leza (1986: 295f.): “This individualization strategy [ = pluralization of abstract nouns, E.S.] is more widespread in the classical languages (Greek, Latin) than in modern German or any other European language; thus it is often difficult to translate an abstract [plural, E.S.] NP without changing its number: [...] Asperitates viarum et angustiae [...] ‘The roughness(es) and narrowness(es) of the ways’”. 8. Cf. Delfitto/Schroten (1991: 177ff.).





Elisabeth Stark

The French noun [ami] is thus not phonetically marked at all for gender or number. By contrast, Modern Standard Italian is different from French in having preserved 3 main declension classes, 2 overtly marked genders as well as overt number marking. However, the declensional endings –a and –e are far from being unambiguous markers of singular or plural, as they can either indicate feminine singular, (rarely) masculine singular or feminine plural (-a) or masculine singular or feminine plural (-e) The morpheme -o unambiguously indicates singular, but both masculine and (rarely) feminine gender: (6) Sg.: -o/ -a/ -e; Pl.: -i/ -a/ -e: a. libr-o – libr-i (m.) ‘book’ – ‘books’ b. cas-a – cas-e (f.) ‘house’ – ‘houses’ c. can-e – can-i (m.) ‘dog’ – ‘dogs

mano – mani (f.) ‘hand’ – ‘hands’ poet-a – poet-i (m.), ‘poet’ – ‘poets’;

bracci-o – bracci-a (m. – f.), ‘arm’ – ‘arms’;

What is marked in bold characters in (6a) is a residue of the original Latin classification potential of the neuter plural in –a, as opposed to a regular plural form in –i (originating in Late Latin, cf. Hofmann/Szantyr ((1997) [1965]: 21)) and reanalysed as feminine (but still plural!), always indicating a collective or at least ‘pair’ reading. Some nouns ending in –o (masculine singular), usually denoting concrete objects like body parts (It.: ginocchio ‘knee’, orecchio ‘ear’ and so on, also muro ‘wall’ etc.), have a plural form in –a when denoting a plurality, body parts or a ‘collective reading’. However, they form a plural in –i when used metaphorically to denote something similar in form, but without a collective denotation (e.g. It.: le braccia denotes both arms of an animate being, whereas i bracci denotes the arms of a river, It. le mura denotes the townwall, whereas i muri denotes the single walls of a building).

3.3

Spanish and Romanian

Just like Italian, Modern (European) Standard Spanish has 3 main declension classes, 2 overtly marked genders and overt number marking. It is also “heterogeneous with respect to gender” (Harris 1992: 66ff.), but unambiguous with respect to number marking (“plurality is manifested consistently with the suffix /-s/”, Harris 1992: 67): (7) Sg.: -o/ -a/ -e/; Pl.: -s: : a. pas-o – pas-os (m.) ‘step’ – ‘steps’ b. pas-a – pas-as (f.) ‘raisin’ – ‘raisins’

man-o – man-os (f.), ‘hand’ – ‘hands’; map-a – map-as (m.), ‘map’ – ‘maps’;



Typological correlations in nominal determination in Romance 

c. jef-e – jef-es (m.) ‘chief ’ – ‘chiefs’

nub-e – nub-es (f.), ‘cloud’ – ‘clouds’

Unlike Italian however, there seems to be no ‘classification potential’ in nominal (declensional) endings in the morphological setup of Spanish nouns; the only slight ‘classification potential’ left in Spanish is a kind of ‘neuter’ (deriving from the Latin neuter singular) in the pronominal system, marking ‘abstract antecedents’, such as quotations, matters of fact, etc. Spanish personal pronouns and demonstratives show a threefold morphological opposition, with forms ending in –e for masculine singular, –a for feminine singular and –o for the so-called ‘neuter’ (e.g. span.: este/esta/esto: lo que me interesa es esto...: ‘what I am interested in is the following:...’ vs. él que me interesa es este hombre, ‘who I am interested in is this man’). Turning finally to Modern Standard Romanian, there exist 4  main declension classes, 2 overtly marked genders and, again, overt number marking. Almost like in Italian/Spanish, Romanian nominal endings are heterogeneous with respect to gender, but relatively unambiguous with respect to number marking and they seem to preserve a certain ‘classification potential’ within the so-called Romanian ‘neuter’ (cf. Windisch 1973, Herslund 1976): (8) Sg.: -u/ -ă [A]/ -e/ -K; Pl.: -i/ -e/ -uri: a. membr-u – membr-i (m.) ‘member‘ – ‘members’ b. coleg-ă – coleg-e (f.) ‘colleague’ – ‘colleagues’ c. frat-e – fraţ-i (m.) ‘brother’ – ‘brothers’ d. coleg – coleg-i (m.) ‘colleague’ – ‘colleagues’

teatr-u (m) – teatr-e (f.), ‘theater‘ – ‘theaters’; sal-ă – săl-i (f.), ‘hall’ – ‘halls’; cart-e – cărţ-i (f.), ‘book’ – ‘books’; caiet (m.) – caiet-e (f.), tren (m.) – tren-uri f. ‘booklet’ – ‘booklets’ ‘train’ – ‘trains’

Singular nouns ending in –u (or consonant) with the respective plural in -uri (derived from the Latin neuter plural in –ora), and, less clearly, in –e, almost without exception indicate inanimate concrete objects or collectives, as opposed to nouns ending in –i (masculine or feminine plural).

4. Interesting correlations (1): Classification inside the noun phrase How can we relate these morphological findings to the problem of the different indefinite determiner systems in the Romance languages? Let us summarize the main differences in the noun morphology of the four Romance languages investigated and look for possible correlations with the respective systems of indefinite determiners:



Elisabeth Stark

Modern standard French shows a complete reanalysis, a complete loss of the Latin neuter plural in –a: Lat. neuter plural folia becomes Fr. la feuille, feminine singular, just like Lat. feminine singular femina becomes Fr. la femme. This loss of the Latin ‘classification system’ via noun morphology is compensated for by the evolution of an obligatory ‘classification system’ via indefinite determiners (cf. Herslund 1998: 70ff.): ‘zero’ is practically excluded in argument position; the indefinite singular article, un, marks ‘contour’, ‘individualized referent’, and thus countability; the ‘partitive article’, du, marks ‘substance’, ‘diffuse’ (mass / abstract), and thus non-countability: (9) a. Lat.: caseus, Fr. un fromage: ‘one single (piece of) cheese’ b. Lat.: caseum, Fr.: du fromage: ‘cheese as a substance’ Modern standard Italian shows some residue of the Latin neuter plural in -a with a certain ‘classification potential’, but it also has a French-like ‘classification system’ via indefinite determiners: ‘zero’ is partially permitted, but exclusively only for abstract / plural noun phrases; the indefinite article, uno, marks ‘contour’, ‘individualized referent’, and thus countability, just as in French; the ‘partitive article’, del, less grammaticalized than in French, marks ‘substance’ (‘mass’, as opposed to ‘abstract’) and thus noncountability. The situation of these central Romance languages differs considerably from the situation in Modern Standard Spanish and Romanian. Here, we find partial preservation of the Latin neuter (in Spanish within the pronominal system, indicating ‘abstract’, in Romanian with the nominal ‘neuter’ meaning ‘inanimate’/’collective’), but no clearcut distinctions in the indefinite determination system. ‘Zero’ can mean ‘abstract’ / ‘mass’, even (rarely) “entity-denoting”, besides the additional possibility of marking non-specificity (cf. Laca 1999); the indefinite article, un(o), less grammaticalized than in French or Italian, marks ‘contour’, ‘individualized referent’ and thus countability, but there is no explicit marking of non-countability and therefore no unambiguous simple obligatory classification system (cf. Herslund 1998: 70ff.). What we can try to formulate now is an answer to the first question in section 2: The rise of indefinite determination in the Romance languages can be related to the loss of the complex Latin nominal morphology which indicated, among other things, the conceptually fundamental difference between a contoured and shaped individual, and diffuse substances/masses and collectives (the former Indo-European genderbased opposition between animate and inanimate entities had been obscured already in the Latin system). In Latin, the same lexical root could appear with different gender and number when denoting one or the other type of entities. The overall Romance indefinite article derived from the Latin numeral unus, ‘one’, originally seems to indicate ‘nominal classification’ in a broad sense, although at a higher level within the nominal’s syntactic structure (probably NumP or PlP, cf. Delfitto/Schroten 1991 and especially Heycock & Zamparelli 2003), which indicates an (ongoing) ‘countability grammaticalization’ in Romance (see also the reduction of the different meanings of the Latin plural to an exclusively additive reading, cf. Meisterfeld 2000). Whereas Latin



Typological correlations in nominal determination in Romance 

nouns could have a phonologically expressed feature not only of (syntactic) PLUR(AL), just like most of the modern Romance languages, they could also have a phonologically expressed feature of ‘semantic PLURAL’ = COUNT, which covers mainly the difference between a countable and a non-countable interpretation of the NP. This second semantic feature has lost its overt phonological realization on the noun (N- or NPlevel) via declensional endings, requiring thus an explicit indefinite determiner in a higher position than N in order to check the syntactic and semantic plural properties of the whole nominal. Now we can also formulate an answer to the second question in section 2: (Un) ambiguous plural and thereby (non-)countability marking in Romance correlates with the presence or absence of a ‘partitive classifier’.9 This is the main difference, for example, between the morphological set-up of Italian and Spanish nouns: Whereas the latter is marked by the overt, independent affix-like and unambiguous expression of [+PLUR], the former unambiguos plural-affix thus requiring explicit ‘determination’ via uno or del at least for the ‘semantic plural-feature’ [COUNT]. (Un)ambiguous plural and thereby (non-)countability marking does not correlate directly with obligatory determination. On the contrary, it is the complete loss of the Latin neuter that correlates with the development of an obligatorily explicit (indefinite) nominal determination (compare French with its necessity to mark both [PLUR] and [COUNT] or ‘syntactic’ and ‘semantic plural’ via determiners as against Italian, Romanian and Spanish). With these findings in mind, we can now turn to the remaining question 3: How does “Differential Object Marking”, which exists in Spanish and Romanian, but not in French and Standard Italian, fit into this picture?

5. Interesting correlations (2): “Correlative typology” and classification outside the noun phrase Körner (1987) observed a clear-cut correlation between the existence of a ‘partitive article’ and the existence of DOM in Romance. Whereas languages without “Differential Object Marking” like French, Occitan and Standard Italian possess a ‘partitive article’, the standard languages of Portuguese, Romanian and Spanish, together with several non-standard varieties and dialects, all of which show DOM, do not have any ‘partitive’ article.

9. Compare (Old) Spanish with (Old) French or Occitan, cf. Buridant 2000: 73, 108, for Old French, Schultz-Gora 61973: 65ff., for Old Occitan, Zauner 21921: 56f., Penny 1993: 116-123, Lloyd 1987: 153f., for Old Spanish; see also certain Central-Southern varieties of Italian with an unambiguous morphological ‘neuter’ = mass declensional ending and without any partitive article, cf. Hall 1968, Delfitto/Schroten 1991: 167.



Elisabeth Stark

“Differential Object Marking” here means the phenomenon of selective prepositional case marking of the direct object according to certain lexical and / or semantic features of the intended referent or noun phrase: (10) He visto (a) un hombre ingles con sombrero. (I saw an English man with a sombrero) In (10), the insertion of a strongly favors a specific reading of un hombre ingles, whereas the noun phrase without a can only have a non-specific reading (cf. Leonetti 2003: 70–76, for a detailed discussion of a as a possible specificity marker in Spanish. Now, based on this empirically valid observation that DOM correlates with the presence or absence of the partitive article, Körner (1987) interprets Sp. a or Rom. pe as markers of potential subject noun phrases (because of their lexical semantics or properties of their referents10) that have the function of direct objects in a concrete sentence, i.e. as a sort of structural or syntactic device of disambiguation (cf. Körner 1987: 42). And assuming a corresponding ‘mirror function’ to DOM, the (French) ‘partitive article’ would act as a marker of noun phrases that cannot be subjects. However, that this second generalization cannot be true is immediately shown by examples (11) and (12): (11) a. b. c.

Il y a de l’argent dans le portefeuille. *De l’argent est dans le portefeuille. (There is money in the wallet) BUT: ? Un franc est dans le portefeuille. (There is one franc in the wallet)

(12) Du beurre était en train de fondre sur la table (Butter was melting all over the table) (11c) demonstrates that the impossibility of putting de l’argent in preverbal subject position is not due to the determiner du, because un franc (with the French indefinite article un) is also odd in this position. Nonetheless, this restriction is not to be related to the kind of indefinite determiner in a subject NP, but to the kind of predication with a stative verb without any temporal specification or anchoring of the described event (compare (11c) to (12) which is perfectly fine, cf. Dobrovie-Sorin 1999: 173ff., Bosveld-de Smet 2000). The following examples illustrate the potential selectional restrictions and/or the triggering factors for DOM in Modern Spanish:

10. Due to identical properties as to animacy etc. of ‘I’ and ‘an English man’ in example (10), both noun phrases, pro in subject position and un hombre ingles... in direct object position could be subjects of a verb like Sp. ver, ‘to see’.



Typological correlations in nominal determination in Romance 

(13) a. b. c.

Busco a un camarero (preferrably specific) Busco un camarero Busco (*a) camarero (preferrably non-specific) (cf. Leonetti 2003: 71) (I am looking for a (new) waiter)

(14) a. b.

Busco (*a) coche (non-specific) (I am looking for a car) Busco (*a) agua (I am looking for water)

(15) a. b. c.

Está buscando a alguien (He is looking for somebody) No está buscando a nadie (He is not looking for anybody) No está buscando (*a) nada (He is not looking for anything) (Leonetti 2003: 73)

(16) Un adjetivo califica a un sustantivo (An adjective modifies a noun) (Torrego Salcedo 1999: 1788) (17) a. b.

*No conozco a candidatos (I do not know any candidates) No conozco a candidatos con esas características (I do not know any candidates with these characteristics)

(18) Este profesor admite (*a) ofensas pequeñas (This teacher accepts small offences) The examples in (13) show specificity effects of DOM in Spanish – in these sentences, the use of a is optional and marks or strongly favours a specific reading of un camerero (‘a certain waiter’). (13c) and (14) – see also (17a) – show that a is impossible whenever an indefinite noun phrase is bare and not even modified by adjectives, relative clauses etc. – bare noun phrases like this are admitted in certain intensional contexts in Spanish, e.g. after the verb buscar, ‘to look for’, for “entity-denoting nouns” and for “mass-denoting nouns”. This could indicate that a would be a specificity marker, but the examples in (15) and (17b) clearly show that (pro)nominals with a non-specific reading (e.g. nadie, ‘nobody’) or nominals without a clear-cut indication of (non-) specificity like candidatos con esas características in (17b) can, or even must, be marked by a – whenever they denote animate (human) beings. The fact that adjectival attribution does not in general save the construction is illustrated in (18): abstract nouns occurring as bare indefinites in direct object position can never be marked by a. Cases like (16) present serious difficulties for all theories that consider ‘animacy’ as the prime feature triggering DOM in Spanish; in fact, they seem to corroborate Körner’s disambiguation hypothesis (see above): adjetivo and sustantivo denote both referents with



Elisabeth Stark

identical positions on any presumed ‘animacy scale’11, and a in front of un sustantivo seems to mark the direct object (recall the relatively free word order in Spanish). Without going into detail (for recents accounts of DOM in Spanish and other languages see Torrego Salcedo 1999: 1784ff., 1790ff., Aissen 2003, Leonetti 2003, Næss 2004), it looks like “differential object marking” is always related to the relative degree of affectedness and control of the nominal arguments in a sentence (cf. Næss 2004). This is indicated by examples (13) to (18) above, and furthermore by several other factors governing DOM: DOM can in fact be obligatory with (animate) direct objects after certain verb classes (as with Sp. atacar/insultar, ‘to attack’, ‘to insult’ vs. optional DOM with Sp. encontrar/ver, ‘to meet’, ‘to see’) and it sometimes reflects specificity and/or topicality of the respective referents in direct object position (cf. Leonetti 2003: 76ff., Şora 2002: 360ff., and especially Farkas / von Heusinger 2003 for Romanian): What is at stake here […] is the emphasis on the individualization of the referent triggered by a, compared to the emphasis on quantity or descriptive content that predominates in unmarked objects. (Leonetti 2003: 80; similarly Torrego Salcedo 1999: 1789+1793ff.).

(19) a. b.

Estaba dibujando a una niña (He was portraying a girl) Estaba dibujando una niña (He was drawing a girl) (Leonetti 2003: 80)

(19) presents a ‘minimal pair’ which clearly shows the semantic contribution of a to the sentence: Only when marked with a, does the direct object NP refer to a single, autonomous entity affected by the action that is denoted by the verbal predicate, whereas the omission of a in (19b) licenses a weak reading relating to something similar to ‘semantic incorporation’ (indefinite NPs with predicative readings, unspecified for animacy etc., cf. Van Geenhoven 1998, Leonetti 2003; for similar remarks see also cf. Torrego Salcedo 1999: 1800, similar remarks concerning Romanian pe can be found in Şora 2002: 359+362f.). Moreover, the diachronic evolution of DOM in Spanish (cf. von Heusinger / Kaiser 2005) seems to confirm this interpretation: DOM starts with highly individualized referents denoted by personal pronouns or proper names, and subsequently (from the 12th century on) spreads towards definite topical NPs, before also marking indefinite specific or topical referents in the sense described above at a later stage. However, the current state of affairs is that DOM never marks non-specific, i.e. not clearly individualized or contoured referents (this also holds true for some modern varieties of American Spanish, which seem to allow DOM also with indefinite NPs referring to inanimate objects – they all appear at least in clearly countable NPs, cf. von Heusinger / Kaiser 2005). But even if DOM in Spanish maybe did not start out 11. Cf. von Heusinger / Kaiser 2005 for a discussion and presentation of possible ‘animacy scales’ for Spanish.



Typological correlations in nominal determination in Romance 

as a marker of contoured referents in the first place, its actual distribution in indefinite object NPs could now be interpreted as a metaphoric shift from the original meaning with definite NPs denoting ‘highly affected referent + maximally identifiable’ (i.e. high on the definiteness scale) towards ‘individualized, contoured referent’ with indefinite NPs, in which classification is crucial, because in contrast to definite NPs the identification of the intended referent is impossible from the hearer’s point of view. Let us summarize: DOM in Spanish marks almost exclusively animate, more precisely non-abstract or non-mass-like referents in definite and / or specific (but not exclusively) and topical direct objects. Moreover, the higher a verb’s transitivity (“other directed”, cf. König 1999, Næss 2004: 1191), the more DOM becomes grammaticalized. In addition, whenever its use is optional, DOM marks single, individualized and autonomous referents with stable referentiality, i.e. referents that have to be considered at least as contoured, shaped entities (recall 16 and 19). These generalizations are also valid to almost the same extent for direct objects in Romanian (marked with pe), albeit exhibiting some syntactic differences (more obligatory clitic-doubling with DOM) as well as some special semantic triggering conditions like the denotation of N: With certain Ns denoting either individuals or social roles, DOM is disallowed with coinciding role-denotation: (20) In America, daca închiriezi un apartament şi ai vreo problema, trebuie să contactezi (?pe) proprietarul. (In America, whenever you rent an apartment and have any problem, you have to contact the owner) This also supports the interpretation of DOM in modern Romance languages as a classification device in a broad sense, indicating a ‘contoured object’. With these generalizations in mind, we can now try to give an answer to question 3 (recall section 2): DOM in peripheral Romance languages seems to be functionally parallel to the complex system of indefinite determiners in central Romance languages, at least in direct object position, which is the most important position for the development of determiners or nominal determination (cf. Leiss 2000). In addition, DOM might be understood – just like the opposition between ‘zero’, ‘partitive’ and ‘indefinite article’, e.g. in Italian12 – in terms of ‘nominal classification’ and surfaces or specializes in this direction in exactly those Romance languages lacking a sufficient complex indefinite determiner system. DOM nowadays marks explicitely individualized referents; in Modern Spanish and Modern Romanian, a / pe encodes the “instruction to process 12. Compare the following examples from Delfitto/Schroten (1991: 160), one without the ‘partitive plural’ and one with the ‘partitive plural’ in the direct object, clearly indicating ‘shaped, individualized objects’: Gianni ha venduto libri solo per cinque minuti (with a possible meaning: ‘Gianni has been a bookseller only for five minutes’, even if he did not sell a single book) vs. Gianni ha venduto dei libri per cinque minuti meaning only ‘Gianni has been selling some books for five minutes’.

 Elisabeth Stark

the object DP as a [...] prominent and referentially autonomous argument” (Leonetti 2003, 84) – via classification and/or determination.

6. Conclusion The beginning of this article raised two points concerning current structural differences within the family of Romance languages despite their common historical development from Latin: Firstly, why is there considerable variation between the systems of indefinite determiners among Romance languages, and secondly, why is “differential object marking” found only in peripheral Romance languages? Since previous approaches based exclusively on overt number marking have to be considered insufficient, this paper presents a different, more comprehensive proposal. At first we reconsidered Latin nominal morphology and discovered a defective, but still functionally valuable system of ‘nominal classification’ via declensional affixes that make particular use of oppositions in gender and number (masculine/feminine vs. neuter, neuter plural in –a vs. other plural endings). The subsequent loss of this ‘classification system’ had different results in the Romance languages investigated: It led either to various complex systems of indefinite nominal determiners (where the Latin neuter has been almost completely lost, e.g. in French, and to a smaller extent, Standard Italian), or to the development of a device to differentially mark direct objects as autonomous, shaped entities vs. abstract, mass-like entities with special emphasis on their descriptive content (in languages that preserve overt number and even gender marking to some degree, e.g. Spanish and Romanian). What remains to be done now is a detailed diachronic description of the different stages of grammatical change from Latin to Romance from this new perspective, re-evaluating data from older stages of French, Italian, Spanish and Romanian, ultimately discovering possible grammaticalization paths related to the conceptually basic dimension of apprehension, or to put it more precisely, nominal classification.

References Aissen, J. 2003. Differential object marking: Iconicity vs. economy. Natural Language and Linguistic Theory 21: 435–483. Allan, K. 1980. Nouns and countability. Language 56(3): 541–567. Behrens, L. 1995. Categorizing between lexicon and grammar. The MASS/COUNT distinction in a cross-linguistic perspective. Lexicology 1(1): 1–112. Bossong, G. 1997. Le marquage différentiel de l’objet dans les langues d’Europe. In Actance et Valence dans les Langues d’Europe, Jack Feuillet (ed.), 193–258. Berlin: Mouton de Gruyter.



Typological correlations in nominal determination in Romance  Bosveld-de Smet, L. 2000. Les syntagmes nominaux en des et du: un couple curieux parmi les indéfinis. In De l’indétermination à la qualification – les indéfinis, L. Bosveld-de-Smet, M. Van Peteghem & D. van de Velde (eds), 17–116. Artois: Artois Presses Université. Buridant, C. 2000. Grammaire nouvelle de l’ancien français. Paris: Sedes. Chierchia, G. 1998. Reference to kinds across languages. Natural Language Semantics 6(4): 339–405. Corbett, G. 1991. Gender. Cambridge: CUP. Corbett, G. 2000. Number. Cambridge: CUP. Delfitto, D. & Schroten, J. 1991. Bare plurals and the number affix in DP. Probus 3(2): 155–185. Dobrovie-Sorin, C. 1999. Le(s) thème(s) entre la syntaxe et la structure de l’information. In La thématisation dans les langues. Actes du colloque de Caen, 9–11 octobre 1997, C. Guimier (ed.), 169–183. Bern: Peter Lang. Farkas, D.F. & von Heusinger, K. 2003. Stability of reference and object marking in Romanian. Talk given at the Workshop on specificity and direct reference (ESSLLI). Vienna. Geenhoven, V. van. 1998. Semantic Incorporation and Indefinite Descriptions. Semantic and Syntactic Aspects of Noun Incorporation in West Greenlandic. Stanford CA: CSLI Publications. Gil, D. 1987. Definiteness, noun phrase configurationality, and the count-mass distinction. In The Representation of (In)definiteness, E.J. Reuland & A.G.B. ter Meulen (eds), 254–269. Cambridge MA: The MIT Press. Hall, R.A. 1968. ‘Neuters’, mass-nouns and the ablative in Romance. Language 44(3): 480–486. Harris, J.W. 1992. The form classes of Spanish substantives. Yearbook of Morphology 1991, G. Booij & J. van Marle (eds), 65–88. Dordrecht: Kluwer. Haspelmath, M. et al. (eds). 2001. Language typology and language universals / Sprachtypologie und sprachliche Universalien / La typologie des langues et les universaux linguistiques. An International Handbook / Ein internationales Handbuch / Manuel international, 2 Vols. Berlin: Mouton de Gruyter. Herslund, M. 1976. Encore le ‘neutre’ en roumain. Revue roumaine de linguistique 13(1): 251–254. Herslund, M. 1998. Le français, langue à classificateurs? In La ligne claire. De la linguistique à la grammaire. Mélanges offerts à Marc Wilmet à l’occasion de son 60e anniversaire, A. Englebert et al. (eds), 65–73. Paris: Duculot. Heusinger, K. von & Kaiser, G. 2005. The evolution of differential object marking in Spanish. In Specificity and the evolution/emergence of nominal determination systems in Romance. Selected papers from the international workshop NEREUS II, Oct. 2004 in Berlin, E. Stark, K. von Heusinger & G. Kaiser (eds), 33–69. Kontanz: Universität Konstanz. Heycock, C. & Zamparelli, R. 2003. Friends and colleagues: Plurality, coordination, and the structure of DP. Ms. University of Edinburgh/Università di Bergamo. Hofmann, J.B. 1997. Lateinische Syntax und Stilistik: mit dem allgemeinen Teil der lateinischen Grammatik. Neubearbeitet von Anton Szantyr. (2nd reprint of the 1st edn of 1965/1972). München: Beck. Iturrioz Leza, J.L. 1986. Individuation and determination III: The concept of verbal plurality and the pluralization of abstractives. Función 1(2): 201–308. Kleiber, G., Laca, B. & Tasmowski, L. (eds). 2001. Typologie des groupes nominaux. Rennes: Presses Universitaires de Rennes. König, E. 1999. From expressions for body parts to reflexive anaphors: Semantic change in the development of intensifiers. In Interdigitations: Essays for Irmengard Rauch, G.F. Carr, H. Wayne & L. Zhang (eds), 503–517. Frankfurt: Lang.



Elisabeth Stark Körner, K.-H. 1987. Korrelative Sprachtypologie. Die zwei Typen romanischer Syntax. Stuttgart: Steiner. Krifka, M. 1991. Massennomina. In Semantik. Ein internationales Handbuch der zeitgenössischen Forschung, A. von Stechow & D. Wunderlich (eds), 399–417. Berlin: Mouton de Gruyter. Kühner, R. & Stegmann, C. 1955. Ausführliche Grammatik der lateinischen Sprache. Satzlehre. Erster Teil. Leverkusen: Gottschalksche Verlagsbuchhandlung. Laca, B. 1999. Presencia y ausencia de determinante. In Gramática descriptiva de la lengua española, Vol.1, Sintaxis básica de las clases de palabras, I. Bosque & V. Demonte (eds), 891–928. Madrid: Espasa. Lehmann, C. 1991. The Latin nominal group in a typological perspective. In New Studies in Latin Linguistics. Selected Papers from the 4th International Colloquium on Latin Linguistics, Cambridge, April 1987, R. Coleman (ed.), 203–232. Amsterdam: John Benjamins. Leiss, E. 2000. Artikel und Aspekt. Die grammatischen Muster von Definitheit. Berlin: Mouton de Gruyter. Leonetti, M. 2003. Specificity and object marking: The case of Spanish a. In Proceedings of the Workshop “Semantic and Syntactic Aspects of Specificity in Romance languages”, K. von Heusinger & G. Kaiser (eds), 67–101. Konstanz: Fachbereich Sprachwissenschaft der Universität Konstanz. Link, G. 1991. Plural. In Semantik. Ein internationales Handbuch der zeitgenössischen Forschung, A. von Stechow & D. Wunderlich (eds), 418–440. Berlin: Mouton de Gruyter. Löbel, E. 1993. On the parametrization of lexical properties. The Parametrization of Universal Grammar, G. Fanselow (ed.), 183–199. Amsterdam: John Benjamins. Lloyd, P.M. 1987. From Latin to Spanish, Vol. 1: Historical Phonology and Morphology of the Spanish Language. Philadelphia PA: American Philosophical Society. Longobardi, G. 2001. The structure of DPs: Some principles, parameters and problems. In The Handbook of Contemporary Syntactic Theory, M. Baltin & C. Collins (eds), 562–603. Oxford: Blackwell. Lyons, C. 1999. Definiteness. Cambridge: CUP. Meisterfeld, R. 1998. Numerus und Nominalaspekt. Eine Studie zur romanischen Apprehension. Tübingen: Niemeyer. Meisterfeld, R. 2000. Die unbestimmte Bestimmung: Zur Entstehung des unbestimmten Artikels in den romanischen Sprachen. In Linguistica romanica et indiana. Festschrift für Wolf Dietrich zum 60. Geburtstag, Bruno Staib (ed.), 303–332. Tübingen: Narr. Morani, M. 2000. Introduzione alla linguistica latina. München: Lincom. Næss, Å. 2004. What markedness marks: The markedness problem with direct objects. Lingua 114: 1186–1212. Penny, R. 1993. Gramática Histórica del español. Barcelona: Editorial Ariel. Ralli, A. 2002. The role of morphology in gender determination: Evidence from modern Greek. Linguistics 40(3): 519–551. Rijkhoff, J. 2002. The Noun Phrase. Oxford: OUP. Schön, I. 1971. Neutrum und Kollektivum. Das Morphem -a im Lateinischen und Romanischen. Innsbruck: Institut für Vergleichende Sprachwissenschaft der Universität Innsbruck. Schroten, J. 2001. L’absence de déterminant en espagnol. In Typologie des groupes nominaux, G. Kleiber, B. Laca & L. Tasmowski (eds), 189–203. Rennes: Presses Universitaires de Rennes. Schultz-Gora, O. 1973. Altprovenzalisches Elementarbuch. Heidelberg: Winter.



Typological correlations in nominal determination in Romance  Seiler, H. 1986. Apprehension. Language, Object, and Order. Part III: The Universal Dimension of apprehension. Tübingen: Narr. Seiler, H. & Lehmann, C. (eds). 1982. Apprehension. Das sprachliche Erfassen von Gegenständen. Teil I: Bereich und Ordnung der Phänomene. Tübingen: Narr. Stark, E. 2002. Indefiniteness and specificity in old Italian texts. Journal of Semantics 19: 315–332. Şora, S. 2002. L’Objet direct pronominal en roumain. In Roma et Romania. Festschrift für Gerhard Ernst zum 65. Geburtstag, S. Heinemann, G. Bernhard & D. Kattenbusch (eds), 359–369. Tübingen: Niemeyer. Torrego Salcedo, E. 1999. El complemento directo preposicional. In Gramática descriptiva de la lengua española, Vol. 2: Las construcciones sintácticas fundamentales, relaciones temporales, aspectuales y modales, I. Bosque & V. Demonte (eds), 1779–1805. Madrid: Espasa. Velinova, M. 1989. Quelques notes sur les déterminants nominaux edin et un en bulgare et en roumain. Contrastive Linguistics (SEZIK) 14(6): 25–29. Wanner, D. 2001. From Latin to the Romance languages. In Language typology and language universals / Sprachtypologie und sprachliche Universalien / La typologie des langues et les universaux linguistiques. An International Handbook / Ein Internationales Handbuch / Manuel international, Vol. 2, M. Haspelmath et al.  (eds), 1691–1706. Berlin: Mouton de Gruyter. Windisch, R. 1973. Genusprobleme im Romanischen. Das Neutrum im Rumänischen. Tübingen: Narr. Zauner, A.2 1921. Altspanisches Elementarbuch. Heidelberg: Winter.

A stranger in the house The French article de Marc Wilmet

Université Libre de Bruxelles The present contribution argues in favour of the recognition of an article de in French which occurs either alone or with another article. The approach adopted calls for an historical recapitulation, i.e. “where does the article come from?”, a theoretical investigation, i.e. “what is an article?”, and the identification of an inventory, i.e. “what qualifies as an article?”. Finally, a table is provided which summarizes the articles in terms of three pairs of oppositions, viz. part vs. whole, mass vs. count, continuous vs. discontinuous. The article de is definable as a partitive article, sometimes mass or count, continuous or discontinuous (where the form is de), sometimes continuous mass (where the form is du, de la), sometimes discontinuous mass or count (where the form is des), sometimes continuous mass and count (where the form is d’un, d’une).

Inspired by the title of a famous novel by Simenon this article sets out to contribute to the recognition of a French article of the type de (i.e. of a type which unites the forms de, du, de la, des, d’un, d’une), which calls for (1) a brief historical recapitulation; (2) a brief theoretical overview; and (3) an inventory of articles. This should give us the conditions for stating with precision (4) how the article de functions within the system.

1. Historical recapitulation The name article is derived from Latin articulus ‘little articulation’, which is the diminutive of artus ‘limb’, which in itself is derived from Greek arthron ‘joint’. The paradox is that the articles are not known in Latin. The first grammarians of French (e.g. Bacon, quoted in Lusignan 1987: 27) had to abandon the idea of a ‘preposed inflection’:

LE maître = nominative dominus ô maître = vocative domine LE maître = accusative dominum DU maître = genitive domini



Marc Wilmet

AU maître = dative domino AVEC/PAR/SANS … LE maître = ablative domino … before they could identify the definite forms le, la, les (Meigret, 1550) by drawing on the Greek articles ο, η, το – which conveniently supported the idea of a ‘noble’ connection of the common language. It took some time for the class of articles to become constituted. The numerals un, une leave the homonymous articles in the shadow (English a vs. one was more fortunate). In the 18th century, Beauzée frankly admitted his confusion (1767, I: 390–391): “I admit that I do not understand why un does not always mark un, nor why sometimes it denotes a definite entity and sometimes an indefinite entity …”. The indefinite articles are not given their official name until 1910. The ‘partitives’ du, de la, des (de) in (1) are still not today consistently separated from the contracted articles in (2), i.e. the merged preposition de + le, la or les. And the zero article in (3), which was long considered to be ‘absence’ or ‘omission’ of the article, further obscured the picture.

(1) Pierre désirerait du poisson, de la viande, des fromages et de bons vins (verb désirer, requires a direct object).



(2) Pierre a envie du poisson, de la viande, des fromages et des bons vins du menu dégustation (the verb construction avoir envie de requires an indirect object).



(3) Pierre a envie de poisson, de viande, de fromages et de bons vins.

Twentyfive years ago Kupferman asked himself: “Is there a partitive article?” (1979), and recently Vandeloise went even further: “Living without a partitive” (2006). The preface of Grammaire française by Noël (1861) is still relevant (except stylistically): “It would be reasonable to say ‘the article is not known in France!’ So what is this strange pixie which has sneaked into the modern languages to torment grammarians [ … ]? What is this spirited joker? What does it do? Why is it here but not there? Why one form here but another there? Why is it sometimes present with a word, sometimes absent?” Whatever underlies these uncertainties, French pre-war analyses of definite and indefinite articles were introduced into to the Anglo-Saxon world by the Danish Anglicist Christophersen, who while working on his doctoral dissertation in Paris became acquainted with work by Guillaume (1919) and Damourette et Pichon (1927); for instance he writes “Guillaume’s book is extremely interesting and fertile in fresh ideas which have profoundly influenced my own view” (1939: 56–57).

2. Theoretical overview What is an article? The writer of this paper has on a number of occasions sought to answer this question, successively in ‘bilan critique’ (1980) and in ‘essai de synthèse’



A stranger in the house 

(1983); then in a book on nominal determination (1986), and, finally, in a long chapter in Grammaire critique du français (2003, 3rd edn. §§ 98–101 and §§ 131–193). With the perspective of distance and a bird’s eye view, the nature of an article will be exposed as four points (for a more thorough treatment – and the vast literature on the subject – the reader is referred to the works quoted above). Point No. 1 The article is an adjective, or, etymologically, an adjunct (from Latin adjicere ‘adjoin’). In this there is nothing new or provocative (in spite of the analysis which incorporates the article into the determiners and thereby excludes them from the adjectives): this has been the position of both Beauzée (1767) and Bloomfield (1933). In order to clarify, let us take the noun menuisier (joiner). This noun encompasses virtually everyone whose occupation is to work with wood. The denotational scope of the word, called its extension, depends on its sense, also called its intension. Constrained by the intension ‘someone who works with wood’ which covers everyone who works with wood or is a joiner, the extension of the word menuisier arises automatically, it is direct. We can now compare this with the words chauve (bare) or courir (to run). Their extension covers everyone who is bare (be it a joiner, a mountain or a mouse) or who is running (be it a joiner, water, rumour, … ): the extension is indirect (obtained through joiners, mountains, water, etc.). Let us hold onto the notion that the extension denotes everyone in the world to whom the words are applicable, directly through extension, or indirectly. On this basis we can characterize nouns as words which have direct extension – in fact the only word type with direct extension – and adjectives, e.g. chauve, and verbs, e.g. courir, as words which have indirect extension. In other words, outside discourse, adjectives are adjoined words which require support. Adjectives share this quality with verbs (which carry their own morphology) and they differ from nouns in terms of this quality (with which they share morphology). Point No. 2 Adjectives are supported in discourse, where they become determiners (DET) of noun heads (NN), together with which they form noun phrases. (It should be noted that NN is not necessarily a noun: some DETs have the ability to nominalize an adjective, a verb, a pronoun, an adverb, a preposition, a phrase, e.g. Souffler le chaud et le froid; Des rires clairs; L’hypertrophie du moi; Il n ‘y a plus d’après à Saint-Germain-des-Prés; Peser le pour et le contre; Se fâcher pour un oui pour un non; … ). Point No. 3 The DET fulfils a determiner function relative to NN. It is a double function. Let us consider a couple of examples:

(4) Les Français d’extrême droite et d’extrême gauche n’ont pas voté Chirac en 2002.



Marc Wilmet



(5) Plus de 80% des Français ont voté Chirac en 2002.

About the French as a whole the DET d’extrême droite et d’extrême gauche supplies information about the extension of the NN Français, or, as mentioned above, all the individuals to whom the word applies: DET reduces the potential referents to the subset of Frenchmen who are political extremists. The DETs les in example (4) and plus de 80% des in example (5) provide information, not about all individuals to whom the word Français applies, but about the individuals about whom it has been used or, with a technical term borrowed from Guillaume, extensité, i.e. the extensity, of NN Français (in the case of the determiner plus de 80% des) or the noun group (GN) Français d’extrême droite et d’extrême gauche (in the case of the determiner les). A DET which signifies extensity, as les or plus de 80% des, is a quantifier. A DET which signifies extension, as d’extrême droite et d’extrême gauche, is a qualifier. We might add that a quantiqualifier simultaneously provides information about the extensity and the extension of NN, e.g. ce or mon: ce livre = ‘le livre [quantifier] + que je désigne [qualifier]’, mon livre = ‘le livre [quantifier] + que je possède [qualifier]’, etc. Point No. 4 The quantifier DET comes from the ‘article’, the ‘cardinal numeral’ and the ‘indefinite adjectives’ of the school grammar (which as a result of the influence of American distributionalism have become determiners in new grammars). Three groups of adjectives are easily identified: a. Numeral quantifiers which signify precise extensity, e.g. {deux, troix, quatre,...} hommes se promènent dans la rue; a number of n men = 2, 3, 4, etc. (and as numbers: Attendre deux minutes; Donner quatre bouts de bois; Voir trente-six chandelles; Faire les quatre cents coups; Souffrir mille morts, etc.) b. Sectional quantifiers which signify a vague zone, e.g. {quelques, plusieurs, …} hommes se promènent dans la rue; the extensity n is higher than 1 and lower than totality t). The totality t may be the determined NN or all the elements which include the determined NN, e.g. Qui sait encore effectuer aujourd’hui une règle de trois? Quelques mathématiciens! = all mathematicians, but actually a small number of individuals, or; Parvenu au terme des W de sa liste et croyant en avoir fini, le professeur tourna une page et s’aperçut qu’il lui restait plusieurs Z à interroger = all candidates with the initial Z. c. Transverse quantifiers (transverse in a strict sense, i.e. cutting across perpendicularly) signify the extensity of the end-points, e.g. L’homme est mortel = all human beings, or Un enfant est toujours l’ouvrage de sa mère = all children: n = t, as against J’ai revu l’homme au chapeau melon or Un homme entra … = an individual: n = 1. Summing up, we define the article as an adjective by nature, and by function a transverse determiner quantifier.



A stranger in the house 

3. Inventory The French articles are divisible into four morphological types: (1) the zero type (form Ø), (2) the LE type (forms le, la, les), (3) the UN type (forms un, une), (4) the DE type (forms de, du, de la, des, d’un, d’une). The zero type The linguistic concept of a zero sign (Ø) is an invention of phonology. We can verify the occurrence of zero as a transverse quantifier with the examples (6) and (7) (under certain conditions commutation with a surface article is possible in this type):

(6) Maison à vendre [refers for instance to the house at 22 rue du Labrador: extensity = 1].



(7) Pierre qui roule n’amasse pas mousse [refers to all rolling stones: maximal extensity = t].

In other types French has a zero article which is: (a) conservative (the result of a negative choice which perpetuates the situation in Latin and Old French, i.e. proverbs, archaisms, proper names, etc.) and (b) progressive (the result of a positive choice which denies the NN its full nounhood, such as the verbal and adjectival expressions avoir besoin, prendre feu, agir avec courage, … ). The LE type Traditionally called the definite article. Like all transverse quantifiers, these are markers of negative or positive polar extensity. The masculine singular le, the feminine singular la and the common gender plural les signify the set of individuals in the world to whom NN can actually be applied and the exhaustive set of individuals in the discourse world concerned. This extensive property, i.e. the ability to equalize the extensity with the extension, distinguishes the LE type from the three other types. We can proceed in three steps: minimal extensity; middle extensity; maximal extensity. With minimal extensity in the singular forms le and la (and the internal plural les: as in les rillettes in (8)), the extensive character creates the effects of (a) familiarity (the NN is assumed to be known); (b) uniqueness (the NN is taken to be unique) – the combination of (a) and (b) undoubtedly shed light on the choice of a definite attributive.

(8) Pierre et Marie se baladaient en auto. La jeune fille suggéra à son compagnon de s’arrêter pour déjeuner. Ils trouvèrent un village, avec un restaurant sympathique situé devant le monument aux morts. On servit en hors d’œuvre des charcuteries. Pierre, maladroit, renversa une bouteille et le vin se répandit sur la nappe, inondant le saucisson, le pâté et les rillettes [la jeune fille, i.e. Marie, properly introduced; le monument aux morts, there is no village without a monument to the dead, and in principle there is one monument for each village; le vin, la nappe, le saucisson, le pâté, les rillettes = the wine of the



Marc Wilmet

overturned bottle, the cloth which covers the table at which the couple were seated; the sausage and the rillette served as starters. However, the familiarity may be overridden at text level, cf. (9):

(9) Tu as vu le Président à la télé? [the interlocutors share knowledge of the identity of the ‘Président’: be it of the Republic, of the university or of their philatelist society].

And the unique character flouts the objectively redundant extension in (10), (11) and (12): (10) Pierre s’est cassé la jambe [Pierre has two legs like everyone else, nevertheless he is not capable of moving – that is the important part of the message – irrespective of whether it is his right or left leg which has been broken]. (11) Marie prendra le bus pour rentrer [a means of transportation with a regular time table which operates in her area; cf. Marie prendra un taxi = any taxi, the driver will follow her instructions]. (12) Le feuilleté d’écrevisses et le gâteau aux trois chocolates [in gastronomical discourse the extensity automatically follows the extension = it is not necessary to look elsewhere, my recipes will make you forget the pale imitations]. With middle extensity, the plural les retains the effect of familiarity which may also be achieved by le and la with minimal extensity (e.g. J’ai revu l’homme au chapeau melon J’ai revu les hommes au chapeau melon: n = 1 + 1 + 1 … ) without encroaching on the singular le and la with maximal extensity: Les hommes sont mortels or L’homme est mortel. With maximal extensity the difference between le, la and les lies in conceptualization: (a) continuous with the singular; (b) discontinuous with the plural. Singular number levels the differences: Les droits de l’homme, La journée de la femme … and is blocked by fractions or enumeration:? Le tigre se divise en deux familles or? Pris un par un, l’homme est bon (vs. non-enumerating Pris individuellement, l’homme est bon = ‘taken out of the class of ’). The sentences Le chat est carnivore or Le chien aboie et le chiot jappe, etc. (all defining features of the classes of chats, chiens and chiots) are more natural than for example? L’enfant s’ennuie le dimanche or? L’ américain a marché sur la lune (two non-defining features of the classes of children and Americans) [as against: L’homme a marché sur la lune: a conquest ascribable to mankind]. The UN type Traditionally called the indefinite article. The masculine singular un and the feminine singular une signify that the number of individuals to whom the NN is applicable does not exhaust all the individuals of the discourse world. This partitive property – i.e. the ability to reduce the extensity under the extension – result in the effects that: (a) unfamiliarity (the NN is not assumed to be known); (b) plurality (the NN is not assumed to be unique).



A stranger in the house 

(13) Pierre et Marie se baladaient en auto. Brusquement, un pneu éclata [ = ‘one of the tyres’, vs. for instance Le moteur toussota]. (14) Marie veut épouser un Chinois [= (a) ‘a certain Tchang who is otherwise unknown’; (b) she explicitly wishes her future husband to be Chinese]. (15) Marie nous a présenté Pierre, un garçon qu’elle aime [ambiguous statement, the plurality could comprise the NN garçon or the NG garçon qu’elle aime (the addition of … et le seul garçon qu’elle aime would have prevented Marie from appearing like a capricious woman); which may be compared with Marie nous a présenté Pierre, le garcon qu’elle aime, in which Pierre is the exclusive beneficiary of Marie’s love]. In order to exceed the minimal extensity un and une require a trigger which presses them some distance from the totality t. Neither unfamiliarity nor plurality are affected. Examples (16) and (17) fall on one side of t, (18) and (19) on the other. (16) Presque tous les romanciers finissent par trouver un éditeur [= ‘editors’ (the quantifier presque tous les has consequences for the quantified un)]. (17) Chaque jour, à chaque heure, à chaque minute, à chaque seconde, une femme met au monde un enfant [ = ‘women give birth to children’ (i.e.: 1 × 24 × 60 × 60 = 86,400 births every day)]. (18) Un enfant est toujours l’ouvrage de sa mère [= ‘the children’ (this individual statement can be verified at will)]. (19) Pierre apprécie un chien quand il est bien dressé [ = ‘trained dogs’ (Pierre’s temper is such that a trained dog inevitably provokes an emotional response in him)]. Our paraphrases show that un and une, which in modern French are deprived of plural form (i.e. the pronouns quelques-uns, quelques-unes), not only borrow their semantic plural from des (de), which is the general state of affairs (e.g. Un livre est un ami/un précieux ami → Les livres sont des amis/de précieux amis), but also from les (e.g. Un livre est un ami → Les livres sont des amis) and – not to be forgotten – from Ø, e.g. Un poème, un roman et un livre sont de précieux amis → Poèmes, romans et livres sont de précieux amis. But what about t and the partitive character? It might be thought that the maximal extensity would make it. That is, however, not the case. Between le, la and un, une = ‘all’, the shades of meaning derive respectively from their extensive character (n ≡ t [i.e. ‘a quantity n equals the totality t’]) and their partitive character (n = 1 < t • x t [i.e. ‘a quantity n equals 1 and is smaller than the totality t multiplied by x so many times that there is 1 in t to complete the totality’]. If we compare for instance Le lapin est proli­ fique and Un lapin est prolifique, le presents the totality of lapin as against the external entities chat, éléphant, raton laveur …; un confronts internally, one by one – it is precisely the distributive mechanism of chaque and tout – all the lapins entities. At the



Marc Wilmet

same time le asserts a statistical truth and un asserts a so-called complete truth: Le Chrétien est charitable (some exceptions are acceptable) vs. Un Chrétien est charitable (an ungenerous soul works against Christian salvation = someone who is not charitable does not deserve to be called a Christian). The DE type Traditionally called partitive articles (masculine and feminine singular du, de la) and indefinites or partitives (masculine or feminine plural des and its variant de). It has been noted that de – in spite of its omnipresence in the series and its extensity scale which goes from the + pole (e.g. De bons vins ne sauraient nuire à la santé = ‘it is true that good wine is not unhealthy’) to the − pole (e.g. Pierre ne boit plus de vin = ‘Pierre doesn’t drink wine at all’) – appears as the eternal poor relation. It is necessary to review the situation.

4. The function of the article DE This time we will continue with three points. Point No. 1 De has the partitive property we also find in un and une. Its distinguishing feature relative to un and une is sometimes to be able to express a number n (e.g. Il n’y a plus de veau à l’étable = ‘no more animals’) and sometimes to be able to express a relative amount q (e.g. Il n’y a plus de veau à la boucherie = ‘no more meat’). Point No. 2 Partitive de distributes in its pure form in front of the pronouns aucuns, aucunes (de is obligatory), autres (d’autres, corresponds with extensive les autres), ça, cela, chacun, chacune (and very informal [boire, manger, goûter, tâter …] de chaque), tel (telle, tels, telles), lequel (contracted to duquel), laquelle, lesquels (contracted to desquels), lesquelles (contracted to desquelles), moi, toi, soi, nous, vous, lui, eux (elided form d’eux), elle (elided form d’elle), elles (elided form d’elles), personne, rien, tout, (n’importe) quoi, e.g. Il y a de ça, Ne manger de rien, Goûter de tout, de quoi amuser la galerie … An example from Émile Zola: (20) … dis-moi s’il est raisonnable qu’un simple magasin de nouveautés se mette à vendre de n’importe quoi. Point No. 3 In front of a noun the quantifying partitive de combines with another quantifier, either (a) zéro (de), or (b) extensive LE (du, de la, des), or (c) partitive UN (d’un, d’une). Here we can exclude the other combinations: (a) with a partitive, de and the part quantifiers Boire de plusieurs vins, Manger de chaque mets, Ne goûter d’aucun



A stranger in the house 

plat …; de and the amount qualifiers certain, quel, tel … e.g. in Comment Pierre ose-t-il proférer de telles âneries? (b) with extensive de and the quantiqualifiers mon, ces … (e.g. Maupassant: ‘Si vous le permettez, monsieur l’abbé, je vais vous offrir de mon parapluie’ = ‘shelter under my umbrella’, or Beckett: ‘La mémoire nous joue de ces tours!’ = ‘strange towers’).

The combination de + Ø: This combination is found in three environments: (a) in front of nouns which precede a qualifying adjective (e.g. Pierre boit de bons vins [and the pronouns which distribute together with partitive en, Pierre en boit de bons]), (b) after an expression of quantity (e.g. Pierre boit beaucoup de vin), (c) in negative phrases or those which refer back to a negative phrase, e.g. Pierre ne boit pas de vin or Marie ne pense pas que Pierre boive encore de vin). These three environments (a), (b) and (c) each results in a partition of the object: les bons vins only represents a sub-set of the wines which are offered on the market, and beaucoup de vin or pas de vin is less than – in terms of upwards or downwards approximation – the total amount of wine available. On the basis of this we can conclude that the step from (+) to (–) of the division blocks the development of an extensive article which matches the (+) to (–) of the extensity to those of the extension. (a) First environment: Singular and plural go in different directions. (i) The singular is almost only found in old expressions, e.g. faire de bonne musique or faire de bonne politique. In other words the existing conception of the singular blocks the division which we established above: Pierre boit du bon vin/de la bonne soupe, etc. A classic example is provided in (21), an actual but deviant example is provided in (22): (21) Eh bien! mon frère, qu’en dites-vous? Cela ne vaut-il pas bien une prise de casse? –Hom! de bonne casse est bonne (Molière). (22) Sa bouche est desséchée par d’autre faim (Marguerite Duras) [normally: d’autres faims or une autre faim). (ii) In the plural usage makes it possible to deviate from the rule (examples 23 and 24): (23) … elle sauve tout par de petites plaisanteries et des petits airs (Jules Vallès). (24) … elle a fait bouillir des pommes de terre; elle en a fait bouillir de vieilles, des grosses … (Jean Giono) [= ‘de vieilles pommes de terre et des grosses pommes de terre’]. The sub-set which corresponds to the GN qualifying adjective + noun head has a tendency to reconstitute itself in wholes, in particular when they constitute independent conceptions, e.g. des jeunes gens = ‘teenagers’, des vieilles filles = ‘seasoned singles’, de gros sabots = ‘ponderousness’ …., and if the qualifier does not have a quantifying sense



Marc Wilmet

which supports the partition: fréquents, nombreux, innombrables, multiples … (de fréquents/nombreux accès de fièvre are to be preferred to des fréquents/nombreux accès de fièvre, etc.), or if it derives its sense outside the SN (des autres/mêmes/pareils/semblables comportements … ). (b) Second environment: The practical and theoretical problems are (i) the separation of the quantifier from the preposition; (ii) the explanation of why the quantifier Ø sometimes does not offer resistance. (i) The influence of the quantifying expressions which has survived the rightwards dislocation of the constituent it forms with the noun: J’ai lu beaucoup de livres J’en ai lu beaucoup, de/des livres, but not leftward dislocation: des/*de livres, j’en ai lu beaucoup, are interpreted as articles; e.g. Pierre en a deux, de voitures; or Marie les a eues toutes, de maladies; or ‘Que j’en trouve encore une, de montre’ (Courteline) …, symmetrically dislocatable in Des voitures, Pierre en a deux; Des maladies, Marie les a eues toutes; Des montres, que j’en trouve encore une (as against for instance de the preposition in La sienne ne marche plus, de montre* Des montres, la sienne ne marche plus). (ii) The quantifier Ø yields its place to LE in three circumstances: – Bien, half way between quantification and qualification (e.g. Les commerçants ont bien travaillé durant les fêtes = ‘they have done a job which is quantitatively and qualitatively satisfactory’) is not treated as a quantifier: Bien du courage, Bien des gens …, in contrast to its quasi-synonym pas mal: Pas mal de courage, Pas mal de gens … (it is worth noting that in Belgian French bien is enforced by quantifying assez: Assez bien de courage, Assez bien de gens …. ). – Qualification of NN allows the combination de + le, la or les. Two literary examples: (25) … beaucoup des combattants que je rencontrai connaissaient de près mes deux livres de guerre (Jules Romains). (26) … bien que là encore beaucoup des horreurs de la fin de la guerre aient été déblayées (Romains).

This is not a question of sub-standard usage (as in Daudet, who uses quotation marks around du: ‘On a parfois beaucoup”du” tourment dans notre métier’. It is also not a mobile adverb (in for instance Nerval: ‘Il y a dans l’attachement à la terre beaucoup de l’amour de la famille’ = ‘[il y a de] l’amour de la famille [qui] entre pour beaucoup dans l’attachement à la terre’). And it is also not a pronoun followed by a preposition as in Plusieurs des combattants connaissaient mes livres or Aucune des horreurs de la guerre n’a été déblayée (the second example by Jules Romains then requires the verb to be singular: Bien que beaucoup des [= ‘among’] horreurs de la guerre ait été déblayé).



A stranger in the house 

So? Simply compare the sentence Beaucoup de combattants connaissaient mes deux livres de guerre and Beaucoup des combatants que je rencontrai connaissaient mes deux livres de guerre with the (a) and (b) examples below (for the purpose of the argument we will assume that socialistes and gagnants are NNs in their respective SNs): (a) En 2002, 80% des socialistes ont voté Chirac (i.e. that 80% of the socialist voters have voted for the Chirac camp) corresponds with En 2002, 35% de socialistes ont voté Chirac (i.e. that the Chirac votes counted a not insignificant share of Jospin supporters); (b) 100% des gagnants ont joué au loto (advertising slogan: in order to win the lotto you have to play, so the winners constitute a subset of the total amount of players) as against 100% de gagnants ont joué au loto (a deceptive claim: in order to win the lotto it is enough to play, i.e. the set of winners equals the set of players). As it appears, if the utterance Beaucoup de combattants connaissaient mes deux livres de guerre assigns a large number of combatants to the total set of Jules Romains readers, then the utterance Beaucoup des combattants que je rencontrai connaissaient mes deux livres de guerre only pertains to the numerous readers who are found in the subset which is constituted by the combatants who have met Jules Romains. The qualifier que je rencontrai avoids invoking the full set of combatants and indirectly a unity of readers where combatants and non-combatants co-exist. – The relative superlatives la plupart (du temps) [la plupart contracts the old la plus part = ‘the major part’], le plus gros (de la troupe) or le plus clair (de la journée) similarly avoid denoting the polar extremes of the whole. (c) Third environment: Negations, such as Pierre ne boit pas de vin affect a left-dislocated noun (e.g. Pierre n’en boit pas, de/du vin) and a right-dislocated noun (e.g. De/Du vin, Pierre n’en boit pas or the following example by Marcel Proust: ‘De blanchisseuse, un dimanche, il ne fallait pas penser qu’il en vînt’). Not that NN cannot sometimes avoid being affected, cf. Pierre n’aurait-il pas bu du vin? (rhetorical negation of a positive suggestion = ‘Pierre has drunk wine, unless I am mistaken’). Marie ne boit du vin qu’aux grandes occasions (negation of exception = ‘Marie does drink wine, but only on special occasions). On ne gaspille pas du vin, on le déguste or N’allez pas boire du vin bouchonné! Or Marie ne boit pas du vin, elle boit des vins … (partial negation, going from the NN to the verb gaspille, the qualifying adjective bouchonné, the quantifying adjective du … ). Ce n’est pas du vin, tout au plus de la piquette (total negation of the equation C’est du vin). Cet ivrogne de Pierre ne boit pas de l’eau (intensional negation which contests the identification of the liquid = ’what Pierre is drinking does not look like water’). Tu ne bois pas du vin? = ‘do not touch the wine which has been put on the table’ (extensional negation confined to a frame).



Marc Wilmet

The combination de + le/la: Du and de la denote a mass (i.e. conflation of objects: du vin + du vin + du vin = du vin; de la bière + de la bière + de la bière = de la bière), which contrasts with the count denotation of un, une (i.e. discrete objects: un homme + un homme + un homme = trois hommes; une bière + une bière + une bière = trois verres de bière or trois sortes de bière) and with the count or mass denotation of le, la (le veau = ‘veal’ or ‘calf ’). Note that mass denotation may either match our collective perception of reality … or turn it upside down: Acheter du dollar, Bouffer du curé, Casser du flic … Some writers exploit this to the point where it makes the reader cringe, cf. examples (27)-(30): (27) … l’autobus, dégorgeant du lycéen et de la dactylo … (Hervé Bazin). (28) … elles font du mioche comme on fait du tricot (ibid). (29) … les pages où, pour faire vibrer la corde à linge, s’accroche de la culotte de princesse (ibid). (30) … elle n’est pas malheureuse, votre nièce, qui secoue ses larmes, qui embrasse du Claude et du Mathilde, tout ce qui se trouve à la portée de sa bouche, au hasard (ibid). [note the masculine du Mathilde which desexualises the person]. A qualifier paves the way from mass to count denotation (e.g. Acheter du vin and Acheter du/un vin de prix) and practically dominates with abstract nouns (e.g. Montrer du courage or Manifester de l’orgueil and Montrer un/?du grand courage or Manifester un/?du bel orgueil). The combination de + les: The discontinous denotation inherently evoked by les results in count denotation in external plurals: des vins (= un vin + un vin + un vin … ) as in des hommes (= un homme + un homme + un homme …), but in internal plurals the mass denotation is retained: des tenailles, des vicissitudes, des rillettes… ≠ ‘une tenaille + une tenaille + une tenaille … ‘, etc. (and, by way of confirmation, the repeated subdivision of des hommes vs. des rillettes end fatally with des hommes on one side and des rillettes always on the other). In case of maximal extensity des is only interesting in respect to les if the original pairing of several elements is justified: Un enfant est toujours l’ouvrage de sa mere Les enfants sont toujours l’ouvrage de leur mère vs. Un frère et une sæur finissent toujour par s’entendre Des frères et des sæurs finissent toujours par s’entendre. The combination de + un/une: Ignored by the grammar books the combination of the two partitives has passed unnoticed through the centuries: ‘… mes que j’ai mangié ançois / D’un mervellos mangier françois’ = ‘après que j’aurai goûté de cette excellente préparation’ (Le jugement de Renart, a text from the thirteenth century [an old testament to the reputation of the French quisine]) and ‘J’ai mis à table d’un petit vin blanc dont vous me direz des nouvelles’ (Simenon). D’un, d’une superimpose the mass and the count denotation: d’un



A stranger in the house 

petit vin blanc + d’un petit vin blanc + d’un petit vin blanc … = d’un petit vin blanc or de trios (sortes de) petits vins blancs.

5. Recapitulation The features of the eight transverse quantifiers may be summed up in three opposed pairs: extensive vs. partitive; mass vs. count; continuous vs. discontinuous. These features may be summed up in a matrix (+ means presence of a feature, – means absence of a feature, and ± means unspecified for the feature):

Ø Le (la) Les Un (une) De Du (de la) Des D’un (d’une)

Extensive

Partitive

Mass

Count

Continu

Discontinu

± + + – – – – –

± – – + + + + +

­± ± ± – ± + ± +

± ± ± + ± – ± +

± + – + ± + – +

± – + – ± – + –

Stated differently: the Ø type is neutral; the LE type is extensive and continous mass or continuous count (le, la) or discountinous (les); the UN type is partitive and continuous count (un, une); the DE type is partitive and continuous mass (du, de la), discontinuous mass or count (des), continuous or discontinuous mass or count (de), continuous mass and count (d’un, d’une).

References Beauzée, N. 1767 [1974]. Grammaire générale. Nouvelle édition en facsimilé avec une introduction par B. E. Bartlett. 2 Vols. Stuttgart-Bad Cannstatt: Friedrich Fromann Verlag. Bloomfield, L. 1933. Language. London: Allen & Unwin. Christophersen, P. 1939. The Articles. A Study of their Theory and Use in English. Copenhagen & Londen: Munksgaard & Alford. Kupferman, L. 1979. L’article partitif existe-t-il? Le Français Moderne 47: 1–16. Lusignan, S. 1986. Parler vulgairement: Les intellectuels et la langue française aux XIIIe et XIVe siècles. Paris-Montréal: Vrin-Les Presses de l’Université. Noël, L. 1861. La grammaire française. Paris: Dutertre. Vandeloise, C. 2006. Vivre sans article partitif. Le Français Moderne 74: 141–158.



Marc Wilmet Wilmet, M. 1980. Le système de l’article français: Un bilan critique. Travaux de Linguistique et de Littérature 18: 53–64. Wilmet, M. 1983. Les déterminants du nom en français. Essai de synthèse. Langue Française 57: 15–33. Wilmet, M. 1986. La détermination nominale. Paris: P.U.F. Wilmet, M. 2003. Grammaire critique du français. 3rd edn. Brussels: Duculot.

Determination in endocentric and exocentric languages With evidence primarily from Danish and Italian Iørn Korzen

Copenhagen Business School In this paper it is argued that there is a correlation between lexico-semantic specificity and a tendency towards textual “promotion” of nouns (and, also, of verbs). Textually promoted nouns serve as “instantiators” of nominal “occurrences”, i.e. first, second or third order entities, and in order for a noun to instantiate an entity in a text, a coding of the feature [± identifiable to the hearer] is needed (possibly along with other features as well). In other words: generally, the noun must appear with a determiner. In fact, there is a general tendency for Romance nouns (which are lexically more specific than Germanic ones) to appear with a determiner, whereas Germanic nouns much more often appear undetermined and, possibly, incorporated in verbal or prepositional structures. The mentioned correlation hypothesis is substantiated with evidence mainly from Danish and Italian.

1. Introduction Recent research on linguistic typology carried out by a group of scholars from Copenhagen Business School1 has revealed fundamental differences between so-called endocentric and exocentric languages. The former are characterised by a relatively high level of lexical specificity and, therefore, informational “weight” in the verb, i.e. in the centre of the proposition (hence the term “en­do­centric” languages), whereas the latter have a higher level of lexical specificity and informational weight in the nominal arguments, i.e. outside the centre of the proposition (hence the term “exocentric” languages). The purpose of this paper is to analyse and, hopefully, shed some new light on the

1. The scholars are Irène Baron, Michael Herslund, Hanne Korzen and Lita Lundquist (on French), Henrik Høeg Müller (on Spanish), Viktor Smith (on Russian), and myself (on Italian).



Iørn Korzen

phenomenon of nominal determination on the basis of the typological distinction between endocentric and exocentric languages.

2. Endocentric and exocentric languages. Lexico-semantic differences Very broadly speaking, Germanic languages are endocentric while Romance languages are exocentric. However, a language such as English, which historically has been under heavy influence from French, exhibits features from both groups, as we shall see. Referring to the publications of the above-mentioned project group for more detail2, I shall here briefly summarise the lexical differences between endocentric and exocentric languages with examples from Danish and Italian, which – unlike English – are relatively straightforward examples of endocentric and exocentric languages respectively.

2.1

Verbs

The particular lexical specificity of endocentric verbs is due to the fact that these verbs generally lexicalise the semantic components manner and therefore select quite restrictively the type of argument involved in the verbal action. By way of illustration, I shall quote one of the motion verbs, namely to enter, which in Danish translations varies according to the manner of “entering” and (therefore often) to the grammatical subject, as is seen in the left-hand column of Figure 1. On the other hand, the equivalent Italian verb, entrare, does not contain any information on how or by which means the motion is carried out, and it may therefore combine with any subject that can perform a movement, as is shown in the middle column of (1). A manner specification may be added in the form of an adverbial satellite, as in the last three examples, or lexicalised by a small group of imperfective manner verbs such as camminare, nuotare ‘to walk’, ‘to swim’, but this only happens in case of a particular emphasis on this semantic component. In contrast to this, with just a handful of exceptions, the endocentric Danish verbs cannot avoid expressing the manner; it is part of their lexicalisation pattern.

2. Cf. e.g. Korzen and Marello (eds) (2000), Herslund (ed.) (2003), Baron (ed.) (2003) and Korzen and D’Achille (eds) (2005). In this research, the main focus is on Danish on the one hand and on French, Italian and Spanish on the other, and it is argued that these languages are good “representatives” of endocentric and exocentric languages respectively. However, there are differences between the various languages in each typological group, and in some respects one or more of the languages may exhibit divergences and variations from what we may call the “prototypical” endocentric or exocentric features.



Determination in endocentric and exocentric languages 

As can be seen in the right-hand column, here we have a good example of both endocentric and exocentric features in English, i.e. a full list of verbs expressing the manner, as in endocentric Danish, and one verb equivalent to exocentric Italian entrare3:

(1)

Endocentric Danish hunden fisken fuglen bilen skibet manden

gik ind svømmede ind fløj ind kørte ind sejlede ind gik ind

Exocentric Italian

English (endocentric / exocentric)

il cane entrò il pesce entrò l’uccello entrò l’automobile entrò la nave entrò l’uomo entrò (a piedi)

the dog walked in / entered the fish swam in / entered the bird flew in / entered the car drove in / entered the ship sailed in / entered the man walked in / entered (walking) the man ran in / entered (running) the man drove in / entered (driving a car)

manden løb ind

l’uomo entrò (di corsa)

manden kørte ind (i bil)

l’uomo entrò (in macchina)

2.2

Nouns

In the case of the nouns, it is the other way around. Exocentric nouns (denoting artefacts) are generally more specific than endocentric ones. This is because exocentric denominations of artefacts tend to lexicalise the semantic component figure, i.e. the external form and structure of the object, whereas endocentric denominations tend to lexicalise the component function. Since different objects may have (more or less) the same function even though they look different, this makes way for a more abstract and generic lexicalisation in endocentric Danish and a more specific and precise lexicalisation in the exocentric Romance languages. The same degree of specification may be expressed in Danish. It is normally achieved by means of nominal compounds as is shown in (2)-(3). However, in line with the exocentric verbs, generally such specification is only used in case of a particular emphasis on this semantic element. The prototype is lexicalised as the generic root on the basis of the function of the object, e.g. in (2), bil: an object used for transportation, and in (3) stol: an object used to sit in or possibly to stand on. By contrast, the Romance prototypes are lexicalised at a hyponymic level compared with Danish, and therefore they cannot avoid specifying the particular 3. The figure also shows the different realisations of the semantic component path (in Talmy’s (1985) terminology), i.e. the point of departure or arrival of the movement. In the endocentric manner verbs it is expressed by a verbal satellite, e.g. ind ‘in’, whereas in the exocentric verbs it is incorporated in the verb. However, this typological difference has no bearing on the argument selection.



Iørn Korzen

subtype. Since Italian lack hyperonyms similar to Danish bil and stol, these nouns cannot be translated into Italian unless the translator somehow knows precisely which subtype s/he is dealing with. Regarding (2), English behaves like Italian, whereas regarding (3), the first four cases are compounds as in Danish, and the last four are independent lexicalisations as in Italian:

Danish (literal translation)

Italian

English

(2) bil personbil lastbil varebil rutebil

[Ø] (‘person –’) (‘freight –’) (‘goods –’) (‘route –’)

– automobile (macchina) camion fur­gone autobus, pullman

– car lorry van bus, coach

(3) stol spisestuestol lænestol liggestol tronstol korstol prædikestol talerstol

(‘chair’) (‘dining room chair’) (‘leaning chair’) (‘lying chair’) (‘throne chair’) (‘choir chair’) (‘preaching chair’) (‘speaking chair’)

– sedia poltrona sdraio trono, seggio stallo da coro pulpito, pergamo podio

chair dining room chair armchair deck chair throne choir stall pulpit rostrum

2.3

Conclusion

The total amount of lexical information in an endocentric and an exocentric sentence may be the same, but the distribution of it on the two word classes will vary. The lexical specificity and weight inherent in endocentric and exocentric verbs and nouns can be illustrated as in (4), where bold and capital letters indicate the lexically more specific word classes4:

(4) Lexico-semantic specificity Endocentric languages (e.g. Danish)

n–V–n

Exocentric languages (e.g. Italian)

N–v–N

4. The figure illustrates “primary lexicalisation” – not e.g. compounds and derivatives.



Determination in endocentric and exocentric languages 

3. Endocentric and exocentric languages. Grammatico-semantic differences I will now argue, and hopefully demonstrate, that these lexical differences are paired with similar differences at a grammatico-seman­tic level, and that indeed lexico-semantic and grammatico-semantic specificity vs. abstractness go together. In other words, my claim is that the endocentric languages show a particular tendency to specify the verbal constituents in discourse, whereas the nominal arguments are often left grammatically unspecified, and the exocentric languages tend to specify the nominal constituents, whereas the verbal arguments are often grammatically more abstract. And I shall argue that the reason for these differences is to be found in the lexicon. The mentioned grammatico-semantic pattern can be illustrated as in (5):

(5) Grammatico-semantic specificity Endocentric languages (e.g. Danish)

np – VP – np

Exocentric languages (e.g. Italian)

NP – vp – NP

The difference between grammatico-semantic specificity and abstractness lies in the linguistic codification, or gram­mati­calisa­tion, of semantic and pragmatic features. The more features that are codified, the more grammatically specific is the constituent. In the verbal system, finite forms are more specified than non-finite forms in that they express features such as tense, aspect, mood, person and number (the precise number of features depending on the language), whereas non-finite forms are completely “unmarked” on this account. In the nominal system, nouns and noun groups with a determiner are more specified than those without one (naturally, in languages that actually have determiners) in that they render explicit the feature [± identifiable].

3.1

Verbs

With respect to the verbs, which I shall treat very briefly since they are not the main topic of my paper5, it is well known that the Romance languages are characterised – ceteris paribus – by a more complex syntax and hierarchic text structure than the Scandinavian languages. One of the main reasons for this is precisely the high frequency of non-finite verb forms and nominalisations. By way of illustration, Figure (6) quotes the results of a survey, in which I counted the non-finite verb forms and nominalisations in a text corpus of Danish and Italian retellings of a Mr Bean-episode:

5. For more detailed comparisons between the Danish and Italian verbal systems, cf. Korzen (1998b; 2004; 2005a/b).



Iørn Korzen

(6) Propositions textualised without a finite verb, % of all propositions (as found in the “Mr Bean corpus”)6

Danish texts

Italian texts

Infinitive

Gerund

Participle

Nominalisation

Total

written oral

12.02 6.40

– –

0.01 0

0.01 0

12.04 6.40

written oral

23.98 20.10

14.39 6.37

5.77 0.62

2.97 0.10

47.11 27.19

As the averaged figures show, non-finite propositions were about four times more frequent in Italian than in Danish, both in the written and the oral texts of the corpus, and in both languages they were about twice as frequent in written as in oral texts. In fact, these results are quite representative for narrative texts in general; in other text types the cross linguistic discrepancy may be even greater. This is partly due to differences in the language systems and morphological inventory (e.g. the gerund and certain participial constructions do not exist in Danish), partly to a rhetorical tradition for hypotaxis in the Neo-Latin languages. But on top of that, it is my claim that these differences could be predicted on the basis of the lexical differences illustrated in Figure (1). I shall return to this claim in the final section of my paper.

3.2

Nouns, noun phrases and determination. General observations

In the case of nouns and noun phrases, grammatico-semantic specificity lies, as mentioned earlier, e.g. in determination. As I have concluded in previous work, the phenomenon of nominal determination is closely linked to the phenomenon of text pragmatic prominence. Before proceeding, I shall briefly summarise some of my earlier findings.7 In both Danish and Italian (and I believe in most other languages as well) there is a clear correlation between the text pragmatic prominence, or topicality, of a nominal constituent and the likelihood that it will appear with a determiner. Vice versa, the less prominent, the higher the likelihood that the constituent will appear without a determiner. The obvious reason for this correlation is that the more important an entity is in the on-going discourse, the more necessary it will be to express whether or not it is supposedly identifiable to the hearer.

6. For more detail on this corpus, which consists of 36 Danish texts (18 oral and 18 written) and 54 Italian texts (27 oral and 27 written), see Skytte et al. (1999). 7.

For more detail, see Korzen (1996; 1998a; 2000b; 2000c; 2002).



Determination in endocentric and exocentric languages 

Text pragmatic prominence may be described using the following four hierarchies: (7) Identifiability [+ identifiable] > [– identifiable] (8) Referentiality Deictic > Specific > Generic > Non-specific > In­ten­sional constituents8 (9) Semantic individuation Countable first order entities, singular > Countable first order entities, plural > Uncount­able entities > Second or­der entities9 (10) Syntactic function/semantic role subject/Agent > dative/Experiencer > object/Patient > non-primary constituents The fur­ther to the left on these four scales, the more text pragmatically prominent is the enti­ty, and con­se­quently the more suitable it is as topic of a sentence or text sequence. The scales are inspired by the topicality hierarchies pro­posed by Givón (1976) and Herslund (ed.) (1996: 79), the transitivity scales discussed by Hop­per and Thomp­ son (1980) (to which I shall return below), and the scales for object marking pro­posed by Bos­song (1991). As it turns out, exactly the same scales are decisive for the likelihood that the noun or noun phrase in question will appear with a determiner. Furthermore, the relative importance of the four hierarchies corresponds with the chronological order above, the most important being the identifiability hierarchy in (7). The feature [+ identifiable] entails the pragmatic instruction to the hearer to identify the designated entity, and this instruction generally requires linguistic codification. The absence of this instruction much more easily allows for the absence of a determiner – depending on the position of the entity in the other hierarchies. For instance, a constituent in the top position of scale (8) always appears with a determiner, while one in the lowest position of (8) never does. In the other positions of scale (8), determination depends on the concurrent position in scale (9): a countable first-order entity in the singular always has a determiner, except in the purely intensional use, as is seen in the left-hand column of Figure (11), whereas determination of plural entities depends on the referential specificity, as is seen in the right-hand column:

8. The intensional constituents designate the pure concept in question and occur, for instance, in noun incor­pora­tions such as John is drinking wine; Susan sells houses; etc. See (13) below. 9. I here follow Lyons (1977: 442ff)’s definitions of first and second order entities.



Iørn Korzen

(11) Referentiality and semantic individuation – scales (8)-(9)10 Scale (9)

a. specific

b. ge­neric

Scale (8)

c. non-spe­ci­fic

d. in­ten­sio­nal

countable first order entities, singular

countable first order entities, plural

Jeg så en bil [*bil] her i går. Ho visto un’automobile [*automobile] qui ieri. ‘I saw a car [*car] here yesterday.’ En bil / Bilen [*Bil] forurener meget. Un’ / L’automobile [*automobile] inquina molto. ‘A car / The car [*Car] pollutes a lot.’10 Hvis jeg ser en bil [*bil], siger jeg til. Se vedo un’automobile [*automobile], te lo dico. ‘If I see a car [*car], I’ll tell you.’

Jeg så nogle ([Ø]) biler her i går. Ho visto delle automobili qui ieri. ‘I saw [some] cars here yesterday.’

Hans har købt [Ø] bil. [not grammatically possible in Italian, see sect. 3.3.1 below] ‘John has bought [a] car.’

Hans sælger [Ø] biler. Gianni vende [Ø] automobili. ‘John sells cars.’

[Ø] Biler forurener meget. Le automobili inquinano molto. ‘[Ø] cars pollute a lot.’

Hvis jeg ser [Ø] (nogle) biler, siger jeg til. Se vedo [Ø] (delle) automobili, te lo dico. ‘If I see (some) cars, I’ll tell you.’

As the figure illustrates, going from left to right and from top to bottom, i.e. “down” on the two scales (8) and (9), the need for determination becomes increasingly smaller. It should be noted that the generic NPs (11b) differ from the rest by being syntactic subjects; the others are all objects. In fact, the role of the syntactic scale, cf. (10), can be illustrated more precisely as in Figure (12): (12) Syntactic function – scale (10) a. (non-unaccusative) subject Scale (10)

b. dative

c. object d. secon­dary constituent

*[Ø] Bambini cantavano in una piazza. ‘Children sang on a square.’ Ho dato delle ca­ra­mel­le a ??[Ø] bam­bi­ni. ‘I gave some sweets to (?)chil­dren.’ Hai visto [Ø] bambini per strada? ‘Did you see children in the street?’ Ho vi­sto un gruppo di [Ø] bambini. ‘I saw a group of children.’

10. The English noun man is an exception since it does not take a defi­nite ar­ti­cle with generic meaning, but may have a zero determiner: [Ø] Man is a horrible beast.



Determination in endocentric and exocentric languages 

The further down the scale, the more easily the determiner may be omitted in Italian. (For Danish, see the following section.) Again, the concurrent position in the other scales will also play a role.11

3.3

Nominal determination in a cross-linguistic perspective

As the examples show, broadly speaking all four hierarchies apply to both Danish and Italian, but nevertheless there are significant differences between the two languages, the main difference being that Danish nouns appear without a determiner much more often than Italian ones. In fact, mastering Romance determination is one of the hardest tasks for a Danish learner. Parallel to the differences in the verbal system, this situation is partly due to differences in linguistic inventory. Unlike the Romance languages, Germanic languages do not have indefinite articles for mass and plural nouns, i.e. for the non-top positions of scales (7) and (9)12, and e.g. in Danish and English, undetermined nouns and noun phrases may be found in all four cases of Figure (12); for a series of authentic Danish examples, see Korzen (2000a: 265ff). Danish quantifiers equivalent to English ‘some’, nogen, noget, nogle, may be used as determiners, but they are considerably less frequent than the Romance indefinite articles (in Italian and French somewhat infelicitously also named “partitive” articles). As with the verbs, I believe that the explanation for this lies in the lexical differences between the endocentric and exocentric languages. In Figs. (2)-(3) we saw how exocentric noun lexicalisation takes place at a hyponymic level compared with endocentric lexicalisation, and in the scales of (8)-(9) we saw that referential and semantic individuation and specificity are factors which favour textual promotion and topicality of a given entity. Vice versa, non-specificity and semantic abstractness will favour textual demotion and backgrounding. In their article on transitivity scales, Hopper and Thompson (1980) define “the notion of Individuation” as “the extent to which O [an entity] is particularized and viewed as a concrete entity distinct from its background” (op.cit. 287; see also op.cit. 253). More precisely, individuation is “a superordinate property defined by sets of binary hierarchies including singular/plural, animate/inanimate, definite/indefinite” (op.cit. 279), and they add that individuation “is more complex and, we suspect, of considerably greater scope and linguistic significance than is implicit in the more limited use we have made of it here.” (op.cit. 287). At this point, I would like to introduce another binary hierarchy in the analysis, one which is closely linked to individuation and crucial for “the extent to which an object is particularized and viewed as a concrete entity distinct from its background”, 11. For more detail, including the role of attributive or other modifiers and special syntactic structures, cf. the references in footnote 7 and examples (27)-(29) below. 12. In Spanish, the indefinite articles are limited to countable nouns in singular and plural; i.e. there is no indefinite article for mass nouns.



Iørn Korzen

i.e. the hierarchy hyponym/hyperonym – or what I have defined above as lexical specificity. As I have shown in Korzen (2005a/b), there is a clear correlation between the lexical specificity of a constituent and the probability that the constituent will perform what Hopper and Thompson (1984: 708) define as “prototypical discourse function”. The prototypical discourse function of both verbs and nouns is to instantiate an “occurrence” of the given category in the discourse, in case of a verb: an action, an activity or a static situation, in case of a noun: an entity of the first, second or third order. In order for a verb to instantiate a verbal occurrence (independently of other linguistic constituents or structures) it must express features such as tense, mood, aspect, subject (depending on the language, cf. section 3.0), in other words: it must appear in a finite form. In order for a noun to instantiate a nominal occurrence, i.e. an entity, it must express features such as identifiability, number, case (again depending on the language). Only with such features rendered explicit, will a verb or a noun fulfil its prototypical discourse function. In fact, according to Hopper and Thompson (1984: 747), verbs and nouns are functionally “acategorial” outside discourse, even though “most forms begin with a propensity or predisposition to become N’s or V’s; and often this momentum can be reversed only by special morphology” (ibid.). In cases of such “reversal”, various degrees of de-categorisation take place. In other words: lexical specificity favours textual “promotion” and instantiation of occurrences of the given category, whereas lexical abstractness favours textual “demotion” and de-categorisation (Korzen 2005a/b). One of the consequences of these correlations is that exocentric languages are, so to speak, systematically programmed to promote their nominal arguments and to use them for instantiation of occurrences, whereas endocentric languages are systematically programmed to demote and background their nominal arguments in various ways. As just mentioned, instantiation of a nominal occurrence requires the marking of identifiability and number (marking of case is irrelevant to the languages involved in the present analysis), and this would appear to be in perfect harmony with the fact that precisely the Romance languages developed a “complete” system of articles fairly early (with the exception mentioned in footnote 12), whereas the article system of the Germanic languages is still “incomplete”. The definite article in Danish appeared in its enclitic form around year 1000, but the indefinite article began to appear as late as in the 14. Century, and not until the 16. Century do we find a usage similar to what we see today; cf. Mikkelsen (1975) and Skautrup (1968a/b). Both the Scandinavian and the Neo-Latin languages derive from languages with very rich inflectional systems. For instance, Old Norse and Latin nouns were inflected for case: nominative, accusative, dative, genitive (in Latin also vocative and ablative), but over the centuries these case distinctions were lost and a system of articles was developed – however, as mentioned, with quite different outcomes. I would like to advance the hypothesis that (at least part of) the explanation for these different outcomes lies in the lexicon, and that lexicon could also be (part of) the reason why Danish, on top of the limited determination possibilities, has at its disposal no less



Determination in endocentric and exocentric languages 

than two different morpho-syntactic ways of demoting nominal constituents in discourse: incorporation (cf. sect. 3.3.1) and a sort of antipassive structure (cf. sect. 3.3.2). Finally, I shall argue (in sect. 3.3.3) that the lexicon is indeed responsible for the great differences in determination of generic NPs in endocentric and exocentric languages. 3.3.1 Noun incorporation In noun incorporation (at primary i.e. sentence level)13, a noun so to speak merges with the verb and the joint structure expresses the effectuation of a specified verbal activity. This is illustrated in examples such as: (13) Noun incorporation

Scale (9)

a. countable first order entities, singular

vaske bil – lit.: ‘wash car’ (= do a car wash) spise æble – lit.: ‘eat apple’ (= do “apple-eating”) bygge villa – lit.: ‘construct villa’ (= do “villa constructing”) læse avis – lit.: ‘read newspaper’ (= do newspaper reading) male hus – lit.: ‘paint house’ (= do house painting)

b. countable first order entities, plural

strikke sokker – ‘knit socks’ (do sock knitting) skrive romaner – ‘write novels’ (= do novel writing) drikke vin – ‘drink wine’ (= do wine drinking) spise is – ‘eat ice-cream’ (= do ice-cream eating) yde hjælp – ‘render help’ (= help) føre krig – ‘make war’

c. uncount­able entities

d. second or­der entities

The incorporated nouns appear as denominalised in different ways. They are incapable of designating individualised or instantiated entities and of functioning as pragmatic peaks. They appear without a determiner and do not have a productive numeral inflection. Semantically they are reduced to expressing the pure and abstract intension of the noun, whereby they end up functioning as a sort of adverbial specification or frame for the verbal activity. A “car wash” in (13a) is a specified kind of “wash”, “apple eating” is a specified kind of “eating”, “villa constructing” is a specified kind of constructing, and so on. The focus is on the verbal content, and a durative verb maintains its full lexical content of durative activity that it expresses when it is not 13. Some compound structures could be defined as noun incorporation at secondary i.e. phrase level, cf. Korzen (1996: 148).

 Iørn Korzen

followed by an object. Only if the object is promoted to instantiating an entity – an operation that requires a determiner – may it “influence” the aktionsart and contribute to composing a telic structure.14 The examples of (13a) illustrate the quite generalised incorporation of countable first-order entities in Danish, a phenomenon which is not found productively in Italian. (In this respect, English appears to be closer to exocentric Italian than to Danish). On the other hand, the further down we move in the figure, i.e. on the scale of (9), the more common incorporation becomes even in Italian (as well as in English). But the differences of the countable first-order entities need to be explained, and for this purpose we return to the lexical differences illustrated in Figs. (2)-(3), which applied precisely to concrete objects. The differences in prototype lexicalisation – at a hyperonymic level in the endocentric languages and at a hyponymic level in the exocentric – may be graphically described as in (14)-(15): ENDOCENTRIC

(14)

PROTOTYPE LEXICALISATION

bil

HYPERONYM :

HYPONYMS :

(15)

personbil

lastbil

varebil

rutebil

‘ car’

‘lorry’

‘van’

‘bus, coach’

EXOCENTRIC

PROTOTYPE LEXICALISATION

HYPERONYM:

HYPONYMS : POSSIBLE SUB _ HYPONYMS :

[Ø]

automobile camion … … … … … …

furgone autobus, pullman … … …

… … …

Now, a very important feature of noun incorporation is the fact that it only happens (at least productively) at a hyperonymic level of the noun, as in ex. (16a). Generally, it does not occur at a hyponymic level, which is proved by the marginality of the examples in (16b). 14. For more detail on noun incorporation in Italian, cf. Korzen (2002). On noun incorporation in French and in the Romance languages in general, cf. e.g. Herslund (1994; 2002).



Determination in endocentric and exocentric languages 

(16) a. Hans har købt bil – ‘Hans has bought [a] [bil].’15 b. *Hans har købt personbil. – ‘Hans has bought [a] car.’ ??Hans har købt lastbil. – ‘Hans has bought [a] lorry.’ ??Hans har købt varebil. – ‘Hans has bought [a] van.’ The reason for this is very clear: lexical specificity will obstruct incorporation, because the nouns become unsuitable for expressing the vague and generic adverbial-like frame that is characteristic of that particular structure.16 In fact, two of the very few Italian exceptions of possible incorporations of singular first-order entities are the nouns casa and bottega, and the very same nouns are also exceptions to the “normal” exocentric lexicalisation pattern we saw in (15). Casa may mean ‘house’, but it also has a hyperonymic meaning of ‘home (where one lives)’, as illustrated in (17). Similarly, bottega may mean ‘shop’, but it also has a hyperonymic meaning of ‘place where one works’, cf. (18): casa (= abitazione – ‘home’)

(17)

villa

appartamento

‘villa’

‘flat’

edificio

camera

casa di campagna

casa rurale

‘room’

‘country home’

‘farm house’

‘building’

bottega (= ‘workplace’)

(18)

negozio ‘shop’

studio ‘study’

laboratorio ‘laboratory’

officina

…

‘workshop, repair shop’

And precisely these two nouns may become incorporated – but as we would expect, only in their hyperonymic meaning:

15. As we saw in Fig. (2), neither Italian nor English has equivalent nouns to the Danish hyperonym bil. 16. Cf. also Sasse (1984: 261-264): “[T]he semantic ingredients of DOs [direct objects] [...] are responsible for the pragmatic status of the noun in question: the more individuated a direct object is, the more suitable it becomes as a candidate for the pragmatic peak of the comment. [...] Those objects that do not qualify as pragmatic peaks of the comment tend to be incorporated. [...] [B]ecause of its lack of inherent pragmatic prominen­ce the non-individuated patient is unsuitable as an information peak and is, therefore, deprived of its grammatical individuality by being incorporated.”



Iørn Korzen

(19)

cercare casa – ‘look for [a] home’ trovare casa – ‘find [a] home’ comprare casa – ‘buy [a] home’ aprire casa – ‘open [your] home’ chiudere casa – ‘close [your] home’

(20) aprire bottega / mettere su bottega – ‘begin a professional activity’ chiudere bottega – ‘end a professional activity’ fare bottega di tutto – ‘deal with things in a dishonest way’ 3.3.2. Antipassive As if noun incorporation was not enough, Danish has another systematic way of textually backgrounding nouns and foregrounding verbs, namely by means of a sort of antipassive structure, where a direct object is reduced to a secondary complement of a preposition. The noun is reduced to designating an entity which is only partially or distantly involved in the action or partially affected by it, and as in noun incorporations, the verb maintains its independent semantic content of durative activity.17 Since these structures do not show any particular behaviour as far as determination is concerned (the noun in question always appears with a determiner or, possibly, otherwise specified, as in (27)-(28) below), I shall only mention them in passing. They are, however, illustrative of the endocentric tendency of demoting nominal arguments. Cf. the examples of (13a-c) with the following: (21)

vaske på en bil – ‘wash prep a car’ (without finishing) spise af et æble – ‘eat prep an apple’ (without finishing) bygge på en villa – ‘costruct prep a villa’ (without completing it) læse i en avis – ‘read prep a newspaper’ (without finishing) male på et hus – ‘paint prep a house’ (without finishing) strikke på en sok – ’knit prep a sock’ (without finishing) skrive på en roman – ‘write prep a novel’ (without finishing) drikke af noget vin – ‘drink prep some wine’ (without finishing)

These structures are highly productive in Danish but much rarer in the Romance languages, especially in Italian. 3.3.3 Generic NPs Another major difference between determination in Danish and in Italian is found in the generic NPs, which always appear with a determiner in Italian. More precisely, they always appear with a definite article, except for the singular first-order entities, which

17. Unlike in the noun incorporations, only imperfective activity verbs may appear in these structures. For a closer account, see Durst-Andersen and Hers­lund (1996).



Determination in endocentric and exocentric languages 

may also appear with an indefinite article, cf. (22a).18 In this respect, Italian and Danish (and English) are alike. But in all other cases, Danish (and English) generic NPs may appear without a determiner, cf. (22b-d): (22) Generic NPs

Scale (9)

a. countable first order, singular b. countable first order, plural c. uncountable entities d. second order en­tities

Danish

Italian

English

En bil / Bilen for­ure­ner

‘A / The car pollutes’

[Ø] Biler forurener

Un’ / L’automobile inquina Le automobili inquinano

[Ø] Vin er en alkoho­lisk drik

Il vino è una bevanda alcolica

[Ø] Misundelse har fanden skabt

L’invidia è una brutta malattia

‘[Ø] Cars pollute’ ‘[Ø] Wine is an alcoholic beverage’ ‘[Ø] Envy is the work of the devil’

Again the different lexicalisation levels can be said to be responsible for these crosslinguistic differences. It is well known that semantic definiteness of generic NPs can consist in a clear lexical contrast to other co-hyponyms and/or to a hyperonym.19 This is demonstrated also intralinguistically in the endocentric languages, where the definite article appears in case of a contrast, as the bottom lines of (23a) illustrate. On the other hand, there will be a zero determiner in cases where such a contrast is either nonexistent or textually irrelevant, as illustrated in (23b): Danskere ‘Danes’ (with the possibility of textually occurring hyponyms). GENERIC NPS

(23)

a.

bøgene

træer – ‘trees’

egene

granerne

befolkninger – ‘populations’

birkene

‘ the beeches’ ‘the oaks’ ‘the spruces’ ‘the birches’

danskerne

englænderne

italienerne

‘the Danes’

‘the Englishmen’

‘the Italians’

18. In very rare cases, an indefinite (“partitive”) article may appear in generic mass or plural NPs; cf. Korzen (1996: 386-395). 19. See e.g. Hansen (1994: 141).



Iørn Korzen

(23)

b.

[Ø] danskere – [Ø] ‘Danes’

kvinderne mændene

børnene

‘the women’ ‘the men’ ‘the children’

Since the exocentric lexicalisation is generally hyponymic, i.e. follows the lexical structure described in the bottom lines of (23a) (cf. (15)), there will always be a relatively close categorical contrast. This explains, I believe, why the definite article has become generalised in Italian generic NPs. And in fact, it would also explain why in many cases equivalent to Danish noun incorporations, we find generic NPs in Italian20: (24) Pia har (købt) bil / villa / cykel / båd / klaver / hund … Pia ha (comprato) l’automobile / la villa / la bicicletta / la barca / il pianoforte / il cane… ‘Pia has (bought) a car / a house / a bicycle / a boat / a piano / a dog / …’ Pia læser avis – Pia legge il giornale – ‘Pia is reading a / the newspaper’ Ole har skæg / paryk / feber / influenza / … Ole ha la barba / la parrucca / la febbre / l’influenza / … ‘Ole has a beard / a wig / a fever / the flu / …’

4. Conclusion Now let us sum up what has been said so far: As Hopper and Thompson (1980) and Herslund (ed.) (1996: 79) have shown, semantic specificity, or “individuation”, is one of the elements responsible for the intrinsic topicality and text pragmatic prominence of a nominal constituent, cf. the scales of (8)-(9). As I have shown in previous work, topicality and text pragmatic prominence have an important bearing on determination: The higher the topicality and text pragmatic prominence of a constituent, the higher the need for determination.

20. Or, rather, “prototype-NPs”. For a discussion on the differences between generic and “prototype-NPs”, see Korzen (1996: 622ff; 1998a: 98ff).



Determination in endocentric and exocentric languages 

Semantic specificity is a very complex phenomenon and may be defined by different parameters, namely referential, grammatical and lexical – lexical as for instance in countable vs. uncountable entities and first vs. second/third order entities, cf. scale (9). But on top of that, we have seen that fundamental differences in the lexicalisation patterns between endocentric and exocentric nouns also play an important role for semantic specificity – and thereby for determination. The lexical specificity of the exocentric nouns provides them with a particular intrinsic topicality, for which reason exocentric languages are, so to say, systematically programmed to promote their nominal arguments and to use them for instantiation of occurrences. This implies that exocentric nouns will typically appear with a determiner. Vice versa, lack of lexical specificity or individuation will hinder an entity from being distinguished from its background and thereby lower its intrinsic topicality. For this reason, endocentric languages are systematically programmed to demote their nominal arguments and to relegate them to various types of background function in discourse. This implies that endocentric nouns will tend to become denominalised, which means deprived of their potential capacity of instantiating entities and e.g. incorporated into verbal units, demoted to expressing the “frame” for the verbal activity, cf. (13), or incorporated into prepositional units and reduced to secondary complements, cf. (21). These typological differences may very well have prompted the development of the complete article system of the Romance languages. On the other hand, in the Germanic languages there was never as strong a need for articles, and these were never developed for nouns that – on top of their general lexical abstractness – were low on text pragmatic prominence for other reasons, i.e. nouns denoting unidentifiable plural or uncountable entities; cf. the scales of (7) and (9). Returning briefly to the verbal systems, we can say that due to differences in lexicalisation patterns, the endocentric verbs are systematically programmed to instantiate verbal situations, which implies that endocentric verbs will typically appear in finite forms. Vice versa, the lexically weak exocentric verbs easily become deverbalised, which means deprived of their potential features tense, aspect, mood, person and number and incorporated into other units – in this case sentences – demoted to expressing background information. This accounts for the cross-linguistic differences we saw in Figure (6), and it may very well have influenced the different evolutions of the verbal systems in the Scandinavian and Neo-Latin languages. Both language groups derive from languages with rich verbal inflectional systems, but many of the deverbalisation possibilities of Old Norse (e.g. the subjunctive, the supine and some participial constructions) were lost over the centuries, whereas most of the Latin deverbalisation possibilities survived in the Romance languages.21

21. For brief outlines on the diachronic aspects and the different evolutions of the deverbalisation possibilities, see Korzen (2005a/b).



Iørn Korzen

Typical examples of non-finite Italian verb forms from the mentioned “Mr Bean corpus” are for instance: (25) a. Andando verso il tavolo dove può accomodarsi [gerund phrase] l’ospite fa molta at­ten­zione a non far scric­chiolare il pavimento (Skytte et al. 1999: ISA1) – ‘Moving towards the table where he can sit down, the guest makes an effort in order not to make the floor creak’. b. Arrivato al tavolo [participle phrase], apre la sua borsa per prendere delle cose che a quanto pare gli serviranno. (ibid.) – ‘[Having] arrived at the table, he opens his bag to take out some things that apparently will be useful to him’. Here, the lexically abstract motion verbs andare, arrivare ‘to go’, ‘to arrive’ are textually demoted to designating the background of the instantiated foregrounded events the guest makes an effort…(25a) and he opens his bag… (25b). In the Danish texts of the same corpus, the motion in question was described with lexically much more specific verbs, such as gå, liste, træde, zigzagge, hoppe, springe, snige sig afsted, ‘walk, tiptoe, step, zigzag, hop, jump, sneak away’, and always in finite forms in order to instantiate the specified action. As is well known, and observed here, nominal determination and verbal finiteness are very parallel phenomena in that they both serve the purpose of instantiating occurrences: in case of a noun a first, second or third order entity, in case of a verb an action, an activity or a situation. What we are witnessing here in terms of linguistic correlations and, possibly, diachronic implications may be illustrated as in Figure (26), where ↓ reads “implies” or “leads to” (see next page). The more semantically individuated and specific a constituent is, the more promoted it will be in the discourse, and the closer it will be to in­stan­tiating an occurrence of the given word class. In fact, this correlation is quite obvious and easy to understand: Human language is always under-specified compared to the extralinguistic reality or mental representation that it is used to convey, but the more specified a constituent is, the closer it will be to the extralinguistic element it represents, and the more suitable it will be as a direct “link” to the extralinguistic world. Therefore, this correlation also works the other way around: an NP or VP which is used to designate a particular extralinguistic occurrence, would tend to be more specific – according to any of the parameters mentioned above – than an NP or VP used in other ways. As said earlier, the correlation accounts for the much more generalised determination in the Romance languages than in the Scandinavian, including the creation of indefinite articles for plural and mass nouns. But what is more, it also accounts for the fact that in some cases Italian noun phrases without a determiner may instantiate a particular entity if they are sufficiently specified in other ways, for instance by attributive constituents:



Determination in endocentric and exocentric languages 

(26) Lexicon, discourse function and morpho-syntax Lexicon Text pragmatics Discourse function Linguistic group / Word class Morpho-syntactic consequences

Diachronic implications

Individuation (lexical specificity) ↓ High topicality, textual promotion ↓

Generalisation (lexical abstractness) ↓ Low topicality, textual demotion ↓

Tendency to instantiate an “occurrence” (prototypical discourse function) Exocentric Endocentric Nouns: Verbs:

Tendency not to instantiate an “occurrence” (atypical discourse function) Endocentric Exocentric Nouns: Verbs:

overt marking of identifiability: determination

overt marking of tense, mood, aspect, person: finiteness

fully developed article system

reduced verbal inflectional system

frequent denominalisation: zero de­ter­ mination and noun incorporation defective article system

frequent deverbalisation: non-finiteness and verb incorporation preserved verbal inflectional system

(27) [...], poiché [Ø] episodi precedenti avvenuti in altri Sta­ti hanno di­mo­strato che adunare migliaia di persone in luoghi circoscritti può tra­sformarsi in tragedia ed ecatom­be. (La Stam­pa 27.7.89, p. II/2) ‘[...] because earlier episodes [which have] occurred in other states have proved that crowding thousands of people into confined spaces may result in tragedies and massa­cres.’ See many more examples of this in Korzen (1996: 197ff; 2000a: 252–254; 2000c). A similar specification may also come from being part of a list (28) or a scene description (29): (28) [...] la voce della madre di Adolfo si schiariva, si face­va solenne, fortissima. Appena finita la colazione, madre e figlio tornavano nella stanza da letto. (Dacia Maraini, Mio marito, Milano, Bompiani 1968, p. 26) ‘the voice of Adolfo’s mother became clear, solemn, strong. As soon as they had finished their breakfast, mother and son returned to the bedroom.’ (29) L’omicidio è stato compiuto con spietatezza: sangue è stato tro­vato nel la­va­bo, sulla porta del bagno e sul viso del sacerdo­te. (Repubblica 4.5.95, p. 16)



Iørn Korzen

‘The murder has been committed with ruthlessness. Blood has been found in the washbasin, on the door to the bathroom and on the priest’s face’. See more Italian and/or Danish examples of this in Korzen (1996: 239–248; 2000a: 254–257). However, in cases like (27) and (29), these possibilities do not apply to constituents placed in the highest positions of the scales (7), (8) or (9). In such “high places”, a good old-fashioned determiner is still obligatory.

References Baron, I. (ed.). 2003. Language and Culture [Copenhagen Studies in Language 29]. Copenhagen: Samfundslitteratur. Bossong, G. 1991. Differential object marking in Romance and be­yond. In New Analyses in Romance Linguistics. Selected papers from the XVIII Lin­gui­stic Sym­posium on Romance Languages Urbana-Champaign, April 7–9 1988, D. Wanner & D.A. Kibbee (eds). Amster­dam: John Benjamins. Durst-Andersen, P. & Herslund, M. 1996. The syntax of Danish verbs. Lexical and syntactic transitivity. In Content, Expression and Structure. Studies in Danish Func­tio­nal Gram­mar, E. Engberg-Pedersen et al (eds), 65–102. Am­ster­dam: John Benjamins. Givón, T. 1976. Topic, pronoun and grammatical agreement. In Subject and Topic, C.N. Li (ed.), 149–188. New York NY: Academic Press. Hansen, E. 1994. Generisk substantiv. In Språkbruk, grammatik och språk­förändring. En fest­skrift till Ulf Teleman 13.1.1994, 137–144. Institutionen för nordiska språk, Lunds universitet. Herslund, M. 1994. La notion d‘incorporation en danois et en fran­çais. Travaux de linguistique et de philologie XXXII: 7–18. Herslund, M. 2002. Incorporation and transitivity in Romance. In Complex Pre­dicates and Incorporation. A Functional Perspective [Travaux du Cercle Linguistique de Co­pen­hague 32], O. Nedergaard Thomsen & M. Herslund (eds), 175–206. Copenhagen: C.A. Reitzel. Herslund, M. (ed.). 1996. Det franske sprog. Kapitel III. Valens og transitivitet. Foreløbig version. Handelshøjsko­len i Køben­havn. Herslund, M. (ed.). 2003. Aspects linguistiques de la traduction. Bordeaux: Presses Universi­taires de Bor­deaux. Hopper, P.J. & Thompson, S.A. 1980. Transitivity in grammar and discourse. Language 56(2): 251–299. Hopper, P.J. & Thompson, S.A. 1984. The discourse basis for lexi­cal categories in universal grammar. Language 60(4): 703–752. Korzen, I. 1996. L’articolo italiano fra concetto ed entità, I-II [Etudes Romanes 36]. Kø­ben­havn: Museum Tu­scu­la­num Press. Korzen, I. 1998a. On nominal determination – with special reference to Italian and com­parisons with Danish. In Nominal De­ter­mination [Copenhagen Studies in Lan­guage 21], G. Hansen (ed.), 67–132. Copenhagen: Samfundslitteratur. Korzen, I. 1998b. On the grammaticalisation of rhetorical satelli­tes. A comparative study on Italian and Danish. In Clause Combining and Text Structure [Copenhagen Studies in Language 22], I. Korzen & M. Herslund (eds), 65–86. Copenhagen: Samfundslit­teratur.



Determination in endocentric and exocentric languages  Korzen, I. 2000a. Reference og andre sproglige rela­tio­ner. In Italiensk-dansk sprogbrug komparativt perspektiv, G. Skytte & I. Korzen (eds), 161–619. Copenhagen: Samfundslitteratur. Korzen, I. 2000b. Nominal determination in translation: A question of topicality and text prag­ matic explicitness. Journal of Philology 3(6): 93–102. Korzen, I. 2000c. Pragmatica testuale e sintassi nominale. Gerarchie pragmatiche, de­ter­mi­ nazione no­mi­nale e relazioni anaforiche. In Argomenti per una linguistica della traduzione. Notes pour une linguistique de la traduction. On Linguistic Aspetcs of Translation [Gli argomenti umani 4], I. Korzen & C. Marello (eds), 81–109. Alessandria: Edizioni dell’Orso. Korzen, I. 2002. Noun Incorporation in Italian. In Complex Pre­dicates and Incorporation. A Functional Perspective [Travaux du Cercle Linguistique de Co­pen­hague 32], O. Nedergaard Thomsen & M. Herslund (eds), 207–266. Copenhagen: C.A. Reitzel. Korzen, I. 2004. Dalla microstruttura alla macrostruttura. In Generi, archi­tet­ture e for­me testuali. Atti del VII Conve­gno SILFI, P. D’A­­chil­le (ed.), 363–376. Firenze: Franco Cesati. Korzen, I. 2005a. Linguistic typology in translation: Endocentric and exocentric languages, as exemplified by Danish and Italian. Perspectives. Studies in Translatology 13: 1: 21–37. Korzen, I. 2005b. Lingue endocentriche e lingue esocentriche: Testo, contesto e identità culturale. In Tipologia linguistica e società. Considerazioni inter- e intralinguistiche. (Linguistic Typology and Society. Inter- and Intralinguistic Reflections), I. Korzen & P. D’Achille (eds), 31–54. Fi­ren­ze: Franco Cesati. Korzen, I. & Marello, C. (eds). 2000. Argomenti per una linguistica della traduzione. Notes pour une linguistique de la traduction. On Linguistic Aspects of Translation [Gli argomenti umani 4]. Alessandria: Edizioni dell’Orso. Korzen, I. & D’Achille, P. (eds). 2005. Tipologia linguistica e società. Conside­ra­zio­ni inter- e intralinguistiche. (Linguistic Typology and Society. Inter- and Intralinguistic Reflections). Fi­ren­ze: Franco Cesati. Lyons, J. 1977. Semantics, 1–2. Cambridge: CUP. Mikkelsen, K. 1975 [1911]. Dansk ordföjningslære. Med sprog­hi­stori­ske tillæg. Copenhagen: Hans Reitzel. Sasse, H.-J. 1984. The pragmatics of noun incorporation in eastern Cushi­tic languages. In Objects. Towards A Theory Of Gram­matical Relations, F. Plank (ed.), 243–268. London: Academic Press. Skautrup, P. 1968. Det danske Sprogs Historie. Copenhagen: Gyldendal. Skyt­te, G., Korzen, I., Polito, P. & Strudsholm, P. (eds). 1999. Tekststruk­tu­re­ring på italiensk og dansk. Resultater af en komparativ under­sø­gelse / Strutturazio­ne te­stua­le in ita­lia­no e danese. ­Risultati di una indagine comparativa. Copenhagen: Mu­se­um Tuscula­num Press. Talmy, L. 1985. Lexicalization patterns: semantic struc­ture in lexical form. In Language Typology and Syntactic De­scrip­tion. Vol. III. Grammatical Categories and the Lexicon, T. Shopen (ed.), 57–149. Cam­bridge: CUP.

Bare predicate nominals in Romance languages Roberto Zamparelli Università di Trento

This paper offers an analysis for the possible absence of determiners in singular predicate nominals that refer to professions, roles and certain relations (e.g. dottore, capo-mafia, figlio di Luigi in Italian). Building on the theory of noun phrases in Heycock and Zamparelli (2005), it argues that while singular count nouns are normally licensed by the presence of a determiner, nouns that form bare predicates have an impoverished set of features (in particular, no set value for gender), and can be licensed by entering in an agreement relation with the subject of the predication. Semantically, the article distinguishes three subclasses of bare predicates, and argues that role / profession nouns ambiguously refer either to sets of individuals or to the activities which can identify these individuals.

1. Introduction In most European languages that have articles, singular count predicate nominals require an overt determiner. (1) a. b.

Carlo è *(un) uomo. Italian Carlo is a man Det hàr ar *(en) stol. Swedish that is (a) chair

However, no determiner is necessary with a restricted class of predicates like those illustrated in (2) for Italian (all original examples will be from this language, unless otherwise noted). (2) a. b.

Carlo è [(un) insegnante]. Carlo is [(a) teacher] Marta è [(una) {parente / cugina} di Marco]. Marta is [(a) {relative / cousin} of Marco]

 Roberto Zamparelli

c. Pia è [(una) vicina di casa di Matteo]. Pia is [(a) neighbor of Matteo] Determinerless singular count predicates like those in (2) (hereafter “bare predicates”) can only be built around a few types of nouns. In Italian, which is typical in this respect, they must contain simple or complex nouns that refer to “roles”: in particular, professions, (e.g. professore associato “associate professor”, macellaio “butcher”, collaudatore di moto “motorcycle test driver”, etc.), family relations (e.g. parente “relative”, fratello “brother”, cugino di secondo grado, “second cousin” etc.), other relations such as vicino di casa “house neighbor”, dirimpettaio “person living opposite” cointestatario del conto “co-holder of the account”, compounds with capo “head / boss” (e.g. capo reparto “sector head” capo cannoniere “bomber (soccer)” capo cordata “roped-party leader”, capo mafia “mafia boss”), etc. I will collectively refer to these classes of nouns as “role nouns”.1 The existence of bare predicates is noted in descriptive grammars for Italian, French, Spanish, European and Brazilian Portuguese, German, Dutch, Danish, Norwegian and Swedish, with some variations: in German, for instance, nouns that express family relations, like Bruder “brother”, cannot form bare predicates. All the grammars I have consulted mention profession nouns and often nouns meaning “inhabitant of X”, X a nation, region or city (Italian, Pole, Swede); some add “believer of Y”, Y a religion (Muslim, Catholic, atheist, etc.).2

1. I am excluding from consideration coordinated bare singular nominals like “he was [judge, jury and executioner]”. See Heycock and Zamparelli (2003) for an analysis. 2. 2Italian does not have morphological distinctions between adjectives and nouns that express nationality or religion (like the pairs Polish / Pole, Swedish / Swede), see (ia vs. c). (i) a. b. c.

La bandiera italiana The flag Italianfem–sn9 “The Italian flag” La bandiera era italiana. The flag was Italianfem–sn9 “The Italian flag” Una italiana cantò “Casta Diva”. An Italian / em–sn9 sang “Casta Diva” “An Italian (woman) sang Casta Diva”

There is no proof that Italian can have bare nominal predicates with names of nationality or religion, and that italiana in (ib) is anything but an adjective. First, note that the subject of the predication (ib) is non-human, a possibility excluded for bare nominal predicates elsewhere. Next, (iia) shows that italiana can be conjoined with an adjective in the predicative post-nominal attributive position. In (iib,c) italiana is modified by the adverbial tipicamente “typically”, by the adjectival suffix -issim- “extremely” and by the adjectival intensifier molto “very”. (ii) a. b.

La cucina italiana ed internazionale di questo ristorante è superba, the cuisine Italian and international of this restaurant is superb Le tue reazioni sono tipicamente italiane. the your reactions are typically Italian



Bare predicate nominals in Romance languages 

English is one exception. The singular count bare predicates that are possible in this language are a subset of the cases considered here: for instance, most profession nouns and all nouns of nationality are impossible without a determiner (3).3 (3) a. * John is {teacher / plumber / architect}. b. *Mary is {cousin / relative} of a good friend of mine. c. *Olaf is {Swede / Pole}. Since in all the languages under consideration predicative adjectives do not need any determiner, the existence of nominal predicates that behave “like adjectives” has implications for the general question of why other predicate nominals must appear with an article. Moreover, any explanation of the phenomenon must account for the difference between English and other bare predicate languages: an explanation based on some semantic feature of the class of nominals that can be bare must be complemented by a syntactic difference between languages. The general idea pursued in this paper is based on the theory of noun phrases developed in Heycock and Zamparelli (2005). According to these authors, singular count nouns need articles at least because they need to receive from them a value for a semantic feature (called Latt, for “lattice denotation”) which would otherwise remain unvalued on N. In plural / mass cases, a value for Latt is provided by pluralization / massification operators, but in the singular an overt element is required. Predicate nominals, which are normally fully specified for agreement features, do not agree with their copular subjects, so they cannot acquire a value for Latt from their subjects. Role nouns, on the other hand, are feature-defective categories. In particular, they lack gender specifications. As such, they force the copular subject DP to merge in a more internal position, from which predicate and subject can enter in an agreement relation. Through this relation the subject can also deliver a value for Latt, obviating the need for an overt determiner. The divergent behavior of English follows from the independent observation that, among the bare predicate languages, English has no grammatical gender. The paper is organized as follows. In the next section I will go through some background considerations, and review the points that an account should cover. Sections 3 and 3.1 will discuss some recent previous analyses. The core of my syntactic analysis c.

Nonostante gli anni in USA, Carla rimaneva {italianissima / molto italiana} despite the years in USA, Carla was {Italian-SUFF / very Italian} nel modo di vestirsi, in her way to dress

All of these possibilities are excluded for role noun bare predicates. This does not prove, of course, that italiano in (ib) could never be a bare nominal predicate, but it certainly tells us that in Italian words of nationality / religion do not allow us to reach any firm conclusion on the status of other bare predicates. 3. The characterization of the remaining possible English bare predicates will be taken up in Section 4.1.

 Roberto Zamparelli

will be given in Section 4. Finally, Section 5 contains my proposal for an informal semantic characterization, followed by the general conclusions.

2. Some preliminary observations Any analysis of bare predicates must start from the observation that the lack of determiners in singular count nouns is not parallel to the lack of determiners which is commonly found with plural or mass nouns, in argumental and predicative position. The possibility of bare plural / bare mass nouns exists in English, which lacks generalized bare predicates, and is missing in French, which does have the bare predicate construction. In Italian, Spanish and European Portuguese any plural or mass noun can be bare, but only when in lexically governed positions (see Contreras 1986). Outside these positions, bare plurals / mass become more acceptable when heavily modified and in certain styles (formal legal style, telegraphic text, etc.). Contrary to this, bare predicates usually become impossible when modified (see (5) below), and are not sensitive to these styles. This strongly suggests that the “bareness” of plural / masses and that of role predicates must come from different sources, though of course, an answer to the question why in many languages determiners are generally obligatory with singular count nouns but not with plural or mass ones is instrumental in understanding the ‘exception’ of bare predicates. Similarly, note that the class of nouns that can form bare predicates is narrower than the class of determinerless appositions. The bracketed elements below make fine bare appositions, but the corresponding bare predicates are impossible. (4) a. b. c.

Carlo, [accanito lettore di fantascienza],... Carlo, [assiduous reader of science fiction]... Cristoforo Colombo, [scopritore dell’ America],... Cristoforo Colombo, [discoverer of America]... Ugo, [principale pilastro della squadra],... Ugo, [main pillar ofthe team]

In fact, none of the syntactic and semantic restrictions discussed for bare predicates in this paper holds for appositions. I will leave an analysis of these interesting structures for another occasion. The possibility of bare predicates also seems unrelated to the availability in a language of bare singular count nouns in argumental position. Brazilian Portuguese, for instance, allows bare singular quite freely (Munn and Schmitt 1998, 2005) and Norwegian allow common nouns as objects of certain main verbs, with an indefinite interpretation (see Delsing 1993, Borthen 2003), yet in both languages predicate nominals can be bare only when formed by role nouns. With these considerations in mind, we can now list some facts that a theory of bare predicates should attempt to explain.



Bare predicate nominals in Romance languages 

I. In the languages where bare predicates are possible, they cannot be generally modified by adjectives, PPs or relative clauses (see (5), (6)). (5) a. b.

Gianni è (*bravo) medico (*che capisce i pazienti). Gianni is ( good) doctor ( that understands the patients) Gianni ist (*guter) Arzt (*der seine Pazienten versteht). German Gianni is ( good) doctor ( that his patients understands)

(6) Carlo è *(un) insegnante {italiano / esperto / anziano}. Carlo is (a) teacher {Italian / expert / old} However, they do sometimes take complements (as in insegnante di religione “religion teacher”, comandante di brigata “brigade commander”), and can be modified by certain adjectives or nouns, when these help to define what kind of N the subject of the predication is. Such Adj+N combinations are probably best regarded as compounds. Examples are infermiere diplomato “licensed nurse”, ricercatore non confermato “nontenured researcher”, insegnante di sostegno “support teacher”, tiratore scelto “sharp shooter”, etc. In Dutch, the adjectives that accompany bare nouns (e.g. werkloz(e) in (7), unlike e.g. klein(e) in (7c)) do not display the normal inflectional ending in -e (Winter, de Swart, and Zwarts 2004). (7) a. b. c.

Jan is een kleine / werkloze visser. Jan is a short[+INFL] / unemployed[+INFL] fisherman. Jan is werkloos / *werkloze visser. Jan is unemployed[-INFL] / * unemployed [+INFL] fisherman. Jan is *klein / *kleine visser. Jan is *short[-INFL] / *short[+INFL] fisherman.

II. Bare predicates are normally predicated of +human subjects (see (8), from Kupferman 1991, and Section 5.1 for some systematic exceptions). It follows that bare predicates cannot appear in existential constructions, (cf. (9)), despite many points of similarity between the nominals in existential sentences and predicate nominals. (8) {Marie / *Cette fourmi} est {reine / ouvrière}. {Marie / *this ant} is {queen / worker} (9) *C’è {insegnante / cugino} di Maria. there is {teacher / cousin} of Maria III. The third problem is how to explain the observation that, when they are possible, bare predicates have greater semantic restrictions than normal predicates. While nominal predicates with a determiner can have a metaphoric interpretation, this is impossible for bare predicates. The (b) cases in (10) – (12) simply state

 Roberto Zamparelli

someone’s (additional) profession, and are not remarks on that person’s ability. Similarly, (13b) excludes the “brother-like figure” interpretation which selects per “to”. (10) a. b.

Pelè era un mago (col pallone). Pelè was a wizard (with_the ball) Pelè era mago. Pelè was wizard.

(11) a. b.

Quel that Quel that

chirurgo surgeon chirurgo surgeon

è un artista (quando opera). is an artist (when he operates) è artista (??quando opera). is artist (when he operates)

(12) a. b.

Quel that Quel that

dentista dentist dentista dentist

è un macellaio. is a butcher è macellaio. is butcher

(13) a. b.

Marco è un fratello {per / di} Carlo. Marco is a brother {to / of} Carlo Marco è fratello {*per / di} Carlo. Marco is brother {to / of} Carlo

The absence of metaphorical or extended interpretations can account for the fact that certain nouns whose semantics would apparently qualify for being the description of a “role” (e.g. pilastro “pillar”, centro “center”, modello “model”) are nevertheless impossible as bare predicates, cf.(14). (14) a. b. c.

Marco è *(il / un) pilastro della nostra azienda. Marco is (the / a) pillar of our firm Francesco è *(il) centro della famiglia. Francesco is (the) center of_the family Carlo è* (un) modello per i suoi amici. Carlo is (a) model for the his friends

Matushansky and Spector (2004) observe that the time argument of a bare predicate must be linked to the utterance time: (15a), unlike (15b), cannot be said now that Bush is president and no longer governor. (15) Scenario: At a fund-raising event all former or current governors must identify themselves. What about Bush? a. #Bush est gouverneur. false Bush is governor. b. Bush est un gouverneur. true Bush is a governor.



Bare predicate nominals in Romance languages 

They suggest that the same applies to the world argument, and that without the article, (16) can only be interpreted if the reader imagines him or herself in Harry Potter’s imaginary world, whereas this is not necessary when the article is present. (16) Harry Potter est (un) magicien. Harry Potter is (a) wizard

3. Previous analyses Bare predicates have often been discussed (for French, see Anscombre 1986 and 1991, Pollock 1983, Kupferman 1991, Munn and Schmitt 2005; for Dutch, Winter, de Swart, and Zwarts 2004; for Italian, Renzi 1995, Vol.1, Section 5.4.1, Zamparelli 1995). One of the first analyses (found in French descriptive grammars and spelled out in Pollock 1983) is that bare predicates behave like adjectives because they have a hidden adjectival syntax. The class of nouns that allows this construction would simply have the exceptional property of undergoing an optional ‘adjectivization’. The adjectival character immediately explains the impossibility of (5) (adjectives cannot be modified by other adjectives or relative clauses) and why many of these nouns agree in gender with their predicate4: (17) a. b.

Marco è {attore / professore / dottore / infermiere}. Marcomasc is {actormasc / professormasc / doctormasc / nursemasc} Maria è {attrice / professoressa / dottoressa / infermiera}. Mariafem is {actressfem / professorfem / doctorfem / nursefem}

One could object that adjectives can be modified by adverbs of degree, like very and more, but bare predicates cannot. However, this discrepancy could be explained away as due to the combination of the non-gradable literal meaning of the putative adjectives (literally speaking, one cannot “be a professor” to different degrees) plus the fact that extended or metaphorical meanings do not seem to be available: a person can be described as “very Italian”, but only in the sense of an amplification of some typical features of this category. 4. Number agreement is more difficult to use as evidence. Predicate nominals in general establish a form of semantic number agreement: they are plural when distributive (ia) but can be singular when they allow a cumulative interpretation (ib). (i) a. b.

Carla e Maria sono {bambine / *una bambina / *attrice}. Carla and Maria are {children / a child / actress} Carla e Maria sono un grosso problema. Carla and Maria are a big problem

The problem of una bambina “a child” in (ia) is probably falsity, not ungrammaticality, and it is difficult to tell whether the number agreement one sees with profession predicates like attrice “actress” is qualitatively different from that of other predicates.

 Roberto Zamparelli

However, the fact that extended or metaphorical readings do not exists for bare predicates does not immediately follow from the adjectival theory. After all, the adjectival predications in (18) have extended / metaphorical senses that are perfectly compatible with Pelè not being, literally, a magician, and Karl being something different from a professor. (18) a. b.

Pelè era magico. Pelè was magical Max è (molto) professorale. Max is (very) professorial

A more serious drawback is that treating bare predicates as adjectivizations runs into problems with those role nominals which do include adjectives, like infermiere diplomato, “certificated nurse” or professore associato “associate professor”. The only solution seems to be to assume that what turns into an adjective is not the head noun, but some DP-internal projection that includes the noun, its arguments and its innermost adjectival modifiers, excluding the determiner. (19) Carlo è [Adjp … [np professore associato]] A model for an operation of this sort might be infinito sostantivato (Skytte, Salvi, and Manzini 1991, Sec. 10, Zucchi 1989, Ch.7), a type of nominalization in Italian where a definite article selects an infinitival VP which can include adverbials (e.g. lentamente in (20)). (20) a. Il parlare lentamente è un segno di educazione. the speak_INF slowly is a sign of education b. [dp il [vp parlare [AdvP lentamente]]]... However, the similarity is more apparent than real: infinito sostantivato can apply to any VP, and return a nominal which can then be modified by adjectives, whereas the “adjectivization” illustrated by (19) should be able to apply only when the NP embeds a role noun, and should block modification by adjectival modifiers.5 Even setting aside the problem posed by modifiers, there are a number of syntactic differences between bare predicates and adjectives, pointed out for French by Kupferman (1991). For instance, bare predicates with a definite interpretation (like est directeur, see Sec. 5) contrast both with definite predicate nominals (est le directeur) and with adjectives in not allowing the extraction of the “partitive clitic” en: (21) a. Paul est {satisfait /le directeur /directeur} de ce college. Paul is {satisfied /the director /director} of this college 5. Treating combinations such as professore associato as a lexical compounds does not help, since such Adj-N heads would have to be reanalized as Adj-Adj compounds, which are not otherwise attested. Notice that the “adjectivization” process would minimally have to apply after inflectional morphology, since in languages where the inflection for N and Adj differs (e.g. Danish) the bare predicate takes the nominal inflection (Alex Klinge, (p.c)).



Bare predicate nominals in Romance languages 

b. Paul en est {satisfait /le directeur /*directeur}. Paul of it is {satisfied /the director /director} Kupferman proposes that the spec of the definite DP constitutes an escape hatch for the clitic, and that this position is not available with bare predicates. Similarly, constructions that select adjectives (e.g. the rendre NP Adj “render NP AdjP” (22a), or the impersonal quelqu ‘un de AdjP “someone of AdjP” (23)) never accept bare profession nouns, while other verbs that never accept adjectives (e.g. elective verbs like faire in (22)) do allow bare profession nouns: (22) a. L’ accident l’a rendu/*fait {borgne/malin/disponible}. the accident him-CL has rendered/made {one-eyed/clever/available} b. Le vote l’ a {*rendu /faite} président. the vote him has {rendered /made} president (23) Quelqu’un de {malin/ blonde / borgne / disponible / *medecin / *concierge / *president}. someone of {clever/blond / one-eyed / available /doctor /doorman /president} While not all of these tests come out the same in a language as close to French as Italian6, there seems to be ground to search for a more satisfactory solution than “adjectivization”. Kupferman observes that the nouns which allow bare predicates have an eventive or time-bound (perhaps “stage-level”, in Carlsonian terms) character, which shows up in the possibility to appear in small clauses and in absolute constructions (see (24a,b) and (25a,b)) in contrast with individual-level adjectives and other predicate nominals ((24) and (25) c, d). (24) a. Il y avait un invité {de malade / d’ivre /d’ absent}. There CL was a guest {of sick / of drunk /of absent} b. Il y avait un invité {médecin / président/père de huit enfants}. There CL was a guest {doctor / president/father of eight children}. c. *Il y avait un invité {de blond / d’ intelligent / de gros}. There CL was a guest {blonde / of intelligent / of fat} d. * Il y avait {un invité / un médecin / un président / un père de huit enfants} There CL was {a guest / a doctor / a president / a father of eight children} (25) a. b.

Avec Luc {de malade / d’ ivre / d’ absent}. with Luc {of sick / of drunk / of absent} “with Luc sick / drunk / absent” Avec Luc {médecin / président / père de huit enfants}. with Luc {doctor / president / father of eight children}

6. In particular, ne-cliticization with bare predicates doesn’t seem entirely out of question in Italian. See the discussion around example (37) below.

 Roberto Zamparelli

c. d.

*Avec Max {de blond / d’ intelligent / de gros}. with Max {of blonde / of intelligent / of fat} *Avec Max {un médecin / un président / un père de huit enfants}. with Luc {a doctor / a president / a father of eight children}

In addition, she observes that it is possible to use bare predicates as a continuation to “here-is-what-happens” remarks (26), just like stage-level adjectives like sick or drunk. (26) a. b.

Ce qui se passe: Luc est médecin / président / père de huit enfants here is what happens: Luc is doctor / president / father of eight children Ce qui se passe: Luc est malade / ivre / furieux / absent, here is what happens: Luc is sick / drunk / furious / absent

Similarly, it is possible to add to bare predicates temporal and spacial specifications which sound odd when an article is present. (27) a. b.

Léa est (??une) enseignante {en ce moment / a Paris}. Léa is ( a) teacher {in this moment / in Paris} Dana été (??un) concierge {par intermittance / a Paris}. Dana has been ( a) housekeeper {intermittently / in Paris}

Though the contrast above is not reliably replicated in Italian, the following pair points in the same direction: bambino “child” (and adolescente “adolescent”) can go without a determiner in Italian when they refer to stages of life (28a), but not when they refer to types of individuals (28b): (28) a. b.

Nel 1945, Carlo era ancora bambino / adolescente. in 1945, Carlo was still child / adolescent A: Il tuo nuovo nato è un maschio o una femmina? B: E’ *(un) bambino. A: your newborn is a male ora female? B: (he) is (a) childmale

These facts clearly show that there is a relation between the possibility to be a bare predicate and the reference to “non stable states” (however formally characterized). However, the problem of linking this feature to the absence of a determiner still stands. Kupferman argues, following Wierzbicka (1986), that nouns do not directly denote properties, but rather classes of individuals which share a certain property.7 Starting from this idea, Kupferman proposes that the “zero determiner” indicates absence of quantification, and that bare predicates are used to refer to a whole undifferentiated class of individuals, which is predicated of the subject as a sort of “tag”. This intuition might have some merits, but it still leaves us in the dark with respect to the fundamental 7. “Le nom determine (spécifié) réfère un ensamble de particuliers qui ont en commun une classe de proriétés. Un adjectif renvoie à une seule propriété, alors qu’un nom spécifié fait plus que réfèrer a une propriété: le particulier est integre à l’ensamble denote par le nom.” Kupferman (1991:59).



Bare predicate nominals in Romance languages 

questions we set out to answer: why this operation should be limited to nouns that denote non-stable states? Why only in predicative positions? It should be added that the definition of “non-stable states” seems insufficient to characterize less prototypical cases of bare nouns, such as, in Italian, family relations like cugino di Francesco “cousin of Francesca” or vicino di casa di Carla “neighbor of Carla”. Munn and Schmitt (2005) (Section 4) build on Kaupferman’s proposal and try to improve the connection between the absence of the article and the class of bare nouns. Their proposal is based on the DP schema in (29). Each layer, DP, NumP and NP, can in principle be used to build a predicative structure (via adjunction of a DP subject), giving (30a-c). (29)

[DP [NumP [NP N]]] a. [np DPsubj [np Npred]] Adjunction to NP (theta-marking) b. [NumP DPsubj [NumP [np Npred]]] Adjunction to NumP (primary predication) c. [dp DPsubj [dp [NumP [np Npred]]]] Adjunction to DP (identification)

(30a) is a configuration of theta marking since the DP is within the lexical domain of N. (30b) is the canonical configuration for primary predication, where (in the terminology of Higginbotham 1985) the subject DP theta binds the open position of the nominal. (30c) is used for identificational constructions. Nouns have an external referential position which needs to be ‘saturated’ (in the Fregean sense) by the subject of the predicate nominal or by the determiner itself. Munn and Schmitt adopt four additional assumptions: (i) that English, but not Italian, has the possibility of a phonologically empty D (from Longobardi 2002); (ii) that when D is empty (De) or the DP layer is absent, a singular NumP must be spelled out as an overt determiner (a(n) in English); (iii) that the NumP layer of Romance languages can be missing, when it is not selected by a quantificational D, whereas the English NumP must always be present8; (iv) that no element which is theta marked by N can also serve as a theta-binder and check its features outside the DP. The combined result of these assumptions is that bare predicates are generally excluded whenever the predicate is construed as something larger than a simple NP. For instance, the structure for a bare predicate construction like (2a) Gianni è insegnante “Gianni is teacher” could not be (31a) (which would be an identity statement), or (31b), where NumP has an empty head Nume, contra assumption (ii). (31c) is ruled out because the Subject starts out from a position adjoined to NP, and is theta-marked

8. This is motivated by the idea that the English NumP doubles as an agreement functional head, which must always be present, whereas Romance languages (with the exception of Brazilian Portuguese) have an independent AgrP projection, so they can avoid projecting NumP without creating a problem for the agreement system. This difference is coded in a parameter, the Split Agr Parameter. See Munn and Schmitt (2005), (1998) for details.

 Roberto Zamparelli

by it: by assumption (iv), no element can satisfy the thematic requirements of the noun and saturate its open (predicative) position. (31) a. Subji be [DP ti [Dp De [NumP Nume [NP N]]]] b. Subji be [NumP ti [NumP Nume [NP N]]] c. Subji be [NP ti [NP N]] Munn and Schmitt propose that role nouns have the option of trading the standard external referential argument for an eventive argument. The structure for Cet homme est médecin can thus be (31c), or more specifically (32), where Cet homme does not receive a theta role from N and is free to check its features outside DP and function as a theta-binder for the bare NP predicate. The thematic event argument of N is saturated by the verb in this case, accounting for the dependence of the time of the nominal from the verb / utterance time. (32) [ip [dp cet homme]j [i’ esti [Vp tj [Agrp tj [Agr’ Agr [NP tj [NP médecin]]]]]]] This line of analysis is picked up by Winter, de Swart, and Zwarts (2004), who subscribe to the idea that bare predicates are bare NPs. Working on Dutch, they attribute three roles to NumP: assigning adjectival inflections, establishing number and hosting operators (in particular the Carlsonian REL(alization) operator, which extracts from a kind the set of its realizations at a given world). The authors assume (with e.g. Krifka 2003 and others) that nouns start out denoting kinds, not properties. Normally, these kinds are converted into properties at NumP, using REL. In bare predicates NumP is missing, so kinds cannot be turned into instances by REL, only by a more specialized covert operator (CAP(acity)), defined as: (33) There is a covert general capacity operator CAP of type < k < e,t >>, mapping a kind Ak to the set of individuals realizing Ak as a particular capacity. (Winter, de Swart, and Zwarts 2004 ex.A5) Unlike REL, CAP is assumed to be able to apply anywhere. The restrictions on modification are obtained by proposing that some adjectives (those that can appear in bare predicates) are functions from kinds to kinds (type ) and can appear below NumP, bearing no inflection when NumP is absent (see above ex. (7)). All the others are property modifiers («e,t>,<e,t»), always take inflections because they must appear outside NumP, and are thus excluded in bare predicates. Winter, de Swart, and Zwarts (2004) do not try to analyze CAP, but this operator must evidently apply only to those kinds which can be realized “as a capacity” (what I called a “role”). Matushansky and Specter (2004) try to provide an independent characterization of this class, and propose that they can be defined by the features [+sentient,-scalar].9 According to their proposal, nouns are often [+scalar], with 9. I have only had access to the work of these authors in the form of an (extended) abstract, so the discussion here must be taken as very preliminary.



Bare predicate nominals in Romance languages 

intensional type >>>: they combine with a degree argument, then with the external subject (<e>), then with a temporal index (), and return a proposition (<s,t>). The degree argument is fixed by an implicit standard of reference; they propose that whenever a nominal argument is satisfied DP-internally, an indefinite determiner appears. If a noun with a degree argument is directly combined with an external subject, a type mismatch arises. [-scalar] nouns, on the other hand, have no degree argument, so they can combine directly with a subject; in this case, since all nominal arguments are set outside the DP, no determiner needs to be present.

3.1

Criticism of previous accounts

The accounts just reviewed rest on a number of assumptions. Munn and Schmitt and de Swart et al. agree that bare predicates are bare NPs (allowing for the possibility of NP-internal positions for some adjectives). This is in agreement with the fact that bare predicates cannot be coordinated with normal ones (Zamparelli 1995, Sec.4.2.1): (34) Sergio è [avvocato e (*un) docente universitario]. Sergio is [lawyer and (a) teacher universitary] Intended meaning: “Sergio is a lawyer and university faculty member” The accounts disagree on the role of the indefinite determiner. Munn and Schmitt (2005) and Winter, de Swart, and Zwarts (2004) take the indefinite to be the spell-out of a functional projection with singular features, NumP; the absence of the article indicates the lack of this projection. Role nouns differ from other nouns simply in that they allow the predication to go through even in the absence of a NumP layer. For Matushansky and Spector (2004), on the other hand, the determiner is the syntactic signal of a semantic operation: an argument of the noun has been filled DP-internally. Both assumptions leave many questions unanswered. Why, of all DP-internal functional projections, NumP should be spelled-out? And why, if D is spelled out, Num does not need to be? Why only in singular count cases? In what other cases is the saturation of an argument signalled by the appearance of a lexical item (distinct from the saturator)? The first approach could at least be modelled on the obligatory presence of an overt Comp when CP is not lexically selected (as in [*(That) John is here] is odd), but independent evidence needs to be presented. Empirically, both proposals fail to account for the lack of determiners in mass / abstract nouns: (35) C’ è (?dell’) acqua/amarezza, there is (some) water/bitterness (I take “happiness” to come in degrees). Take, moreover, nouns such as lavastoviglie “dishwasher” estremità “extremity” bronzo “bronze”. They are not scalar in any intuitive sense of this term, yet determiners are obligatory with them just as well. What would their role be in this case?

 Roberto Zamparelli

A final point is that the system of Matushansky and Spector (2004) is tailored to capture the fact that predicates without determiners and predicates with determiners are roughly in complementary distribution. In particular, they report a contrast in French in small clause contexts: (36) a. b.

Pierre croit Marie physicienne. Pierre believes Marie physicist “Pierre believes that Marie is a physicist.” * Pierre croit Marie une physicienne. Pierre believes Marie a physicist

M&S 2004, ex. 10,13

This contrast is not replicated in Italian, another bare predicate language. In Italian, the meaning and distribution of bare predicate nominals is essentially a subset of the meaning and distribution of predicate nominals with the singular indefinite article: un+Npred can always be interpreted as a bare predicate.10 Matushansky and Spector account for this (and for a dialect of French which seems to behave like Italian) by assuming that (in some dialects / languages) the presence of an article can coerce a [-Scalar] noun into a [+Scalar] one. This implies that saying of someone that he is “a professor” should be a statement about degrees of professorhood, which seems to me to go too far. Note that even bare predicates can be qualified by degrees of lexical adequacy: (37) a. b.

Pablo era artista nel pieno senso del termine. Pablo was artist in the full sense of the term Con questa vittoria, Maria è campionessa a tutti gli effetti. with this victory, Maria is champion to all intents and purposes

4. Analysis The alternative I would like to propose builds on the theory of DPs developed in Heycock and Zamparelli (2005) (hereafter H&Z ), which I summarize for the relevant parts. H&Z start from the observation that the common distributional properties of plural count and singular mass / abstract nouns can only be explained by appealing to a common semantic property, cumulative / divisive reference (Quine 1960, Cheng 1973). In the semantic literature, this property has usually been captured by adopting a common structure, i.e. a semilattice (see e.g. Link 1983). Rather than assuming that this structure is associated with the lexical entry of plural and mass / abstract nouns, H&Z propose that the semilattice is generated only at a certain point in the DP structure, by 10. Notice that there are cases in Italian where the indefinite determiner must be absent, e.g. as-constructions (Come (*un) giudice, Marco è severo lit. “as (a) judge, Marco is stern”). This means that one cannot simply analyze Italian as a language where the indefinite article can optionally be disregarded as “empty of meaning”. On the difference between the Italian and the English indefinite article, see Heycock and Zamparelli (2005).



Bare predicate nominals in Romance languages 

the application of two semantic operators (one for mass, one for count nouns) to the noun’s original denotations (sets of atomic objects). The operators (OP) are associated with the head of a functional projection above NP, called PlP: (38) [DP D [NumP Dindef / cardinal [plp OP [NP (adj) N (Compl)]]]] The rationale for this choice is to force modifiers which crucially require a semilattice structure to produce a meaning (in particular, numerals and amount modifiers) to appear outside PlP. Cross-linguistically, modifiers of this type appear at the periphery of the DP, a distributional constraint which would otherwise have to be stipulated at the syntactic level. In English, semantic pluralization entails syntactic pluralization when N is count but not when it is mass. Once noun denotation and pluralization become structurally distinct, the need arises for a way to transfer information between N and the Pl head: for the semantics to work out correctly, Pl needs to “know” whether N is count or mass, and N needs to “know” whether Pl will perform semantic pluralization or not. A mechanism for sharing grammatical information is already in place in current syntactic theory: it is the feature percolation system put forth in Chomsky (1995) and refined in Chomsky (2000) and (2001). H&Z modify some of its assumptions (in particular, the idea that unvalued features are always [-interpretable]), to make it apt to transmit semantic information throughout the DP. To this effect, the authors assume that Pl has a syntactic feature Latt(-ice denotation) (set positively when semantic pluralization is performed, and negatively otherwise) and N a feature Plur (set positively when the noun has plural morphology); in addition, N has an unvalued feature Latt whose value is provided by the Pl head, and Pl, an unvalued feature Plur whose value comes from N. The mechanism of Agree will percolate feature values from valued to unvalued heads (as in e.g. (39)). The precise semantic operations performed at N and Pl will ultimately depend on the feature values present at both heads at the end of this process (see Heycock and Zamparelli 2005 for details). (39) a. [DP … [PIP Pl[PLUR=unvalued,+LATT] [NP N[+PLUR,LATT=unvalued]]]] b. [dp … [pi Pl[+plur,+latt] [np N[+PluR,+Latt] ]]]

→ agree

With this system in place, the need for an overt determiner can be derived from the necessity for N to obtain a value for its unvalued feature Latt when the operation of semantic pluralization does not take place, as it is the case with syntactically singular count nouns. Contrast (40a) and (b): in (a) one of the semantic pluralization operators transmits to N the feature value [+Latt] (i.e. “semantic pluralization is active”) and turns the denotation of dogs, a set of animals, into the set of all possible collections of dogs. In (b), “a dog”, PlP has no semantic role to play (the denotation is and must remain the set of all dogs), so a support element – in English the indefinite a(n) – is inserted in order to pass N the value [-Latt] (i.e. “no semantic pluralization”). If a determiner is not inserted, N has an unvalued feature Latt at LF and the derivation crashes.

 Roberto Zamparelli

(40) a. [dp... [pip OP[+Latt] [np dogs[latt=unvalued]] b. [dp … [pip a[-LATT] [np dog[LATT=unvalued]]]]

before Agree before Agree

The underlying assumption is that a head cannot assign feature values if it is both phonologically and semantically empty. One advantage of a feature-based system is its flexibility. We no longer need to ask why Num (as opposed to D, or Pl, or one of the many other DP-internal projections that abound in the literature, see e.g. Valois 1991) should be overtly spelled out. Any functional head would do, and if economy rules, exactly one will suffice. In practice, the -Latt feature value can come to N from an article in Pl (as H&Z argue to be the case for English a(n)), or from material at NumP or DP, as long as there is evidence of grammatical agreement between N and the DP-internal position which hosts the -Latt source. Consider in this light the relation between the subject DP and the predicate nominal in a normal predication. No subject-predicate syntactic agreement is enforced, as shown by cases of mismatched gender and number in (41a). In many languages, this contrasts with adjectives, which must agree in gender and number (41b).11 (41) a. Quelle donne sono un problema. thoseplur,fem womenplur,fem are asing,masc problemsing,masc b. * Quelle donne sono problematico. thoseplur,fem womenplur,fem are problematicsing,masc Since there is no gender / number agreement between the subject and a predicate nominal, the noun inside the predicate cannot derive a feature value from its subject. Some determiner or other must be inserted. Let’s focus on gender agreement. The notion of grammatical gender across languages is notoriously very complex (see e.g. Corbett 1983, Corbett 1991), but for the range of languages under consideration it can be reduced to (the combination of) a limited set of cases. Gender can be derived from the male / female distinction (with a neuter value in some languages), from the human / non-human distinction (the common / neuter gender opposition of Dutch, Norwegian, etc.), or from a purely lexical distinction (of a suffix, as in the neuter gender of diminutive suffixes in continental Germanic languages, or of a root, as for German Sonnefem vs. Italian solemasc, “sun”); in the latter case, gender can be distinctive (two roots with different meanings may be formally distinct only for their gender specification, see Italian finemasc “goal”, finefem “end”). All this applies to abstract gender, which in Italian can in turn have no morphological manifestation (as in certain proper names), appear as a single default morpheme (typically the masculine -o), as distinct morphological suffixes (Italian

11. “Problema” is a collective predicate. See footnote 4 for the difference between distributive and collective predicates.



Bare predicate nominals in Romance languages 

figli-o / figli-a “son / daughter”), or as allomorphic roots (fratello / sorella “brother / sister”). Pre-theoretically, role nouns seem to refer to roles which can be acted or taken up by human beings (more on this in Sec. 5). Due to this semantics, they don’t seem to be lexically specified for abstract gender like Italian sole or fine; rather, they adopt the syntactic gender specification of the noun they apply to. To see this, consider the following scenario. In Italy, Letizia Brichetto Arnaboldi Moratti, a woman, is currently the Minister of Education (Ministromasc della Pubblica Istruzione). There is no female version of ministro: the default masculine form is used for both sexes, so it is possible to say (42a). However, when we turn to role nouns which have a double gender form (e.g. cugin-o / -a “cousin” dottor-e / -essa “doctor”), we see that when an article is present the gender of the predicate nominal can follow the sex of the person who occupies the position of minister, but when the article is absent, as in (42c), the gender is sensitive to the syntax, and the female forms become impossible.12 (42) a. Letizia Brichetto Arnaboldi Morattii è Ministromasc della Pubblica Istruzione. Letizia Brichetto Arnaboldi Morattifem is ministermasc of Education b. Il ministroi è {una cugina di Carla / una dottoressafem in scienze politiche} themasc ministermasc is {afem cousinfem of Carla / a doctorfem in political science} c. * Ilmasc ministroi è {cugina di Carla/dottoressa in scienze politiche}.

the ministermasc is cousinfem of Carla / doctorfem in political science}

d. Letizia Morattifem è (una) {cugina di Carla / dottoressa in scienze politiche} Letizia Morattifem is afem {cousinfem of Carla / doctorfem in Political Science} Suppose, with Munn and Schmitt, that the subjects of role noun predication can start out from a position closer to NP, close enough to trigger obligatory feature agreement between the predicate and the subject, much like with adjectives. Suppose, for consistency with the system of Munn and Schmitt, that this position is an NP adjunct (alternatively, it might be a PlP adjunction – nothing crucial hinges on this choice). (43) Giannii è [np ti [np dottore]] Like other grammatical features, ±Latt can percolate up to the DP projection. We can thus assume that a DP is marked +Latt if plural or mass, -Latt otherwise. From its 12. In this case, the masculine form of the bare predicate is also felt unnatural. I take it that this is due to a semantic clash with the sex of the person behind the post. The correct characterization is thus that normal predicates are only sensitive to the semantics, bare predicates are sensitive to both syntax and semantic.

 Roberto Zamparelli

NP-adjunct position, the DP subject can thus transmit ±Latt (among other features) to the noun in the predicate nominals. A determiner is no longer necessary, as desired. The question that remains to be considered is why not all predicates allow NPadjunct subjects. Munn and Schmitt’s answer is that the predicate discharges a theta role on the NP-adjunct position, and theta marked elements cannot serve as external subject. This idea, however, equates “being in a position which can be theta marked” with “receiving a theta role”, disregarding the possibility that there may be nouns with no thematic role to assign. Take the contrast between: (44) a. John’s portrait b. John’s planet The preferred interpretation for (44a) is that John is depicted by, owns or has painted the portrait, though other relations are possible. In contrast, (44b) is open to any contextually salient relationship between John and the planet. The standard way to capture this contrast is that portrait has one or more theta roles to assign to the possessor phrase whereas planet has none. If this is the case, it is not clear what kind of theta-role a subject adjoined to a planet-type NP might possibly receive. An alternative explanation starts from the assumption made above that role nouns are underspecified for feature values and in particular, for (abstract) gender. A value for gender cannot be acquired from the indefinite articles which are inserted to provide a value for Latt: these articles come in masculine / feminine versions and, just like adjectives, adjust to the gender normally associated with the noun. A gender value for a feature-defective predicate nominal can only be obtained from the DP subject, or from a pro-form coindexed with it. The latter case is probably realized in bare appositions such as13: (45) L’ Alfa Romeoi, ora proi società del gruppo FIAT, è del 1910 The Alfa Romeo, now pro firm of the group FIAT, is from 1910 The former case – I propose – is what we find in the bare predicate construction: the DP subject is in an adjunct position, close enough to transmit agreement information to the noun. Suppose, specifically, that in a normal predicational small clause the DP subject originates outside the domain of Agree (for instance, adjoined to NumP, with numerals and indefinites in Num), but it can originate inside (i.e. adjoined to NP) if and only if this is necessary to transmit feature values which couldn’t be obtained otherwise. When a predicate has no gender specification, the DP subject must appear low, i.e. NP-adjoined, and transmit a value for gender. From this position it can also transmit Latt values. Determiner insertion becomes unnecessary. 13. In example (45), I take pro to be generated in the position normally occupied by a determiner (Num or D), and to be able to transmit ±LATT specification as well. As a result, the range of predicates and predicate modifiers in appositions is much wider than in predicative constructions.



Bare predicate nominals in Romance languages 

The rationale for the view that subjects are closer to predicates only when this is necessary to transmit agreement information on gender (and syntactic number, which I have set aside in the discussion) is that Agree should not be vacuous: two nouns fully specified for features within the same Agree domain would trigger an ‘unnecessary’ agreement operation. This hypothesis has some implications for the distribution of elements inside DP. If a noun cannot appear within the Agree domain of another noun unless it actually agrees with it, multiple Ns in the same DP should be impossible unless they are in positions from which agreement is blocked. One case is coordination. The fact that an attributive adjective in a conjunct cannot modify a nominal element in the other conjunct ((46), contrasted with “John is stupid and a liar”) shows that conjuncts are not in an agreement relation with each other (though each conjunct can agree with material outside the coordination). As a result, multiple nouns are possible (47). (46) *[dp This [&p [AdjP stupid ] and [np liar]]] Intended meaning: “This stupid liar” (47) [dp This [&p [np liar ] and [np cheat]]] Another case is N-N compounds. These constructions are possible because the nonhead noun is “locked in” the morphological component, thus invisible to syntactic processes (e.g. pronominalization, cf. *A donkeyi owner beats iti/onei/itselfi), and excluded from the domain of agreement. A third case might be predicative adjectives. In German, adjectives agree (to various degrees) when appositive, but do not agree when in predicative position. In the present system this could be captured by saying that in this language adjectives can shed their inflectional suffixes and the unvalued gender feature along with them, if this can spare them from agreeing with a DP subject (in other terms, removing the element that must agree is less ‘costly’ than adopting the marked syntactic configuration, with an internal DP subject, which is required for agreement). In the attributive position the adjective is already inserted in an Agree domain, so dropping the inflection would not buy anything. Leaving these sketchy suggestions unexplored, we can now return to the main ideas on the syntax of bare predicates. The proposal is that the crucial syntactic difference between role nouns and other nouns is their lack of inherent abstract gender, which forces the subject to appear quite close to the predicative noun and leads to small-sized predicates. Schematically: (48) be [NP [dp[-Latt,-Gender] Gianni] [NP dottore[GENDER=unvalued, LATT=unvalued]] This structure inherits two virtues of Munn & Schmitt’s proposal: first, it preserves the possibility (but not the necessity) of a theta-marking relation between the predicate and the copular subject. Second, since the predicate is structurally small, modifiers that appear outside the attachment point of the subject are automatically excluded. A case in point might be relatives, which are never compatible with our bare predicates. The absence of the affix in Dutch modified predicates (ex. (7) above) could also be

 Roberto Zamparelli

linked to the absence of a functional head which normally licenses the suffix. However, the size of the predicate alone cannot exclude many modifiers which appear after the noun, a position which – following Cinque (1994) – I take to be very DP-internal. I will return to this problem in Section 5.3.14

Bare nouns in English

4.1

Let’s now consider the behavior of English predicates. Bare singular count predicates are possible in this language only if they have a presupposition of uniqueness (Stowell 1989) (see (49) vs. (50), whose equivalent would be grammatical in Italian). (49) a. George is {president of the United States. / head of school. / deputy general.  / treasurer of the association.} b. The queen appointed her lover treasurer of the realm. c. They elected John king of Marvin’s gardens. (50) * George is employee / doctor / manager at an important firm / cousin of one of my dearest friends Why does English, unlike its continental Germanic cousins, have no generalized bare predicates with role nouns? The key is that English has no grammatical gender (i.e. gender is present in this language only as a notional category: female being – and ships – require “she”, male beings “he”, inanimates “it”). If the assumption in the previous section is correct and nouns cannot appear in syntax within the Agree domain of another noun unless one of them needs to transmit feature values via Agree, the lack of gender features in English automatically entails that the subject of the predication must start out in an external position even with predicative role nouns. Syntactically, in this language role nouns are just like all other predicate nominals. This leaves out the cases in (49) as an exception. I will return to these cases in Section 5.1. 14. The restriction to [+HUMAN] subjects might also be an effect of the small size of the predicate, if this setting is taken as a default value which appears whenever the DP lacks certain features. The relevant features are normally associated with the presence of a noun, as shown by the following Italian minimal pair:

(i) a. Nessuno strumento era pronto.



No-one tool



“No tool was ready”



was ready

b. Nessuno era pronto.



No-one was ready.



“No-one (=no human being) was ready”

However, they could also be associated with a functional projection immediately outside NP, which would be missing in bare predicates.



Bare predicate nominals in Romance languages 

5. Role nouns The syntactic account presented in the previous section is not fully satisfactory until the class of what I have called “role nouns” is semantically characterized. Without a semantic side of the explanation, one could wonder why common nouns which are arbitrarily specified for gender (e.g. the words for “sun”) couldn’t simply lack a gender specification, and behave exactly like our role nouns. The first thing to consider, having reviewed the range of nouns which can be bare, is that a unitary characterization of all role nouns seems unlikely. To say that they are “eventive”, with no further qualification, goes too far: nouns that indisputably denote (the nominalization of) events, such as arrivo “arrival”, distruzione “destruction”, etc. cannot form bare predicates. Saying that they are “stage-level” also seems too strong. (51) illustrates a well-known contrast between canonical stage- and individual-level adjectives: only the former can have their event variable bound by an unselective adverb of quantification. (51) a. Un Norvegese è spesso disponibile / presente / ubriaco. a Norwegian is often available / present / drunk. Possible meaning: “A Norwegian is available/present/drunk many times “ b. Un Norvegese è spesso alto / poliglotta / biondo, a Norwegian is often tall / polyglot / blond Only meaning: “Many Norwegians are tall/polyglot/blond” (52) shows that in the same construction bare predicates pattern more like individual than like stage-level objects. It not easy to obtain a reading saying that a (perhaps specific) Norwegian is a professional fisherman repeatedly / various times through his life.15 (52) Un Norvegese è spesso pescatore / operaio petrolifero, a Norwegian is often fisherman / worker in petrol plants Hard-to-get meaning: “A Norwegians is a fisherman /petrol worker many times” The characterization [+sentient,-scalar] is also too broad: nouns like scopritore “discoverer”, condottiero “leader”, or inquilino “tenant”, etc. have the right features, and correspond to our intuitions for what a “role” might be, but cannot form bare predicates. (53) a. Colombo fu *(lo) scopritore dell’ America. Colombo was (the) discoverer of America b. Pietro fu *(un) apostolo. 15. I should make clear that this meaning is not impossible, but surely harder than in e.g. Gianni è spesso un pescatore fortunato “Gianni is often a lucky fisherman” or Ugo è spesso un modello per suoi amici “John is often a model for his friends”, which cannot drop the determiner in Italian. The point is that this effect seems to depend more on how stable we perceive the noun meaning to be than on the presence / absence of the determiner.

 Roberto Zamparelli

c. d. e. f..

Pietro was (an) apostle S.Agata fu *(una) martire. S.Agata was (a) martyr. Cesare è *(un) condottiero / eroe. Cesare is (a) leader / hero Carlo è *(un) lettore di fantascienza. Carlo is (a) reader of science fiction Ugo è *(un) inquilino / malvivente. Ugo is (a) tenant / criminal

Kupferman observes that sometimes bare predicates are possible to the extent we can imagine their denotation to identify a stable class of individuals. It seems that the property which is predicated of the subject must express an important classification, and not be reducible to an attitude or habit. This could rule out inquilino “tenant”, which does not seem to be a defining feature. Similarly, having the habit of walking is not a person’s defining feature, whereas being a racing car pilot can certainly be. (54) a. Michael è *(un) camminatore. Michael is (a) walker b. Michael è (un) corridore automobilistico. Michael è (un) runner automobilistic “Michael is a racing car pilot” Yet, subkinds of human beings (like ragazzo “boy”, bambino “child”) which satisfy this requirement cannot appear bare unless they refer to stages of life. At the same time, family relations like brother, daughter which have the look-and-feel of stable states can form bare predicates in some languages.

5.1

Classes of role nouns

To sort out these puzzling data, I will start by dividing role nouns into three (potentially overlapping) classes: “pure relational”, “unique descriptions” and “professions”. We can take them in turn. “Pure relational” cases include family relations (fratello di “brother of ”, gemello di “twin of ”, but also (at a higher stylistic register) more abstract cases such as padrone di “owner of, signore di “master of ”, causa di “cause of, motivo di “reason of ”, etc. They differ from other cases in three respects. First, they require that all the arguments of the relation be saturated by overt material. Contrast (55b) with (b’), where the argument of “sister” can be easily inferred from the question in (a), yet the bare predicate is impossible. (55) a. Question: Ada, in che relazione stà con Francesco? Ada in which relation is with Francesco? b. Answer: Beh, Ada è {una /la / 0} sorella di Francesco, well, Ada is {a / the / 0} sister of Francesco



Bare predicate nominals in Romance languages 

b’. Answer: Beh, Ada è {una/la / *0} sorella. well, Ada is {a / the / 0} sister Second, they are not always restricted to humans (56), and do allow certain extensions of usage (57) (from Google). (56) La notizia fu {motivo / causa / fonte} di grande preoccupazione. The news was {reason / cause / source} of great worry (57) Il palazzo era gemello di altri sette edifici di nove piani. The building was twin of other seven buildings with nine floors Third, they are certainly the least ‘eventive’ or ‘stage-level’ of the group. Typologically, at least family relations are systematically impossible as bare predicates in some languages (e.g. German). The idea that bare predicates are bare NPs can help shed some light on the peculiarities of this class. Suppose that, as relations, these nouns have an obligatory internal argument to theta-mark. When no overt complement is present, the theta role must be assigned to a pro-possessor outside NP: (58) Marco è [NumP un pro [NP fratello]] Marco is [NumP a pro [NP brother]] When the predicate is a bare NP, the pro has no functional projection to be merged in, and the internal argument can only be assigned to the PP complement. When this is missing, the bare predicate is ill-formed. In turn, the presence of a complement could help license the bare noun, if DP complements are allowed to raise and adjoin to the nominal that selects them, as I have proposed for daughter in (59) in Zamparelli (1995), Ch. 5 (these are the “weak definites” of Poesio 1994). (59) a. I met the daughter of a linguist No presupp. that the linguist has only 1 daughter b. I met the daughter of the linguist Uniqueness both for linguist and for daughter c. LF: [DP [DP the/a linguist]i the daughter of ti] This additional licensing possibility could explain the relative freedom of this subclass of bare nouns.16 The second group, “unique descriptions” contains relations with a presupposition of (relative) uniqueness, like presidente, “president”, re “king”, capo dell’istituto “head of school”, etc. In Italian, the role nouns introduced by capo “head / boss” probably fall in this group. This is also the type of bare predicates which tends to 16. It is tempting to enroll in this group nouns of nationality and religious beliefs, treating them as relations whose internal argument (X and Y in “citizen of Y”, “adept of X”) has incorporated. On the other hand, these are also the cases for which it is more difficult to exclude an adjectival derivation. I leave the question open.

 Roberto Zamparelli

be possible in English (see Stowell 1989 and Longobardi 1994 for discussion), and partially overlaps with the class of “titles” (social roles one can be formally appointed to). The complement is possible but in general not obligatory. Compare: (60) a. Ada è figlia di Giovanna. Ada is daughter of Giovanna b. Ada è figlia unica. c. *Ada è figlia.

“Pure relational “ case unique description: “only child” no complement, no uniqueness and not a profession

The uniqueness presuppositions, similar to those of proper names, suggest that raising of the head noun to D (or of the NP to [Spec,DP] as in Heycock and Zamparelli 2003) might play a part in their licensing, but I will not attempt an analysis here. The third class is formed by the profession nouns proper, which I have examined in some detail in the previous sections, plus a number of activities (e.g. being a ropeparty leader, an Alpinist, etc.) which are not technically professions, but “sufficiently characterizing” in some respect. I will now try to spell out in more details the features of this class.

5.2

Role nouns as activities

Let’s start from the assumption that nouns denote properties, but that some properties (sometimes called sortal properties) can be used to identify classes or kinds of objects with some particular cognitive salience. In the literature on genericity and kinds, many have proposed the existence of “natural” or “well-established” kinds of objects17, sometimes treated as cognitive or ontological primitives. The relation between sortal properties and well-established kinds is a matter of much philosophical discussion (see e.g. Moltmann 2004). Perhaps well-established kinds are just another name for sortal properties, or for ‘definitional’ properties (if they may in turn be defined), or just properties whose persistence through time gives us a tool to consistently identify an object. Whatever approach is taken, there are linguistic phenomena which are sensitive to this notion, in particular the possibility of using the singular definite determiner with a generic meaning. Consider examples such as (61) (attributed to Partee) and (62), discussed in Lawler (1973) and Carlson (1977). (61) a. The Coke bottle has a narrow neck, b. ??The green bottle has a narrow neck. (62) a. ??The container is common, b. * The object is everywhere.

17. See e.g. Lawler (1973), Carlson (1977), Chapt.7, Krifka et al. (1995), Dayal (2004).



Bare predicate nominals in Romance languages 

Coke bottles are a well-established kind, green bottles are not. Moreover, in the absence of special (usually technical) contexts, natural kinds cannot be ‘too general’, so “Coke bottle” in (61) can be a natural kind for the purpose of the definite generic, but nouns like “container” or “object” in (62) cannot. Linguistic examples of properties which define well-established kinds are almost invariably concrete, but one can of course imagine that such kinds may be identified by different means. Suppose, in particular, that activities which are typical of natural classes of entities could also be used to individuate them (let’s call them well-established activities), and that the third class of bare predicate nominals (thus excluding ‘pure relational’ and ‘unique description’ cases like fratello “brother” and re “king”) can denote such activities.18 One point of similarity between the nominals that can appear in bare predicates and definite generics is the exclusion of categories which are ‘too general’: out of special contexts (in which, for instance, “worker” is contrasted with the class of “unemployed”), (63) suffers the same problem as (62a,b). (63) a. b.

Ugo è *(un) lavoratore. Ugo is (a) worker Ugo è *(un) professionista. Ugo is (a) professional

What I want to propose is that the profession nominals that can make up bare predicates are ambiguous: they can denote classes of human beings, or they can denote the abstract well-established activities which identify those classes. In the first case, they are just like all other nouns: lexically set for gender values, requiring a determiner to deliver Latt as detailed above, and taking a wide range of modifiers. In this case, the copular construction is simply interpreted as membership (arguably the default value for copular predication). (64) Ada è una dottoressa. Ada is a doctorfem Ada ∈ (doctor’ ∩ female’) In the second case, the role nominal denotes an activity (an abstract notion), which in turn defines a natural / well-established class of people. In this case the noun has no gender value of its own, and assigns a theta role to the subject of the predication in the NP-adjunct position. The theta-marked argument is identified with the agent that participates in the activity (indicated as acting-as-N in a Parsons-style semantics). 18. What I have in mind is a formal characterization of activities (and other types of eventualities) which treats them as ontological primitives, or at least does not reduce them to functions from worlds to the set of people who are engaged in the activity in that world. I believe that the activities and the sets of their participants should both be analyzed intensionally, but not collapsed together. Indeed, I think that the very existence of bare predicates can be seen as an argument to keep the two notions distinct. This said, I will not give a formal ontology for activities in this paper.

 Roberto Zamparelli

(65) Ada è dottores. Ada is doctorfem ∃e[acting-as-doctor'(e) ∧ Subj(e,Ada)] Restricting the denotation of this class of bare predicates to activities excludes cases like martire “martyr” or scopritore “discoverer”, which are achievements (in Vendler’s 1957 sense). It also excludes inquilino “tenant”, apostolo “apostle” or malvivente “criminal”, which are probably classified as states, more than activities. The exclusion of obiettore di fantascienza “science fiction reader” is due, I suggest, to the fact that the classes these activities individuate are not sufficiently well-established. This is of course far from a hard-and-fast constraint, but it goes with the intuition that such activities are not primary occupations (imagine asking a child what she would like to do when she grows up: answers like “the tenant” or “the science fiction reader” sound not only lazy, but whimsical). The exclusion of metaphorical and extended meanings can now be attributed to the fact that metaphorically interpreted nominals no longer refer to the appropriate types of activities. “Butcher” in “That surgeon is a butcher” means something along the line of “person who (in dealing with bodies) has the sloppiness, carelessness and violence characteristic of the stereotypical professional butcher”. This is most certainly not an activity. The idea is that metaphorical meanings can be extracted from the stereotypes of the classes of people identified by the activities, not from the activities themselves. These remarks apply to the third class of bare predicate that I have distinguished, but it is possible for a noun to be a member of more than one class of bare predicates. Take for instance padre “father”: it can be used without a complement (66a), and in this case is has an eventive ring; or it can come with a complement, this time with no particular eventual connotation (66b). Apostle, unsuitable as a pure activity, makes a fine relation and thus a bare predicate when a complement is present (at a high stylistic register, and with an extended meaning). (66) a. b.

Aldo è padre (da appena mezz’ora). Aldo is father (since just half an hour) Aldo è padre di cinque figli. Aldo is father of five children

(67) Gandhi fu apostolo della non violenza. Gandhi fu apostle of non violence

5.3

Modification

As we have seen, Winter, de Swart, and Zwarts (2004) and Kupferman (1991) assume that nouns natively denote kinds, rather than properties (sets of individuals at a given world), and that they are converted into properties at a certain point in the DP



Bare predicate nominals in Romance languages 

structure. The modifiers which would be acceptable in bare predicates would be just those that denote functions from kinds to kinds. One of the main advantages of the view that nouns denote kinds19 is that the treatment of kind-denoting bare plurals like dogs in dogs are mammals becomes straightforward: bare plurals would simply be cases where the native denotation of nouns, which is normally turned into a property and used to restrict the determiner, appears in its original form, perhaps with the help of a silent movement of the kind-denoting element (N or NP) to a high “referential” position (D or [Spec,DP]), as proposed in Longobardi (1994). The Achilles’ heel of this elegant theory is precisely the treatment of adjectives and modifiers. As is well known, bare plurals in a language like English can be fully modified by adjectives, PPs and relatives, and still remain kind-denoting: (68) [Large carnivorous dinosaurs with the ability to control their system temperature] appeared on Earth 200 million years ago. If dinosaurs denotes a property (the set of all dinosaurs at a certain index), the semantics of the modifiers can simply work by intersections, once some scalar parameters are fixed (e.g. what counts as “large”). The kind can then be formed by association with the set of its instances in all possible worlds: (69) ιxk[∀y ☐ [dinosaur(y) ∧ carnivorous(y) ∧ large-for-a-dinosaur(y) ∧ abilityto-control-body-temperature(y) ←→ REL(y,xk)]] Suppose instead that dinosaurs denotes a kind (the proper name of a class of animals), carnivorous a property (the set of carnivorous animals) and carnivorous animals again a kind. To construct the denotation of carnivorous animals one needs to ‘unwrap’ from the dinosaur kind the set of dinosaur instances, intersect them with the modifier and then ‘rewrap’ the result into another kind: (70) ιxk[∀y ☐ [REL(y,dinosaurs) ∧ carnivorous(y) ∧ REL(y,xk)]] Though technically feasible, this semantics flies in the face of the idea that the decomposition of a kind into the set of its instances is a restricted operation, performed only at a specific functional projection. Assigning the unwrapping / rewrapping process to the semantics of adjectives of type , as Winter, de Swart, and Zwarts (2004) implicitly propose, is not a more elegant solution, since it does not address the general problem of modified bare plurals like (68), which can contain any modifier whatsoever with the sole exception of those which are anchored to the utterance point or time (as pointed out by Carlson, dogs in the next room or parts of this machine can hardly refer to kinds). Even looking at modified bare predicates, which are far more limited, 19. This approach has been put forward in Krifka 1995, Zamparelli 1995, and it has recently been adopted by various authors working on the syntax / semantics interface. Chierchia (1998) takes it to be a parameter among the world languages. See also Krifka (2003).

 Roberto Zamparelli

the rigid division between -type and <<e,t>,<e,t>>-type modifiers seems too strong. Take anziano “senior” vs. scelto “chosen” in (71). Each can make bare predicates only with a certain choice of nouns. Are they or <<e,t><e,t>> (71) a. b.

Carlo è membro {anziano / *scelto} del club. Carlo is member {senior / chosen} of the club “Carlo is a senior / chosen member of the club” Carlo è tiratore {*anziano / scelto}. Carlo is shooter { senior / chosen} “Carlo is a sharpshooter / an old shooter”

The alternative semantic approach I am pursuing here is based not on the distinction between kind- and property-modification, but between modifiers which apply to the abstract activities or relations that bare predicates can denote and modifiers which apply to the people identified by those activities / relations. Only the former are allowed in bare predicates, provided their application returns an activity / relation which is also “well-established”. In addition, modifiers which only occur outside NP are excluded, for structural reasons. These are however a minority in Italian, so additional semantic constraint are certainly needed to exclude many NP-internal adjectives. A case in point is Italian diminutive suffixes such as -in-, -ett- “small / little”. They are N-internal, and always block bare predicates: (72) a. b.

Carla era (una) cugina / sorella di Salvo. Carla was (a) cousin / sister of Salvo Carla era *(una) (cugin-ett-a / sorell-in-a) di Salvo. Carla was (a) little-cousin / little-sister of Salvo

The point is that these suffixes do not modify the property “being a sister”: -in- “small / little” in (72b) does not mean that Carla’s sisterhood is in any sense small or minor, only that the person identified as a sister is small (of age or size). The presence of this suffix forces sorella to refer to the set of human being who are sisters of someone, not to the relation of sisterhood itself. The bare predicate is thus excluded.

6. Conclusions In this paper I have proposed a syntactic and semantic account for the appearance of certain nominals without the article in predicative position. The main idea is that the nouns at issue denote activities or relations, and are defective for the feature Gender. The function of the article is taken up by the DP these nouns theta-mark. If correct, the account points to the existence of three fundamental ways in which nouns can avoid taking articles: a noun can be licensed by a semantic operator (when mass or plural), by movement to D or [Spec,DP] (proper names, coordinated Ns, maybe kind-denoting



Bare predicate nominals in Romance languages 

bare plurals in Germanic), and by coindexation with a DP which is independently licensed (in bare predicate nominals and probably in appositions). This classification can help frame the next natural research question: what is the exact typological status of the article / no article distinction? Can these same types be detected even in languages that have no articles? In how many ways can N be licensed in these languages?

References Anscombre, J.-C. 1986. L’article zéro en francais: Un imperfait du subjonctif. Langue francaise 72: 4–39. Anscombre, J.-C. 1991. La détermination zéro: Qualques propriétés. Langages 102: 103–123. Borthen, K. 2003. Norwegian Bare Singulars. PhD dissertation, Norwegian University of Science and Technology. Carlson, G.N. 1977. Reference to Kinds in English. PhD dissertation, University of Massachusetts at Amherst. Cheng, C.-Y. 1973. Response to Moravcsik. In Approaches to Natural Language, J. Hintikka, J. Moravczik & P. Suppes (eds), 286–288. Dordrecht: D. Reidel. Chierchia, G. 1998. Reference to kinds across languages. Natural Language Semantics 6: 339–105. Chomsky, N. 1995. The Minimalist Program. Cambridge MA: The MIT Press. Chomsky, N. 2000. Minimalist inquiries: The framework. In Step by Step, R. Martin, D. Michaels & J. Uriagereka (eds). Cambridge MA: The MIT Press. Chomsky, N. 2001. Derivation by phase. In Ken Hale: A Life in Language, M. Kenstowicz (ed.), 1–52. Cambridge MA: The MIT Press. Cinque, G. 1994. Partial N-movement in the Romance DP. In Paths Towards Universal Grammar, G. Cinque et al. (eds), 85–110. Washington DC: Georgetown University Press. Contreras, H. 1986. Spanish bare NPs and the ECP. In Generative Studies in Spanish Syntax, I. Bordelois, H. Contreras & K. Zagona (eds). Dordrecht: Foris. Corbett, G. 1983. Resolution rules: Agreement in person, number, and gender. In Order, Concord and Constituency, G. Gazdar, E. Klein & G. Pullum (eds), 175–206. Dordrecht: Foris. Corbett, G.G. 1991. Gender. Cambridge: CUP. Dayal, V. 2004. Number marking and (in)definiteness in kind terms. Linguistics and Philosophy 27(4): 393–450. Delsing, L.-O. 1993. The Internal Structure of Noun Phrases in the Scandinavian Languages. PhD dissertation, University of Lund. Heycock, C. & Zamparelli, R. 2003. Coordinated bare definites. Linguistic Inquiry 34(3): 443–169. Heycock, C. & Zamparelli, R. 2005. Friends and colleagues: Plurality, coordination, and the structure of DP. Natural Language Semantics 13: 201–270. Higginbotham, J. 1985. On semantics. Linguistic Inquiry 16: 547–593. Krifka, M. 1995. Common nouns: A contrastive analysis of English and Chinese. In The Generic Book, G. Carlson & F. Pelletier (eds), 398–411. Chicago IL: University of Chicago Press. Krifka, M. 2003. Bare NPs: Kind-referring, indefinites, both or neither? In Proceedings of Semantics and Linguistic Theory (SALT) 13 [CLC Publications]. Ithaca NY: Cornell.

 Roberto Zamparelli Krifka, M. et al.  1995. Genericity: An introduction, 1–124. Chicago IL: University of Chicago Press. Kupferman, L. 1991. Struture événementielle de l’alternance un / 0 devant les noms humains attributs. Languages 101: 52–75. Lawler, J. 1973. Studies in English Generics. University of Michigan Papers in Linguistics 1. Link, G. 1983. The logical analysis of plurals and mass terms: A lattice-theoretical approach. In Meaning, Use, and the Interpretation of Language, R. Bauerle, C. Schwarze & A. von Stechow (eds), 302–323. Berlin: Walter de Gruyter. Longobardi, G. 1994. Proper names and the theory of N-movement in syntax and logical form. Linguistic Inquiry 25: 609–665. Longobardi, G. 2002. How comparative is semantics? A unified parametric theory of bare nouns and proper names. Natural Language Semantics 4(9): 335–369. Matushansky, O. & Spector, B. 2004. Tinker, tailor, soldier, spy. Talk presented at Sinn und Bedeutung 9 (University of Nijmegen). Moltmann, F. 2004. Properties and kinds of tropes: New linguistic facts and old philosophical insights. Mind 113(449). Munn, A. & Schmitt, C. 1998. Against the nominal mapping parameter: Bare nouns in Brazilian Portuguese. In Proceedings of NELS-29. Munn, A. & Schmitt, C. 2005. Number and indefinites. Lingua 15(6): 821–855. Poesio, M. 1994. Weak definites. Proceedings of SALT-4. Pollock, J.-Y. 1983. Sur quelques propriétés des phrases copulatives en Francais. Langue Francaise 58: 89–12. Quine, W. 1960. Word and Object. Cambridge MA: The MIT Press. Renzi, L. 1995. L’articolo. In Grande Grammatica Italiana dì Consultazione, Vol. I., L. Renzi, G. Salvi & A. Cardinaletti (eds). Bologna: Il Mulino. Skytte, G., Salvi, G. & Manzini, R. 1991. In Grande Grammatica Italiana di Consultazione, Vol. 2: Frasi subordinate all’infinito, 483–570. Bologna: Il Mulino. Stowell, T. 1989. Subjects, specifiers and X-bar theory. In Alternative Conceptions of Phrase Structure, M. Baltin & A. Kroch (eds). Chicago IL: The University of Chicago Press. Valois, D. 1991. Internal Syntax of DP. PhD dissertation, UCLA. Vendler, Z. 1957. Verbs and times. The Philosophical Review 66: 143–160. Wierzbicka, A. 1986. What’s in a noun? (Or: How do nouns differ in meaning from adjectives). Studies in Language 10(2): 353–389. Winter, Y., de Swart, H. & Zwarts, J. 2004. Bare predicate nominals in Dutch. In Proceedings of “Sinn undBedeutung”, Vol. 9. Zamparelli, R. 1995. Layers in the Determiner Phrase. PhD dissertation, University of Rochester. (Published by Garland, 2000). Zucchi, S. 1989. The Language of Propositions and Events: Issues in the Syntax ond Semantics of Nominalization. PhD dissertation, University of Massachusetts at Amherst.

Definiteness effects and the role of the coda in existential constructions Manuel Leonetti

Universidad de Alcalá The Definiteness Effect (DE) in existential contexts appears as a robust constraint in some languages (Spanish) while it seems to be absent in others (Italian, Catalan). However, a closer inspection of Italian and Catalan data shows that the DE has some presence in those languages as well, when the coda of the existential is explicit and occurs inside the VP. This paper investigates the effects produced by the coda on definiteness, and connects such effects to other constraints on the licensing of postverbal subjects, all ultimately tied to information structure. I suggest that a clash between definiteness and Focus structure is at the origin of the DE, when definite expressions resist insertion into pure thetic or Broad Focus sentences.

1. Introduction The so-called “Definiteness Effect” (from now on, DE) in existential contexts is usually considered one of the basic diagnostics for definiteness. In fact, checking the possibility of inserting a nominal expression after there be (or its equivalents in different languages) is the most direct way to assess whether that expression behaves as a definite. It is uncontroversial that making some progress in our knowledge of the DE also means making some contribution to our knowledge of the semantic property of definiteness and, more indirectly, to our understanding of how determination works in natural languages. Thus, in this paper I intend to make a (modest and mostly descriptive) contribution to our knowledge of the semantics and pragmatics of definiteness by examining certain factors involved in the crosslinguistic distribution of the DE. My main concern is the relationship between definiteness and information structure. My immediate goal will be that of accounting for a cluster of data, mainly from Romance languages, which may throw some light on the nature of the DE. In doing this I hope to provide evidence for a number of claims: (a) a classification of the different subtypes of existential constructions is needed, because only some of them should be associated with the DE; (b) some manifestation of the DE seems to emerge even in

 Manuel Leonetti

languages that apparently do not show such a constraint, like Italian and Catalan; (c) the DE is essentially a semantic and pragmatic phenomenon, but syntax obviously plays a decisive role, particularly in the encoding of information structure; (d) all general constraints having to do with definiteness or specificity are ultimately related to information structure. Hopefully the discussion will provide some ideas for the evaluation of current approaches to the DE. I intend to limit my observations to the behaviour of definite articles, thus excluding demonstratives as well as universal or strong quantifiers.1 I will consider only existential and unaccusative constructions as the main contexts for the DE, without extending the analysis to other grammatical phenomena such as Extraposition of relative clauses or modifiers in English, or constructions with have in different languages. The paper is organized as follows. Section 2 is a brief presentation of the DE in general terms. The original problem that attracted my interest, the apparent lack of DE in Italian and Catalan, is introduced in section 3, and the main facts concerning the role of the coda in existential contexts are treated in section 4, with some crosslinguistic evidence for what I will call the Coda Constraint: in section 4 it is argued that, under closer scrutiny, the DE is not absent in Italian and Catalan existentials. Section 5 is devoted to a constraint on definite postverbal subjects that is clearly related to the effects of the coda on definiteness. An informal proposal is advanced in section 6 concerning the connection between definiteness and Focus structure that is at the basis of the DE: at this point all the data presented in the preceding sections are brought together under a unified account. Finally, section 7 develops one of the issues that are directly involved in the discussion of the counterexamples to the DE: the existence of different kinds of “existential constructions”, of which only one is associated with the DE. Given the resulting typology of existentials, the three classical proposals for a syntactic analysis of there be clauses in English happen to be correct in some sense. Section 8 contains a summary and some concluding remarks, together with some speculations on definiteness / specificity constraints in grammar.

2. The Definiteness Effect The examples in (1) and (2) present the well known basic facts of the DE in English and Spanish: there be and haber constructions are compatible with indefinite DPs and bare nouns, but incompatible with definite DPs. (1) a. There are {some / two / many / few / no / Ø} dogs. b. *There is {it / the dog / that dog / Fido}.

1. Following McNally (1992, 1998), I assume that a unified account of the DE covering both definites and universal quantifiers is not necessarily an optimal solution.



Definiteness effects and the role of the coda in existential constructions 

(2) a. b.

Hay {algunos / dos / muchos / pocos / Ø} perros. Have {some / two / many / few / Ø}dogs *Hay {él / el perro / ese perro / Fido}. Have {it / the dog / that dog / Fido}

There are basically two lines of thought in current accounts of the DE: one is based on the presuppositional nature of definite / strong determiners and the clash between this presupposition and the felicity conditions for existential contexts (Lumsden 1988, Zucchi 1995); the other is based on the predicative, non-referential, property-denoting or incorporated nature of weak determiners, as a way to obey the licensing conditions imposed by existentials (Milsark 1977, McNally 1992 and 1998, Zamparelli 1996, Van Geenhoven 1998, Bende-Farkas and Kamp 2001, Landman 2004). Most current accounts rely on a Novelty Condition associated with existential sentences: the referent of the internal DP must be hearer-new (McNally 1992, Ward and Birner 1995). It is an open issue whether the Novelty Condition can be derived from some basic property of the existential construction. It is also an open issue whether some construction-specific condition has to be included for a full account, or, on the contrary, an explanation can be obtained on the basis of the interaction of general principles only. I do not intend to fully work out the theoretical consequences of the data under consideration for current hypotheses concerning the nature of the DE (see section 6 and section 8 for some remarks). In a naive and informal way, the DE in existential sentences boils down to a requirement that the referential or denotational properties of the internal DP are provided by the existential predicate, and cannot be independently obtained by the DP itself; this may be translated into different formal devices, such as binding of the discourse referent by the predicate and semantic incorporation, or property denotation. It is usually accepted that only indefinites may be licensed by such devices. Definites, on the contrary, are referentially independent, because of the instruction encoded in definite determiners to search a uniquely identifiable referent. Definiteness is thus, in principle, incompatible with a syntactic position that constrains nominal expressions to be bound by the predicate. Nevertheless, as the literature on the topic had begun to make it clear that several definite expressions are in fact not incompatible at all with existential contexts, it appeared to be necessary, on the one hand, to investigate under what conditions definites were acceptable in certain kinds of existentials, and on the other hand, to distinguish ‘strong’ definites from ‘weak’ definites, the latter being characterized as definites whose use is not strictly associated with a uniqueness condition (in fact weak definites may appear in existentials, as in There is [the outline of a human face] hidden in this puzzle). Both topics have been studied in depth, at least for English, and I will rely on current research on them in what follows (see Lyons 1999: 239 for a brief overview). The two basic ingredients one needs to be able to deal with the DE are thus a theory of definiteness, and some assumption concerning the conditions that an existential

 Manuel Leonetti

context imposes on the postverbal DP. As for definiteness, I assume that it consists of a uniqueness presupposition (see Abbott 1999 for a uniqueness approach to definiteness), and that such content acts as an instruction that has to be contextually satisfied. To build the second ingredient into a grammatical theory, Focus structure has to be considered. There is ample consensus about the inherently thetic nature of existential contexts2: this means that they are Broad Focus or Sentence Focus. Such a property will reveal itself as crucial for a discussion of the facts (see section 6). Unfortunately, it is not possible to derive the DE straightforwardly from theticity, and I will show that certain lexical properties have to be included among the necessary notions. It goes without saying that operating with a minimum of theoretical notions is the desirable option, but at the moment there are still many things about theticity that we do not understand.

3. The apparent lack of DE in Italian and Catalan 3.1

Italian

It is usual to assume that the DE is not present in languages like Italian3 and standard Catalan (Moro 1997, Rigau 1994 and 1997, Brucart 2002). The relevant constructions are characterized by the use of the copula essere with the locative clitic ci, in Italian, and the verb haver ‘have’ with the locative clitic hi, in Catalan. The examples in (3) and (4) illustrate the apparent lack of DE in such languages (notice that definite DPs and proper names are perfectly grammatical in the postverbal position); none of them includes a locative coda, for reasons that will become clear later. (3) C´è un cane. / C’è il cane. / C’è Gianni. Cl-is a dog / Cl-is the dog / Cl-is John (4) Hi ha un gos. / Hi ha el gos. / Hi ha en Joan. Cl has a dog / Cl has the dog / Cl has the John This is nonetheless a surprising fact, given that related languages such as Spanish and French, as well as the Germanic languages, show relatively robust manifestations of the constraint. An obvious way to capture the difference between the two groups of languages is treating the DE as a grammatical constraint subject to parametric variation. This is the position defended in Moro (1997), as part of a suggestive analysis of English and Italian existential sentences as inverted copular structures. Moro (1997: 133) traces 2. See Rosengren (1997) and Kuroda (1995) for an overview of the thetic / categorical distinction. 3. A manifestation of the DE in Italian unaccusative constructions is analysed in Belletti (1988), but the problem of existential structures is not directly addressed in this paper. See section 6 for a unified approach to both cases.



Definiteness effects and the role of the coda in existential constructions 

the crosslinguistic difference between Italian and English back to an independent syntactic parameter, the pro-drop parameter. Moreover, he tries to derive the difference without relying on any specific semantic notion. I am not going to dwell on a detailed criticism of the proposal, which is quite technical, but at least a couple of remarks are in order here. The first one has to do with data, and with the simplistic assumption underlying Moro’s approach that the DE is to be seen as a steady grammatical constraint in English. As I already mentioned, a huge amount of research done on the distribution of definite DPs in English, from Milsark (1977) and Napoli and Rando (1978) to the most recent contributions4, offers now solid evidence against such an assumption: several prima facie violations of the alleged constraint are in fact acceptable strings in English, once properly contextualized, and even the basic contrast between English *There is John and Italian C’è Gianni, which is the starting point for Moro’s account, dissolves as soon as we consider the so-called ‘list’ reading, or the ‘reminder’ reading, of the English sentence. Not only the ‘counterexamples’ to the constraint have to be properly accommodated into the theory, instead of being excluded from the discussion; as we will see in section 7, the very notion of existential construction has also been used too rigidly and is in need of refinement. The second remark has to do with theories, and is simply a consequence of the aforementioned situation. As a number of semantic and pragmatic factors are involved in the licensing of certain definites in existential contexts, purely syntactic approaches like the ones proposed until now (Safir 1985, Belletti 1988, Moro 1997, Basilico 1997) run into serious difficulties when confronted with a representative set of data. I believe that these approaches have essentially failed to specify the role of the semantic notion of definiteness in its interaction with the existential context, sometimes discarding it as a source of explanation, and sometimes implicitly relying on it, but without exploring its nature. A widely accepted position in recent research is the idea that semantic definiteness, and not formal definiteness, is the central notion for an account of the DE. Lyons (1999: 246) concludes his presentation of the topic stating that the DE, whatever it is, “is more likely to be a semantic or pragmatic constraint than a syntactic one”. Thus, it seems that the distribution of definite and indefinite DPs in existentials cannot be explained by means of a single syntactic principle (i.e. in terms of conditions on chains, or case assignment), and has to be related to the semantic constraints imposed on definiteness by the construction.5 This leaves us with a need to find an alternative solution for the contrast between English and Italian, which is surely not as clear-cut as has often been thought, covering the case of Catalan as well, if possible. A crucial issue that is in need of clarification is the status of the Italian construction C’è Gianni in (3). According to Moro (1997: 154), 4. See Lumsden (1988), McNally (1992), Abbott (1993), Ward and Birner (1995), Cann (2007), among many others. 5. For some criticism of syntactic approaches to the DE, see for instance Abbott (1993) and Lyons (1999).

 Manuel Leonetti

(3) is grammatical simply because Italian is a Null Subject language that allows subjects to occur in postverbal position (a position that is not the internal DP or pivot position of existential sentences, independently of one’s favourite analysis for Italian inverted subjects): in a nutshell, C’è Gianni would be a construction that escapes the constraint on definiteness because it is no longer an existential, at least in the canonical sense, its subject DP being located in a higher position than the VP-internal one typically associated with the DE.6 As the possibility of Subject Inversion is traditionally considered a property of Null Subject languages, this is how the absence of the DE is indirectly derived from one of the values for the pro-drop parameter. Sentences like (3) are then characterized as locative sentences (Moro 1997:  138): Italian would differ from English in the systematic reanalysis of every case of ‘esserci + definite DP’ as a locative (non-existential) construction.7 I partially agree with Moro’s statement about the special status of sentences like C’è Gianni, but not with his claim that such a status is related to the absence of the DE in Italian. In fact, what I intend to demonstrate is that the DE is present in Italian too. Let me address first the characterization of our alleged locative sentences with esserci. On the one hand, an analysis of (3) as a locative construction is not totally convincing, because (a) the sentence still has a presentative meaning and conforms to the existentials’ grammatical pattern; (b) its counterparts in Catalan and Spanish are not locative predications, as one can see in (5); (c) the DP is in focus, as in existentials, and (d) it is essere, and not esserci, the verb that Italian resorts to for real locative constructions, as in (6). (5) a. Catalan: Hi ha en Joan. (Cf. En Joan hi es – locative predication-) b. Spanish: (Ahí) está Juan. (Cf. Juan está – locative predication-) (6) Gianni è in giardino. ‘John is in the garden’ On the other hand, there is one fact, as far as I know first pointed out in Moro (1997: 280, fn. 29), which supports the claim that (3) does not have the properties of an existential

6. As Moro (1997:155) puts it, “The crucial difference is that Italian provides an escape hatch for the “offending” class of DPs” (i.e. definites). His analysis is in line with an idea that is clearly stated in Rizzi (1986), Belletti (1988) and Vangsnes (1994), among many others: only VP-internal DP positions are subject to the DE, VP-adjoined (or higher) positions being excluded from the range of the constraint. 7. Following Freeze (1992) and Zamparelli (1996: ch. 5), I am assuming that existentials are the opposite of locative predications: while in existentials the location may be interpreted as a topic, even when it is implicit (cf. the notion of stage topic in Erteschik-Shir 1997) and the entity is in focus, in locatives we find the opposite pattern. See also Borschev and Partee (2002) for a similar statement in terms of Perspectival Structure.



Definiteness effects and the role of the coda in existential constructions 

context: it is the possibility of assigning wide scope to the indefinite DP in examples like the one in (7). (7) Non c’erano molte ragazze. Not Cl-were many girls (7) is ambiguous: it can be paraphrased with the sentences in both (8) and (9) in English. In the first case, the indefinite expression has narrow scope with respect to negation, as expected in truly existential contexts; this is, in fact, the only possible reading of the English sentence in (8) and its equivalents in Spanish and several other languages. In the second case, the indefinite escapes the well-known restriction against wide scope of the postverbal DP in existentials.

(8) There weren’t many girls.



(9) Many girls weren’t there.

This shows that the Italian sentence in (7) may be assigned a straightforward existential interpretation as well as another reading that we might well call ‘locative’. A non-existential analysis of C’è Gianni is then justified. Following Zamparelli (1996: 200), it could be designated as a “pseudo-existential locative construction”. Zamparelli (1996) correctly points out that the reason why esserci constructions in Italian do not obey the DE is that they conflate two different structures, locative predications and existence statements.8 As locative predications are exempt from the DE, any apparent counterexample to the DE in esserci constructions is likely to be considered a case of locative or “pseudoexistential locative” predication, being no longer neither exceptional nor unexpected. A simple way to look at the general problem of counterexamples to the DE is to suppose that when a definite (or strong) DP appears in an existential sentence, either the sentence is not actually existential, or the definite is not behaving like a prototypical definite (i.e. it is a “weak definite”, or a “new mention definite”, or it has a “kind reading”). Most counterexamples to the DE in English are instances of the second possibility, as shown in McNally (1992), Abbott (1993), Vangsnes (1994), Ward and Birner (1995) and Zamparelli (1996). For Italian, on the contrary, one should consider the first possibility too, i.e. that C’è Gianni should not be considered a proper existential construction. This situation somehow blurs the original issue of the presence or absence of the DE in Italian: is this language exempt from the restriction because of the conflation of existential and “pseudo-existential locative” constructions, or is there still some evidence for the DE in Italian existentials, given that it seems to emerge in other contexts (cf. the unaccusative constructions studied in Belletti (1988))? Before offering an answer in section 4, it is worth introducing the Catalan data into the picture.

8. See Ziv (1982) for a similar argument regarding Colloquial Modern Hebrew.

 Manuel Leonetti

3.2

Catalan

As pointed out by Brucart (2002), the DE appears in Catalan in examples like those in (10): (10) a. b.

Hi ha {una / *la} solució al problema. Cl has {a / *the}solution to the problem Hi ha {un noi / *ell} al pati. Cl has {a boy / *he} in the courtyard

While (10a) is not representative of a systematic constraint (see the data in (11)) and is maybe amenable to a pragmatic explanation, (10b) illustrates a strong restriction against personal pronouns in haver-hi constructions, which is something one can expect, if it is assumed that pronouns are the highest elements in a definiteness scale and thus the least prone to be inserted in an existential context.9 But leaving aside the special case of personal pronouns, which I will not discuss, the examples in (11), again from Brucart (2002), show that definite DPs are in fact acceptable in haver-hi constructions, contrary to what occurs in English or Spanish, but similar to what occurs in Italian: (11) a. b.

Al pati hi ha el noi i la noia.10 In the courtyard Cl has the boy and the girl Hi ha la policia al pati.11 Cl has the police in the courtyard

We are thus confronted with another language that seems to lack a DE, or at least shows only a minimal trace of it. The same problem arises as with Italian C’è Gianni: what is the status of sentences like Hi ha en Joan ‘There is the John’ in (4) or the ones in (11)? Are they true existentials? Although the Catalan construction with haver-hi does not exhibit the marked scope properties observed in the equivalent Italian construction12, its use and interpretation are very close to it. The main factor in determining the possible combinations of haver-hi with definites and indefinites seems to be the existence of a clearly locative construction with the verb esser-hi ‘be + locative’. The contrast between haver-hi and esser-hi is the formal way of encoding the existential / locative distinction in Catalan (I will come back to such a paradigmatic contrast in section 4). If this is correct, then Catalan provides a genuine counterexample to the DE.

9. Unexpectedly, clitic pronouns are acceptable in existentials in a language with as strong a DE as Spanish: cf. Los hay lit. ‘Them Cl+has’. See Longa, Lorenzo and Rigau (1998) for an account of this fact in terms of “clitic recycling”. 10. Cf. English *There are the boy and the girl in the courtyard and Spanish *En el patio hay el chico y la chica. 11. Cf. English *There is the police in the courtyard and Spanish *Hay la policía en el patio. 12. This was pointed out to me by Vicky Escandell-Vidal.



Definiteness effects and the role of the coda in existential constructions 

One way to account for the Italian and Catalan facts is by resorting to parametric variation in the syntax, in the sense of Moro (1997). I think it is beyond discussion that the crosslinguistic distribution of the DE is ultimately tied to parametric syntactic variation, but it is so only indirectly, in its dependence on the encoding of information structure. Taking into account the principles of information structure in each language is essential for our understanding of the link between syntactic positions and definiteness. This seems to me the only way to obtain an explanation that relates the DE to the semantics of definiteness. Moreover, such a perspective is compatible with the search for some stable manifestation of the DE beyond superficial variation among languages, i.e. with keeping the DE as a general semantic/pragmatic constraint, although it may surface in different ways. To show why such a path is worth following, I will present some facts that have not received much attention in the literature and have not played any role in recent theoretical discussions in the field, as far as I know.

4. The Coda Constraint 4.1

Codas and definites

The data I want to focus on involve the behaviour of the postverbal locative XP in existential contexts in Italian and Catalan. I assume that such a constituent, which I will refer to as the Coda, following common use, has to be analysed as a VP adjunct, and not as an NP adjunct or as a predicate in a small clause (see McNally 1992, Abbott 1993, Zucchi 1995, Moro 1997 for detailed argumentation). This is crucial for an appropriate understanding of available interpretations. However, I do not exclude a small clause analysis of the DP-XP sequence in certain cases, as the following discussion will make clear. The basic syntactic structure of an existential is then something like (12): (12) [IP... [VP V [ DP ] [ XP ]] As argued by Zucchi (1995:  56), Bende-Farkas and Kamp (2001) and Keenan (2003: 194), the semantic role of the coda is to provide the contextual domain for the interpretation of the postverbal DP. This means that there-sentences incorporate the coda property into the denotation of the postverbal DP (Zucchi’s Coda condition): in There is a dog in the garden, the intersection of the coda and the NP denotation (the denotation of dog in the garden) is the set that plays a central role in assessing the truth value of the proposition expressed and the compatibility between the determiner and the felicity conditions of the construction. There are different ways of implementing the basic intuition of Zucchi’s analysis. Later on (in section 6) I will suggest a syntactic motivation for the Coda condition. For the purposes of this paper, I am just interested in the idea that the coda plays a relevant role in obtaining the denotation of the postverbal DP and deriving the DE.

 Manuel Leonetti

With such ideas in mind, it is possible now to introduce the facts on which I wish to concentrate. They are illustrated in the contrasts in (13) and (14), in Italian and Catalan: (13) a. C’ è la statua di Michelangelo, in Piazza della Signoria. b. ??C’ è la statua di Michelangelo in Piazza della Signoria.13 ‘There is the statue by Michelangelo(,) in Piazza della Signoria’ (14) a. Hi havia el degà, a la reunió.14 b. ??Hi havia el degà a la reunió. ‘There was the dean(,) at the meeting’ The contrasts show that dislocation of the coda is obligatory when a definite DP occurs in existentials. This is unexpected, because Italian and Catalan easily accept definites in such constructions, as we already saw in section 3. The option of leaving the locative inside the VP while keeping the definite is ungrammatical. The appearance of definites is also acceptable if the locative is implicit, as in (15): (15) a. C’ è la statua di Michelangelo. b. Hi havia el degà. Crucially, the contrast disappears when the internal DP is indefinite: (16) a. C’ è {una / qualche} statua di Michelangelo in Piazza della Signoria. b. C’ è {una / qualche} statua di Michelangelo, in Piazza della Signoria. (17) a. Hi havia un estudiant a la reunió. b. Hi havia un estudiant, a la reunió. Thus, indefinites are compatible with the presence of the locative coda, be it dislocated or not. Definites, on the other hand, are not compatible with a coda inside the same VP. The contrasts in (13)-(14) are quite subtle and have gone largely unnoticed. They are strongly dependent on intonation patterns, as well as on heaviness factors that I will not investigate here, but I believe that they cannot be discarded as marginal or non-systematic facts. It may be worth including some additional data in order to get a more accurate description: so (18)-(19) confirm the phenomenon observed in the previous examples, and at the same time introduce a new problem.

13. This sentence would be acceptable in case la statua di Michelangelo in Piazza della Signoria is taken as a DP (or even, marginally, as an eventive small clause), i.e. in case the PP in Piazza della Signoria is not interpreted as a VP adjunct. 14. The Catalan examples are from Rigau (1994) and (1997). She already noticed that in haver-hi constructions there is no DE, but the locative XP has to be a topic, be it left or right-dislocated:

(i) A la reunió hi havia el degà. (ii) Hi havia el degà, a la reunió.



Definiteness effects and the role of the coda in existential constructions 

(18) a. ?C’è Chomsky al MIT. b. C’è Chomsky, al MIT. / Al MIT c’è Chomsky. Cl-is Chomsky at MIT (19) a. ?Hi ha en Chomsky a MIT. b. Hi ha en Chomsky, a MIT. / A MIT, hi ha en Chomsky. Cl has the Chomsky at MIT The pattern is again the same found in (13)-(14): in Italian and Catalan, if the usual existential reading is tested, the definite DP, in this case a name, is incompatible with the coda when the coda is inside the VP. But it is important to notice that there is a possible reading for (18a) and (19a) that does not make them deviant: it is the possibility of taking the sequence corresponding to Chomsky at MIT as the description of an event, of something that is happening at some moment, for instance on a screen where some activity of Chomsky at MIT is being shown. This kind of eventive reading appears quite naturally too in the following examples, where the exclusion effect of the coda on the definite DP is not perceived: (20) C’è Gianni {in giardino / al telefono / che aspetta} Cl-is John {in garden / at the phone / that waits} ‘There is John {in the garden / on the phone / waiting}’ (21) Hi ha la Maria {molt enfadada / al telèfon / que espera} Cl has the Mary {very angry / at the phone / that waits} ‘There is Mary {very angry / on the phone / waiting}’ A plausible analysis for such sentences involves positing a small clause as the complement of esserci and haver-hi, because a small clause is the syntactic structure that naturally fits the sequence formed by a name and an XP predicate. Moreover, the XP predicate has to be a stage-level one (cf. *C’è Gianni genovese ‘There is John genoese’; *Hi ha la Maria profesora ‘There is Mary a teacher’), thus conforming to a well known restriction operating in many languages against individual-level predicates as secondary predicates in existential sentences15; this is also in accordance with a small clause analysis of the postverbal string. Then, as already announced, the initial assumption about the adjunct status of the coda in existentials has to be slightly revised: the default analysis of the coda in an existential construction is to treat it as a VP adjunct, but in some cases, when the coda behaves like a non-verbal secondary predicate, it seems natural to consider it as a predicate in a small clause.16 At least, such an analysis is able to 15. See Milsark (1977), McNally (1992), Basilico (1997) for different accounts of this constraint. 16. A small clause analysis seems plausible for the following English examples too, which Lyons (1999:239) presents as perfectly good for many speakers:

(i) There is John waiting at the door for you. (ii) There is that man on the phone again. (iii) There is the postman coming up the drive.

 Manuel Leonetti

capture the essential intuition in the interpretation of sentences like (20) and (21), i.e. that the postverbal string denotes an event, a state of affairs that can be located in time. Moreover, a crosslinguistic survey of the appearances of definite DPs inside existentials will reveal (see section 4.2) that considering the small clause analysis as one of the options is crucial for an adequate understanding of the facts. In any case, putting aside the cases where the existential verb selects an eventive small clause, the generalization that emerges from the data in (13) to (18) can be formulated as the Coda Constraint: (22) Coda Constraint The presence of the locative coda inside the VP blocks the insertion of definite DPs: these are excluded unless the locative coda is itself (right- / left-) dislocated (or removed). The first interesting consequence we can draw from (22) is an answer to the question of whether Italian and Catalan are really exempt from the constraint on definiteness: the answer is obviously negative. As soon as some attention is paid to intonation and information structure, and the restrictions imposed by the locative coda are taken into account, the good old DE shows up again in Italian and Catalan as well, i.e. in languages where it was believed to be missing. In a sense, this is good news for a semantic approach to the DE, although it is an indication that such an approach must take detailed phonological and syntactic considerations into account. Once this has been established, further questions arise. On the one hand, the nature of the Coda Constraint has to be explicitly addressed: is it a universal principle, or just a generalization that holds for some languages? On the other hand, one must pose the question of whether the Coda Constraint could be considered an epiphenomenon, and ultimately derived from the interaction of general principles. In fact, as it is in (22), the constraint states that there exists an incompatibility between definites and codas in the VP, but it does not offer any clue about why that incompatibility should arise. In sum, the constraint in (22) immediately calls for some principled explanation. I will defer the discussion of this second question and its implications until section 6, and I will proceed now to address the first question concerning the empirical support for (22).

4.2

Some crosslinguistic evidence

4.2.1 English The insertion of definite DPs in existentials is much more constrained in English than in Italian or Catalan. However, it is possible to find evidence that the Coda Constraint has some presence in English too. There is an interesting asymmetry, already mentioned in Lumsden (1988: 216–219), Abbott (1993: 43–44) and Vangsnes (1994: 116), that reminds us of the interplay between definiteness and the presence of the coda: the two examples in (23) show a minimal contrast superficially, but a deeper difference in their constituent structure.



Definiteness effects and the role of the coda in existential constructions 

(23) a. There’s a unicorn in the garden. b. There’s the unicorn in the garden. In (23a) the locative PP in the garden is interpreted as a separate constituent – a VP adjunct – , but in (23b) it has to be taken as a part of the focussed DP the unicorn in the garden. This seems to be a general condition in so-called “list” or “enumerative” readings for definites in existentials: if a postnominal constituent appears, it must be parsed as a nominal modifier, i.e. it is not a real coda (as noticed by Rando and Napoli (1978: 304), “List sentences do not freely allow material outside the focused NP”). Abbott takes this to be the most salient difference between “non-contextualized existentials” – the standard ones, with indefinites in postverbal position and a coda usually following them – and “contextualized existentials” – the cases with a definite in the Focus position which have no codas and require special contextualization. The following are two of her examples of contextualized existentials: (24) A: Is there anything to eat? B: Well, there’s the leftover chicken from last night. (25) A: I guess we’ve called everybody. B: No, there’s still Mary and John. Abbott (1993: 44) suggests an explanation for the difference: “With proper names and anaphoric definites, the predicational slot (the coda, ML) must be fixed before they are introduced. Since the referents for such NPs are, by definition, part of the discourse context, it is only appropriate to include them in an existential which has the kind of reminding function noted above. But this presupposes some purpose or issue for which the entity in question might be suitable.”

Her remarks are in line with the Coda Constraint. Abbott (1993: 47) points out that the impossibility of interpreting the PP in the room as a part of the postverbal DP is the reason why (26) sounds deviant (which means that as a real coda it is incompatible with a definite like my sister): (26) ?There is my sister in the room. Abbott (1993: 44–45) supports a pragmatic view of the counterexamples to the DE, by which any definite DP which can be interpreted as introducing a new entity into the discourse should be possible in a non-contextualized existential (see also Ward and Birner 1995), while names and anaphoric definites are only possible with marked interpretations: “On the pragmatic view the effect of (some) definite NPs in focus position in an existential sentence is a sentence which requires special contextualization” (1993: 47). I am sympathetic to her overall approach, although in my opinion something more specific has to be said about the role and the effects of the coda. I expect to offer some ideas in section 6. Here the relevant point is that the Coda Constraint seems to be in force in English too.

 Manuel Leonetti

It is worth recalling that, according to Abbott (1993: 44), the condition on the coda does not hold for cataphoric definites – the ones that do not introduce familiar referents into the discourse – , as one can check in classical examples like (27a, b, c): (27) a. There was the usual crowd at the beach last Sunday. b. There was the smell of pot all over the apartment c. There weren’t the funds necessary for the project we had in mind. In English, cataphoric and weak definites do not obey the condition on the coda, while strong definites (for instance, anaphoric ones) have to comply with it, thus being incompatible with a locative VP adjunct and forcing the interpretation of the DP-XP string as a single constituent, whenever possible.17 This divergent behaviour is confirmed by data from other languages, like Spanish (see section 4.2.3). I will return to this issue later on, in section 6. 4.2.2 French Additional evidence for the systematic nature of the facts under analysis comes from French. Beyssade (2004), elaborating on a proposal in Lambrecht (2002), demonstrates that three different types of existential constructions have to be distinguished, each of them subjected to a particular semantic constraint, and, moreover, that the presence of the coda plays a central role in the licensing of definites. The basic il y a + DP + XP construction in French (“la construction proprement existentielle”) is described, following McNally (1992, 1998), as an instantiation predicate that selects a property-denoting argument (i.e. an argument of type <e,t>). It is the predicate that introduces existential quantification, as in Milsark (1977) and McNally (1992). Here the coda is optional (although some implicit location is always inferred when the coda is not explicit), and the DE appears regularly: (28) a. b.

Il y a Cl Cl-has *Il y a Cl Cl-has

un questionnaire que je n’ai pas. a questionnaire that I don’t have le questionnaire que je n’ai pas. the questionnaire that I don’t have

A second type of construction is the eventive existential (“la construction événementielle”). It introduces a new event into the discourse, instead of a new entity, and the

17. Abbott (1993:43) notices some apparent exceptions to the generalization on anaphoric definites:

(i) There are the dishes to wash, and the laundry to bring in.

I think there are reasons to analyse the coordinated constituents in (i) as small clauses, conforming to the pattern in (20) and (21). If that is correct, (i) would not be a counterexample to the condition anymore.



Definiteness effects and the role of the coda in existential constructions 

new event is presented as focal information.18 Here are some examples from Beyssade (2004: 69): (29) a. b.

Il y a le téléphone qui sonne. Cl Cl-has the phone that sounds Il y a le chat qui meurt de froid dehors. Cl Cl-has the cat that dies of cold outside

In eventive existentials the coda is obligatory, as shown in (30). This, together with the eventive interpretation, makes it clear that they are to be identified with the cases where a small clause based on a stage-level predicate is selected by the predicate (Beyssade 2004: 70; Côté 1999 for Quebecois French). (30) a. ?Il y a le téléphone. b. ?Il y a le chat. Eventive constructions allow both definites and indefinites as subjects of the embedded small clause, hence no DE is expected in them. Some interesting evidence supporting the notion of ‘eventive existential’ can be found in Rando and Napoli (1978), where it was pointed out that (31) is actually ambiguous between an eventive reading and a purely existential reading, and that two different intonational contours correspond to the two readings. (31) There is a woman in the house. The third type of existential construction is termed enumerative by Beyssade, and is the counterpart of the well known “list contexts” and “reminders” in the literature about English. In fact a representative example like (32) is simply the translation of the English sentences in (25). (32) Je crois qu’on a appelé tout le monde. Non, il y a encore Marie et Jean. Enumerative existentials don’t pose any restriction on the semantic type of the postverbal DP, hence there is no DE. The coda is optional: it is usually absent or implicit, and when it appears it is not pragmatically asserted (i.e., it is not in Focus, as in the proper existential subtype) but presupposed, thus showing the pragmatic role of Topic, while the internal DP is in Focus. It is precisely this feature of the enumerative construction that reminds us of the Coda Constraint: if a definite is inserted, the coda must be dislocated or topicalized (or be implicit). It is exactly the same condition that holds for list 18. Most of the examples of the eventive subtype correspond to the Presentational Relative Construction in Lambrecht (2002), exemplified in (i), where the coda is represented by a predicative relative clause following the postverbal DP: (i) Il y avait une jeune fille qui fumait. Cl Cl-had a young girl who smoked ‘There was a young girl smoking’

 Manuel Leonetti

there-constructions in English and, mutatis mutandis, for equivalent cases in Catalan and Italian. The same connection between definiteness and locative codas shows up in all these languages. From now on I will adopt Beyssade’s classification and the terms proper existential, eventive existential and enumerative existential. The typology of existentials will be taken up again in section 6. 4.2.3 Spanish Interestingly, Spanish does not offer contrasts as clear as those found in Italian, French and Catalan in the interaction of definiteness and the structure of existentials. This is mainly due to the fact that the DE is quite robust in Spanish. Names are absolutely excluded from haber constructions, and only certain kinds of definites (superlatives, complex DPs) are allowed, as shown in (33)19: (33) a. b. c.

No hay el menor indicio de culpabilidad. Not has the slightest sign of guilt Incluso hay la perspectiva de que el viaje se pueda retrasar. Even has the perspective of that the trip Cl may be delayed Había el problema de los recursos hidrológicos. Had the problem of the resources hydrological

As for the DE, Spanish is without doubt the most restrictive of all the languages under consideration. Different factors interact to give rise to this situation: a. the “division of labour” between the verbs haber and estar is the following: haber is strictly used in proper existentials, but the interpretations corresponding to eventive and enumerative existentials in other languages are always rendered by means of estar. For instance, There is John should be translated as (Aquí) está Juan, and There is John at the phone as Está Juan al teléfono. Since estar can be used in thetic utterances with a definite subject, this severely limits the possibility of having definites with haber: such a combination would in most cases be outranked by some optimal construction with estar. I consider the competition with estar as the main factor limiting the use of haber with definites. b. haber seems to exclude the selection of embedded small clauses, contrary to what has been observed for Catalan haver-hi, Italian esserci, English there be, French avoir and Chinese you. As a consequence, one of the main contexts responsible for DE violations is excluded. c. last but not least, Spanish syntax is not as transparent with respect to information structure as the syntax of Italian or Catalan, and the grammatical mechanisms for Topic and Focus marking are certainly not the same. This leads to different possibilities for word order patterns (see section 6 for some comparative remarks). 19. The examples are from Leonetti (1999). The classical reference for definiteness in haber sentences is Suñer (1982).



Definiteness effects and the role of the coda in existential constructions 

However, I believe that certain subtle “coda effects” can be found in some cases with definites after the verb haber: (34) a. [Había unos vinos muy caros en la tienda] / [Había unos vinos muy caros], en la tienda Had some wines very expensive in the shop ‘There were some very expensive wines in the shop’ b. ?[Había sólo los vinos más caros en la tienda] / [Había sólo los vinos más caros], en la tienda Had only the wines most expensive in the shop ‘There were only the most expensive wines, in the shop’ Several native speakers of Spanish have expressed doubts on the contrasts in (34), hence I prefer not to rely on those examples as solid supporting evidence for the Coda Constraint. For some reason the presence of the coda inside the construction seems almost irrelevant for definiteness in Spanish. I will suggest an explanation for this difference in section 6. The data collected from other languages constitute, in any case, an acceptable starting point for theoretical discussion.20

5. Inverted subjects All the previous facts concerning the effects of the coda on definiteness are still waiting for a global explanation, but they could not be properly understood if yet another cluster of data was not introduced. The rationale for enlarging our set of basic phenomena a bit more is quite simple: there are further manifestations of the DE outside the restricted area of existential sentences, and they are not unrelated to those already mentioned.21 In fact, they will provide important clues for a global account. In Romance linguistics, in the case of French and Italian in particular, the oddness of definite postverbal subjects in certain constructions has often been pointed out. I will take some representative Italian examples from Benincà, Salvi and Frison (1988), Rizzi (1986) and Belletti (1988) as the basis of my argumentation. In (35) there is evidence for a contrast between definite and indefinite postverbal subjects in unaccusative and passive sentences: 20. Chinese provides additional evidence supporting the Coda Constraint: see Huang (1987), Li (1996) and Chang (2004) for discussion. 21. Another fact that is worth analyzing as an extension of the Coda Constraint is the ban on topicalization of the DP in existential contexts, due to the presence of the coda: internal DPs cannot be topicalized unless the coda is itself dislocated (or removed) – at least in Romance languages. The same “coda effects” reappear when the DP position is relativized, in non restrictive relatives. Unfortunately, space limitations do not allow me to include a discussion of such correlation of topicalization and definiteness (see Leonetti (2005) for a survey of the data).

 Manuel Leonetti

(35) a. b.

È entrato un ladro dalla finestra Is entered a thief through the window ‘A thief has entered through the window’ ?È entrato il ladro dalla finestra. Is entered the thief through the window

(36) a. b.

È affondata una nave alle cinque. Is sunk a ship at five ‘A ship has sunk at five’ ?È affondata la nave alle cinque. Is sunk the ship at five

(37) a. b.

È stato messo un libro sul tavolo. Is been put a book on the table “There was put a book on the table.” ?È stato messo il libro sul tavolo. (Cf. Il libro è stato messo sul tavolo.) Is been put the book on the table (The book has been put on the table.)

The crucial fact that relates this kind of DE to the “coda effects” is that definite postverbal subjects are odd when the constituent order is VSX, as in (35)-(37), but they do not produce any anomaly in simple VS constructions, as shown in (38)22: (38) a. È entrato il ladro. b. È affondata la nave. Rizzi (1986) and Belletti (1988) already noticed that as soon as the last constituent is dislocated, topicalised, destressed or marginalized, and correspondingly an intonational break is placed after the postverbal subject, the definite DPs become fully acceptable, with a narrow focus reading, and the DE disappears. The following examples prove that the deviance vanishes in such conditions: (39) a. È entrato il ladro, dalla finestra. / Dalla finestra, é entrato il ladro. b. È affondata la nave, alle cinque. / Alle cinque, é affondata la nave. c. È stato messo il libro, sul tavolo. / Sul tavolo, é stato messo il libro. It is quite clear that the acceptability pattern is the same as that which emerges in existential constructions: with respect to definiteness, the final constituent plays the role that the coda plays in existentials. As Italian and Catalan are so similar to each other as

22. I omit the sentence corresponding to (37) because it is ungrammatical due to independent reasons: the locative complement sul tavolo ‘on the table’ is an argument of mettere and cannot be removed.



Definiteness effects and the role of the coda in existential constructions 

far as the DE in existentials is concerned, it is expected that they are similar too in unaccusative constructions, and in fact this expectation is borne out: (40) a. b. c.

Ha nascut una nena dins un taxi. Has born a girl inside a taxi “A girl was born inside a taxi.” ?Ha nascut la nena dins un taxi. (Cf. La nena ha nascut dins un taxi.) Has born the girl inside a taxi Dins un taxi, ha nascut la nena.

(41) a. b. c.

Ha arribat un estudiant al periòdic. Has arrived a student to the newspaper ‘A student has arrived to the newspaper’ ?Ha arribat el estudiant al periòdic. Has arrived the student to the newspaper Ha arribat el estudiant, al periòdic.

(42) a. b. c.

Es va enfonsar una nau a les cinc. Cl sunk a ship at five ‘A ship sank at five’ ?Es va enfonsar la nau a les cinc. Cl sunk the ship at five Es va enfonsar la nau, a les cinc.

What Italian and Catalan share is a ban against VSX order when the subject is definite. We are not dealing with existential sentences anymore in (35)-(42), but still we have thetic structures that typically introduce new referents into the discourse, with passive and unaccusative verbs. The DE reappears, though maybe in a weaker fashion. As attested in Kampers-Mahne et al. (2004: 572–574), subject inversion in French confirms the systematic nature of this pattern, both in subjunctive clauses – (43)- and in indicative unaccusative clauses – (44)-: inverted definite subjects may not be followed by any constituent, while the restriction does not hold for indefinite ones. (43) a. b.

*Je veux que parte Paul {immédiatement / aux Etats-Unis}. ‘I want that Paul leaves immediately / for the United States’ Je veux que partent trois étudiants {immédiatement / aux Etats-Unis}. ‘I want that three students leave immediately / for the United States’

(44) a. b.

*Ce jour-là fut assassiné César sans pitié. ‘That day Caesar was assassinated without pity’ Aussitôt entrérent dix policiers avant même qu’on les ait annoncés. ‘Immediately ten policemen came in even before they had been announced’

 Manuel Leonetti

Kampers-Mahne et al. (2004) suggest that the contrasts in (43) and (44) are related to the fact that definites are less naturally focused than indefinites. In fact a basic condition for subject inversion to be felicitous is that the subject is in Focus. For the moment, it is worth recalling that some cases of inverted definite subjects followed by another constituent are judged to be grammatical in French, and they are always nonanaphoric definites, as in the examples in (45): (45) a. b.

Je veux que soit convoqué le tribunal avant demain. ‘I want that the tribunal is convoked before tomorrow’ Je regrette qu’ait été impliquée la classe politique dans cette affaire. ‘I regret that the political class has been implicated in this case’

Thus, “it is only strong definite subject DPs that cannot be inverted when not in sentence-final position, the default focus position” (Kampers-Mahne 2004: 574).23 It seems clear that the constraints operating on definites are essentially the same as in Italian and Catalan and, more importantly, the same as in existential sentences.24 Interestingly, there is no DE in inverted subject constructions in other Romance languages, such as Spanish and Romanian, where sentences like those in (46) and (47) are judged to be acceptable.25 (46) a. b.

Ha entrado el ladrón por la ventana. Has entered the thief through the window Se ha hundido el barco a las cinco. Has sunk the boat at five

23. In my opinion non-anaphoric or non-referential definites are acceptable as inverted subjects in VSX sequences in Italian too (for instance, in È entrata la speranza nel suo cuore ‘Hope has entered his / her heart’ or Sul più bello è arrivato il marito in casa ‘... her husband got home’). No reference to this particular fact can be found in the literature, as far as I know. It simply strengthens the parallelism between inverted subject constructions and existential constructions. 24. Although Brazilian Portuguese shows no systematic DE in existential contexts, certain subtle contrasts can be found in sentences with inverted definite subjects. The relative position of the subject DP and the locative PP seems to be relevant for acceptability in (i)-(iii): the first sentence, with VSX order, is slightly degraded in comparison with (ii), where there is no locative complement, and (iii), where the order is VXS (thanks to Helena Guerra and Aroldo Leal de Andrade for their judgements). (i) ?Chegou o homem na sala. Arrived the man to the room (ii) Chegou o homem. (iii) Chegou na sala o homem. As already observed for other Romance languages, this is a form of the DE: inverted definite subjects are odd when another constituent prevents them from being in an unmarked Focus position (typically at the end of the sentence). 25. The Romanian examples in (47) are from Alboiu (2002: ch. 3).



Definiteness effects and the role of the coda in existential constructions 

(47) a. A cumpArat Ion {inelul / un inel} Has bought John the ring / a ring b. A cumpArat {inelul / un inel} Ion Moreover, as already stated in section 4.2.3, the effects triggered by the presence of the coda in existential sentences do not seem as clear in Spanish as in other languages. Some factor interacting with the DE must be responsible for the contrast between Spanish (and Romanian) on one hand and Italian, Catalan and French on the other. This raises at least two questions. The first one is how to capture the role of the coda / final constituent in its interaction with definiteness in Italian and Catalan. The second one is why the central facts are present in certain languages but absent in others. Let me first address the question concerning the motivation for the “coda effects”, in order to face the problem of crosslinguistic distribution later on a more solid basis (cf. section 6). The classical way to deal with the data in (35)-(37) is by means of Case Theory.26 Belletti (1988) claims that, if it is assumed that unaccusative verbs may assign partitive case to their internal argument, the DE follows automatically from a semantic clash between partitive and definiteness. This explains the deviance of (35b) or (36b). When the final constituent is dislocated or removed, the DE disappears and definiteness is licensed in the subject DP; Belletti (1988) and Rizzi (1986) argue that in these cases the final complement has been extracted from the VP and the subject DP is in a VP-adjoined position where it receives nominative case, and not partitive. As nominative case is perfectly compatible with definiteness, the examples where the subject occupies the final position – presumably a position external to the VP- are exempt from the DE. All the examples in (35)-(42) fit the analysis, which correctly predicts that only those postverbal subjects occurring in an object position will be constrained by the DE. In fact the postverbal definites in (48) are acceptable, with a narrow focus interpretation27: (48) a. È entrato dalla finestra il ladro. b. È arrivato al giornale lo studente. In spite of its simplicity and elegance, such an account of the DE in terms of partitive case assignment does not seem totally convincing to me. On the one hand, there is no compelling evidence for partitive case at least in Romance languages, and the formal mechanism is related to the semantics of definiteness only by stipulation. On the other hand, if every instance of the DE is reduced to a clash between case and definiteness, it is hard to understand why Spanish has a strong manifestation of the DE in existentials,

26. See Brassil (2004) for a recent reappraisal of this approach to the DE in Spanish. 27. As reported in Frascarelli (2000:108), Cardinaletti (2001) and Belletti (2001:78), among others, VSX sentences with definite subjects in Italian are deviant when pronounced with unmarked intonation, but perfect when pronounced with narrow focus on the subject. The dislocation of the final complement automatically projects narrow focus on the inverted subject.

 Manuel Leonetti

where it is plausible to assume that partitive is assigned, but not with inverted subjects of unaccusative constructions (cf. (46)).28 Instead of trying to further elaborate the Partitive Hypothesis, I prefer to explore a different approach to the distribution of the DE, drawing on focus structure.

6. Focus and definiteness My starting point is an uncontroversial observation concerning word order and focus assignment in Italian. While VOS / VXS is an acceptable pattern with Narrow Focus on the inverted subject, VSO / VSX is possible only with a marked intonation, more precisely with the final constituent dislocated or marginalized, and Narrow Focus on the subject. French and Catalan show similar tendencies, while other Romance languages, like Spanish and Romanian, allow both VXS and VSX quite naturally (Zubizarreta 1998, Alboiu 2002). The characteristic feature of Italian and Catalan information structure, as stated in Frascarelli (2000) and Vallduví (1995), is an extraposition / dislocation process that takes place in Narrow Focus sentences: any constituent that is not focal is extraposed, so that the sentence is partitioned into a single constituent containing the verb and the Focus, and a number of emarginated constituents on either side. What Italian and Catalan encode by means of syntactic extraposition is marked in other languages by other defocusing mechanisms, such as scrambling, special case morphology or intonational contour. With this in mind, it is possible to perceive a striking correlation between information structure and definiteness in Italian and Catalan. Briefly, the DE manifests itself when a postverbal definite subject does not receive a Narrow Focus interpretation, in particular in constructions where the unmarked interpretation is one of Broad Focus (Sentence Focus) or where another constituent following the subject is assigned Narrow Focus. Two of the preceding examples are repeated here in order to illustrate the generalization: (13) ?C´è la statua di Michelangelo in Piazza della Signoria. (35) ?È entrato il ladro dalla finestra. Let’s review the case of (13) first. Being an existential construction, hence a prototypical instance of thetic construction, (13) requires the assignment of a Broad Focus interpretation. Being definite, the postverbal DP gives rise to an anomaly, unless it is assigned Narrow Focus, forcing the dislocation of the coda, as predicted by the Coda Constraint. As for (35), it should be compatible both with Broad Focus and with Narrow Focus on the last constituent, but again the postverbal DP is incompatible with those readings, 28. I do not intend to deny the relevance of the well attested interactions between case and definiteness. I refer the reader to Leonetti (2004) for some speculations on accusative and specificity in Spanish.



Definiteness effects and the role of the coda in existential constructions 

and the sentence has to be interpreted with Narrow Focus on the subject and the concomitant dislocation of the locative PP. The generalization, then, is valid for the two DE contexts in Italian (and Catalan). Anyway, the basic question still needs an answer: why are only definites, and not indefinites, constrained by Focus assignment? Frascarelli (2000: 182, fn. 32) offers a valuable clue when she states that in presentative structures a definite “cannot be part of ‘all new’ information. So, narrow focalization is needed to exclude its interpretation as a Topic. In other words, we can either say “È affondata LA NAVE alle cinque” or “È affondata # la nave # alle cinque” (in which ‘la nave’ is an internal Topic).” So, according to Frascarelli’s remark, in certain constructions definites must be Topic or Narrow Focus, but they resist their incorporation into Broad Focus. In case they get interpreted as Topics, the resulting construction may be locative; when they are interpreted as Narrow Focus, not necessarily contrastive, the result is a “pseudo-existential locative construction”, a “list” or “reminder” reading, or simply a presentative construction (in the languages that permit this option). So maybe the basic question must be reformulated as in (49): (49) Why do definites need to escape from Broad Focus in certain constructions? This truly seems to be the basic question because it points to the heart of the DE problem. Let’s concentrate on proper existentials, leaving eventive and enumerative constructions aside until next section. First of all, notice that the need to be interpreted as a part of Broad Focus only affects postverbal DPs followed by some other constituent; in other syntactic positions DPs are assigned a Topic reading or a Narrow Focus reading. Intuitively, definites (at least names and anaphoric definites) are too “prominent”, in a sense that is to be further specified, to be “pressed” into a thetic construction. There seems to be a clash between definiteness and the need to compute an appropriate thetic reading of the sentence. This suggests that a valuable theoretical option would be trying to derive the DE from theticity, but serious difficulties arise with thetic contexts that allow definites, for instance sentences like Llega el tren or Arriva il treno ‘The train arrives’ in Spanish and Italian. So maybe the most adequate strategy is to make Broad Focus / theticity interact with some other property. Here I see no way to avoid resorting to some construction-specific condition, lexically associated with the verb or the construction. After all, this is what most theories do when they derive the DE from some Novelty Condition, or some felicity condition regarding the denotation of the postverbal DP. Assuming that the lexical requirements of the verbs that characterize existential constructions have to be taken into account (and I will devote the next section to this issue), the central property of existentials is that they have to be interpreted as Broad Focus sentences, in the unmarked case. This is not incompatible with the hypothesis that there is some kind of implicit spatio-temporal topic (a stage topic).29 Both the coda and the DP must be inside the Broad Focus. This triggers a sort of integration of 29. See Basilico (1997) for a similar proposal about there.

 Manuel Leonetti

the two constituents into each other; definites are excluded from such an integration process when their reference obtains from contextual information outside the construction. I take this integration to be the grammatical counterpart of Zucchi’s (1995) so-called Coda Condition. As already mentioned in section 4, Zucchi (1995) suggests that the role of the coda is to act as the contextual domain for the interpretation of the postverbal DP. In fact some assumption about the interaction between the coda and the DP seems to be necessary for an account of the DE. On the one hand, the felicity conditions of there-sentences require that the common ground be neutral about the (non)emptiness of the intersection of the set denoted by the DP constituent with the set denoted by the coda; on the other hand, the semantics of definite and strong determiners demand that the common ground entail that the denotation of DP should not be empty (this is the presuppositional basis of Zucchi’s analysis). An interpretive conflict arises when the coda provides a part of the denotation of the definite DP: in a few words, the intersection of the DP set with the XP set is subject to contradictory requirements by the existential predicate and the strong determiner, i.e. the construction requires that the (non)emptiness of the intersection should not be previously established in the context, but definiteness demands the DP denotation to be contextually established. This works as a derivation of the DE only if the coda is given an interpretive role along the lines of Zucchi’s proposal. Presenting the coda as a pragmatically presupposed constituent or a dislocated Topic prevents the occurrence of interpretive conflicts with definites. Zucchi’s analysis has been criticized on different grounds, and I am not sure that his theory works adequately for weak / cataphoric definites, but it certainly offers some interesting ideas on the role of codas. Moreover, it is in accordance with all the previous considerations on the interpretation of existentials, since the reason why the denotation of the coda plays such a crucial role in his account (according to the Coda Condition) is to be found in the particular information structure of existential sentences and the requirement on Broad Focus as a default interpretation. Otherwise, one wonders why the coda should intervene in the licensing of (in)definiteness in the construction. In any case, my observations concerning the role of the coda are also compatible, as far as I can see, with an account of the DE in terms of the property-denoting condition on postverbal DPs or semantic incorporation, as in McNally (1992, 1998) and Bende-Farkas and Kamp (2001). Suppose that the characteristic property of existential sentences is a constraint on the semantic type of the internal DP: it has to be of type <e,t>. At least a part of the classical facts covered by the DE are predicted in this way. The motivation for the constraint could be the necessity of “compacting” the DP and the coda into a focal constituent, as if the presence of the coda should force semantic incorporation of the nominal in order to obtain a thetic utterance (recall that semantic incorporation is not supposed to affect definites, or at least strong / anaphoric definites). As the semantic analysis, whatever version we choose, is entirely to be worked out and cannot be developed here, I do not intend to present a complete account of the DE, but just to specify some features of its dependence on information structure. My



Definiteness effects and the role of the coda in existential constructions 

claim is simply that DE contexts are Broad Focus structures where the internal DP must be in focus and, at least in a number of languages, it cannot be definite unless it is the only constituent in focus, apart from the verb (these are the cases where the DE reduces to the clash between definiteness and the coda or final constituent). All DE contexts involve focal DP positions, which suggests that Topic-Focus articulation always plays a role in (in)definiteness constraints. As for the “coda effects”, the major underlying problem we face here is that we still don’t know how informationally complex a thetic structure can be, nor even how to pose such a question in a precise way.30 On the one hand, if a thetic utterance is an allnew utterance, it must be within some limits in its complexity, since an increase in complexity will produce some Topic / Focus partition. On the other hand, languages differ in the extent to which they allow such an increase without introducing informational partitions. In particular, what is relevant for the DE is the fact that some languages prohibit the insertion of definites in internal positions in presentational sentences (Italian, Catalan, French) while others allow such an operation (Spanish, possibly Romanian). This is surely related to general features of the syntax of Topic and Focus in the two groups of languages, as I have tried to show in the previous sections. The first group is transparent and highly restrictive in the marking of information structure, the second one is more opaque and quite permissive, so that in certain cases a constituent that has to be obligatorily emarginated / dislocated in a language of the first group can be easily integrated into Broad Focus in a language of the second group. Not only is this difference crucial for predicting the distribution of the DE, it should be taken as one of the basic parameters of information structure from a comparative point of view.

7. A typology of existentials The claim that there are different kinds of “existential constructions”, each of them associated with different conditions on the internal DP, is not new. It has been presented in one version or other in works like Huang (1987), Abbott (1993), Li (1996), Lambrecht (2002), Paducheva (2003) and Beyssade (2004). It is undeniable that whatever classification of existentials we choose, it will be relevant for an analysis of the DE. But it also seems clear that such a classification is only a descriptive tool that in turn raises new questions and new problems (for instance, why are the subtypes just these ones? Is it possible to decompose the typology into a combination of basic features?). In this section I intend to clarify the place that a typology of existentials occupies in a discussion of the DE.

30. Lambrecht and Polinsky (1997) provide a list of basic properties that characterize thetic sentences.

 Manuel Leonetti

Suppose there are three basic types of existentials, as assumed until now: the proper existential, the eventive and the enumerative. Pure existential sentences are the prototypical instance of thetic / rhematic structures, they typically include a locative adjunct (although it is not obligatory), and the internal DP lacks autonomous reference. This means that its interpretation has to be solved without accessing contextual information from outside the construction; it is within the limits of the existential construction that the referential properties of the DP are to be established, which is in line with the Novelty Condition – moreover, I would claim that the Novelty Condition actually is a result of this requirement, since it bans access to given information and thus to anaphoric readings. The requirement of non-autonomous reference is dependent on the presence of the coda. It can be satisfied by expressions that do not encode any referential procedure, like indefinites – they only encode information about quantities and their interpretation can be determined by the surrounding elements (the verb, negation, quantificational adverbs) –, and, in some cases, by expressions that do encode some kind of referential procedure, like definites – they encode a uniqueness condition, but such a condition is not necessarily checked on contextual information. When definite expressions rely on contextual information for their interpretation, thus being assigned a deictic or an anaphoric reading, they are typically excluded from proper existentials (for instance, pronouns); in case they satisfy the uniqueness condition by means of linguistic information that is accessible in the sentence, in principle they should not be excluded from existentials (it is the case of “weak” definites and “cataphoric” definites, where definiteness is solved inside the DP). The second basic type is the eventive existential. I believe that the only thing that changes here, with respect to proper existentials, is that the postverbal expression is propositional and of the Stage-Level kind. The syntactic format of the propositional content is a small clause with its own Topic / Focus articulation. Inside the small clause the DP acts like an internal topic, thus eliminating any motivation for the DE: we noticed that names and definite DPs are possible in these contexts. The only additional assumption that is needed to deal with eventive existentials is this: in some languages the existential verb is able to select a small clause as its argument (Catalan, Italian, French, English) and in others it is not (Spanish). Finally, the third type is the enumerative one, with the classical “list” or “reminder” reading. Here the coda is implicit or absent, and definite DPs are allowed: the presupposed or implicit nature of the coda poses no constraint on the referential status of the postverbal DP, so that there is no DE. As a result, only pure existentials with codas are affected by the DE. If enumeratives are just a third type of structural pattern for existential sentences, we could say that the three classical analyses that were proposed in the literature on existentials, namely the NP-analysis, the adjunct-analysis and the small clause-analysis, are adequate. Each one of them is adequate for one of the three patterns described. In some sense, all of them are correct, although I consider the adjunction analysis to be the default one.



Definiteness effects and the role of the coda in existential constructions 

Some additional comment on enumerative existential sentences is in order. In English, they seem to be just an exceptional and marked use of the proper existential construction, instead of being an independent construction. The main motivations for viewing them as a special use are their marked status and the strong contextual restrictions that govern their appearance in discourse – recall that they are characterized as “contextualized existentials” in Abbott (1993). As pointed out in Abbott (1993) and Lumsden (1988:  214–223), so-called “list” and “reminder” interpretations are pragmatically inferred assumptions, triggered by the exceptional insertion of a definite DP in postverbal position; such readings may be seen as “last resort” inferential strategies involved in the interpretive process of an otherwise ill-formed construction. There is no semantic distinction between existential and enumerative interpretations, and the relevant examples seem to retain a literal existential interpretation along with an enumerative one. I refer to Lumsden (1988), Abbott (1993) and Cann (2007) for some more details on the pragmatic approach to sentences like There is John. Now the question that an approach like this may raise is why There is John forces its enumerative reinterpretation and happens to be finally acceptable – though as a marked and strongly contextualized use- while other ill-formed strings remain odd, and no inferential mechanism is there to rescue them. I think that two factors conspire to license enumerative existentials in English: one is the already mentioned absence of the coda, which changes the conditions for inserting a definite, since without the coda the definite is not constrained to be a part of Broad Focus; the other one is crucial and has to do with the absence of lexical alternative ways to convey an enumerative reading in English. In a few words, there are no competitors that might be better solutions than a there be construction. This opens the possibility for a contextual reinterpretation of the marked string. The same happens in French with enumerative uses of il y a DP. Spanish, on the contrary, absolutely excludes sentences like *Hay Juan, mainly because it can resort to the use of estar (Está Juan is perfect, though pragmatically constrained too). Thus, haber is used only in strictly existential sentences, and correspondingly the DE shows up in a strong version in Spanish. Catalan and Italian present different situations. Catalan allows sentences like Hi ha en Joan, though they are not limited to enumerative interpretations. As already pointed out, haver-hi is in competition with esser-hi, but esser-hi is limited to locative uses: this implies that the range of possible presentational uses with postverbal DPs that cannot be expressed by means of esser-hi is covered by haver-hi. With no competing lexical alternatives, haver-hi admits a number of contextual reinterpretations when followed by definite DPs. Italian allows any kind of definites in the postverbal position with esserci. This is due to the conflation of existential and quasi-locative constructions into a single pattern, as observed by Zamparelli (1996). Purely locative sentences are based on the verb essere, but the remaining possibilities are covered by esserci. Being not limited to existential use, esserci does not give rise to the habitual DE unless the coda is inside the VP.

 Manuel Leonetti

Sentences like C’è Gianni are not exceptional, marked, uses of an existential pattern, like its counterparts in English. Much more work is still to be done on the lexical means that languages rely on for the expression of the different kinds of existentials, but for my purposes it is enough to point to the central role of this factor in the distribution of the DE. As for enumerative uses, then, they are marked reinterpretations of existential sentences, as a result of last resort inferential strategies (the case of English or French), or otherwise they are possible readings of non purely existential constructions (the case of Italian). Summarizing, there are two basic patterns that syntax generates: a. the verb is followed by DP + optional XP coda b. the verb is followed by an eventive small clause Some languages only use the first one, some languages use both. Only in the first one is some variant of the DE expected. In addition, the presence and distribution of the DE is dependent on (a) the way each language lexicalizes the expression of existential and related interpretations; and b) the extent to which it allows for complex structures and referentially independent DPs inside Broad Focus sentences. Therefore, the possible types of existential constructions are only one of the ingredients of the complex cluster of notions needed for a crosslinguistic approach to the DE.

8. Concluding remarks Many questions still lack an adequate answer for a global view of the DE, but I hope that this discussion has highlighted at least a number of partial results that could contribute to a better understanding of (in)definiteness marking. Among them are the following ones: 1. The DE is usually presented as a diagnostic for definiteness and as a defining general property of existentials, but a look at different types of existential constructions shows that only pure existentials exhibit the DE as a core property. This may lead a grammarian into error if (s)he fails to pay attention to the relevant set of data or ignores the distinction between proper and eventive existentials. 2. There are grounds to believe that the DE manifests itself in some way in most, if not all, languages, though its presence may be obscured by a combination of factors, partly lexical, partly syntactic. The fact that it is such a widespread phenomenon supports the accepted view that it is a semantic / pragmatic constraint “involving something broader than grammatical definiteness” (Lyons 1999: 246). 3. Just identifying some condition associated with the sequence there be as the origin of the DE (for instance, the Novelty Condition) is not enough. The presence of the coda must be taken into account and given a central role, possibly assuming that the coda requires some kind of semantic incorporation of the internal DP into the



Definiteness effects and the role of the coda in existential constructions 

predicate in order to obtain a thetic reading of the construction. The basic factors involved in the DE that I have tried to highlight are (a) the role of the internal composition of the existential construction, and in particular the presence of the coda in final position; (b) the effects of different lexical predicates competing for the expression of existential / presentative meanings (haber and estar in Spanish, esserci and essere in Italian, haver-hi and esser-hi in Catalan); (c) the possibility that the existential verb selects an eventive small clause, thus cancelling the conditions for the DE, and finally (d) the crosslinguistic variation in Focus structure and information packaging as a source of contrasts linked to the DE both in existential sentences and in inverted subject constructions. Much work is still needed to tie all these factors together in a fully developed theory. 4. The variety of discourse values that existentials show (pure existence, location, enumeration, presentation...) is the result of the interpretive mechanisms triggered by the combination of the existential predicate with different kinds of DPs and codas. Some violations of the DE in languages like English are only acceptable as contextual reinterpretations of marked constructions (the ‘list’ / ‘reminder’ cases). 5. The DE is one of the constraints affecting the syntactic distribution of definites and indefinites. While it favours indefinites over definites in certain contexts, other constraints favour definites or specific DPs in other contexts (typically, in subject position). As argued in Lyons (1999: ch.6), they are crosslinguistic general tendencies, stricter in some languages than in others. I believe that we have enough evidence to consider all of them as ultimately derived from the principles of information structure, since definiteness / specificity requirements are always favoured by Topic positions, and indefiniteness / non-specificity requirements are favoured by Focus positions. It is information structure that mediates between (in)definiteness and syntactic structure. In this sense the DE is only a small part of a broader mechanism connecting reference and grammar.

Acknowledgements I am grateful to the audiences at the Copenhagen Determination Symposium (August, 2004) and the Second NEREUS Workshop in Berlin (October, 2004) for comments on earlier versions of this paper, and especially to Chris Barker, Ignacio Bosque, Luis Eguren, Louise McNally, John Moore, and Maria Polinsky for their ideas and suggestions. Vicky Escandell-Vidal deserves a very special mention for her crucial help with the Catalan data, her intuitions and her encouragement. Thanks to Claire Beyssade, Susann Fischer, Helena Guerra Vicente, Aroldo Leal de Andrade, Carmen Marchante, Gemma Rigau and Elisabeth Stark, who provided me with materials and judgements, and to Aoife Ahern for correcting my English. All mistakes and obscurities are mine. This paper was completed during my stay as a visiting scholar in the Department of

 Manuel Leonetti

Linguistics of the University of California at San Diego (February-March, 2005), and I am grateful to the people in the Department for their warm hospitality. This research has been partially supported by the Spanish Ministry of Education through the grant HUM2006–06630 for the project “SPYCE: Semántica procedimental y contenido explicito”.

References Abbott, B. 1993. A pragmatic account of the definiteness effect in existential sentences. Journal of Pragmatics 19: 39–55. Abbott, B. 1999. Support for a unique theory of definite descriptions. In Proceedings from Semantics and Linguistic Theory IX, T. Matthews & D. Strolovitch (eds), 1–15. Ithaca NY: Cornell University. Alboiu, G. 2002. The Features of Movement in Romanian. Bucharest: University of Bucharest Press. Basilico, D. 1997. The topic is ‘there’. Studia Linguistica 51(3): 278–316. Belletti, A. 1988. The case of unaccusatives. Linguistic Inquiry 19: 1–34. Belletti, A. 2001. Inversion as focalization. In Subject Inversion in Romance and the Theory of Universal Grammar, A. Hulk & J.-Y. Pollock (eds), 60–90. Oxford: OUP. Bende-Farkas, A. 2001. Verb-object Dependencies in Hungarian and English. PhD dissertation, University of Stuttgart. Bende-Farkas, A. & Kamp, H. 2001. Indefinites and binding: From specificity to incorporation, Ms. Benincà, P., Salvi, G. & Frison, L. 1988. L’ordine degli elementi della frase e le costruzioni marcate. In Grande grammatica italiana di consultazione, Vol. I, L. Renzi (ed.), 115–225. Bologna: Il Mulino. Beyssade, C. 2004. Les constructions existentielles. In Définir les indéfinis, C. Beyssade & C. Dobrovie-Sorin (eds). Paris: CNRS. Brassil, D. 2004. On the Distribution of the Definiteness Effect in Spanish: Relativizing the Partitive Case Hypothesis. Ms. Brucart, J.M. 2002. Els determinants. In Gramàtica del català contemporani, Vol. II, J. Solà (ed.). Barcelona: Empuries, 1435–1516. Cann, R. 2007. Towards a dynamic account of be in English. In Existence: Semantics and Syntax, I. Comorovski & K. von Heusinger (eds), 13–48. Springer. Cardinaletti, A. 2001. A second thought on Emarginazione: Destressing vs. ‘Right Dislocation’. In Current Studies in Italian Syntax, G. Cinque & G. Salvi (eds), 117–135. Oxford: Elsevier. Chang, H.H. 2004. Definite NPs in Mandarin you-existentials. Proceedings of the Chicago Linguistic Society, 40. Chicago: The Chicago Linguistic Society. Côté, M.-H. 1999. Quantification over individuals and events and the syntax-semantics interface: The case of the existential constructions. In Proceedings of the Seventeenth West Coast Conference on Formal Linguistics, K. Shahin, S. Blake & E.-S. Kim (eds), 147–161. Stanford CA: CSLI. Erteschik-Shir, N. 1997. The Dynamics of Focus Structure. Cambridge: CUP.



Definiteness effects and the role of the coda in existential constructions  Frascarelli, M. 2000. The Syntax-phonology Interface in Focus and Topic Constructions in Italian. Dordrecht: Kluwer. Huang, C.T.J. 1987. Existential sentences in Chinese and (in)definiteness. In The Representation of (In)definiteness, E. Reuland & A. ter Meulen (eds), 226–253. Cambridge MA: The MIT Press. Kampers-Mahne, B. et al. 2004. Subject NP inversion. In Handbook of French Semantics, F. Corblin & H. de Swart, 553–579. Stanford CA: CSLI. Keenan, E. 2003. The definiteness effect: Semantics or pragmatics?. Natural Language Semantics 11: 187–216. Kuroda, S.-Y. 1995. Sentences, judgments and propositions. University of California at San Diego, Ms. Lambrecht, K. 2002. Topic, focus and secondary predication. The French presentational relative construction. In Romance Llanguages and Linguistic Theory 2000, C. Beyssade et al. (eds), 171–212. Amsterdam: John Benjamins. Lambrecht, K. & Polinsky, M. 1997. Typological variation in sentence-focus constructions. In Proceedings of the Chicago Linguistic Society 33, 189–206. Chicago IL: The Chicago Linguistic Society. Landman, F. 2004. Indefinites and the Type of Sets. Oxford: Blackwell. Leonetti, M. 1999. El artículo. Gramática descriptiva de la lengua española, Vol. I, I. Bosque & V. Demonte, 787–890. Madrid: Espasa-Calpe. Leonetti, M. 2004. Specificity and differential object marking in Spanish. Catalan Journal of Linguistics 3: 75–114. Leonetti, M. 2005. Notas sobre tópicos y definitud en construcciones existenciales, Ms. Li, Y.-A. 1996. Definite and indefinite existential constructions. Studies in the Linguistic Sciences 26(2): 175–191. Longa, V., Lorenzo, G. & Rigau, G. 1998. Subject clitics and clitic recycling: Locative sentences in some Iberian Romance languages. Journal of Linguistics 34: 125–164. Lyons, J. 1999. Definiteness. Cambridge: CUP. McNally, L. 1992. An Interpretation for the English Existential Construction. PhD dissertation, University of California at Santa Cruz. (Published as A Semantics for the English Existential Construction. New York NY: Garland, 1997). McNally, L. 1998. Existential sentences without existential quantification. Linguistics and Philosophy 21: 353–392. Mikkelsen, L. 2001. Reanalyzing the Definiteness Effect: Evidence from Danish, Ms. Milsark, G. 1977. Toward an explanation of certain peculiarities of the existential construction in English. Linguistic Analysis 3(1): 1–31. Paducheva, E. 2003. Definiteness effect: The case of Russian. In Reference and Anaphoric Relations, K. von Heusinger & U. Egli (eds), 133–146. Dordrecht: Kluwer. Moro, A. 1997. The Raising of Predicates. Cambridge: CUP. Rando, E. & Napoli, D.J. 1978. Definites in there-sentences. Language 54: 300–313. Rigau, G. 1994. Catalan presentational sentences and the properties of Agr nodes. In Paths Towards Universal Grammar, G. Cinque et al. (eds), 343–359. Georgetown DC: Georgetown University Press. Rigau, G. 1997. Locative sentences and related constructions in Catalan: ésser / haver alternation. In Theoretical Issues at the Morphology-syntax Interface, A. Mendikoetxea & M. UribeEtxebarría (eds), 395–421. Bilbao: Universidad del País Vasco.

 Manuel Leonetti Rizzi, L. 1986. On the status of subject clitics in Romance. In Studies in Romance Linguistics, O. Jaeggli & C. Silva-Corvalán (eds), 391–419. Dordrecht: Foris. Rosengren, I. 1997. The thetic/categorical distinction revisited once more. Linguistics 35(3). Safir, K. 1985. Syntactic Chains. Cambridge: CUP. Suñer, M. 1982. The Syntax and Semantics of Spanish Presentational sentence-types. Washington DC: Georgetown University Press. Van Geenhoven, V. 1998. Semantic Incorporation and Indefinite Descriptions. Stanford CA: CSLI. Vallduví, E. 1995. Structural properties of information packaging in Catalan. In Discourse Configurational Languages, K. Kiss (ed.), 122–152. Oxford: OUP. Vangsnes, Ø. 1994. The Syntactic Source and Semantic Reflexes of the Definiteness Effect. MA dissertation, University of Bergen. Zamparelli, R. 1996. Layers in the Determiner Phrase. PhD dissertation, University of Rochester. Ziv, Y. 1982. Another look at definites in existentials. Journal of Linguistics 18: 73–88. Zubizarreta, M.L. 1998. Prosody, Focus and Word Order. Cambridge MA: The MIT Press. Ward, G. & Birner, B. 1995. Definiteness and the English existential. Language 71: 722–742.

Determination of N2 modifiers in Spanish nominal syntagmatic compounds Henrik Høeg Müller

Copenhagen Business School To many informants the semantic difference between the following two types of nominal syntagmatic compounds (NSCs) in Spanish seems almost undetectable or maybe even non-existent. a. Crisis de la energía (energy crisis), crisis del petróleo (oil crisis), puesta del sol (sun set) [internal structure: N1prep. def.N2] b. Fuente de energía (energy source), pozo de petróleo (oil well) día de sol (day of sun) [internal structure: N1 prep. ØN2] However, I will claim that the two structure types are far from being synonymous. A fundamental assumption in the article is that conceptualisation of basic entity types as either heterogeneous, delimited and internally structured (bounded) or homogeneous, amorphous and without delimitation or internal structure (unbounded) interacts with the grammatical determiner system in attributing interpretational values to NSCs. The point of departure is that the definite article attributes to N2 either a referential reading or a prototype reading, while the zero determiner brings about an interpretation as either a mass or a concept. In the last part of the article it is argued that the semantic effects of the N2 configuration are essentially different from genericity. Nominals in existential or generic expressions denote, or are sometimes even said to refer to, a class or type of objects, whereas N2 gives access to characteristic properties of the entity type.

1. Introduction This article will focus on the determination or lack of determination of N2 modifiers in nominal syntagmatic compounds (NSCs) in Modern Spanish. NSCs are here understood as the following two types of construction:

 Henrik Høeg Müller

(1) a. [internal structure: N1 prep. def.N2] Crisis del petróleo (oil crisis), consumo del cigarro (cigarette consumption), puesta del sol (sun set) b. [internal structure: N1 prep. ØN2] Fuente de petróleo (oil well), alquiler de coche (care hire), día de sol (day of sun) In the standard analysis of NSCs, it is characteristic of these structures that N2 appears without any determinative or quantificational modifiers, i.e. it occurs as a bare noun, for which reason it cannot be referential in the sense of pointing out a specific individual.  The bare-noun modifier, i.e. N2, is incorporated into the morphosyntactic head, i.e. N1, and the two nouns act referentially as one in the sense that the whole complex NP denotes one entity type only. By this incorporation process, N1 and N2 merge conceptually, but there is no direct form-meaning iconicity because the nouns are still formally separated by a preposition, i.e. the structure of the compound is phrasal.1 There is a clear distinction between the structural and the semantic aspects of the relation between N1 and N2. According to the traditional description of phrasal compounding, only the examples of (1b) qualify as NSCs. However, the examples in (2) illustrate that in many cases, although introduced by the definite article, N2 is not singling out a specific object in the discourse domain, but should rather be described as conveying some kind of nonreferential, attributive meaning dimension to the head of the NP. (2) a. ¿Qué características debe tener la celebración del matrimonio? El matrimonio cristiano requiere una celebración litúrgica que […] (What characteristics should a marriage celebration have? The Christian marriage requires a liturgical celebration which […]) b. ¿Qué hacemos con la colilla del cigarro cuando fumamos por la calle? (What do we do with the cigarette stub when we smoke in the street?)

It is, however, obvious from the analogous examples in (3) that an appropriate context can change the interpretational value of N2 from referential to non-referential, or vice versa. In general, this means that the definite article can stimulate at least two completely different interpretations of N2 inside an NP, one which is compatible with a compound reading and another which is not. (3) a. Dicho documento, que no fue elevado a escritura pública antes de la celebración del matrimonio, fue firmado por las partes en presencia de una notario en la ciudad de Baltimore en Maryland. (The document in question, which had not been vested the form of a notary document before the celebration of the marriage, was signed by the

1.

In this article I shall only comment NSCs which include the preposition de.



Determination of N2 modifiers in Spanish nominal syntagmatic compounds 

parties in presence of a notary public in the city of Baltimore in Maryland.) b. Él también cruzó la calle y al pasar junto a ella arrojó la colilla del cigarro a sus pies. (He also crossed the street, and when he passed her, he threw the stub of the cigarette in front of her feet.) To many informants the semantic differences between the two types of NSCs, i.e. [N1 prep. det.N2] and [N1 prep. ØN2], which together with the [N + adj.]-structure, as in filtro solar (sun filter), largely constitute the productive basis of the nominal wordformation system in Spanish, seem almost undetectable or maybe even non-existent. Apparently, both types of NSCs license non-referential readings of N2. In the grammatical literature the phenomenon of presence or absence of the definite article inside NSCs to my knowledge is left undescribed maybe because, consciously or unconsciously, it is attributed to some sort of ad hoc language development not particularly worthy of attempts at scrutiny and systematic description. However, contrary to the intuition of native speakers, I will claim that the two structure types are far from synonymous. The point of departure is that the zero-determiner and the definite article are capable of configuring the meaning of N2 inside NSCs in two basic ways each. The definite article attributes to N2 either a referential reading or a prototype reading, while the zero determiner brings about an interpretation as either a mass or a concept. In addition to these basic meaning configurations, under certain circumstances the definite article can convey to the prototypical ground meaning a connotation of singular individual. The same kind of singular individual interpretation can also be activated by zero determination in combination with some deverbal N1 nouns. In both cases, of course, the singular individuals are to be interpreted as non-specific. Finally, together with a handful of deverbal head nouns ‘det.N2’ obtains a generic reading, which, however, is rather marginal. These different meaning configurations are shown in (4). (4) [N1 prep. det.N2] Specific a. Media hora después la madre se acerca a la cama del niño y melosamente le dice: (Half an hour later the mother approaches the bed of the child and she says to him in a sweet voice: […]) “Pure” prototype b. El consumo del cigarro puede provocar disfunción eréctil. (Cigarette consumption can provoke erecteil dysfunction.) Singular individual prototype c. ¿Qué hacemos con la colilla del cigarro cuando fumamos por la calle? (What do we do with the cigarette stub when we smoke in the street?)

 Henrik Høeg Müller

Generic d. Un sistema de cultivo permite el crecimiento de la fresa sin necesidad de que esté en contacto con el suelo. (A cultivation system permits the growth of the strawberry without it being necessary that it is in contact with the soil.) [N1 prep. Ø.N2] Mass e. El atacante lanzó una bomba de petróleo. (The attacker threw a petrol bomb.) Concept f. No había duda: aquel tipo llevaba un hacha de bombero en las manos. (There was no doubt: that guy had a fireman’s axe in his hands.) Singular individual g. Cada vez se requiere un mayor sacrificio para la compra de piso. (It becomes more and more expensive to buy an apartment [buying of apartment].) The meaning variations of N2 illustrated in (4) exhibit different alternation patterns. The alternation between ’ØN2’ and ’det.N2’ can, for instance, stimulate a focus shift from a concrete-entity meaning dimension of N2 with zero determination to a scenario in which the entity denoted by N2 with a definite determiner obtains a metonymic status, cf. the examples in (1). Speaking very generally, we might say that the interaction between the grammatical system in the form of ’ØN2’/’det.N2’ and our perception of the noun denotations, together with the situations in which they are usually involved, determine which interpretations and alternation patterns are relevant in a given context. This theoretical position concerning the basic meaning of N2, or any noun for that matter, builds on the generally accepted insight (see e.g. Herslund 1997 and Vangsnes 2001) that nouns are not specified as mass or count as part of their lexical information, but are coded as such by syntax. The first part of this article will concentrate on identifying the different meanings of N2 and contrast them with each other. The last part of this article will be dedicated to broadening a specific perspective of the N2 meaning configuration triggered by the definite article, i.e. the difference between a prototype and a generic reading of N2. By drawing parallels and distinctions between the sentence level and the nominal level, I will substantiate the claim that the different non-specific meaning facets that can be ascribed to N2 by use of the definite article cannot be adequately accounted for on the basis of the concept of genericity alone. Nominals in existential or generic expressions denote, or are sometimes even said to refer to, a class or type of objects, whereas definite NPs in N2-position typically give access to characteristic properties of the entity type thus creating a prototype reading of the noun. In other words, the definite article introducing N2 in compounds, or structures similar to compounds, should not be reduced to functioning as a so-called generic article.



Determination of N2 modifiers in Spanish nominal syntagmatic compounds 

2. The definite article inside compound structures First of all, we can observe very clearly that the Spanish NSCs in the example in (5) below find a natural counterpart in English and Danish compound structures (cf. the translations in parentheses of the Spanish highlighted structures), although in some of the cases, genitive or prepositional phrases would also constitute adequate translation possibilities. (5) a. Cuando un periodo solunar cae dentro de los 30 minutos de la salida o la puesta del sol es posible pronosticar buena pesca. (sun set/solnedgang) A la luz de la luna le era muy fácil seguir el rastro de sangre. (moon shine/ måneskin) Desde las naves espaciales, la superficie de la Tierra muestra una continua presencia de nubes en la atmósfera. (surface of the Earth2/jordoverflade) b. Cada vez que la presión del aire, la temperatura o cualquier otro parámetro de perturbación ocasione un cambio de posición, el controlador tendrá la […] (air pressure/lufttryk) El nivel del agua puede aumentar en varios metros, incluso en casos extremos el nivel del agua ha aumentado en 30 metros. (water level/vandstand) Durante décadas, la barrera del sonido fue una impenetrable pared contra la que […] (sound barrier/lydmur) Así, según la teoría de la relatividad, la velocidad de la luz es una barrera fundamental de la naturaleza que no puede ser superada. (relativity theory/relativitetsteori) ¿Cuales son las formas de celebración del matrimonio? (marriage contracting/ægteskabsindgåelse) c. La pulpa macerada llega a una prensa discontinua donde se obtendrá por medio de presión el zumo de la manzana. (apple juice/æblesaft(juice)) […] tratan aquí las características de cultivo del pepino, del melón, de la [...]. (cucumber growing – melon growing/agurkedyrkning – melondyrk­ ning) Y si tener un pequeño municipio les ha impedido disponer de tierras suficientes para la plantación del melón, han sabido sobreponerse y continuar […] (melon planting/melonplantning) Elaboración: Vaciar los melones y reservarlos para utilizarlos de recipiente. Con la pulpa del melón y el resto de los componentes hacer un gazpacho. (melon pulp/melonkød) Una iniciativa prevé promocionar el consumo del champiñón y divulgar sus propiedades saludables. (mushroom consumption/champignonforbrug)

2.

In this case the English translation does not correspond to a compound.

 Henrik Høeg Müller



El consumo del cigarro, puede provocar disfunción eréctil. (cigarette consumption/cigaretforbrug)

The fact that the Spanish structures translate nicely into Danish compounds with solid ortography and English compounds with separate orthography of course does not in isolation constitute a very good argument in support of the assumption that the Spanish structures should be compounds too. On the other hand, it also seems difficult to reject this observation as an irrelevant coincidence when, almost without exception, the Spanish examples actually find perfect translational matches in Germanic compound structures. The examples in (5) have been divided into three groups. From a superficial glance at the examples in (5a), it appears acceptable to recognise that N2 in these cases has specific reference, i.e. refers uniquely to our sun, moon and earth, which is also corroborated by the fact that sol, luna and tierra in such contexts are sometimes spelled with capital letters, cf. the third example in (5a). It is commonly claimed that the lexemes in question, when occurring at the sentence level, are often introduced by the definite article because of some shared cultural background knowledge or conventional conception of the world according to which the denotations of the lexical items constitute prominent and unique physical entities (see e.g. Givon 1984: 399). It would probably also be possible to argue convincingly in favour of the supposition that the N2-modifiers in (5b) are introduced by the definite article because they denote general concepts, which due to a stereotyped world view occupy salient positions within many language communities. However, one also has to admit that there is a difference between (5a) and (5b) in that the lexemes aire, agua, sonido, etc. do not designate unique first order entities but denote more abstract phenomena. Finally, the examples in (5c) differ substantially from the ones in (5a) and prob­ ably a little less from those in (5b). The modifier nouns of the structures in (5c) denote common, every day physical items, i.e. they do not denote entities that hold culturally prominent positions or can be considered as fundamental phenomena in society. Drawing an analogy to the sentence level, in these last cases the presence of the definite article could be tentatively explained with reference to the idea that N2 has some kind of generic status, cf. e.g. El champiñon se consume …/El pepino se cultiva … (mushrooms are eaten …/cucumbers are grown …). This is of course not without merit, but the problem is that in general an array of different vague concepts and explanations are used to account for these seemingly quite similar constructions. It appears that the possibilities of systematisation and generalisation have not been challenged or developed sufficiently, especially not with reference to contrasting data with zero determiner counterpart constructions. At the same time, none of the conventional approaches to the function of the definite article, such as e.g. uniqueness vs. non-uniqueness, familiarity vs. novelty, strength vs. weakness and specificity vs. non-specificity (for a survey of the functions that are widely taken to characterise the use of determiners at the sentence level, see



Determination of N2 modifiers in Spanish nominal syntagmatic compounds 

e.g. Abbott’s (2004) article about definiteness and indefiniteness), seem to be capable of pinning down its use and meaning in compounds. However, Vangsnes (2001: 271) introduces a distinction between substantial and functional elements that to a certain extent proves useful in understanding the role of the definite article in NSCs: “By a ‘substantive element’ we understand an element which contributes substantial denotative information of its own, whereas by a ‘functional element’ we understand an element which does not contribute substantial denotative information of its own.” As regards how the articles relate to this distinction, he states the following: “Articles are the most typical functional elements of the nominal system, […]. As for most other determiners they provide substantial enough denotative information of their own to be considered substantive elements.” This means that apart from functioning as determiners in the sense that they provide procedural information about how the entity denoted is to be accommodated in the discourse, the other determiners supply extra information to the noun. For example, demonstratives convey information concerning the speaker’s position in relation to the referent of the NP, and the possessive pronoun has an independent referent which is different from the NP it determines. Vangsnes’ description of the difference between these two types of determiners explains why it is only the definite article – in this connection I disregard the status of the indefinite article – which can form part of compounds. It is due to the fact that the independent, denotative content of the substantial elements blocks the incorporation of N2 into N1. In order to regard a constellation of nouns as a phrasal compound, it is necessary that N2 should have the status of a modifier. This means that N2 cannot be actualised independently in the syntax, i.e. N2 is embedded under the NP and has no direct contact with syntax at the sentence level. This negative definition may provide an answer to why the substantial determiners cannot introduce N2 in compounds, but it does not explain what it means to be a functional element, i.e. which function the definite article is supposed to have and subsequently what it is that in some situations allows ‘det.N2’ to be semantically merged with N1. Vangsnes (ibid: 271f.) himself is not very precise in his description of what he thinks the function of the definite article might be. Comparing the demonstrative and the definite article he writes the following: “The [definite] article, on the other hand, entails something like a default deictic reading: it points to a given referent, but rather than picking one which is located in certain ways spatially, it merely picks out the most salient one in the discourse.” And he adds: “The distinction between substantive and functional elements as conceived of here may appear more scalar than strictly dichotomous, and it seems difficult to give a clear definition of where on the scale the dividing line should be drawn, i.e. what is means to contribute sufficiently substantial denotative information to be considered a substantive element.” So Vangsnes introduces an important distinction between two types of determiners which may be able to explain why it is the definite article and not other determiners that can form part of composite structures, but his conception of the function of the

 Henrik Høeg Müller

definite article as a kind of abstract deictic marker does not contribute to throwing light on its role inside compounds. Therefore it is necessary to look for alternative explanations. As mentioned in the introduction to this section, the difference of meaning between a determined and a non-determined N2 can be hard to detect, but in the following I will argue that the two constructions reflect different conceptualisations of basic entity types.

3. Alternations and interpretations Mass reading vs. prototype reading

3.1

The distinction between the mass reading and individual entity reading of nouns reflects two radically different ways of conceptualising entities: (1) either as inhomogeneous, delimited and internally structured (bounded), or as (2) homogeneous, amorphous entities without delimitation or internal structure (unbounded) (Langacker 1991, Jackendoff 1991 and Herslund 1997). It is not a part of the lexical make-up of nouns to be specified in advance in the lexicon as either mass or count, i.e. as denoting masses or individuals. This value is encoded into the noun via the grammatical system when the noun is inserted in a syntactic environment (see Herslund ibid. and Vangs­ nes 2001). The position argued for here is then that conceptualisation of basic entity types interacts with the grammatical determiner system in attributing interpretational values to NSCs. The examples below illustrate how the noun in N2-position receives different interpretations according to whether it is instantiated with or without determiner.

(6)

Mass reading

Prototype reading

Pozo de agua (water well) Depósito de agua (water deposit) Pozo de petróleo (oil well) Refinería de petróleo (oil refinery)

Guerra del agua (war of the water)

Depósito de petróleo (oil deposit) Pulpo de melón (melon pulp) Cultivo de melón (melon growing)

Crisis del petróleo (crisis of the oil) Política del petróleo (politics of the oil) Magnate del petróleo (magnate of the oil) Pulpo del melón (pulp of the melon) Cultivo del melón (growing of the melon)



Determination of N2 modifiers in Spanish nominal syntagmatic compounds 

In the first examples in which the lexeme in N2-position is agua or petróleo, the absence of a determiner stimulates a mass reading of the N2-nouns, i.e. a focus on the liquids water and oil conceptualised as substances. This reading of the nouns is in good keeping with the semantic or denotational features of the head nouns. Refinería, depósito and pozo give access to semantic fields which involve the comprehension of petróleo and agua as substances. The lack of determiner encourages a mass interpretation, which implies that we aim our focus on water and oil as concrete materials or products in the form of liquid substances. We could tentatively say that the zero determiner has an inclination towards triggering the reading ”made of ” because of the concrete, material interpretation it entails. In the parallel examples with determined N2, the definite article causes us to make an abstraction from the substance or mass reading to a more intangible and conceptual level. It is a mental process where we go from the concrete substance to all its implications. Instead of focussing on the concrete, material meaning dimension of the noun, the definite article allows us to view the whole mental space or scenario in which the substance occupies a central position, i.e. it opens up the possibility to draw on all the associations to which the noun in question gives access. You could say that this abstraction permits us to conceptualise agua and petróleo in an institutionalised manner, an observation which is also in this case corroborated by the semantics of the heads with which they combine. When the lexemes agua and petróleo are united with the head nouns crisis or guerra, they do not just denote the relevant substances but they imply a series of related concepts and factors concerning for example extraction, political and commercial aspects, etc. By adding the determiner we create a prototype and thus a sort of metonymy reading of agua and petróleo in the sense that these nouns represent a larger meaning domain than their “immediate” denotations. The examples with the modifier melón illustrate that in many cases both types of construction will be possible. However, in line with the above discussion, I would argue in favour of the hypothesis that the two types trigger different interpretations, i.e. a material reading where focus is on the substance per se, and a prototype reading where the noun melón denotes a kind of idealised abstraction. When examples are identical apart from the presence or absence of the article so that N1 cannot contribute to the disambiguation of the construction, this claim remains an interpretation of data which is difficult to verify empirically, though it can be readily discerned in the contrasting examples pozo de petróleo as against crisis del petróleo.

3.2

Singular individual prototype vs. concept

The examples in (7) below show that in certain situations the prototype reading is accompanied by an interpretation of N2 as denoting one single individual. Partitive constructions like the ones in (7a and b) imply an understanding of the denotation of N1 as forming part of one and only one entity which constitutes the whole. Equally, the constellations in (7c, d and e) entail in our normal conceptualisation of the world a

 Henrik Høeg Müller

one-to-one relationship between the entities denoted by N1 and N2 (e.g. following the logic that to one axe corresponds one fireman). (7) a. ¿Qué hacemos con la colilla del cigarro cuando fumamos por la calle? (What do we do with the stub of the cigarette when we smoke in the street?) b. Para que las puertas del coche no se congelen en invierno, cubre con aceite de cocina los burletes de goma. (In order to prevent that the doors of the car freeze in winter, cover the rubber weather stripes with cooking oil.) c. – Mire vuestra merced – respondió Sancho –, que aquellos que allí se parecen no son gigantes sino molinos de viento, y lo que en ellos parecen brazos son aspas que volteadas del viento hacen andar la piedra del molino. (”Look, your worship,” said Sancho. ”What we see there are not giants but windmills, and what seem to be their arms are the vanes that turned by the wind make the millstone go.”) d. Como ejemplo citamos la hacha del bombero, usado en el combate al fuego y la hacha del guardabosques utilizado para derrumbar árboles.3 (As an example we mention the fireman’s axe used in the fight against fire and the woodman’s axe used to fell trees.) e. Si todavía no conoce Portugal, sepa que siete días en Lisboa le supondrán 51.700 pesetas en un hotel céntrico como el Miraparque. La cama del niño es gratis. (If you do not know Portugal yet, you should be aware that seven days in Lisbon will cost you 51.700 pesetas in a centric hotel like Miraparque. The cot is free of charge.) In corresponding constructions N2 can, of course, be referential, cf. the examples in (8). (8) a. Él también cruzó la calle y al pasar junto a ella arrojó la colilla del cigarro a sus pies. (He also crossed the street, and when he passed her, he threw the stub of the cigarette in front of her feet.) b. El jugo que sale mediante la presión de la piedra del molino va hacia las tinas de fermentación por un sistema de canales subterráneos. (The juice that exits because of the pressure of the milestone runs to the fermentation barrels through a system of underground channels.)4

However, it is characteristic of the singular individual prototype constructions that they cannot occur with cardinals (*dos hachas del bombero, *dos colillas del cigarro) without 3. Rightfully the expression should be el hacha del bombero but I have chosen to reproduce the example in its original form. 4. This is a description of a process in a specific mill.



Determination of N2 modifiers in Spanish nominal syntagmatic compounds 

losing the prototype interpretation, and that the alternation with zero-determiner (co­ lilla de cigarro, piedra de molino) does not provoke a mass interpretation, but instead what was in the introduction referred to as a conceptual reading.5 It is distinctive about the conceptual reading when negatively defined that the denotation of N2 can neither be interpreted as a mass nor a non-specific individual. In examples such as pulsera de mujer (bracelet for woman) and torre de iglesia (tower of church/church tower) it would simply be misleading to claim that the lack of a determiner causes the denotations of mujer or iglesia to be taken as masses, or a reading as non-specific individuals for that matter. In these cases N2 only denotes the features that constitute the concept. When N2 denotes a concept, the whole NP is countable, cf. dos torres de iglesia, dos pulseras de mujer, a fact which clearly points towards the existence of a higher degree of connection between the nominal constituents. For N2 to be interpreted as a prototype in the form of a singular individual, it is required that the whole NP should be interpreted as a prototype at the sentence level, cf. the examples in (7). This requirement does not apply when N2 denotes a prototype which does not carry the co-meaning of singular individual, cf. the examples in (10bcd). The pragmatic restrictions concerning the possibilities of obtaining prototype reading in general imply that it may seem less obvious to interpret prototype NPs as close-knit units, and perhaps it is even debatable to what extent ‘det.N2’ in these cases is syntactically incorporated into N1. However, I will still maintain that the NPs in question denote single concepts, i.e. that they have unitary meaning.

Referential reading

3.3

In the examples below the absence of a determiner has the result that N2, the noun sol, is interpreted as denoting a mass. Rayo de sol means a beam made out of the abstract material sun – the photons – día de sol a day filled with the substance sun, and reloj de sol a clock which functions on the basis of sun beams. We might say that a sort of constitutive reading is established between N1 and N2.

(9)

Mass reading

Referential reading

Quemadura de sol (sun burn) Rayo de sol (sun beam) Día de sol (day of sun)

Quemadura del sol (burn of the sun) Rayo del sol (beam of the sun) Día del sol (day of the sun) Puesta del sol (set of the sun)

Reloj de sol (sundial) 5. The prototype interpretation of N2 can also be achieved by letting both N1 and N2 occur in the definite plural form, cf. the following constructed example: ¿Qué hacemos con las colillas de los cigarros cuando fumamos por la calle? In this way, a distributive interpretation arises where every stub, so to speak, has its cigarette counterpart.

 Henrik Høeg Müller

The tricky question is, however, what happens in these last cases when N2 is introduced by the definite article. It is difficult to reject that N2 in the right column examples should not be interpreted as referring uniquely to the one sun that exists in our daily conception of how the world is organised. This view-point implies that examples such as quemadura del sol and puesta del sol actually mean ”burning caused by the sun” and ”event where the sun disappears in the horizon”. That the determined nouns in N2-position in these cases establish unique reference and do not activate a prototype interpretation is indirectly corroborated by the examples in (10). (10) a. [...]: por ejemplo la carencia de energía por la crisis del petróleo. Se sabe, además, que no hay nada más parecido a una salida de sol que una puesta de sol. ([…]: for example the shortage of energy caused by the oil crisis. Moreover, we know that there is nothing more similar to a sunrise than a sun set.) En ese mundo había dos enormes soles, dos puestas de sol y dos amaneceres. (In that world there were two enormous suns, two sun sets and two dawns.) Debido a la rápida rotación de Bespin, los visitantes pueden disfrutar de dos puestas de sol y dos amaneceres cada día “estándar”. (Due to the rapid rotation of Bespin, the visitors can enjoy two sun sets and two dawns every day, as standard.) Según explica Antoine de Saint-Exupéry (1979), un día el Pequeño Príncipe vio cuarenta y tres puestas de sol solo moviendo un poco la [...] (As Antoine de Saint-Exupéry (1979) tells us, one day the Little Prince saw fortythree sun sets just moving a little bit the […]) La mejor vista se tendrá unos 45 minutos después de la puesta de sol. (The best view you get apr. 45 minutes after (the) sun set) Las fotografías son de las puestas de sol en partes diferentes de Noruega. (The photos show the sun sets in different parts of Norway.) b. Con la llegada de los años setenta, el sector se vio afectado por dos crisis del petróleo, en 1973 y 1979. (In the seventies, the sector was affected by two oil crisis, in 1973 and 1979.) La crisis de las materias primas nos es presentada como una crisis del petróleo. (The raw material crisis is presented to us as an oil crisis.) Esto es lo que parece haber sucedido, por ejemplo, con las crisis del petróleo entre 1973 y 1984. (This is what seems to have happened, for example with the oil crisis between 1973 and 1984.)



Determination of N2 modifiers in Spanish nominal syntagmatic compounds 

c. Parece que algunos pocos prefieren que estalle una guerra del agua en la zona mediterránea española […] (It looks as if a few people prefer that a water crisis breaks out in the Spanish Mediterranean zone […]) d. Albert Einstein desarrolló dos teorías de la relatividad bastante diferentes. (Albert Einstein developed two quite different theories of relativity.) The six examples in (10a) show that when N2 is undetermined, the expression puesta de sol – often in connection with quantification by cardinals – refers to the phenomenon ”sun set” independently of our own sun. If the definite article had introduced the N2-lexemes in the examples of (10a), they would have held a uniquely referential relation to the sun, and consequently the examples would have meant a certain amount of sets of the sun, our sun, behind the horizon. However, this is not the intended meaning as focus is on the phenomenon itself in an abstract and independent sense. In this connection it is necessary to take into account that the definite article is certainly not generally blocked in ”pluralised” expressions such as “Las fotografías son de las puestas de(l) sol en partes diferentes de Noruega.” (viz. the last example in (10a)), as long as N2 in such constructions can be taken to allude to our sun. The claim I have argued in favour of here is that there is a meaning difference between the two types of constructions which is substantiated by the fact that it would be impossible to insert the definite article in the second and the fourth example of (10a), and still achieve the intended meaning: ”En ese mundo había dos enormes soles, dos puestas de sol y dos amaneceres.” and Según explica Antoine de Saint-Exupéry (1979), un día el Pequeño Príncipe vio cuarenta y tres puestas de sol solo moviendo un poco la [...]”. Had N2 been introduced by the definite article in these cases, the lexeme sol would have referred uniquely which is incompatible with the meaning of the sentences.6 Apparently the meaning variation which exists between puesta de sol and puesta del sol does not represent a fully generalised phenomenon, and for example it is not relevant in connection with (10b, c, and d). In these cases the internal structure of the NPs does not react to quantification, because the alternation between a unique interpretation of N2 and its incorporation, i.e. merging of the two nominals, is not relevant. In other words, the N2-nouns in (10b, c and d) maintain their prototype status, which is dictated by the definite article and whose counterpart is the mass interpretation triggered by zero-determination. However, these considerations also imply that the definite article in connection with puesta del sol is never capable of encoding the prototype interpretation into N2; it 6. I have run into examples as the following: ”Para notar el movimiento de la Luna en su órbita, hay que tener en cuenta su ubicación en el momento de la puesta de Sol durante algunos días.” (To notice the movement of the moon in its orbit, you have to take into consideration its position at the moment of sun set for some days.), where N2 is undetermined and at the same time introduced by capital letter, which according to my assumptions should be ruled out. As examples of this type seem to be few, they will be disregarded.

 Henrik Høeg Müller

will always result in a referential reading. Consequently, the next problem is to account for the interpretative status of the N2-nouns in (10a). Here it would hardly be reasonable to claim that lack of determination leads to mass interpretation. So in certain cases zero-determination can both bring about a mass reading and something else which has not yet been identified in the discussion so far.

3.4

Singular individual reading7 (N1=nominalised action verb)

Certain action verbs and their nominalisations can incorporate first order entity denoting nouns in the singular without enforcing on them a mass interpretation. Usually the incorporation of singular count nouns is described as a process of recategorisation (Bosque 1999) or grinding (Jackendoff 1991) according to which the denotation of the count noun undergoes a massification. However, in some cases the incorporated singular noun apparently denotes a non-specific singular individual instead. It is not a very frequently observable fact as this interpretation possibility only arises as an option if correspondence can be established with a stereotyped real world situation in which an action is directed towards or involves only one individual. In the constellation puesta de sol, N1 consists of the lexeme puesta which is derived from the action verb poner, and the situation denoted by the whole expression puesta de sol can be characterised as extremely stereotyped and only involving one individual, namely the sun. The following examples illustrate the same point. (11) a. Cada vez se requiere un mayor sacrificio para la compra de piso. (It becomes more and more expensive to buy a flat [flat buying].) b. El alquiler de coche es muy aconsejable en sus estancias en las islas grandes, […] (Car-hire is very advisable when you stay on the large islands, […]) Car-hire and flat-buying are conventionalised situations in which human beings carry out actions aiming at achieving temporary or permanent control of an object, in these cases objects of the types denoted by the nouns piso and coche. This way of expression is clearly extremely restricted pragmatically. Native speakers of Spanish indicate quite unambiguously that constructed examples such as *compra de periódico (buying of newspaper), *compra de perro (buying of dog) or *detención de terrorista (detention of terrorist) seem strange although these situations in some way could qualify as being conventional or stereotyped. Instead of denoting masses, the N2-nouns in (10a) and (11) denote non-specific singular individuals of a certain type. When N1 is derived from an action verb, zero-determination of N2 can both result in a mass reading and an interpretation as a singular individual, but since the last 7. This reading of N2 should not be confused with the reading of N2 as ‘singular individual prototype’, although the two readings are clearly related in the sense that both imply a non-specific singular individual.



Determination of N2 modifiers in Spanish nominal syntagmatic compounds 

possibility is heavily constrained pragmatically, the general alternation pattern is between mass and prototype denotation. Finally, it should be emphasised that insertion of the definite article in constructions where N2 denotes a singular individual often results in the conversion of N2 to a uniquely referential element. In other words, normally alternation takes place between the reading as singular individual and the referential one. In concrete terms this means that the explanation why N2 in for instance puesta del sol establishes unique reference is not only the special status of the N2-lexeme sol as a culturally prominent phenomenon, but it is also due to the linguistic fact that, under distinctive pragmatic circumstances, some nouns derived from action verbs can select an undetermined, non-specific individual denoting noun in the singular as N2. However, in certain cases the prototype interpretation actually constitutes an alternation possibility, cf. (12). (12) La compra del coche – Consejos legales prácticos de los abogados de iAbogado. com (The buying of the car – Legal, practical advice from the lawyers of iAbogado. com) The semantic difference between compra de coche and compra del coche, i.e. between the non-specific individual reading and the prototype interpretation, is very hard to perceive, because both constructions imply the understanding of a non-specific singular individual. It could be suggested that the zero-determiner construction represents the “transaction” as a process, while the construction with determined N2 pictures the “transaction” as a singular event. However, it is not possible to substantiate this intuitional understanding, although it is in fairly good keeping with what is usually considered the semantic consequences of the syntactic alternation between incorporation and non-incorporation.

3.5

Summarising

The above observations and comments concerning the interpretational possibilities and alternation patterns of ’ØN2’ and ’det.N2’ can be synthesised in the following way. Zero-determination gives rise to two different meanings, namely the mass interpretation and the concept interpretation, while the definite article, apart from its referential use, can bring about the ”pure” prototype reading, and the prototype reading which implies the understanding of a single individual. These interpretations are connected with each other in the sense that the mass reading finds its counterpart in the “pure” prototype reading while the concept reading alternates with the reading of singular individual prototype.

 Henrik Høeg Müller

‘det.N2’ Prototype

Referential Specific

“Pure” prototype

Singular individual prototype

Figure 1.

Specific: Él también cruzó la calle y al pasar junto a ella arrojó la colilla del cigarro a sus pies. (He also crossed the street, and when he passed her, he threw the stub of the cigarette in front of her feet.) “Pure” prototype: Albert Einstein desarrolló dos teorías de la relatividad bastante diferentes. (Albert Einstein developed two quite different theories of relativity.) Singular individual prototype: ¿Qué hacemos con la colilla del cigarro cuando fumamos por la calle? (What do we do with the cigarette stub when we smoke in the street?) ‘ØN2’

Mass

Concept

Singular individual (N2=nominalised action verb)

Figure 2.

Mass: Acaba de explotar una refinería de petróleo en Nueva York. (An oil refinery in New York has just exploded.) Concept: Puedes llevar tantas como consigas, pero solo una de cada tipo. Por ejemplo: No puedes llevar dos hachas de bombero a la vez. (You can take as many as you can, but only one of each type. For example you cannot take two fireman’s axes at the same time.) Singular individual: Cada vez se requiere un mayor sacrificio para la compra de piso. (It becomes more and more expensive to buy a flat [buying of flat].)



Determination of N2 modifiers in Spanish nominal syntagmatic compounds 

The most important alternations between meanings encoded by ’ØN2’ and ’det.N2’ are the following: 1. Mass (ØN2) – “Pure” prototype (det.N2) 2. Concept (ØN2) – Singular individual prototype (det.N2) The referential interpretation of ’det.N2’ is always a possibility, provided of course that the context supports it. 3. Non-specific individual (ØN2) – Specific [referential] (det.N2) According to (12) the non-specific reading can also alternate with the prototype interpretation in which case it is difficult to spot a difference of meaning between the two constructions.

4. Prototypicality vs. genericity In continuation of the above discussion it would be natural to pose the question: why has the generic use of the definite article not formed part of the considerations presented so far, especially taking into account that the concept of genericity seems so straightforward as a possible way of explanation. I have rendered the meaning of ‘det. N2’ in terms of prototypicality. Genericity and prototypicality are related but also distinct categories. In the following I will substantiate the position that the definite article that introduces N2 in compositional structures is not a generic article, but an article that prompts a prototype reading of N2. As indicated in the introduction to this article, the prototype and the generic way of interpreting NPs are clearly related, but at the same time they differ from each other.

4.1

Constructions with predicative N1

Traditionally the term generic is used when an NP does not refer to specific entities, i.e. entities that can be singled out and identified separately, but to a category of objects or class of individuals (see e.g. Leonetti 1999: 870). It is well recognised that the term generic NP covers a wide range of different forms of expressions in different languages, and that the concept ”generic reference” is far from unambiguous. Leonetti (ibid.), who builds on a solid foundation of works about generic use of articles, distinguishes, in accordance with for example Gerstner and Krifka (1993: 968), between definite and indefinite generic reference. The typical expressions encoding these two kinds of reference are the definite and the indefinite article. According to Leonetti and others, the definite article is used to refer to a ”generic individual”, and consequently the definite article has the same invariable function whether it occurs in a generic or a non-generic context. Thus Leonetti supports a reductionist, monosemantic approach to the

 Henrik Høeg Müller

function of the definite article, according to which sentences with quite different semantic contents, cf. (13) (ibid.: 872–73), basically all express some sort of genericity. (13) a. b. c. d. e. f.

El mamut está extinguido. (The mammut is extinct.) El Rover 620 ha obtenido un notable éxito de ventas. (The Rover 620 has had substantial sales success.) El kiwi tiene un sabor ligeramente ácido. (The kiwi has a slightly acid flavour.) Los españoles entienden de vinos. (The Spaniards understand about wine.) El hombre alcanzó la Luna en 1969. (Man reached the moon in 1969.) Aquella mañana tuvimos nuestro primer encuentro con el gorila. (That morning we had our first encounter with the gorilla.)

However, Korzen (1998: 39ff. and 2000: 269ff.), on whose work on NP semantics in sentences this presentation is founded, convincingly argues in favour of the existence of a fundamental difference between the generic and the prototype interpretation of NPs. At the sentence level this distinction can be illustrated with the following Spanish examples: (14) a. El haya crece y supera a especies pioneras. (The beech grows and outdoes other pioneer species.) El tiburón ballena come plantón y peces pequeños los cuales son filtrados en su aparato digestivo al ingerirlos. (The whale shark eats plankton and small fish, which are filtered in its digestive system when they are swallowed.) El tigre es un gran carnívoro de la familia de los felinos (gatos). (The tiger is a large predator of the feline family (cats).) El tigre es audaz, valiente y temible. (The tiger is bold, daring and frightening.) Aparentemente el vino tiene un efecto protector, pero no sabemos si esta reducción de riesgos de cáncer […] (Apparently {the} wine has a protective effect, but we do not know whether this reduction of cancer risks […]) El elefante tiene una gran memoria. (The elephant has a great memory.) b. Abunda en esta zona el rabilargo. (The magpie is abundant in this zone.) La patata llegó a Francia demasiado tarde para impedir la Revolución Francesa.



Determination of N2 modifiers in Spanish nominal syntagmatic compounds 



(The potato came to France too late to prevent the French Revolution.) El 20 de julio de 1969 el hombre puso el pie en un cuerpo celeste distinto del suyo, demostró que podía superar trampas que le tendía el Universo. (On July the 20th 1969 {the} man put foot on a heavenly body different from his own, he showed that he could overcome the obstacles of the Universe.) Muchos estudiantes cogen el tren para volver a casa en Navidad y vacaciones de Pascua. (Many students take the train to go home for Christmas and Easter holidays.) […] cuando los hijos están enfermos al menor síntoma le dan medicamentos fuertes o llaman al médico, […] ([…] when the kids are ill at the slightest symptoms they give them strong medications or call the doctor.) Afirmar que el coche se inventó en 1886 es, por tanto, muy arbitrario. (To state that the car was invented in 1886 is then very arbitrary.)

According to Korzen (1998: 42), it is characteristic of the generic interpretation of NPs that the speaker formulates an assertion whose descriptive content is claimed to hold over all members of a category, even though some exceptions to the general rule can be accepted without causing the sentence to be regarded as a false description on that ground. Both Leonetti and Korzen emphasise that in this respect the generic sentences radically distinguish themselves from sentences containing the all-quantifier, as the allquantifier asserts that the statement in question applies to each member of the set quantified over. Leonetti illustrates the difference with the sentences: Los holandeses hablan muy bien inglés. (The Dutch speak very good English.) vs. Todos los holandeses hablan muy bien inglés. (All Dutch speak very good English.). The first statement will not be considered falsified because a few Dutchmen speak English poorly, while to be true the second statement requires that every single Dutchman speaks the language well. The all-quantifier asserts that a property holds for every member of a set, whereas the generic sentences only contain this information as an implication. Therefore the generic sentences are often interpreted pragmatically as ”usually it will be the case that …” However, an exception to the above observation is generic sentences where the predi­ cate as in El tigre es de la familia de los felinos expresses membership of a category, i.e. the so-called analytic sentences that are inherently true in virtue of the meanings of its elements (the predicate is contained in the subject). There are no tigers that do not belong to the feline family, but possibly some specimens of the race are not covered by the content of the sentence: El tigre es audaz, valiente y temible. Analytic sentences and sentences containing the all-quantifier share the same truth value conditions, for which reason quantifiers like todo (all) and cualquier (every) can be inserted in analytic sentences and gain scope over the subject argument without changing the truth value conditions.

 Henrik Høeg Müller

However, the sentences in (14b), which contain the prototype-denoting NPs, cannot be interpreted in the same way. In these cases, the speaker does not assert, neither implicitly nor explicitly, that the description of the predicate holds for all the elements of the set. It does not apply to all magpies that they are abundant in a certain zone, neither did all or almost all men put foot on the moon. The prototype interpretation does not entail that the predicate assertion holds for all individuals of the type, but what it involves is that the relevant NP functions as a representative of a category. From the description of the difference between generic and prototype readings it follows logically that what by the generic statement is claimed to be true of the whole extension of the NP also holds for subsets in the form of subtypes or single elements, but what is claimed to be true of the prototype NP is not necessarily true of any subtypes or single elements that can be associated denotatively with the prototype noun. As pointed out by Korzen (ibid.: 46), the generic statements entail that what is true of the hyperonym also holds for the hyponym, but not necessarily the other way round, whereas the prototype statements imply that what is true of the hyponym also applies to the hyperonym, but not necessarily vice versa. The generic statement: El elefante tiene una gran memoria. implies that both Indian and African elephants have good memories, while the generic statement: El elefante africano tiene grandes orejas. (The African elephant has large ears.) does not amount to saying that all types of elephants have large ears. Contrary to this, prototype statements such as Muchos estudiantes cogen el tren para volver a casa en Navidad y vacaciones de Pascua. (Many students take the train to go home for Christmas and Easter holidays.) and […] cuando los hijos están enfermos al menor síntoma le dan medicamentos fuertes o llaman al médico, […]([…] when the kids are ill, at the slightest symptoms they give them strong medications or call the doctor, […] do not necessarily imply that the students take the Talgo (Spanish train) or that the parents call a paediatrician. However, statements like Los estudiantes cogen el Talgo. and Los parientes llaman al pediatra. unambiguously imply that the students have taken a train and that the parents have called a doctor. In other words, the NP described by the predicate in generic statements functions extensionally, because in principle it includes all subtypes or single members of the denoted category, while prototype NPs function intensionally by making one single idealised element representative of a category. It is important to emphasise in this respect that this so-called representative member does not exist as a real individual, but it functions as an abstraction over the features considered typical for the members of the category in question. With the prototype statement we do not generalise over elements of a set, but we create an abstract, idealised ”model” which contains all the features typical for the category, and therefore this model can act as a representative of the whole category. This also explains the implicatures concerning hyponymi/hyperonymi-relations, or at least it casts light on them from another angle. The extension of a lexeme includes all entities of which the lexeme is a true description, and consequently all groups, subtypes and individuals covered by the general description. The intension of a lexeme includes all the lexical or



Determination of N2 modifiers in Spanish nominal syntagmatic compounds 

semantic features relevant to the definition of the category denoted by the lexeme. The more specific the definition of the set or category is, the more meaning components the lexeme must contain to describe it, and that is why a hyponym contains more features than its hyperonym. The prototype-denoting NP is as indicated an abstraction over the features which are characteristic of exactly the entity type in question, and as the denotation of a subtype requires more features, the prototype NP can never comprise hyponyms. If we apply Carlson’s (1977) well-known tripartition of predicate types on the complex of problems presented here, it is evident that the generic reading of the determined NP correlates with the use of individual level predicates, whereas the prototype interpretation is activated when the speaker employ stage level and kind level predicates. Individual level predicates express/describe the inherent, permanent features or properties, including category membership, that an individual possesses, and when a statement in a given context is taken to refer not to an individual but to a kind, the features in question are asserted of the whole kind. In such a case, the NP to which a property is ascribed always functions as subject in the sentence. Stage level predicates express accidental factors which apply to prototype denoting NPs in certain situations. They predicate something which holds for an individual, in this case a prototype, at a certain time or at a certain place as in: La patata llegó a Francia demasiado tarde para impedir la Revolución Francesa. (The potato came to France too late to prevent the French Revolution.) or they establish a stereotyped event where the prototype denoting NP functions as an object, cf. Muchos estudiantes cogen el tren para volver a casa en Navidad y vacaciones de Pascua. (Many students take the train to go home for Christmas and Easter holidays.) The relation between predicate types and interpretation of NPs is by no means surprising. A predicate that grounds an individual in an event or predicates something about this individual in a certain situation, will not be suited for describing all members of a kind for the simple reason that no real world situations involve a whole kind. It will always be particular members of a kind that participate in an event, and then of course this event can be more or less typical or conventional of the kind. Contrary to that, individual level predicates describe individuals’ habitual, sometimes permanent, features and therefore they can be upgraded to express general features common to the whole kind. What applies to an individual as a permanent and characteristic feature is likely also to hold for the rest of the kind. As noted by Korzen (2000: 303) among others, kind level predicates differ from the others in that invariably they force a kind-denoting interpretation upon the singular subject NP, i.e. they are not capable of conveying individual reference to the NP as the two other predicate types are. Kind level predicates stimulate a prototype reading of the NP, because, contrary to the individual level predicates, they do not predicate anything of each specimen of the kind but of the kind as such, understood as an abstraction. As mentioned earlier it is not the case in connection with the example: Abunda en esta zona el rabilargo. that each member of the category (the magpie family) is abundantly

 Henrik Høeg Müller

represented in an area, or in relation to Leonetti’s example: El mamut está extinguido. that every single specimen of the kind has become extinct. The kind level predicates in question mean that the type of entity that possesses the necessary features to be described as magpie or mammut is abundantly represented or extinct, respectively. The next step is to transfer these descriptions and observations to the nominal level. Is it by any means possible to make the same distinctions at the nominal level considering that what characterises the NP compared to the sentence is that, regardless of level of complexity, an NP can never assert a description? An NP does not comply with the condition of ascribing truth value. It seems more or less straight forward to assume that to the extent that the different predicate types allow nominalisation there will be an analogy between the interpretation possibilities at the sentence level and at the nominal level. This means that in relation to NPs like crecimiento de la haya (growth of the beech) and llegada de la patata (arrival of the potato), N2 must be interpreted generically and as a prototype, respectively, naturally disregarding the possibility of a referential reading.8 The predicates capable of ascribing properties to the subject referent are either state or activity verbs. The state verb ser (be) conveys the meaning of category membership of the subject referent either by assigning to the referent a role or a property, cf. El hombre es médico. (The man is {a} doctor) and El coche es rojo. (The car is red.). The basic semantics of the state verb tener (have) is to establish a locative relation between object and subject referents, which can often be interpreted as a permanent possessive relation (see for example Herslund & Baron 2001). Neither the verb ser nor tener, which are considered the two main property ascribing state verbs in Spanish, allow nominalisation. Not surprisingly, activity verbs, such as mentir (lie), caminar (walk), and trabajar (work), can also be seen as ascribing a property to the subject referent. This is due to the fact that a type of individuals which more or less habitually realise an activity is likely to be associated with that activity as a distinctive feature. The activity/feature contributes to the identification of the type of individual. In many cases the type of individual can be constituted in such a way that it is capable of carrying out the relevant activity, i.e. possibly there is a relation between the physical design or shape of the type of individual and the activity it carries out habitually. However, most activity verbs are inergative verbs, which do not nominalise. These verbs are themselves generally derived from nouns, and partly as a consequence of that they cannot incorporate the agentive subject argument. Summing up, this means that individual level 8. It should be noticed that crecer (grow) can both function as an inergative activity verb meaning ”grow larger”, either in circumference or in length, and as unaccusative state verb meaning ”be in a place”, cf. the contrast between El haya crece mucho cada año. (The beech grows a lot every year.) and El haya no crece en zonas demasiado calurosas. (The beech does not grow in warm areas.). In its unaccusative variant crecer can be compared to vivir (live). However, vivir selects an animated subject referent, while crecer selects a subject referent from the vegetable kingdom. It is only the inergative activity meaning that can be transferred to the nominal level.



Determination of N2 modifiers in Spanish nominal syntagmatic compounds 

predicates that at the sentence level can stimulate a generic reading of the subject referent only in very few cases permit transposition to the nominal level. For that reason the ascription of a generic reading to N2 in NPs with a deverbal head noun is not a very frequent phenomenon. Stage level predicates, which, contrary to the two former types, as mentioned, do not assign properties to the subject referent, are predominantly based on action verbs, such as matar (kill) and destruir (destroy). Action verbs are frequently subject to nomi­ nalisation and typically, both in their nominal and verbal form, they incorporate the Patient-argument, i.e. the direct object of a transitive predicate or the subject of an intransitive, unaccusative predicate. Action verbs combine a process and a state description (see Durst-Andersen 1992), and therefore they can generally be qualified as denoting transitory situations, i.e. events where the referent is involved in a process that results in a state. In general action verbs predicate something of the referents that holds in a certain situation, and consequently they are not suitable to prompt a generic reading of the arguments, because specific situations, as mentioned earlier, are not about types of individuals but involve concrete specimens of a kind. Kind level predicates are formed by unaccusative state verbs that permit nominalisation and incorporation of the subject argument both at the sentence level and at the nominal level, cf. Abundan rabilargos en esta zona. and La abundacia de rabilargos en esta zona. Based on the current evidence, we may conclude, then, that the generic interpretation of N2 is only triggered in very few cases, and conversely that ’det.N2singular’ – as an alternative to the referential reading – is generally interpreted as a prototype in NPs with deverbal head noun.

4.2

NSCs with non-predicative head noun

Some of the examples which have been used in this article to illustrate the meaning differences activated by ’ØN2’ and ’det.N2’ do not have a deverbal noun as head, cf. nivel del agua (level of water), barrera del sonido (sound barrier), teoría de la relatividad (relativity theory), pulpa del melón (melon pulp), etc. Although it is not possible in these cases to build up an argumentation favouring the prototype reading on the basis of N1’s verbal origin, it is still possible to substantiate the assumption of the prototype interpretation being the most plausible explanation. With our point of departure in random examples like pantalla del televisor (screen of the television) and hacha del bombero (axe of the fireman) in which the N2-nouns denote objects that are usually understood as individuals, we could turn the perspective around and ask the question: What argument could be put forward to support the claim that ’det.N2’ in these cases should be interpreted generically? The ascription of generic value to an NP – in the sense that genericity has been presented in this article – implies a reading according to which ’det.N2’ includes in its denotation all, or at least almost all, members of the category. In spite of the fact that NPs do not assert their

 Henrik Høeg Müller

descriptive content, and that the head nouns are non-predicative, it seems evident that pantalla del televisor and hacha del bombero do not carry the meaning ”screen R(elator)9 all the entities for which it holds that they fit the description of being televisions” or ”axe R(relator) all the entities for which it holds that they fit the description of being firemen” (Here I disregard the difference between generic expressions and expressions with the all-quantifier.). Rather they should be paraphrased into ”screen R(elator) imaginary entity for which it holds that it meets the requirements of being able to be considered a typical television” or ”axe R(relator) imaginary entity for which it holds that it meets the requirements of being able to be considered a typical fireman”. When the context dictates that N2 cannot be referential, N2 must denote an abstract entity, namely a prototype of how a television or a fireman normally looks. The reason why the expressions pantalla del televisor and hacha del bombero can be used without N2 being interpreted referentially is exactly that a screen is one of the constituting components of a prototypical television, and an axe is one of the tools a fireman prototypically uses in his job. As opposed to that, N2 in an example like la pantalla del coche (the screen of the car) would most likely be interpreted referentially and not prototypically, because screens do not form part of our prototypical conception of which elements constitute cars. Without doubt, it is even more difficult to substantiate with formal evidence that when N2 denotes entities that we normally conceive of as masses or abstract entities, it also will not be interpreted generically but as a prototype. On the other hand, it is not intuitively satisfactory to reconcile oneself with the idea that N2 in for instance crisis del petróleo and crisis del agua should denote kinds, with the implication of totality that is entailed by generic structures. It must be rejected that the semantics of the head nouns should contribute to the stimulation of a reading according to which N2 is conceived of as denoting all manifestations or types of oil or water. Rather crisis could be compared to a stage level predicate that places N2 in a situation/event which by analogy with the sentence level and the NSCs with deverbal head noun can be seen as an indication of prototypicality. By way of examples crisis del petróleo and crisis del agua in my opinion do not imply conceptions of ”commonness of all kinds of oil or water” but rather an understanding of oil and water as abstract phenomena. However, when N2 induces a mass interpretation it could be objected generally that the prototype meaning dimension is somewhat harder to identify. In the cases where N2 denotes an abstract entity from the beginning as in celebración del matrimonio or teoría de la relatividad the difference between mass and prototype interpretations neutralises. N2 appears with and without determiner at random: searches in Google show that both variations are richly represented (however, with an over representation of ’det.N2’-structure). It is not possible on the basis of different types of contexts systematically to detect a semantic difference between the two 9. The term ”relator” expresses neutrality towards the semantic relation holding between the two nouns.



Determination of N2 modifiers in Spanish nominal syntagmatic compounds 

types of constructions. This idea of neutralisation is also in good keeping with the well-substantiated fact that abstract nouns and nouns denoting masses act similarly with respect to incorporation.

5. Conclusion At the nominal level it is not possible to provide formal evidence in support of the hypothesis that in general ’det.N2’ should not be attributed a generic reading but rather an interpretation as prototype. Although the argumentation favouring the prototype approach to a certain extent rests on intuition and contains elements of speculation, I still believe that there are important indications that the relevant constructions are described most appropriately by taking point of departure in the view that the determiner stimulates an interpretation of N2 as a prototypical, imaginary representative of the type of individuals denoted by the noun. This approach also creates an understanding of why the structure [N1 prep. det.N2] under certain circumstances can be said to denote a unitary concept and thereby qualifies as a compound.

References Abbott, B. 2004. Definitieness and Indefiniteness. In The Handbook of Pragmatics, L.R. Horn & G. Ward (eds), 123–149. Oxford: Blackwell. Bosque, I. 1999. El nombre común. In Gramática descriptiva de la lengua española 1, I. Bosque & V. Demonte (eds). 3–75. Madrid: Espasa Calpe. Carlson, G. 1977. A unified analysis of the English bare plural.  Linguistics and Philosophy 1: 413–456. Durst-Andersen, P. 1992. Mental Grammar: Russian Aspect and Related Issues. Columbus OH: Slavica Publishers. Gerstner, C. & Krifka, M. 1993. Genericity. In Syntax: An International Handbook of Contemporary Research, J. Jacobs et al. (eds). 966–978. Berlin: De Gruyter. Givón, T. 1984. Syntax. A Functional-Typological Introduction. Vol. I. Amsterdam: John Benjamins. Herslund, M. (ed.). 1997b. Det franske sprog. Kapitel II. Konsistens. Preliminary version. Copenhagen Business School. Herslund, M. & Baron, I. 2001. Introduction: Dimensions of possession. In Dimensions of Possession, I. Baron, M. Herslund & F. Sørensen (eds). 1–25. Amsterdam: John Benjamins. Jackendoff, R. 1991. Parts and boundaries. Cognition 41: 9–45. Korzen, I. 1998. At tale om ting [Copenhagen Working Papers in LSP]. Copenhagen Business School. Korzen, I. 2000. Italiensk-dansk sprogbrug i komparativt perspektiv. Vol.II. Referencer og andre sproglige relationer. Copenhagen: Samfundslitteratur. Langacker, R.W. 1991. Foundations of Cognitive Grammar. Vol.II. Descriptive Application. Stanford CA: Stanford University Press.

 Henrik Høeg Müller Leonetti, M. 1999. El artículo. In Gramática descriptiva de la lengua española 1, I. Bosque & V. Demonte (eds). 787–890. Madrid: Espasa Calpe. Vangsnes, Ø.V. 2001. On noun phrase architecture, referentiality, and article systems. Studia Linguistica 55(3): 249–299.

Reference to individuals, person, and the variety of mapping parameters* Giuseppe Longobardi Università di Trieste

The paper addresses two related questions: whether the principle that nominal arguments must be projections of D, apparently holding in many Romance languages, holds in English and in other languages as well; and why such a principle should exist at all. The answer empirically suggested to the first question is that the principle does hold in English, contrary to Chierchia (1998) and supporting the N-movement approach to the nominal RomanceGermanic parametrization proposed by Longobardi (1996), but that certain other languages, specifically Japanese, are likely to exhibit nominal arguments without D, as expected in Chierchia’s (1998) framework. Following the restrictive approach to phrase structure proposed by Chomsky (1995, ch 4.), the second question will be addressed by identifying D with the Person head and by arguing that the latter feature is crucial to allow type-shifting from property- to individual-denotation. Under a minimalist theory of parameter formats, it will be argued that all the three possible polymorphic realizations of the feature Person admitted by such a theory are crosslinguistically instantiated, precisely by Japanese, English, and Italian.

1. Background assumptions In this paper I try to sketch some speculative lines of approach to a few questions left open in Longobardi (2005a), mostly centering around crosslinguistic variation in the mechanisms governing the syntax/semantics mapping of nominal arguments. In so doing, I will explore to what extent such mechanisms can be reconciled with two (sometimes conflicting) restrictive requirements: constraints on the general form of parameters (Longobardi 2005b) and the need to prune phrase structure by eliminating categories which are not indispensable for the computation nor adequately justifiable * I am indebted for valuable comments to audiences in Copenhagen, Los Angeles, Newcastle upon Tyne, Berlin, and Leipzig. Also thanks to Alex Klinge, Henrik Müller, Philippe Schlenker, and especially to Hajime Hoji for useful suggestions and discussion.

 Giuseppe Longobardi

as primitives of the two interface levels (in the spirit of Chomsky 1995, ch. 4). I will argue that such a line of inquiry, originally rooted in purely minimalist considerations of simplicity and conceptual necessity, may naturally lead us to also attain a higher degree of the classical descriptive and explanatory adequacy in one intricate area of the syntax/semantics interface and it is thereby empirically supported. In order to proceed, it will be necessary to recall and slightly refine some basic assumptions of the mapping theory worked out in Longobardi (2005a) and previous work.

1.1

A (mental) ontology and its syntax

As a background, I will essentially accept Carlson’s (1977) idea that among the entities presupposed by the human language faculty (more precisely, by its Conceptual-Intentional system C-I, in Chomsky’s 1995, 2005 terms) are two types of individuals, objects and kinds. Kinds will be understood as maximal sets of objects sharing some properties across all possible worlds, while objects will be regarded as primitives. Properties, the other relevant type of primitive entities (cf. Chierchia 1984), will be distinguished from individuals by the fact, among others, that unlike the latter they fail to be located with respect to speech act roles (therefore, distinctions like +speaker/ hearer will not apply to them). Next, I will use the general term denotation for any type of relation holding in an utterance between an expression (usually a phrase, e.g. an argument or a predicate: the man, a man, every man, no man, men, John, Man) and the entity or entities (individuals or properties, respectively) it stands for in the ontology of the C-I system. I will instead reserve the term reference (as opposed primarily to quantification) for a subcase of denotation, namely a one-to-one relation between an expression (usually, prima facie, a single word, e.g. John, Man, men (in one sense, at least)) and an entity (thus a constant relation, not the product of an operator-variable structure). In this paper, for simplicity, I will focus mainly on reference, understood in this meaning and saliently instantiated by determinerless proper or common nouns, and less so on quantificational (operatorvariable) types of nominal denotation, though the fundamental insights can be argued to apply to both categories (cf. Guardiano and Longobardi in prep.). A further rough intuition is that nouns, as part of their lexical meaning, can name individuals and, particularly, that so-called common nouns name kinds and so-called proper names of traditional grammar name objects.1 Let me interpret this as meaning that, minimally, nouns (including, in particular, superficially determinerless ones) merged in an appropriate structure can refer to an individual’s property of being named 1. Borer (2004) suggests that the difference may be just one of derivational history, not a lexical one. Although this happens to be largely true on the surface, it does not exhaust the properties distinguishing ‘proper names’. Cf. Longobardi (2005a, sect. 4) for a more fine-grained empirical analysis of such a distinction.



Reference to individuals, person, and the variety of mapping parameters 

in a certain way. This feature of nouns is especially apparent in predicative usages, of course (cf. ‘I am John’, ‘This is gold’…). Now, it seems that such items can be used in some languages also to refer to individuals (objects or kinds) they name (cf. ‘I met John’, ‘Gold is precious’): in this work I will precisely explore and qualify this conclusion, arguing that no ambiguity nor parametric variation is involved in the referential properties of nouns. I will anyway suppose that no other lexical categories (e.g. articles, quantifiers,…) can refer to individuals, except for pronouns, which may stand for nouns, and demonstratives, which can denote by ostension. Let me call all these items (nouns, pronouns, and demonstratives) ‘potentially individual-referring’. Against this background of assumptions, Longobardi (2005a), especially building on Longobardi (1994 and 2001), proposed, for the nominal arguments of Italian and several other Romance languages, a generalization like (1a) and complemented it with (1b), designed to explain why bare common nouns never undergo the same N-to-D movement as proper names: (1) a. Core Generalization: N-to-D chain/CHAIN iff reference to individuals (i.e. the two types of constant interpretation: that of proper names and that of common nouns when used as kind-denoting names) b. Economy principle: Avoid an (overt) chain/CHAIN except as a last resort

In the case of nouns with object reference, (1)a is supported by paradigms like (2), where (2a) exemplifies N-raising, (2b) lack of raising, and (2c) might instantiate a presumably expletive article: (2) a. b. c.

Roma antica t era una città potente Rome ancient was a powerful city *Antica Roma era una città potente Ancient Rome was a powerful city L’antica Roma era una città potente The ancient Rome was a powerful city

chain no chain/CHAIN CHAIN

as for kind reference, it is supported by examples like (3) in conjunction with the observation, derived from (1a), that common head nouns can never raise to the D position: (3) a. b.

Madame Curie ha scoperto il radio Madame Curie discovered the radium *Madame Curie ha scoperto radio Madame Curie discovered radium

CHAIN2

2. Again possibly with an expletive, in the sense proposed in Vergnaud and Zubizarreta (1992). However, even if the article in (3) did not instantiate an expletive but a particular type of definite operator, along the promising lines of Dayal (2003), the core content of generalization (1)a would remain unchanged: a nominal argument without a phonetically spelt out D cannot be referential cf. (2) and (3), in the intended sense.

 Giuseppe Longobardi

the bare noun in (3b) could only have an indefinite reading, i.e. as a variable, bound in this case by a default existential operator and ranging over amounts of ‘radium’, producing an interpretation obviously inappropriate for the object of the predicate discover, which in the pragmatically salient sense (first discovery of the new element) should denote the kind.

1.2

The type-shifting nature of the Core Generalization

The fundamental intuition behind (1a) is that nouns are never sufficient, by themselves, to refer to individuals, e.g. when used as arguments in actual utterances, and that this is equally true for both common and proper nouns. Even simple reference to individuals (in our technical, restricted, sense above), then, turns out to be like other forms of individual denotation: an essentially syntactic, computational, property of nominal arguments, relying on, but not being exhausted by, lexical properties of nouns; it is never satisfied through just a lexical head but requires at least a functional position (the head D, hence a full phrase DP), no less than usual quantificational structures with overt operators. Let me finally recall and stress that all the constraints and effects following from generalization (1a) (e.g. N-to-D of proper names, indefinite reading of bare nouns) necessarily hold of proper names and bare nouns whenever functioning as arguments, but are suspended, again for both categories of nouns, when they are used as non-arguments, e.g. predicates (cf. Longobardi 2005a sect. 5). This seems to mean that nouns, by themselves, are instead sufficient to refer to properties, minimally the property of being named that way.

1.3

A topological mapping theory (TMT)

In Longobardi (2005a), (1a) is then derived from a more general theory of the denotation of nominal arguments, based on the DP-internal position (hence the qualification ‘topological’) of the items involved, sketched below:

(4) Denotation Hypothesis: Individuals are denoted in D



(5) Licensing condition: Arguments denote individuals, as constants or variables

(6) Definitions: a. Constants have a fixed referential value, thus denote one and only one individual (kind or object) b. Variables are bound by (coindexed with) an operator and range over a set of values, thus denoting a set of individuals (kinds -for taxonomic readings- or objects) The conjunction of (4)-(5) already yields (7) (which is independently well supported by argument/non-argument asymmetries in Romance: cf. Longobardi 2005a) as a theorem:



Reference to individuals, person, and the variety of mapping parameters 



(7) A ‘nominal expression’ is an argument only if it is introduced by a category D (cf. Szabolcsi (1987), Stowell (1989, 1991) (N may denote properties, not individuals)

From the conjunction of (4)-(5) with (6) one deduces the theorem (8), among others, which is largely equivalent to the Core Generalization (1a):

(8) An argument is a constant if and only if D contains α, α a potentially individual-referring expression3 (otherwise: an argument is a variable)

A constant will denote whichever kind/object α names or may anyway refer to; otherwise, denotation will be secured by binding the variable either selectively through an actual operator in D (definite/indefinite descriptions, quantified phrases…) or unselectively (bare nouns, with empty D bound e.g. by Ex or Gen operators: Diesing 1992, Longobardi 2001).4 Finally, perhaps not too surprisingly, their empty D, acting as a (non-locally bound) variable, subjects Romance argument bare nouns to what can be termed an ECP-like distribution (in the sense of descriptively obeying Chomsky’s 1981 Empty Category Principle, like wh-traces or syntactic variables), as first pointed out by Contreras (1986)5: (9) a. b.

Gianni ha trovato uranio in questa miniera Gianni found uranium in this mine *Uranio è stato trovato in questa miniera da Gianni Uranium was found in this mine by Gianni

Notice that, on the whole, in a language like Italian the effects of (4), namely generalization (1a), are well detectable off phonetic representations, i.e. point to a so-called early, pre-SpellOut, application of (4) itself.

3. I.e. α can be an actual noun, able to refer to the individual it lexically names, a pronoun or a demonstrative (or equivalently an expletive article linked to the noun in a CHAIN in Chomsky's 1986 sense). 4. Such two subcases of the variable interpretation might fall together under Dobrovie Sorin’s (1994) DR, which could derivationally create an empty category in D (as suggested to me in various forms by B. Laca, S. Rothstein, D. Delfitto, H. Kamp p.c.) by moving out the operator. A technical alternative to be explored is that the empty category functioning as a variable, as well as the target of N-movement, is the Spec of D, not its head, perhaps more in line with Higginbotham’s (1985) notion of saturation. 5. This third property may be less detectable or invisible in languages which anyway escape the constraint producing classical subject-object asymmetries in case of movement.

 Giuseppe Longobardi

1.4

Parametric extensions

As noticed at least since Longobardi (1994), English systematically contrasts with Italian with respect to all the three properties mentioned above (raising of proper names, interpretation and distribution of bare nouns): (10) a. Ancient Rome was a powerful city b. *Rome ancient was a powerful city (11) Madame Curie discovered radium (12) Uranium was found in that mine In other words, the three following crosslinguistic generalizations on determinerless nominal arguments hold: (13) a. Romance proper names must raise to D b. English proper names cannot raise to D (14) English bare nouns may be kind-referential names (i.e. have the interpretation of expressions like ‘that type/species of object’) in all the environments where Italian bare nouns fail to achieve it (cf. Longobardi 2001, Delfitto 2002 for a detailed analysis of various contexts) (15) English bare nouns, unlike the Italian ones (Contreras 1986 and much subsequent work), have ECP-free distribution, like proper names (14) and (15) can be descriptively unified as (16) (cf. Longobardi 2001, Delfitto 2002): (16) a. Italian bare nouns: only quantificational expressions (syntactic variables, i.e. indefinites, like overt indefinites and unlike proper names) existentially or generically bound b. English bare nouns: always potentially ambiguous between a referential interpretation (constants, precisely kind-referential names, unlike overt indefinites and somewhat like proper names) and the quantificational interpretation above On such grounds, Longobardi (1994, 2001) proposed a second generalization of this domain: (17) The two differences (13) and (16) are typologically related In other words, no such a thing as (1a) as a whole seems to be visibly observed in English. If the syntactic unification of object- and kind-reference proposed with (1a) is correct, then (17) is exactly what we ought to expect. One must then formulate some parametrization of roughly the following form: (18) Parameter (informal): Italian: +(1a) (call it ‘strong reference’)/English: -(1a) (‘weak reference’)



Reference to individuals, person, and the variety of mapping parameters 

Guardiano and Longobardi (in prep.), expanding some suggestions in Longobardi (1996), propose that the distinction between these two language types apparently extends, mutatis mutandis, beyond Italian/English6: (19) a. English side: Norwegian, Icelandic, Welsh,...(weak reference languages) b. Italian side: Spanish, Rumanian, Greek, Bulgarian, Arabic (cf. Fassi Fehri 2005),…(strong reference languages).

2. Two questions Thus, in certain languages (strong reference ones), the derived position of nouns in referential arguments (denotational constants) seems to be the same (modulo an overt chain/CHAIN), namely D, as that of lexical articles in say, definite descriptions, for principled reasons. Now, two questions arise, among others: (20) a. How, exactly, do the other languages differ? b. Why is precisely the D position involved in the denotation of individuals (and not, say, of properties)?

2.1

Determinerless nouns crosslinguistically

Addressing the first question minimally implies raising issues like the following: (21) How can English proper names and English bare nouns escape the constraints holding on Italian bare nouns? Recall, to illustrate the problem, that English and Italian, but most likely Germanic and Romance more generally, seem to differ in the following, prima facie curious, way: (22)

Germanic

Romance

PNs

a.

[ø

N]

b.

[N

t]

BNs

c.

[ø

N]

d.

[ø

N]

Formally, it is (22a,c,d) that have roughly the same internal structure, i.e. a non raisedto-D noun (though so far only Romance has provided some evidence for ø being an actual empty category [D e], precisely through the pattern (23a) below); but with respect

6. Longobardi’s (1996) hypotheses were in turn built on Borer’s (1994) and Siloni’s (1994) insight that definiteness checking is crucially involved in the raising of Hebrew construct state nouns.

 Giuseppe Longobardi

to meaning and external distribution it is (22d) which is isolated and (22a,b,c) that pattern together: (23) a. (22d) is subject to obligatory variable interpretation and an ECP-like distribution b. (22a,b,c) are free from such constraints It has been argued (since Longobardi 1994) that Romance proper names (22b) indeed escape the constraints holding on bare common nouns (i.e. variable interpretation, ECP-like distribution) via N-raising, neutralizing the relevant empty category otherwise present in D. Since no such displacement is overtly visible in Germanic, we are faced with questions like (21).

2.2

Two alternatives

The basic issue behind (21) is whether (7) does indeed hold in English, as well as in Romance, the hypothesis argued for in Stowell (1989,1991) and then extended to other Germanic varieties in Crisma (1999). If this hypothesis is correct, then the TMT sketched in (4)–(5) above can be universal, but the ø of structures (22a and c) must then be a true empty D like that of (22d). Therefore something must prevent this superficially empty D from being subject to the Romance constraints. A viable solution was suggested in Longobardi (1994) and especially (1996), namely that N-raising can be only overt in Romance and only covert in Germanic (and that the relevant constraints apply after such covert movement), with the final consequence that both (22a and c) could end up having the same logical representation as (22b). This approach maintains exactly the same mapping theory for both language types and reduces the differences to a single one of a very well known format, overt/ covert movement dependency. Briefly, according to this theory, the LF of (10a) would look like (24), i.e. like the PF of (2a), and the LF of English bare nouns could analogously be as in (25): (24) [Rome [ancient [...t...]]] was a powerful city (25) a. [Dinosaurs [...t...]] have become extinct b. [Dogs [...t...]] are sitting on my lawn Thus, LF (25) could be understood à la Carlson (1977), i.e. as anyway kind-referential, with a prevailing generic or existential meaning according to the interpretation of the predicate7: in either case the ECP-like constraint on null Ds should be circumvented by

7. After all, an existentially flavored reading is a possibility admitted also for many Romance definites, which often properly translate English bare nouns: cf. ‘Le formiche erano dappertutto/hanno distrutto la mia viola da gamba…’ vs. ‘Ants were everywhere/destroyed my viola da gamba…’.



Reference to individuals, person, and the variety of mapping parameters 

means of the same strategy (N-to-D) used by Romance proper names.8 Of course, this suggests that the distributional constraint in question applies to LF representations and, at the same time, that Romance nouns must be unable to take advantage of covert movement to avoid its effects. Then, four questions arise in this framework: (26)

a. b. c. d.

Why can Germanic proper names raise covertly? Why can Germanic common nouns raise at all? Why can’t Germanic nouns raise overtly? Why can Romance nouns only raise overtly?

(26a) is answered by formulating parameter (18) roughly as (27): (27) Parameter: in Romance (4) applies also at PF, in Germanic only at LF (26b) is answered by refining the Economy condition (1b) into (28), along lines proposed by Reinhart (1995), Fox (2000): (28) Economy: overt chains are a Last Resort for convergence, covert chains are licensed if they affect the interpretation (hence may involve common nouns, thus shifting from property- to kind-reference) (26c) is answered by (27) in conjunction with Chomsky’s principle Procrastinate, (26d) by the conjunction of (27) and (28). As a consequence, the TMT and the lexical semantics of nouns remain basically invariant but the interpretation of D (hence the denotation of DP) is ‘frozen’ on the surface in Romance, not in Germanic, so that it is unambiguously readable off PF in Romance but only off LF in Germanic The other conceivable approach may seem especially appealing under the impetus of minimalist theories of phrase structure (e.g. Chomsky 1995) since it tries to reject Stowell’s hypothesis altogether and prune the syntactic tree by eliminating D from those nominal arguments where it is not obviously manifested by phonetic, semantic, or distributional properties (i.e. from Germanic object- and kind-referential arguments (22)). Under this approach, however, it is necessary to account for the differences between Romance and Germanic (manifested in (13) and (16)) by means of two more radically distinct mapping theories ((4) for Romance and something roughly like ‘Individuals are denoted sometimes by D, sometimes by N alone’ for Germanic). Nouns would be of a different semantic type in Romance (+predicates, -arguments) as opposed to Germanic (+predicates, +arguments), a parametric difference not immediately reducible to consolidated parameter schemata. A first proposal precisely along

8. It should also remain possible for an English bare common noun, in the same syntactic and lexical environments in which its Romance correspondents may survive, to optionally fail to covertly raise to D, being thus treated as an unselectively bound indefinite.

 Giuseppe Longobardi

these lines has been attempted in Chierchia (1998), though not yet addressing all the differences above. Apart from conceptual considerations, the issue in (21) is ultimately a (subtle) empirical one. The mentioned works of Stowell (1989, 1991: cf. subcases (29a, b)) and Crisma (1999: for subcases (29c, e, f, g)), among others (especially cf. Zamparelli 1995 for subcase (29d)), suggest that Modern English and the other Germanic varieties indicated below display several constructions subtly but almost unequivocally pointing to the necessity of a high functional category (presumably of distributional type D) with a wide range of argument nominals though not with some of their non-argument correspondents: (29) a. Presence of overt operator-like items (quantifiers, definite articles) b. Presence of overt count-checking items (the same as above, indefinite articles) c. Distribution of Saxon genitive constructions (German) d. Definite reading of Saxon genitives e. Definiteness suffixes cooccurring with adjectives (Mainland Scandinavian) f. Lexically sporadic N-raising (Old and Middle English, Swedish) g. Supposed expletive articles with proper names (Old English, German) In these respects, Germanic acts rather like Romance, then, seemingly supporting (7). However, not all of this evidence is equally relevant: most of these cases can be construed as strong arguments that D is required in Germanic by argument nominals where the marked positive value of the features definite or count must be checked (precisely in D, presumably; cf. Zamparelli 1995, Crisma 1997), i.e. singular or definite common nouns; but the crucial question is whether D is required for argumenthood itself in all cases (i.e. if (7) holds in Germanic as well), independently of the checking of feature values like +definite or +count. In fact, the last two cases of (29) have already been construed by Crisma (1999) as suggesting that D is anyway required for argument proper names, whose count and essentially definite interpretation is likely to be intrinsic. Unfortunately, such arguments cannot be reproduced in Modern English. Furthermore, no proof yet has been provided concerning bare mass and plural common nouns, where the negative values for definite, if relevant at all (presumably in the supposed indefinite, not kind-referential readings: Diesing 1992), and count might be assigned by default, and which have so far shown no asymmetries. In the second part of this article, however, I will present some observations which seem to speak in favor of (7) holding in Modern English as well. Let me, for now, just stress one point which becomes relevant if (7) holds in Germanic: if this turns out to be true, in fact, a category D will be required for argumenthood even in cases where it is neither manifested by overt phonological features relevant for the sensory-motor system S-M (again, in Chomsky’s 2005 terminology) nor motivated by substantive properties of the position, relevant to the C-I system (precisely with proper names, kind names, and perhaps some definite descriptions too, in



Reference to individuals, person, and the variety of mapping parameters 

Scandinavian languages where definiteness is marked by lower affixes9). Therefore, under a strict minimalist approach, e.g. in the spirit of Chomsky (1995, 378: ‘…the only functional categories are those with features that survive through the derivation and appear at the interfaces, where they are interpreted’) one may regard (30) as remaining an unresolved issue: (30) Why is the category D often manifested by nothing else than a syntactic operation (overt or covert Move), without apparently any content of its own either on the C-I or on the S-M level? I will now focus on question (20b); nonetheless, from this some further suggestions on certain aspects of both (30) and (20a), hence of issue (21) and generalization (7), will arise.

2.3

Arguments and D

Addressing question (20b) implies further deepening our understanding of SzabolcsiStowell’s generalization (7), based on the two concepts of ‘argument’ and of ‘D’. Such relation between the notion of argument and that of D, however, is a rather opaque and stipulative one. Notice that already deducing (7) from (4)-(5) allows one to eliminate the direct relation between argumenthood and D, resorting to the hypothesis that the categorial status of arguments derives from their denoting individuals, rather than properties. For, (5) seems to be independently needed as a definition of argument. But the whole move may have empirical, in addition to conceptual, support, since (4) appears anyway more accurate than (7). Crisma (1997, ch. 3) has shown that vocatives, being non-arguments, should behave like predicates in disallowing a relative clause built on a singular count head noun (also cf. Longobardi 2005a, sect. 5), but actually do not; this suggests that vocatives are likely to be full DPs in such examples, with a peculiar way to license an empty D. Crisma sketches such a licensing theory and supposes that vocatives are always introduced by a D and that this property does not characterize arguments as opposed to non-arguments, as hypothesized by Szabolcsi (1987 and subsequent work), but individual-denoting expressions versus expressions denoting properties. She proposes, therefore, that simple NPs can only denote properties, not individuals, and that arguments and vocatives are both individual-denoting. Crisma’s insights are essentially those built into the TMT above (cf. (4)-(5)). This approach already eliminates the term ‘argument’ from the original generalization (7). However, it still leaves open question (20b), i.e. the issue of why, now, such a relation should hold precisely between denotation of individuals and D, as postulated in (4). 9. Cf. Guardiano and Longobardi (in prep.). In the case of definite arguments the value of the count feature becomes irrelevant, according to Crisma (1997, 1999). Hence, even a +count interpretation of a definite description with a suffix lower than D would not be a sufficient reason by itself to motivate the presence of D.

 Giuseppe Longobardi

2.4

Pronouns

One notable exception to the parametrization in (18) concerns the other usually object-referential category, personal pronouns. Longobardi (1994) noticed that in that domain Italian and English cease to contrast, pronouns being obligatorily in D in both languages10: (31) Noi ricchi/*I ricchi noi stiamo trascurando certi problemi (32) We rich/*The rich/*Rich we are neglecting certain issues (33) I poveri imitano noi ricchi/*i ricchi noi (34) The poor imitate us rich/*the rich us/*rich us (35) Noi due/*I due noi…. (36) We two/*The two we/*Two we…. This property of pronouns is not a peculiarity of English, but is found in other languages as well. For example it was carefully argued to characterize even articleless Slavic languages such as Serbo-Croat (Progovac 1998) and Polish (Rutkowsky 2002) .11 They observe, among other things, that such languages pattern exactly like English with respect to Longobardi’s (1994) generalizations: proper names and common nouns share the same distribution, but pronouns occur higher. Why then must personal pronouns always surface in D in Italian and why must (and of course may) they do so in English as well? Which properties of pronouns can be responsible for this difference between proper names and pronouns? A tentative answer may lie in the fact that in both languages pronouns are overtly specified for the full range of φ-features, namely gender, number, person and Case. In several Romance and Germanic languages nouns can be specified for gender, number, or Case (e.g. in German), but in none do they bear person distinctions. Let me then suppose that the head D, in addition to being available as the checking position for interpretive properties like for instance Definiteness and Count (cf. the discussion in Crisma 1997, among others), intrinsically irrelevant in the case of object-referential expressions, is also the only one where the interpretable exponence of person morphology is admitted; we can hypothesize, indeed, that, crosslinguistically, the so-called D category minimally consists of the person feature: 10. Bernstein (this vol.) notices that, in addition to 1st and 2nd person plural pronouns, the construction with a pronoun followed by A or N is available for 3rd person plural them, in sub­ standard English, and then proceeds to suggesting an explanation for its absence with singular pronouns, based on the lack of proper person or number specifications. In Italian the conditions explaining this latter property are likely not to hold and correspondingly the construction, though more natural with 1st and 2nd person plural, is sometimes acceptable even in the singular: i.

Io professore non posso chiedere a lui studente di fare il mio lavoro gratis I professor cannot ask him student to do my job for free

11. Veneeta Dayal (p.c.) kindly informs me that the same apparently holds in Hindi.



Reference to individuals, person, and the variety of mapping parameters 

(37) D is the Person head (37) seems in essential agreement with Bernstein’s (this vol.) hypotheses and in partial agreement with Platzack (2004). It also naturally complements proposals like Picallo (1991), Ritter (1991), Bernstein (1993) supposing that Number and Gender heads occur in the functional projection of nouns. From this assumption, it is possible to immediately draw a further conclusion: English overtly determinerless nominals do, at least in some cases, involve a D position, after all. Of course, that a head D is present with pronouns in English does not imply that it needs to be present with all nominal arguments, as – we assume – in Italian. The latter is still an empirical question, solvable in principle on the grounds of evidence like that discussed directly below; but notice that the suggested hypothesis, that D actually coincides with (at least) the Person feature, undermines one reason of appeal for any attempt to prune the category in a minimalist spirit: its presumed lack of substantive content. The hypothesis itself may remain neutral with respect to whether pronouns may/ must be merged in the Person position or are moved to it.12 Reasons to suppose that they are merged essentially like nouns and then raised are provided by the correct prediction that, rather like Italian proper names, in at least some non-argument occurrences they seem to surface postadjectivally, as in the following exclamation: (38) Poveri noi! Poor us! the impossibility of determiners in this construction, even with proper and common nouns, suggests that the whole exclamation consists of a bare NP, and not of a verbless predicate+argument constituent: (39) a. b. c. d.

Povero cane! Poor dog! ?*Povero il cane! *Poor the dog! Povera Maria! Poor Maria! ?*Povera la Maria! *Poor the Mary

This pattern confirms that non-arguments are not necessarily DPs but arguments are, and can be interpreted as follows: Person, if and only if present, overtly attracts pronouns. Notice, finally, that if pronouns are indeed universally moved there, the overt nature of this movement in, say, languages like Polish and English, where proper names 12. Rutkowsky (2002) suggests some interesting evidence from agreement phenomena that movement could be involved even in a language where other nouns do not apparently surface in D, like Polish.

 Giuseppe Longobardi

instead do not overtly raise to D, can be imputed not to semantic reasons, as suggested in Progovac (1998), but precisely to the fact that the exponence of person on pronouns is itself an overt phenomenon, one which needs to feed phonetic interpretation. On the whole, however, the paradigm in question entails a much more far-reaching consequence: Italian and English completely share not only the pattern in (31)-(36), but also in (38)-(39). The obvious conclusion is that if this complex paradigm provides a further argument for the DP/NP asymmetry opposing arguments to (some) nonarguments in Italian, the same reasoning goes through for English: therefore, the contrast e.g. between (34) and (38) eventually provides a first direct empirical argument for Stowell’s hypothesis that D is required for argumenthood itself even in Modern English. In turn, this conclusion supports the need for an approach like Longobardi’s (1996) to the parametrization here in question.

2.5

Person and TMT

Now, under hypothesis (37), principle (4) of the TMT is naturally reinterpreted as: (40) Denotation Hypothesis revised: Individuals are denoted through the Person feature (40) appears superior to (4) because a natural interpretation of it answers question (20b): denotation of individuals (of which reference to individuals is a subcase) basically consists of associating lexical material, e.g. the individual-naming content of nouns, with person specification, i.e. grammatical person; hence the head Person is required to search the ontology for an individual to be denoted. The fact that it does not apply to property denotation is now no longer mysterious: we have assumed properties, by definition, to be inherently personless (i.e. neutral with respect to specifications like for instance +speaker or +participant in the speech act), on the basis of the intuition that only individuals can occur as speakers/hearers of utterances, so that the roles specified by Person are intrinsically irrelevant for properties. This assumption seems also supported by some empirical observations: if postcopular adjectival and (pro)nominal predicates can be prototypical expressions of properties, then the well known fact that in various languages they can only be resumed by morphologically 3rd person masculine or neutral singular clitic pronouns (Italian lo, French le, Catalan ho: cf. Bartra 1986, 1988) suggests that they are deprived not only of interpretable number or gender features (cf. (41a)), but also of the person feature ((41b)), sharply contrasting with individual-denoting expressions (arguments, as in (41c)): (41) a. Gianni è monarchico e anticlericale; se anche Maria lo/*la/*li/*le fosse… Gianni is monarchical and anticlerical; if also Maria CL(3SgM)/CL(other forms)-were… ‘…; if Maria were too…’



Reference to individuals, person, and the variety of mapping parameters 

b. c.

Se Gianni fosse te o anche solo se Maria lo/*ti fosse…13 If Gianni were you or even just if Maria CL(3SgM)/CL(2Sg)-were… ‘…; or even simply if Maria were…’ Se Maria lo/ti vedesse… If Maria himCL(3SgM)/CL(2Sg)-saw… ‘If Maria saw him/you…’

3rd person masculine singular appears to be the unmarked uninterpretable form to spell out clitic pronouns, rather like it is the verbal agreement form for sentential (even coordinated) or impersonal subjects. Individual-denoting expressions, even though in non-pronominal cases are often assigned a default 3rd person value, are not universally bound to this assignment: in some languages, e.g. the Ibero-Romance ones, definite descriptions and other quantified subjects may impose 1st or 2nd person agreement on the verb: (42) Las mujeres salimos con vosotros The women go-out (1pl) with you (2pl) ‘We women go out with you people’ More generally, the default 3rd person of most definite and indefinite descriptions, which can be either singular or plural, must not be confused with truly impersonal 3rd singular of the type in question. Thus, even empirically, properties seem to be featureless (though forced to be encoded by means of the intrinsic features of the individual bearing the relevant property when expressed by a non-clitic pronoun), individuals are not. Then, under these assumptions, the distinction between the syntax of individual denotation, crucially involving D, and that of property denotation, often satisfiable by N alone, is automatically drawn in a principled way: nouns name individuals, but, alone, can only refer to (sets of) properties, minimally the property of being named that way; it is the association with a Person category which turns such entities into individuals (recall, e.g., that no N-to-D movement is required for N to function as a predicate: cf. Longobardi 1994, 2005a). This empirical hypothesis connects two apparently unrelated phenomena, both singling out at the same time the category D and the concept of grammatical person: the morphophonological property of D concerning pronouns (the exponence of person) and the semantic property of D concerning predicates (not being required for denoting entities deprived of person, i.e. properties). Of course, expressing person is different from being just associated to it by movement: hence pronouns in English occur overtly in the Person head, while proper names do not. But two apparently disparate phenomena are unified. Attraction of proper names to D (now Person) is an interpretive phenomenon, relevant for C-I, not to be confused with the morphological nature of Person’s attraction 13. The sentence with lo sounds quite acceptable at least in the interpretation in which ‘Maria’ should share with ‘you’ all ‘your’ properties though not (counterfactually) her individual identity.

 Giuseppe Longobardi

of pronouns, necessarily relevant at the S-M level. Therefore, for names, it is licit to expect potential parametrization on the S-M level. We can now understand the intuitive content of parameter (27) as (43): (43) Parameter: certain languages refer to individuals (Romance,…) by overtly associating the lexical content of nouns to Person (strong Person), others (English,…) do not (weak Person) Such a parameter must be a very deep and consequential one in the structure of DPs: only in the Romance-Germanic domain, it is likely to be responsible for at least the three original types of contrasts pointed out in Longobardi (1994, refined 2001), i.e. (13), (14), and (15) above, and the two additional ones introduced in Longobardi (1996) (Genitive licensing and headless possessives), and the Rumanian-Scandinavian contrasts in the positioning of definiteness suffixes (Guardiano and Longobardi in prep.). Furthermore, it may account for the same contrasts and for various further intricate types of contrasts between Greek, Semitic, Bulgarian on one side and several Germanic varieties on the other (cf. Longobardi 2001, Guardiano 2003, and especially Guardiano and Longobardi in prep.).

2.6

The person feature

The conclusions arrived at induce one to look for the source of all this deep DP-internal parametrization (27), now conventionally labeled weak/strong Person distinction, in some more general (i.e. potentially manifested also outside the nominal domain) difference in the status of the person feature. This issue is too complex to be addressed here, but some exemplification of the possible perspectives will be provided: a possible place of manifestation of contrasts in the person feature is definitely person agreement. Now, all European languages so far known to display the possibility of non-3rd person agreement for non-pronominal DPs, as exemplified by Spanish in (42), appear to fall among those which can be considered strong Person languages in the sense above according to their nominal properties (cf. Guardiano and Longobardi in prep.): Spanish, Catalan, Greek. This may suggest that an implication exists between the parametric status of D as Person in nominals and its ability to control full-range (i.e. not necessarily 3rd person) agreement: namely, the latter property would be an option only among strong Person languages. On the other side, it is perhaps not irrelevant that only among weak Person languages are varieties found with no apparent exponence of the person feature on verbal agreement, such as Mainland Scandinavian or, according to Kayne (2000), even Modern English: in these languages, in fact, it is the pronominal system and especially the obligatory antecedent-anaphor agreement in person (I…myself, He…himself, etc.) that show clear morphosyntactic effects of such a feature. It is conceivable that other potential direct or indirect manifestations of the strong/ weak Person parameter like these might be discovered outside the nominal domain.



Reference to individuals, person, and the variety of mapping parameters 

For example, languages with strong Person in the nominal system might display traces of this status of the feature in the verbal system as well: most or perhaps all the languages with supposedly strong Person, if they have verbal tenses licensing null subjects, treat the latter as fully personal empty pronouns, while, among weak Person languages, there definitely exist some whose null subjects are only impersonal (German, Icelandic). Whether these scattered remarks are typologically significant I will leave to further inquiry.

3. Further parametrization? If (7), wherever clearly observed, can be conceptually motivated along the lines of the previous sections, i.e. ultimately derives as a theorem from (40), it will remain to be seen whether it is equally at work in languages where person is not a formal feature of grammar triggering syntactic operations. From now on, I will pursue some highly preliminary and speculative remarks on this further aspect of question (20a).

3.1

Grammaticalization of person

Though I agree with Benveniste (1971) that one cannot imagine a natural language where the meaning of person (i.e. the role of individuals talked about with respect to the speech act) is really ineffable, it is the case that some languages have been argued to be deprived of syntactic effects of φ-features altogether, including person (e.g. person agreement on predicates or anaphors), a case in point being e.g. Japanese (cf. Kuroda 1992 among others). Furthermore, it has also been proposed (cf. Siewierska 2004, and examples and references cited) that Japanese so-called pronouns would not be full equivalents of personal pronouns in many European languages, but essentially nouns (with some meaning suggesting reference to speech act roles, presumably), because they (44)

a. b. c. d.

do not form a limited class display the same number morphology as normal nouns tolerate most of the same modifiers (e.g. including demonstratives) cannot function as bound variables even in the non-speaker, non-hearer form kare.

Whether some or all such properties of Japanese are related is unclear but plausible. Under the approach of section 2. we should expect such languages to survive without syntactic effects of anything like (40), i.e. we might minimally expect the following implication:

 Giuseppe Longobardi

(45) If a language has no grammaticalized φ-features, (ceteris paribus with respect to European languages) it will have no head Person (=D) in its syntactic representation of nominal arguments Hence, (40), if indeed a semantic universal, will have to be satisfied by free association of the lexical individual-naming meaning of a noun with the semantic concept of person necessarily present in the C-I system. A few predictions ensue for such a language type: (46) a. proper names will have the same surface distribution as common nouns b. bare nouns will be able to achieve kind-referential interpretation c. expressions translating Indoeuropean pronouns will have the same distribution as nouns and, more generally, also (47) a. no person agreement will ever appear on verbs b. no person agreement will ever appear on anaphors Recall, next, that Bernstein (this vol.) argues that definite determiners in English (and perhaps, by parity of reasoning, other languages as well) express a person feature, hence under hypothesis (37) above they should surface in D (=Person) for principled reasons. Furthermore, Guardiano and Longobardi (in prep.) argue that various sorts of operators in different European languages are attracted to D overtly or covertly according to the same parameter governing overt/covert raising of individual-referring nouns, i.e. (27). Therefore, if these hypotheses are correct, a further consequence will be predicted for languages with no Person: (48) no obligatory collocation of any operator (such as quantifiers and definiteness markers, if any14) in the equivalent of a fixed D (i.e. Person) position within nominal arguments15 At least properties (46c), (47b), and (48) should superficially distinguish such a language from English (and all of them should distinguish it from Italian, of course). Now, all such predictions are fulfilled in Japanese: (46a, b) are observed, though perhaps irrelevantly, owing to the fact that Japanese displays no overt equivalent of articles at all: for this suggests that the possibility of a generalized null expletive determiner occupying D (a situation analogous to that presumably found in Latin, Russian, Polish: cf. Crisma 1997 and Gianollo 2005) as an alternative explanation cannot be entirely dismissed. However, (46c) and (47a, b) are also respected: Japanese ‘personal pronouns’ distribute in the nominal phrase exactly like nouns and no person agreement phenomena 14. If Bernstein’s association of person with definiteness holds crosslinguistically and can be shown to be principled, it can even be the case that languages not grammaticalizing person systematically fail to grammaticalize definiteness as well. 15. Such items could then occur either in lower NP-internal or in NP-external positions.



Reference to individuals, person, and the variety of mapping parameters 

occur anywhere in the language (Kuroda 1988). Finally, prediction (48) is exactly the conclusion independently arrived at by Hoji and Ishii (2004): all sorts of operators may appear in floating positions outside the nominal phrase. The success of the predictions above lead to a more full-fledged formulation of the parametrization in (43): (49) Languages with grammaticalized φ-features may associate referential and operator expressions with Person overtly (strong Person: Romance) or covertly (weak Person: Germanic): (40) applies in narrow syntax (50) Languages without grammaticalized φ-features (e.g. Japanese) may associate person to the relevant expressions freely (or at most under pragmatic constraints): (40) only applies in conceptual or pragmatic representations The main consequence of these speculations is that languages without grammaticalized φ-features may indeed have bare (i.e. without D) NPs in argument position, precisely as argued for Japanese by Hoji and Ishii (2004). Such entities would thus exist among the phenotypes admitted by UG, as suggested in Chierchia (1998), though not in Germanic languages, contrary to what had been proposed in Chierchia (1998). If the generalized parametrization is tripartite as suggested here, the differences between English and Japanese with respect to (46)-(47)-(48) provide further support to the hypothesis that even English (and all of Germanic: Crisma 1999) has D, now Person, in every argument nominal, as proposed by Stowell (1989, 1991) and that the contrasts with Romance as to the properties of bare nouns and proper names are due to the overt/covert nature of the N-to-D dependency (Longobardi 1994, 1996), not to optional lack of D. Tentatively trying to add a few more languages to the picture just from superficial evidence (cf. Guardiano and Longobardi in prep.), the tripartite parametrization pattern might be supposed to be as follows: (51) Generalized nominal mapping parameter (in Chierchia’s 1998 perspicuous terminology) Grammaticalized person –

+ Strong Person

Japanese –

Germanic Celtic?

+ Romance Greek Bulgarian Arabic

 Giuseppe Longobardi

3.2

A minimalist critique: features and categories

The conclusions tentatively reached in this article have several consequences for the theory of parameters and that of UG. First of all, as noted, they support the idea proposed by some scholars that also Person, perhaps on a par with Number and Gender, occurs as a distinct head in the functional projection of nouns. More importantly, notice that all such heads have immediately understandable semantic content, while at the same time features and heads with dubious or indirect interpretation in the C-I system (and often no obvious exponence in the S-M one) such as +referential, +argument, D, widely used in the literature in the past (cf. e.g. Longobardi 1994, Chierchia 1998 among many others), are eliminated altogether: at best, such notions are a byproduct of the computation, not substantive features of the lexicon and their elimination arises exactly in the same minimalist spirit as that of the heads Agr (vs. T) in Chomsky (1995) in the system of verbal functional projections. In general, such an approach executes the program of category pruning hinted at in Chomsky (1995, 378). Furthermore, the conclusion that the locus of all parametrization of nominal mapping is D, i.e. Person, rather than N, reconciles this domain of facts with Borer’s (1984) restrictive conjecture that parameters are properties of functional heads, not of lexical ones like nouns, whose semantic type could thus be regarded as universally invariant and just property-denoting.

3.3

Parametric Minimalism

However, the most capital insight suggested by the present approach has to do with the theory of parameters itself and the way of investigating it: it turns out from the present approach that mapping parameters not only respect Borer’s (1984) conjecture, but also seem to fall into just two pervasive parameter schemata: (51) a. An interpretable feature may be: grammaticalized/extragrammatical (essentially Kuroda’s 1988 schema: for a DP-internal case study cf. the treatment of +definite, +count in Crisma 1997, 1999) b. If grammaticalized, a feature may trigger: overt/covert Move (essentially Huang’s 1982 schema) The very fact that the whole parametrization required within this framework to account for the variation examined reduces to independently attested parameter formats conceptually supports the approach. Actually, trying to subject a wider corpus of parameters to a minimalist critique in this sense, i.e. to reduce them to a restricted number of conceptually necessary forms, as envisaged in Longobardi (2003), they may turn out to fall into few abstract parameter schemata: for a very preliminary example, notice that over 40 of the 49 parameters tentatively collected in Gianollo, Guardiano, Longobardi (2004) for a comparative analysis of DP structures fall into only 4 recurrent



Reference to individuals, person, and the variety of mapping parameters 

parameter schemata, two of which are those identified in (52). If this kind of approach reveals productive, then minimalist considerations can begin to apply not only to principles but also to parameters of UG. In this spirit, Longobardi (2005b) has suggested that the crucial notion in acquisition may precisely become ‘parameter schema’, not ‘parameter’. The initial state of the language faculty S0 would not be endowed, at least for the core grammar, with a finite though enormous list of parameters, but with a much more limited number of parameter schemata. According to the evidence contained in primary corpora, only the parametric choices required by each target language, in terms of specific observable properties of the (functional) items of its lexicon, would be elaborated and, with a value set, would become part of successive states, including the steady state Sn. An approach of this type, well supported in a domain of robust variation like that of nominal mapping, might be a first step toward simplifying and, hopefully, better understanding the fundamental evolutionary problem of why grammatical variation exists at all in natural languages.16

References Baker, M. 2001. The Atoms of Language. New York NY: Basic Books. Bartra, A. 1986. Ho, predicat i argument. Els Marges 35: 77–85. Bartra, A. 1988. Reflexions sobre els pronoms febles predicatius. In Actes del vuitè Col.loqui internacional de llengua i literatura catalanes, 189–200. Publicacions de l’Abadia de Montserrat. Benveniste, E. 1971. La structure des relations de personne dans le verbe. Problèmes de linguistique générale. Bernstein, J. 1993. The syntactic role of word markers in null nominal constructions. Probus 5: 5–38. Borer, H. 1984. Parametric Syntax. Dordrecht: Foris. Borer, H. 1994. Deconstructing the Construct. Ms, University of Massachusetts at Amherst. Borer, H. 2004. Structuring Sense Vol. I: In Name Only. Oxford: OUP. Carlson, G.N. 1977. A unified analysis of the English bare plural. Linguistics and Philosophy 1: 413–456. Chierchia, G. 1984. Topics in the Syntax and Semantics of Infinitives and Gerunds. PhD dissertation, University of Massachusetts at Amherst. Chierchia, G. 1998. Reference to kinds across languages. Natural Language Semantics 6: 339–405. Chomsky, N. 1981. Lectures on Government and Binding. Dordrecht: Foris. Chomsky, N. 1986. Knowledge of Language. New York NY: Praeger. Chomsky, N. 1995. The Minimalist Program. Cambridge MA: The MIT Press. Chomsky, N. 2005. On Phases, Ms, MIT. Contreras, H. 1986 Spanish bare NPs and the ECP. In Generative Studies in Spanish Syntax, I. Bordelois, H. Contreras & K. Zagona (eds) 25–49. Dordrecht: Foris.

16. Cf. Baker (2001).

 Giuseppe Longobardi Crisma, P. 1997. L’articolo nella prosa inglese antica e la teoria degli articoli nulli. PhD dissertation, Università di Padova. Crisma, P. 1999. Nominals without the article in Germanic languages. Rivista di Grammatica Generativa 24:105–125. Dayal, V. 2003. Genericity and (In)definiteness: A Cross-Linguistic Study, Ms, Rutgers University. Delfitto, D. 2002. Genericity in Language. Alessandria: Edizioni dell’Orso. Diesing, M. 1992. Indefinites. Cambridge MA: The MIT Press. Dobrovie-Sorin, C. 1994. The syntax of Romanian: Comparative Studies in Romance. Berlin: Mouton de Gruyter. Fassi Fehri, A. 2005. Nominal classes and parameters across interfaces and levels, with particular reference to Arabic, ms. Université Mohammed V. In press in Linguistic Variation Yearbook. Fox, D. 2000. Economy and Semantic Interpretation. Cambridge MA: The MIT Press. Gianollo, C. 2005. Constituent Structure and Parametric Resetting in the Latin DP: A Diachronic Study. PhD dissertation, Università di Pisa. Gianollo, C., Guardiano, C. & Longobardi, G. 2004. Historical implications of a formal theory of syntactic variation, Ms, Università di Trieste-Pisa-Modena & Reggio Emilia. In press in S. Anderson & D. Jonas (eds), Proceedings of DIGS VIII. Guardiano, C. 2003. Struttura e storia del Sintagma Nominale nel Greco antico: Ipotesi parametriche. PhD dissertation, Università di Pisa. Guardiano, C. & Longobardi, G. In preparation. The underlying unity of reference and quantification, Ms. Università di Trieste-Modena & Reggio Emilia. Paper presented at the XXXII Incontro di Grammatica Generativa, Florence 2006. Higginbotham, J. 1985. LF, binding, and nominals. Linguistic Inquiry 16. Hoji, H. & Yasuo Ishii. 2004. What gets mapped to the tripartite structure of quantification in Japanese. In WCCFL 23 Proceedings, B. Schmeiser, V. Chand, A. Kelleher & A. Rodriguez (eds). Somerville MA: Cascadilla. Huang C-T. J. 1982. Logical Relations in Chinese and the Theory of Grammar. PhD dissertation, MIT. Kayne, R.S. 2000. Parameters and Universals. New York NY: OUP. (Chapter 8: Person morphemes and reflexives). Kuroda, S.-Y. 1988. Whether we agree or not: A comparative syntax of English and Japanese, Linguisticae Investigationes 12: 1–47. Longobardi, G. 1994. Reference and proper names: A theory of N-movement in syntax and logical form. Linguistic Inquiry 25(4): 609–665. Longobardi, G. 1996. The Syntax of N-raising: A Minimalist Theory [OTS Working Papers]. Utrecht: Research Institute for Language and Speech, University of Utrecht. Longobardi, G. 2001. How comparative is semantics? A unified parametric theory of bare nouns and proper names. Natural Language Semantics 9: 335–369. Longobardi, G. 2003. Methods in parametric linguistics and cognitive history. Linguistic Variation Yearbook 3: 101–138. Longobardi, G. 2005a. Toward a unified grammar of reference. Zeitschrift für Sprachwissenschaft 24: 5–44. Longobardi, G. 2005b. A minimalist program for parametric linguistics? In Organizing Grammar: Linguistic Studies for Henk van Riemsdijk, H. Broekhuis, N. Corver, M. Huybregts, U. Kleinhenz & J. Koster (eds.), 407–414. Berlin: Mouton de Gruyter.



Reference to individuals, person, and the variety of mapping parameters  Picallo, C. 1991. Nominals and nominalization in Catalan. Probus 3(3): 279–316. Platzack, C. 2004. Agreement and the person phrase hypothesis. Working papers in Scandinavian Syntax 73: 83–112. Lund: Department of Linguistics. Progovac, L. 1998. Determiner phrase in a language without determiners. Journal of Linguistics 34: 164–179. Reinhart, T. 1995. Interface Strategies [OTS Working Papers]. Utrecht: Research Institute for Language and Speech, University of Utrecht. Ritter, E. 1991. Two functional categories in noun phrases: Evidence from Modern Hebrew. In Perspectives on Phrase Structure: Heads and Licensing [Syntax and Semantics 25], Susan Rothstein (ed.) 37–62., San Diego CA: Academic Press. Rutkowski, P. 2002. Noun/pronoun asymmetries: Evidence in support of the DP hypothesis in Polish. Jezikoslovlje 3:1–2: 159–170. Siewerska, A. 2004. Person. Cambridge: CUP. Siloni, T. 1994. Noun Phrases and Nominalizations. PhD dissertation, University of Geneva. Stowell, T. 1989. Subjects, specifiers, and X-bar theory. In Alternative Conceptions of Phrase Structure, M. Baltin & A. Kroch (eds), 232–262. Chicago IL: University of Chicago Press. Stowell, T. 1991. Determiners in NP and DP. In Views on Phrase Structure, K. Leffel & D. Bouchard (eds), 37–56. Dordrecht: Kluwer. Szabolcsi, A. 1987. Functional categories in the noun phrase. In Approaches to Hungarian, Vol. 2, I. Kenesei (ed.), 167–189. Szeged: Jate. Vergnaud, J.-R. & Zubizarreta, M.L. 1992. The definite determiner and the inalienable constructions in French and English. Linguistic Inquiry 23: 595–652. Zamparelli, R. 1995. Layers in the Determiner Phrase. PhD dissertation, University of Rochester, NY.

English th- Forms Judy B. Bernstein

William Paterson University This paper develops the idea that English words like the, they, this, and existential there share an initial th- morpheme, which is identified as a 3rd person marker unspecified for number and gender. Also developed is the proposal that person is a property of D (head of the functional projection “Determiner Phrase”). Not adopted is the idea that definiteness or deixis is inherently encoded in D, although the proposal is compatible with an approach that takes these features to be derivationally associated with D. The proposed analysis therefore departs from that of Lyons (1999), who argues that the features person and definiteness are conflated and simultaneously associated with D.

1. Introduction* This paper focuses on the similarities of English th-initial forms such as pronouns, demonstratives and definite articles. Greenberg (1978: 75), for example, has noted that “in many languages the third person pronoun is identical with a demonstrative, and often an article is identical with one or the other.” In (1), I illustrate with English examples the pattern Greenberg observed.1 (1) a. pronoun: b. demonstrative: c. definite article:

them them the

I ate them. I ate them apples. (vernacular) I ate the apples.

* Thanks to the organizers and participants of the Copenhagen Symposium on Determination for their feedback on an earlier version of this work. Thanks especially to Paola Crisma, Alex Klinge, Giuseppe Longobardi, Henrik Müller, and Jan Rijkoff for their comments and suggestions. All errors and shortcomings are my own. 1. More recently, Siewierska (2004: 249-251), based in part on Bhat (2004), discusses the connection between 3rd person forms and demonstratives crosslinguistically. Although she does not include English, I believe the connection is indeed there, though mediated through th- and not the demonstrative per se.

 Judy B. Bernstein

The accusative pronoun them in (1a) is isomorphic with the demonstrative form (1b) found in many varieties of vernacular English. Although the definite article the in (1c) is not identical to the forms in (1a,b), the correspondence is apparent. Referring specifically to th- forms, Greenberg (1978: 77) further noted that “we will have at best suspicious similarities among some forms as in English ‘this,’ ‘that,’ ‘they,’ ‘them’…” I will claim that these similarities are in fact systematic, and furthermore that these th- forms share syntactic properties. Following Kayne (2000), Déchaine and Wiltschko (2002), Harley and Ritter (2002), Bernstein and Tortora (2005) and others, I take pronominal (including pronominal th-) forms to be morphologically complex, and th- to be an independent morpheme. I will extend this idea to the other relevant th- forms (see also Klinge this volume and Bernstein, Cowart, and McDaniel 1999), which (with the exception of the) I also assume to be at least bimorphemic. The fact that th- forms tend to be definite might suggest that the th- morpheme is an (abstract) definiteness marker. Déchaine and Wiltschko (2002: 422), for example, treat decomposed th- as a D element, which is also definite. Similarly, Bloomfield (1933: 252) treats non-speaker/hearer “substitutes” (i.e., pronouns) as definite forms. More recently Klinge (this volume) argues that English (and more generally Germanic) th- is involved in deixis.2 In this paper I claim that th- is a 3rd person (agreement) marker, and more generally that person is associated with D (see Longobardi this volume).3 The similarities across th- forms in English have been obscured by the disparate grammatical categories these forms have traditionally been assigned to (on this point, see Klinge this volume). The analysis is not inconsistent with previous ideas (see, in particular, Longobardi 1994) about D’s association with definiteness. In particular, I will adopt the idea that definiteness can indeed be associated with D, at least derivationally. I therefore depart from Lyons (1999), who develops the idea that person and definiteness are conflated and that both features are simultaneously associated with D. The paper is organized as follows. In section 2, I examine the inventory of English th- forms and suggest that many, despite appearances and traditional classifications, probably share an initial morpheme. I also demonstrate that th- forms traditionally identified as definite (the, that, etc.) are frequently used as indefinites. Section 3 develops the main thesis, which is that English th- is a person marker unspecified for number and gender. The data examined in section 4 provide support for the proposal. Section 5 considers various claims that 3rd person, unlike 1st/2nd person, is default, unmarked or not a natural class. Finally, section 6 summarizes the major proposals and suggests a direction for further research. 2. On the [+deictic] or [+referential] nature of D associated with demonstratives see, among others, Panagiotidis (2000), Bernstein (1997), Giusti (1997), Brugè (1996), Roca (1996). On forms with the [+TH]-feature which are associated with specificity or definiteness, see Campbell (1996). 3.

Lyons (1999: 27) identifies the as 3rd person, which he aligns with definiteness.

English th- Forms 



2. Inventory of English th- forms Initial th- appears across grammatical categories in English, some examples of which I provide in (2). It appears with definite articles (2a), personal pronouns (nominative and accusative) (2b,c), demonstratives (2c,d), possessive pronouns (2e), relative pronouns (2f), complementizers (2g), locatives (2h), and existential subjects (2i). (2)

a. b. c. d. e. f. g. h. i.

the (definite article) they (pronoun) them (pronoun; demonstrative) this, that (demonstrative) their (possessive pronoun) that (relative pronoun) that (complementizer) there (locative) there (existential subject)

the book, the sun, the hospital they are ready I saw them; them books this/that book their books the book that I read I believe that you are right the book over there there’s a book on the table

If I am correct in taking the th- of these forms to share a common initial morpheme (see section 3), the question arises as to how to identify this morpheme. Consideration of only the definite article, demonstratives, and personal pronouns could reasonably lead to the identification of th- as a definiteness marker as in Déchaine and Wiltschko’s (2002) work, or as a morpheme identified with deixis as in Klinge’s (this volume) work. However, consideration of a wider range of th- forms and their associated functions raises questions about such a conclusion. For example, “existential there,” which I argue displays the (relevant) th- morpheme, is associated with indefiniteness, judging by the now famous “definiteness effect” discussed by Milsark (1977) and others, whereby there’s associate can involve an indefinite noun phrase introduced by a (in (3a)), but not easily a definite one introduced by the (in (3b)) or the demonstrative this (in (3c)).4, 5 (3) a. There’s a book on the table. b. *There’s the book on the table. c. (*)There’s this book on the table. Similarly, the relative pronoun that in (2f) introduces a clause headed by what is arguably a [-definite] CP (Bernstein 1997: 104, building on work of Kayne 1994).6 It is for

4. The example in (3c) is ungrammatical on the relevant (deictic) interpretation of the demonstrative. The example becomes possible on an indefinite interpretation of the demonstrative (see discussion around example (6)). 5. I avoid the issue of the category status of existential there’s associate (e.g. NP, DP, small clause), and instead employ the neutral and non-committal label noun phrase. 6. For a comparison of relative pronouns and determiners, see Wiltschko (1998).

 Judy B. Bernstein

this reason that deictic demonstratives (as in (3c)) generally do not appear as the heads of relative clauses: *this picture that Bill saw.7 Even those th- forms that have traditionally been treated as definite or even deictic forms display uses that cannot be uniformly identified in such a way. Consider the definite articles in (4). (4) a. I went to the hospital. b. I went to the beach. c. I went to the circus. In each of these examples, there can be an interpretation associated with a specific hospital, beach, or circus (known to speaker and hearer), but there does not have to be. These sentences can also be uttered quite naturally in contexts where no specific hospital, beach or circus is implied. In those types of contexts the definite article can be substituted by the indefinite article. In (5), extra context facilitates such an indefinite interpretation. (5) a. When I visit Egypt next year, I hope I won’t have to go to the/a hospital. b. When I visit Egypt next year, I plan to go to the/a beach. c. When I visit Egypt next year, I plan to attend the/a circus. Similarly, demonstratives can be used in non-specific contexts, as illustrated in (6) (Bernstein 1997; see also Postal 1969: 205, fn.10).

(6) I met this strange guy on the subway last night.

(= a strange guy)

As the parenthetical indicates, the interpretation of the demonstrative is parallel to that of the indefinite article.8 English th- pronouns may also be used non-referentially. Example (7) (from Bloomfield 1933: 254) illustrates a th- pronoun with a non-specific denotation and

7. The fact that a definite article, unlike the deictic demonstrative, may introduce a relative clause (the picture that Bill saw) is explained by Kayne’s (1994) approach, since the definite article is generated outside (and above) the relative clause structure. Bernstein (1997) argues that the demonstratives start out within the relative clause CP and then raise to SpecCP. The example with a deictic demonstrative is barred because the [+definite] feature of the demonstrative would be incompatible with the [-definite] feature of relative clause C. Indefinites and other quantifiers are therefore predicted to be fine, consistent with the facts: a picture/some pictures that Bill saw. 8. Giuseppe Longobardi (personal communication) suggests that what I label “indefinite this” differs from a true indefinite (with a). In particular, he suggests that indefinite this introduces a noun phrase that must then be followed up in the subsequent narrative, and that this requirement does not hold for true indefinites. I’m not sure I share this intuition.

English th- Forms 



example (8) (see Déchaine and Wiltschko 2002) illustrates that th- pronouns participate in bound variable anaphora.9, 10

(7) They say Smith is doing very well.

(8) a. Somebodyi lost theiri hat. b. Everyonei thinks theyi are right. c. Peoplei have lost theiri minds. The point of these examples is to show that while th- forms tend to be definite, they do not have to be and some cannot be. This weakens the thrust of proposals that take thforms, or even the th- marker, to encode or display definiteness directly. As the examples in (2) illustrate, th- forms have been assigned to various grammatical categories, which do not form an obvious homogeneous group. The disparate nature of these categories obscures what I will claim is a shared grammatical property of the forms. I have discussed and illustrated that the most common features associated with many th- forms--deixis and definiteness--do not exceptionlessly characterize some of the forms and do not characterize others at all. My specific claim will be that what characterizes English th- forms is the encoding of 3rd person. This idea will be developed in the next section.

3. English th- as a person marker Although I am unaware of any previous treatment of th- as a person marker, there is a precedent for the idea that th- is a separate morpheme. Déchaine and Wiltschko (2002: 422) provide some independent support for the treatment of th- as a separate morpheme. As they point out (see (9)), th- may be absent in phonologically reduced accusative forms.

(9) I know them. → I know ’em.

Why initial th- may not be suppressed in other forms (e.g. that, their, these), even in accusative contexts, is not clear.11 Similarly, Klinge (this volume) takes th- to be a separate morpheme.

9. Anaphoric th- (they/them) is replacing anaphoric he/him as the default bound variable in examples like (8a,b). This is probably related to the gender neutrality characterizing the thforms (see section 3). 10. Siewierska (2004: 10-11) uses very similar examples to make an analogous point; however, she will ultimately appeal to the overwhelmingly definite nature of th- forms. 11. That the is never, to my knowledge, phonologically reduced is less mysterious. This is because this form is probably monomorphemic, made up of the th- morpheme and an epenthetic vowel (usually [ә]).

 Judy B. Bernstein

Unlike Déchaine and Wiltschko, however, I don’t take th- to be directly involved in definiteness. The specific proposal I make about English th- is given in (10). (10) English th- is a (3rd) person marker unspecified for gender and number; th displays the person of an associated noun or noun phrase, or that of a DPexternal referent. In the case of a lexical DP consisting of th- determiner element + noun, the th-initial determiner element expresses the DP’s 3rd person feature, never overtly expressed on the noun, as the examples in (11) illustrate. (11) a. [DP the3rd book(s)] b. [DP that/this3rd book] c. [DP these/those/them3rd books] I take as indirect the fact that these examples may also convey definiteness (cf. Lyons 1999: 313, where definiteness and person are conflated and associated with D). Some support for this assumption comes from the range of data presented and discussed in section 2, where I illustrated indefinite use of the th- forms. The proposal in (10) may be compatible with Vergnaud and Zubizaretta’s (1992: 613) claim that the complementation relationship between the and N is an agreement relation of token (D) to type (N). What I am suggesting then is that the token-type relationship may be one of 3rd person agreement (on D) matched up with N, and that this person marking is associated with, and then spelled out in, D. With th- (as well as non th-) pronominal elements, person agreement must be with a DP-external referent, as illustrated schematically in (12) and (13). (12) a. DP3rd…………. [DP their3rd books] b. DP3rd…………. [DP them3rd] c. DP3rd…………..[DP they3rd ] (13) a. DP1st (speaker)...[DP my1st books] b. DP1st (speaker)...[DP me1st] c. DP2nd (hearer)….[DP your2nd books] As the examples in (12) show, 3rd person pronouns display a th- person marker. The examples in (13) show that 1st person pronouns display m- for 1st person (corresponding to the speaker) and y- for 2nd person (corresponding to the hearer), also discussed in Kayne (2000) for French. Note that the pronominal elements in (12) and (13), unlike the determiner elements in (11), refer directly. This is because (personal) pronouns have an anaphoric or a deictic function, a point I return to in section 5. Also noteworthy is the fact that the th- forms in (12) either agree with or directly refer to plural entities. Indeed, th- pronominal forms have traditionally been treated as “3rd person plural” forms. This does not seem to be true for the th- forms in (11). As

English th- Forms 



(11a) illustrates, the definite article may co-occur with a singular or a plural noun; (11b) illustrates a singular demonstrative with a singular noun; and (11c) shows a plural demonstrative with a plural noun. If, as I claim, th- is a 3rd person marker, why is it that most of these forms (with the exception of the) may not be used interchangeably as singular and plural forms? Further, why are the th- personal (possessive) pronouns (recall (12)) grammatically plural?12 What the examples in (11) nicely illustrate, in fact, is that th- on its own is not directly associated with singularity or plurality. This is clearly demonstrated in (11a), where the definite article consists of only th- plus what I take to be an epenthetic vowel. Because there is no marking or specification beyond th-, the definite article is free to occur with singular and plural nouns. In (11b,c), on the other hand, further (overt or covert) morphological specification yields information about number: this, that as singular forms, these, those as plural forms. Similarly, in the pronominal forms in (12), the plural referent required by these forms is obtained from additional specifications of the pronoun, not from the th- morpheme itself.13, 14 Although it is not obvious which component, if any, of the (plural or singular) th- forms encodes number, it is clear that it cannot be th- itself, given a form like definite article the. Existential there, which strongly favors indefinites (recall discussion from section 2), further illustrates the point about number. That is, it provides further evidence for the idea that th- is unspecified for number. Consider the examples in (14). (14) a. There’s a book on the table. b. There are many books on the table. What these examples show is that the associate of existential there may be singular as in (14a) (a book), or plural as in (14b) (many books).15 If I am correct in assuming that no number specification is encoded in th-, then the absence of a number association with the forms in (2f-i) is unproblematic. In fact, it is expected if th- encodes nothing more than 3rd person. This raises the question of word-initial h-, which might be associated with 3rd person in pronominal forms like

12. I distinguish grammatical plurality from plural reference. Although the th- personal (possessive) pronouns do not require plural reference (recall (8)), they would trigger plural forms of verbs. 13. The necessity for a plural demonstrative form with plural nouns is also illustrated by the vernacular demonstrative them, which is isomorphic with the 3rd person plural accusative pronoun: them books (cf. *them book). 14. Bernstein and Tortora (2005) take word-final -r in possessive their to be the plural form of the copula, so equivalent to -r in they’re. 15. Further, in colloquial English even the verb may not agree with a plural associate: There’s many books on the table.

 Judy B. Bernstein

him, her(s), he, and his.16 In other words, if the claim that th- is a 3rd person marker is on the right track, what is h- in these other (3rd person) forms?17 That h-, like th-, is a separate morpheme is supported by the existence of reduced forms, as seen in (9) for th-. As (15) shows, initial h- may be deleted in accusative contexts. (15) a. I saw him. → I saw ’im. b. I saw her. → I saw ’er. I’d like to suggest that what distinguishes the h- pronouns from the th- forms is gender. In other words, in contemporary varieties of English the pronominal h- forms, but not the th- forms, encode gender (masculine or feminine).18 These forms are then the only pronominal forms in the language to display morphological gender overtly.19 What this means is that although the h- forms function as 3rd person forms, there is no 3rd person marker per se. If initial h- encodes gender in the singular forms, then gender is neutralized in the plural (th-) forms, which never display h- or any other morpheme identifiable as a gender marker. Crosslinguistically, plural gender neutralization is extremely common. In French, for example, gender is robustly displayed on the singular forms of the definite article (16a,b) but neutralized in the plural (16c). (16) a. b. c.

le the (m.sg.) la the (f.sg.) les the (m.pl. or f.pl.)

16. Inclusion of here in this group of h- forms would be on analogy with locative there. I will not address the locatives here, but I note the potential assimilation (see Klinge’s work on deixis, this volume). Similarly, I have nothing interesting to say about suppletive she. 17. I am essentially ignoring diachronic factors related to the co-occurrence of th- and hforms, and focusing on the synchronic patterns. The editors raise an interesting question about why English has generalized neither th- nor h- as the sole marker of 3rd person. Although I have nothing insightful to offer, I refer the reader to fn. 9 (about the generalization of th- in cases of singular bound variable anaphora). 18. If h- is (masculine or feminine) gender in contemporary English, then a question that arises is why the masculine and feminine pronouns are not isomorphic (cf. her, him). This may relate to the original Case endings, which displayed a gender alternation. 19. A question is whether English displays any overt grammatical gender marking (beyond my proposal for h-). I do not consider gender to be involved in (an isolated case like) blond/blonde, which displays only an orthographic distinction, or other cases of what I take to be discrete lexical items or perhaps lexically-derived gender: prince, princess.

English th- Forms 



Gender-neutral pronominal it is interesting since historically, and even in contemporary conservative varieties like Appalachian English, the form displayed an audible word-initial h- (i.e., [h]it).20 The gradual loss of initial h- from this form is consistent with the idea that h-, as a gender marker, is currently only a masculine/feminine marker in most English varieties.21 Under this account, the varieties that residually display h- with gender-neutral forms represent vestiges of a three-way gender distinction.22 In this section I have claimed that English word-initial th-, despite appearances, is not a definiteness marker. I have argued instead that it is a 3rd person morpheme unmarked for number and gender. Initial th- may display the 3rd person of an accompanying common noun (the [N book]), noun phrase (there is [NP a book]…), or DP-external referent (I saw [DP the girls]. They…). The approach defended here contrasts with that of Lyons (1999: 316), which conflates the categories definiteness and person, both directly associated with D. As an alternative, the preceding discussion defends the ideas that what characterizes and unifies 1st/2nd personal pronouns and th- forms is the display of person, and furthermore, that person marking does not necessarily imply definiteness. Rather, person marking is a necessary condition for definiteness, though not a sufficient condition. In other words, person is an inherent feature of personal pronouns and th- forms. Although not an inherent feature, definiteness may become associated with these forms derivationally (Longobardi personal communication, this volume). A question arises about indefinites. That is, if th- forms like the, these, and them encode 3rd person, what is the proper treatment of the indefinite examples in (17), which are also traditionally taken to be 3rd person forms? (17) a. a book b. some books c. three books For these examples, I follow Lyons’ (1999: 316) claim that indefinites are “personless,” and so not DPs. Interestingly, as Lyons points out, the so-called 3rd person verbal 20. Hit was the sole neuter form in Old English; Middle English and Early Modern English displayed both hit and it (Smith 1999). Wright (1898-1905) provides examples of pronominal hit in English varieties of Scotland, Northumberland, and the U.S. See Montgomery and Hall (2004) for examples from varieties of Appalachian English and Robinson (1997) for examples from varieties of Ulster-Scots. 21. Occurrences of both it and hit within a particular dialect may be phonologically and/or syntactically conditioned. 22. Old English exhibited a grammatical gender system corresponding to that found in many contemporary Germanic and Romance languages. By Middle English, the system had changed to one where gender corresponded to biological sex (Smith 1999). The presence of hit in varieties of Appalachian English probably stems largely from a Scots-Irish influence, but may also have been imported by English and Scottish settlers (Michael Montgomery personal communication).

 Judy B. Bernstein

agreement found with these forms might be better viewed as singular/plural agreement, which is consistent with proposals made by Kayne (1989 and subsequent work) and others about English verbal agreement (see section 5).

4. Support for the proposal In the previous section, I have outlined the basic proposal, which is to treat th- as a person morpheme. In particular, I have suggested that th- encodes 3rd person and is unspecified for gender and number; a form’s number marking must be expressed independently, if at all. In the following subsections I provide some support for the basic analysis. In section 4.1, I argue that English th- may convert 1st/2nd person pronouns to 3rd person forms. In section 4.2, I take up the treatment of pronouns as determiners, an approach that goes back to Postal (1969). In section 4.3, I discuss the fact that English th- forms, unlike explicit/implicit 2nd person forms, may not appear in vocative contexts and I argue that a predicate/argument distinction cannot straightforwardly account for the full range of facts.

4.1

Person conversion

Interestingly, insertion of a th- form can render a 1st or 2nd person pronoun useable as a 3rd person form.23 Consider the examples in (18). (18) a. I found that [the me]3rd in the photo is more mysterious than [the me/the person]3rd of real life. b. They didn’t get to see [the real us]3rd. The insertion of the definite article in (18a,b) basically converts the 1st/2nd person pronouns, which also display person marking, into (3rd person) common nouns.24 Note that the two instances of [the me] in (18a) refer to two different images of the speaker, at least according to his/her perception. Siewierska (2004: 10, fn.8) discusses examples parallel to those in (18) and similarly takes them to involve cases of conversion, in particular conversion from a personal pronoun to a (common) noun.25 This is consistent with the idea above, because 23. These facts in (18) challenge the general view that 1st/2nd person pronouns are always definite and must refer directly (see Lyons 1999). 24. Note that examples like (18) would not be grammatical without modification by an adjective, prepositional phrase, or relative clause: the me *(in the photo). 25. Siewierska’s (2004: 10) parallel examples in (i) and (ii) are taken from Noguchi (1997: 778-779).

(i) This is not the real me. (ii) Do you know the real you?

English th- Forms 



a common noun must always be 3rd person even though there is no morpheme that can be identified as such. What my proposal contributes then is the idea that the thform itself is responsible for the conversion while also displaying an overt manifestation of the result. Without such an accompanying th- form, personal pronoun conversion would be impossible.26 And so any explanation for examples like (18) must consider the licensing contribution of th-, in particular its status as a 3rd person marker. The contribution of th- in (18) cannot be one related to definiteness, since the personal pronouns may, on their own, function as definite forms. Under an analysis such as Lyons (1999), where person and definiteness are conflated and so person is always definite, examples like (18) are unexplained, since two separate arguably definite forms representing two different persons co-occur within the same simple DP.

4.2

Pronouns as determiners

Déchaine and Wiltschko (2002) distinguish 1st/2nd person pronouns, which may be used as determiners, from 3rd person pronouns, which apparently cannot (see also Postal 1969, Abney 1987, Roehrs 2005, Lyons 1999; cf. Bhat 2004). Consider the examples in (19). (19) a. we students b. you students c. *they students Déchaine and Wiltschko account for the ungrammaticality of (19c) via a structural distinction between 1st/2nd person forms and 3rd person forms.27, 28 Although I will not enter into a discussion here of the internal structure of these forms, I believe that consideration of the full range of facts shows that the 1st/2nd vs. 3rd distinction they make based on these examples does not hold up (see also fn. 27).

26. Results with (th-) demonstratives are equivalent to (18) (that me in the photo); results with indefinites are sharply degraded (*a me in the photo; *a real me). If indeed indefinites are “personless” (see section 3), this may offer an explanation for why indefinites cannot convert 1st/2nd forms into 3rd person forms. 27. Wales (1996: 201, fn. 17), citing Milroy and Milroy (1993: 65), does provide an example equivalent to (19c) from Devon (UK) speech, where it is apparently grammatical: Look at they spiders. 28. In particular, they claim that 1st/2nd person pronouns are full DPs, and 3rd person pronouns are “φPs,” which involve less structure (see also Ritter 1995).

 Judy B. Bernstein

In particular, the examples in (20) show that the paradigm is more regular with accusative pronouns.29, 30 (20) a. us students b. you students c. them students

(vernacular)

Note that the example in (20c) is found in vernacular varieties of English, and so contrasts somewhat with the examples in (20a,b), which I expect are found across English varieties. Furthermore, the interpretation of them in (20c) corresponds to that of a demonstrative. This should not be problematic, however, since the 1st/2nd pronouns in (19) and (20) are similarly deictic.31 Returning to the examples in (19), I note that the gap in the nominative paradigm in (19c) may be filled by another th- 3rd person form, namely the true demonstrative (these/those students), which is in fact the interpretation that (19c) would have, were it grammatical. Like the nominative pronoun they, the demonstratives display person (th-) and number marking. I return to this observation directly, since it will play a crucial role in what I will propose about licensing of these constructions. The overlap between nominative pronoun and demonstrative goes both ways, since the demonstrative form can often be used in place of the 3rd person nominative, as the examples in (21) and (22) illustrate. In both sets of examples, either (b) or (c) is a legitimate response to the speakers’ remarks in (21a) and (22a). (21) a. speaker 1: Have you seen the new Hondas? b. speaker 2a: Yes! They are my favorite cars! c. speaker 2b: Yes! Those are my favorite cars! (22) a. speaker 1: White peaches are available in the market this week. b. speaker 2a: Oh, I love them! c. speaker 2b: Oh, I love those! A final relevant point about the 3rd person plural pronouns comes from English dialects. Both 2nd and 3rd person pronouns form special complex pronouns in some 29. If my approach is on the right track, then Déchaine and Wiltschko’s (2002: 421-422) distinction between two dialects, only one of which admits them linguists, becomes unnecessary. 30. Interestingly, as Rijkhoff (2002: 83) points out (based on Bruce 1984: 96), in a language like Alamblak (Indo-Pacific) an obligatory person-number-gender marker appearing phrase finally on noun phrases “shows great similarity with a corresponding member of the set of personal pronouns”: yima-m (person-3pl. ‘people’). This is of course parallel to the pattern displayed for English in (20). The Alambak facts further support Postal’s original claim that these examples involve D+N, rather than a competing idea that treats them as cases of apposition (see Roehrs 2005 for discussion). Thanks to Jan Rijkhoff (personal communication) for directing me to the Alambak facts. 31. See Roehrs (2005: fn. 17) for similar observations about examples like (20c).

English th- Forms 



varieties of UK and American English. In particular, the Ulster-Scots example in (23b) would not be expected, alongside (23a), if 3rd person pronouns could not be used as determiners. The examples are from Robinson (1997: 71). (23) a. yous’uns b. thaim’uns

(formed from: (formed from:

yous + ones) them + ones)

Often observed is the fact that the singular pronouns may not appear as determiners, as shown in (24). (24) a. *I/me student b. *you student c. *(s)he/him/her student Note that both the nominative and accusative forms of the pronouns are ungrammatical, so Case cannot be relevant here. To my knowledge, these facts have never been accounted for. I suggest that in order for a pronoun to function as a determiner in English, it must meet two criteria: a) it must display overt person marking, and b) it must agree in number with the noun. These two criteria are satisfied by the accusative pronominal forms shown in (20): (20a) is 1st person and plural, agreeing with the plural noun students; (20b) is 2nd person and plural, agreeing with the plural noun; (20c) displays 3rd person and is plural, agreeing with the plural noun. Unlike the examples in (20), the two criteria (overt person marking and number agreement) are not met in the examples in (24), which accounts for their ungrammaticality. Although the pronoun I in (24a) displays person, I claim that it does not agree in number with the noun. Following Kayne (1989: 44; see also Bernstein and Tortora 2005: 1224), I adopt the idea that I is unmarked for number, which is consistent with irregularities like I am (1st person copula), I walk(*s) (pl. verb), I was (sg. copula), and Aren’t I? (pl. copula; cf. *Amn’t I?). For (24b) as well, person is not the issue, since 2nd person is displayed. The problem with (24b) is a number mismatch. So although modern English you (like French vous), may refer to a singular or plural referent, it is grammatically plural: you are vs. *you is (see Kayne 1989, Bernstein and Tortora 2005).32 Unlike (24a) and (24b), I claim that the trouble with (24c) is person, not number. Although not obviously marked as such, these forms are singular, matching the singularity of the noun. My proposal in section 3 was that the h- pronouns, unlike the thpronouns, fail to display (3rd) person marking. And for this reason they fail to function

32. I am not including the exclamative cases, which in English are possible with singular you (You fool!, You moron!), but not with 1st person I (*I fool!, *I moron!). Note, however, that both the exclamative and the equivalent of (24a) are possible in German, as Lyons (1999: 27) illustrates: Ich Esel! (I donkey, ‘silly me!’); Ich Vogelfänger bin bekannt… (I birdcatcher am known…). See also Roehrs (2005).

 Judy B. Bernstein

as determiners, contrasting with the th- forms.33 This is consistent with the ideas that person is encoded on D and that a (pronominal) form lacking person marking may not function as a determiner.

4.3

Vocatives

Vocatives generally involve, explicitly or implicitly, 2nd person singular or plural (see Szabolcsi 1987). This is illustrated straightforwardly in the examples in (25). (25) a. You (guys)! b. (Hey) you! c. (You) boys! (You) girls! In (26) I provide examples of implicit 2nd person. In other words, although not explicitly marked as such, the intended referent in these examples is 2nd person. (26)

a. b. c. d. e.

Mary! Girls! Children! Waiter! Driver!

Longobardi (1994: 626, fn. 20; see also Szabolcsi 1987: 181) proposes that vocatives, like predicates, are non-arguments and so would not take D. This idea can apply straightforwardly to examples like (26b,c,d,e). The examples in (25) and (26a) would require additional explanation, since both personal pronouns (as determiners) in English and proper names generally have been associated with D and in non-vocative contexts are indeed arguments (see also Szabolcsi 1987: 181). If proper names raise to D only at LF in English (Longobardi 1994), then (26a) could be a case of a non-argument proper name, hence no movement to D involved.34 Examples like (25a,b) are more difficult to explain, particularly if person (here, 2nd) is encoded in D (see section 4.2). Crisma (1997: 135–139) offers a viable solution to this problem. She explicitly argues, based on data from several languages, that vocatives may have a D position as part of their structure even though the position is not always lexicalized.35 Crisma also 33. Roehrs (2005) discusses examples from northern Scandinavian dialects (from Delsing 1993), where 3rd person pronouns may appear with proper names, as in the following Norwegian example: hun Mari (she Mary, ‘Mary’). Roehrs suggests that these pronouns function as expletive elements. If that is the case, then the absence/presence of person marking would not be relevant, since the pronouns are not used referentially. 34. This does not settle the issue of proper names in Romance, which Longobardi (1994) argues raise to D in the syntax. 35. An important piece of evidence in favor of her account is the fact that vocative noun phrases may sometimes appear as heads of relative clauses in Italian.

English th- Forms 



develops the ideas that DPs denote individuals (NPs denote properties), and that only nominal expressions that denote individuals (DPs) may function as arguments. A consequence of the proposal is therefore that vocatives, which denote individuals, may in fact involve arguments, consistent with the data she presents but contrary to the earlier hypotheses of Szabolcsi and Longobardi. Crisma’s proposals are also consistent with the English data presented in (25) and (26), including the examples in (25a,b) that were problematic under the earlier approaches. Although proper names and 2nd person pronouns are not barred in vocative contexts in English, th- forms clearly are, as illustrated in (27).36 (27) a. *The waiter(s)! b. *That/that waiter! c. *Them (guys)! A predicate/argument distinction, which was already found to be problematic for data in (25) and (26), cannot adequately explain the ungrammaticality of the examples in (27). In particular, if the ungrammaticality of the examples in (27) reduces to their status as arguments (which, under this approach, are barred in vocative contexts), why is the 2nd person pronoun in (25a,b), but not the 3rd person pronoun in (27c), allowed? Similarly, it is not clear to me how Crisma’s distinction between properties (NPs) and individuals (DPs) would apply to these examples either. If individuals (as DPs) can appear in vocative contexts, the examples in (27) might all be predicted to be grammatical. Even if there is a mechanism responsible for the suppression of definite articles and demonstratives in vocative contexts in English and other languages (as in examples (27a,b)), I see no obvious way to distinguish between 2nd person pronouns (allowed) and 3rd person pronouns (prohibited). I suggest that the problem with the examples in (27) does not reduce to a predicate/ argument distinction, and cannot straightforwardly be explained by a properties/individuals distinction either. Instead, I claim that vocative, which necessarily involves an explicit or implicit 2nd person, is inconsistent with 3rd person th- forms. So the offending component of the examples in (27) is the appearance of a th- form, which I have claimed encodes 3rd person in English. Apparently, th- forms in English may not be used as, or understood to represent, 2nd person forms. On the other hand, I have proposed (in section 4.1) that th- can convert 1st and 2nd person forms into 3rd person. My explanation for the impossibility of th- forms in vocative contexts in English (and presumably other languages) may actually complement Crisma’s approach to vocatives. That is, we can adopt Crisma’s claim that vocatives involve D, which accounts for the otherwise problematic cases of (25) and (26). At the same time, my proposal about the relevance of person in vocatives, specifically that 2nd person must 36. Although vocatives cannot involve definite articles in English, this is not always so in some languages, such as special cases in French (e.g. les enfants! ‘children!’). On this and related facts see Crisma (1997); see also Longobardi (1994).

 Judy B. Bernstein

be explicit or implied, accounts for the general absence of (lexicalized) 3rd person thforms in English. The idea I have developed in this paper is that person is directly associated with D and that definiteness may be indirectly (or derivationally) associated with D, so unlike Lyons’ (1999) idea that person and definiteness are conflated and simultaneously associated with D. If conflation of person and definiteness were the right approach, it would not be clear how to account for the distinction between (25a) and (27c) in vocative contexts, since these examples should be parallel in terms of definiteness and the display of person. The separation of person and definiteness that I have advocated in this paper allows for an explanation of the grammaticality of one and the impossibility of the other.

5. The status of 3rd person It is often observed that 1st and 2nd person forms seem to share features and paradigms, and that these are either absent in 3rd person forms, or at least the paradigm is distinct from that shared by 1st/2nd forms. For example, in recent work on Romance clitic and possessive forms Kayne (2000) distinguishes the m- (1st person) and t- (2nd person) forms from the l- forms, which he argues do not encode 3rd person. In fact, building on ideas in Benveniste (1966: 228), Kayne claims that 3rd person forms do not form a natural class, unlike 1st/2nd person forms.37 For Forchheimer (1953; see discussion in Harley and Ritter 2002), 1st and 2nd person (speaker and hearer, respectively) are marked with respect to 3rd person, which is unmarked.38 Similarly, Lyons (1999: 316) argues that 3rd person is unmarked or default. Based on the discussion and proposals made in this paper on English th-, I’d like to address the idea of 3rd person as unmarked or default, in particular in terms of encoding within the DP. If my approach is on the right track, English 3rd person forms are robustly marked: th- pronouns, th- possessives, th- demonstratives, th- definite articles, th- existential there, etc.39 On the other hand, 3rd person is frequently unmarked with respect to the expression of verbal agreement.

37. Kayne’s conclusions draw support from Romance 1st/2nd vs. 3rd person forms, such as in the possessive paradigm, where m-/t- encode 1st/2nd person (French ma/ta ‘mine’/‘yours’ - sg., mes/tes ‘mine’/’yours’ - pl.), but l- does not participate (la, les function as definite articles and accusative clitics, but not possessive forms; cf. sa, ses ‘his/her’ - sg., ‘his/her’ - pl.). 38. Harley and Ritter (2002: 487) attribute the difference between 1st/2nd vs. 3rd “to the fact that reference of the former is determined by the changing discourse roles, whereas the reference of the latter is fixed.” 39. Unlike Kayne (2000), I believe that this may extend to Romance l- forms, but won’t pursue the topic here for reasons of space.

English th- Forms 



As Lyons (1999: 316–317) discusses and illustrates, an absence of verbal agreement characterizes 3rd person singular in many disparate languages such as Spanish, Biblical Hebrew, Turkish, and Warlpiri. I illustrate the general point with the Spanish verb hablar (‘talk’, ‘speak’) in (28).40 (28)

a. b. c. d.

hablar hablo, hablamos hablas, habláis habla, hablan

(infinitive) (1st person sg., pl.) (2nd person sg. pl.) (3rd person, sg. pl.)

The form used for 3rd person singular (habla) displays no special marker that may be identified as a person marker; the terminal vowel -a corresponds to the stem class of the verb and so also appears with most of the verbal forms, including the infinitive. In this respect, English seems to be a counterexample to the generalization since the only verbal agreement displayed in modern English is the verb-final -s associated with 3rd person singular subjects. But as Kayne (1989) argues (see also Lyons 1999, Bernstein and Tortora 2005), the verbal agreement facts are more consistent with the idea that -s does not represent person, but rather (singular) number. Building on Kayne’s work, Bernstein and Tortora (2005) further argue that a zero marker on verbs in English (e.g. they know-Ø) is similarly an indicator of number (i.e., plurality), not person.41 If so, then English never displays person agreement on verbs, and so does not constitute a counterexample to the crosslinguistic generalization about the default status of 3rd person verbal agreement. Lyons (1999: 316–317) makes a similar point about languages like German and Spanish. In particular, he argues that the verbal marker -t in German that appears with 3rd person singular subjects is really a singular marker. Similarly, the -n verbal marker in Spanish that appears with 3rd person plural subjects (see (28d)) is a plural marker, not a person marker. This is not to deny that there are differences between 1st/2nd vs. 3rd person forms. Compatible with Harley and Ritter’s (2002) characterization of the differences in terms of discourse roles (see fn. 38), I observe that 1st and 2nd person forms are uniformly deictic. This is not true for 3rd person forms, which I assume may acquire an anaphoric or deictic function derivationally. This idea should also extend to other semantic functions that may (or may not) be associated with 3rd person forms, such as definiteness, referentiality, and argument (vs. predicate) status.

40. The examples in (28c) are sg. and pl. informal forms. The formal counterparts (usted, ustedes) appear with the same verbal forms as 3rd person subjects (illustrated in (28d)). 41. See Bernstein and Tortora (2005) for the further claim that not only plural verbs in English carry a zero marker. In particular, they claim that a zero marker also characterizes lexical DP plural possessives. In this case, -s indicates the plurality of the possessor, and the possessive feature itself carries a zero marker (e.g. the girls-Ø mother).

 Judy B. Bernstein

While these other semantic functions may have no impact on the morphological shape of the accompanying verb, they contribute information relevant to interpretation. This array of possible semantic functions is not characteristic of 1st/2nd person forms, which are more restricted with respect to the semantic information they convey. The distinct and varying semantic information that may be conveyed by 3rd person forms is arguably derived, not inherent. This might suggest that 3rd person forms are derivationally more complex than 1st/2nd person forms, an idea that would also be consistent with greater morphological complexity. The patterns displayed across the Romance languages may be telling. In these languages, many 3rd person forms (e.g. definite articles, subject pronouns, and pronominal clitics) display gender agreement, whereas the corresponding 1st/2nd person forms never do. Details of this idea of greater derivational complexity for 3rd person forms would need to be further developed and then compared with the approach of Déchaine and Wiltschko (2002; recall fn. 28), since their proposal is that these forms involve less structure than 1st/2nd person forms.

6. Conclusions In this paper, I have argued that what unifies English th- forms is not a feature encoding definiteness or deixis, but rather person. Specifically, I have claimed that th- is a morpheme that encodes 3rd person in English and that person is associated with D, the head of the functional projection DP (see Longobardi this volume). Consistent with its distribution, I have further suggested that th- is unspecified for number and gender. Evidence against an account of English th- as an inherent definiteness or deixis morpheme came from several sources. First, many DPs with definite articles or even demonstratives may have a non-definite interpretation in certain contexts. Second, existential there, which I argued patterns with the other th- forms, may not as a rule be associated with a definite DP. Finally, th- pronouns may be used non-referentially and with non-specific denotation. Although these facts cast doubt on an account of th- as an inherent definiteness marker, nothing precludes definiteness (or deixis) being derivationally related to th-, and to D. This is in fact the approach I have adopted, without developing the details. Also presented were data supporting the proposal that English th- encodes 3rd person. One source of support involves what I label “conversion.” In particular, thforms can convert 1st/2nd person forms to 3rd person, probably by converting the 1st/2nd pronouns into common nouns. Another type of support comes from the function of 3rd person pronominal forms as determiners, something shared with 1st/2nd person forms. I claimed in fact that pronominal determiners, in order to assume that function, must display both person and number features. In this respect, th- pronominal forms (but not h- pronominal forms) meet the necessary requirements. A third source of support comes from data involving vocatives. Interestingly, th- may never

English th- Forms 



appear in vocative contexts, which necessarily involve (explicit or implicit) 2nd person. I argued that th- 3rd person forms are incompatible with the 2nd person requirement. This paper has focused on a very narrow range of facts particular to one language, English. Questions naturally arise about how the account extends to other Germanic languages, as well as to other languages more generally. These crosslinguistic issues will form the basis of future work.

References Abney, S.P. 1987. The English Noun Phrase in its Sentential Aspect. PhD dissertation, MIT. Benveniste, E. 1966. Problèmes de linguistique générale. Paris: Gallimard. Bernstein, J.B. 1997. Demonstratives and reinforcers in Romance and Germanic Languages. Lingua 102: 87–113. Bernstein, J.B., Cowart, W. & McDaniel, D. 1999. Bare singular effects in genitive constructions. Linguistic Inquiry 30(3): 493–502. Bernstein, J.B. & Tortora, C. 2005. Two types of possessive forms in English. Lingua 115(9): 1221–1242. Bhat, D.N.S. 2004. Pronouns. Oxford: OUP. Bloomfield, L. 1933. Language. New York NY: Holt, Rinehart and Winston (reprinted 1984, Chicago IL: University of Chicago Press). Bruce, L. 1984. The Alamblak language of Papua New Guinea (East Sepik) [Pacific Linguistics C-81]. Canberra: The Australian National University. Brugè, L. 1996. Demonstrative movement in Spanish: A comparative approach. University of Venice Working Papers in Linguistics 6(1): 1–61. Campbell, R. 1996. Specificiy operators in SpecDP. Studia Linguistica 50: 161–188. Crisma, P. 1997. L’articolo nella prosa inglese antica e la teoria degli articoli nulli. PhD dissertation, Università degli Studi di Padova. Déchaine, R.-M. & Wiltschko, M. 2002. Decomposing pronouns. Linguistic Inquiry 33(3): 409–442. Delsing, L.-O. 1993. The Internal Structure of Noun Phrases in the Scandinavian Languages. PhD dissertation, University of Lund. Forchheimer, P. 1953. The Category of Person in Language. Berlin: Walter de Gruyter. Giusti, G. 1997. The categorial status of determiners. In The New Comparative Syntax, Liliane Haegeman (ed.), 95–123. London: Longman. Greenberg, J.H. 1978. How does a language acquire gender markers? In Universals of Human Language, Vol. 3, J. H. Greenberg (ed.), 47–82. Stanford CA: Stanford University Press. Harley, H. & Ritter, E. 2002. Person and number in pronouns: A feature-geometric analysis. Language 78(3): 482–526. Kayne, R.S. 1989. Notes on English agreement. CIEFL Bulletin, 41–67. Hyderabad. Kayne, R.S. 1994. The Antisymmetry of Syntax. Cambridge MA: The MIT Press. Kayne, R.S. 2000. Parameters and Universals. New York NY: OUP (Chapter 8: Person morphemes and reflexives). Longobardi, G. 1994. Reference and proper names: A theory of N-movement in syntax and logical form. Linguistic Inquiry 25(4): 609–665.

 Judy B. Bernstein Lyons, C. 1999. Definiteness. Cambridge: CUP. Milroy, J. & Milroy, L. (eds). 1993. Real English: The grammar of English dialects in the British Isles. London: Hatchards. Milsark, G. 1977. Toward an explanation of certain peculiarities of the existential construction in English. Linguistic Analysis 3: 1–29. Montgomery, M.B. & Hall, J.S. (eds). 2004. Dictionary of Smoky Mountain English. Knoxville TN: The University of Tennessee Press. Noguchi, T. 1997. Two types of pronouns and variable binding. Language 73: 770–797. Panagiotidis, P. 2000. Demonstrative determiners and operators: The case of Greek. Lingua 110: 717–742. Postal, P. 1969. On so-called pronouns in English. Modern studies in English, D. Reibel & S. Schane (eds), 201–224. Englewood Cliffs NJ: Prentice-Hall. Rijkhoff, J. 2002. The Noun Phrase. Oxford: OUP. Ritter, E. 1995. On the syntactic category of pronouns and agreement. Natural Language and Linguistic Theory 13: 405–443. Robinson, P. 1997. Ulster-Scots: A Grammar of the Traditional Written and Spoken Language. Belfast: The Ullans Press. Roca, F. 1996. La determinación y la modificación nominal en español. PhD dissertation, Universitat Autònoma de Barcelona. Roehrs, D. 2005. Strong pronouns are determiners after all. In The Function of Function Words and Functional Categories, M. den Dikken & C. Tortora (eds), 251–285. Amsterdam: John Benjamins. Siewierska, A. 2004. Person. Cambridge: CUP. Smith, J. 1999. Essentials of Early English. London: Routledge. Szabolcsi, A. 1987. Functional categories in the noun phrase. In Approaches to Hungarian 2, Theories and Analyses, I. Kenesei (ed.), 167–189. Szeged: JATE. Vergnaud, J.-R. & Zubizarreta, M.L. 1992. The definite determiner and the inalienable constructions in French and in English. Linguistic Inquiry 23: 595–652. Wales, K. 1996. Personal Pronouns in Present-day English. Cambridge: CUP. Wiltschko, M. 1998. On the syntax and semantics of (relative) pronouns and determiners. Journal of Comparative Germanic Linguistics 2: 143–181. Wright, J. 1898–1905. The English Dialect Dictionary. London: H. Frowde.

Stating the case for þ- root and hw- root determiners Alex Klinge

Copenhagen Business School In this article I will show that the existence of coherent classes of determiners based on pan-Germanic þ- and hw- roots, such as English ‘this’ and ‘which’ and German ‘diese’ and ‘welche’, make it necessary to reanalyse the syntactic classes and paradigmatic contrasts in the functional domain of Germanic DPs. It will be argued that the two þ- and hw- morphemes contain procedural semantics which encodes two contrasting ways that a referent may be identified for an index in a propositional form. I will focus on English and German, but reference will also be made to Danish. As a first step it will be argued that there is no coherent syntactic class of articles. Then I will draw on etymological and comparative data, and on morphological and distributional facts to show that the two morphemes have been remarkably resilient across Germanic languages for more than a thousand years. Finally, I will anchor their resilience in their semantic and pragmatic raison d’être.

1. Preliminaries When Abney (1987) formulated a coherent analysis of nominal projections in terms of the DP-analysis, he not only meet his own goal of bringing the structure of phrasal projections in the nominal domain into line with phrasal projections in the clausal domain, he also laid the foundation of a reanalysis of a whole range of formal and distributional facts in nominals. Abney’s phrase structure consisted of a sequence of head-complement projections, together producing an internally multiheaded structure of DP, QP, AP and NP projections. With Abney’s analysis it became possible to divide the nominal projection into functional and lexical subdomains in the same way as the clausal domain. One crucial development was that Abney’s framework made it possible to show that most of the members in the general syntactic class of pronouns, which had until then been treated as proN-heads, could and should be accommodated in the analysis as higher

 Alex Klinge

functional heads, including D-heads. Not only did such an analysis fit the distributional facts of pronouns in a higher projection than N, as illustrated through well-known types of data, such as those in (1), but it also landed most so-called ‘pronouns’ in a functional syntactic domain which semantically encodes various aspects of reference assignment. (1a) below reveals that we and you take the same distribution as for instance the, this and that; (1b) reveals that something belongs in a projection which is higher than the NP, and; (1c) below reveals that in spite of a very long tradition for calling them ‘pronouns’, forms such as it and this do not distribute at all as they should if they are supposed to be proforms for nouns, whereas a true pronoun such as ‘one’ does. (1) a. We logicians do not necessarily disagree with you pragmaticists b. Something terrible has happened. c. Could you hand me the red {*it, *this, one}? Abney’s insights gave rise to a range of studies of possible functional heads inside the nominal domain (for an overview of some of these, see Bernstein 2001, and cf. also Zamparelli 2000, Bernstein, this volume, Fukui and Zushi, this volume, and Longobardi, this volume). While the term ‘pronoun’ is still being infelicitously applied as a cover term, it has become clear that traditional ‘pronouns’ are highly heterogeneous (see for instance Déchaine and Wiltschko 2002). However, the majority of studies have focussed on syntagmatic relations involved in projection, feature-checking processes and movement and merger operations. In this article, I will address some crucial paradigmatic relations which are found in D-heads across the Germanic languages across more than a thousand years of language development. More specifically, I will argue that the forms listed in (2a-c) and (3a-c) are related, not only in terms of some trivial and obscure origin in proto-Germanic, but also in terms of shared synchronic morphological roots, viz. þ- and hw- roots, with similar distributional and semantic properties. (2) a. Modern English: the, this, that, there, then, thus, thence, thither, thou, they b. Modern German: das, dieses, da, dort, dann, denn, du c. Modern Danish: det, dette, da, der, did, de, du,

(3) a. Modern English: what, which, who, where, when, whence, whither, why b. Modern German: was, welches, wer, wo, wann, wie, warum c. Modern Danish: hvilket, hvo, hvad, hvor As preliminary validation of a level of unitary morphological analysis of the forms listed in (2), note that with the possible exception of ModE thou and the, ModG/ModD du, all the forms in (2) may be paraphrased by demonstrative constructions, e.g. there = at that place, then = at that time, thus = in that way, thence = from that place, thither = to that place, and they = those people/things. And as preliminary validation of the complementarity of the two morphemes, note that all the forms in (3) may be



Stating the case for þ- root and hw- root determiners 

paraphrased by an interrogative construction in which what or which is simply substituted for that, e.g. where = at what place, when = at what time, whence = from what place, whither = to what place, and who = what people. (2) and (3) offer only an incomplete picture of the distribution of the two þ- and hw- morphemes. As we shall see, for some of the forms there are contrasting case forms, such as thou, thee, thine, and gender variants in ModG and ModD. ModG has the threeway gender contrast in der[masc], die[fem] and das[neut]. ModD has a dual gender system with an opposition between neuter det, marked by the suffix {-et[neut]} and the merged masculine-feminine ‘common gender’ den, marked by the suffix {-en[com]}. Moreover, while technically belonging to ModE, thou and all its case forms are obsolete in contemporary English, and, similarly, while technically belonging to ModD, the nominative form hvo (who) is obsolete in contemporary Danish, in which the OD dative case form hveim has become generalized as ModD hvem (whom) in non-genitive cases. However, in spite of not constituting exhaustive lists, the forms provided in (2) and (3) suggest that the two þ- and hw- morphemes have a fairly wide distribution across the three languages. It should also be clear that the forms listed in (2) and (3) are highly frequent and that therefore the þ- and hw- roots are highly frequent. One crucial observation is that the forms listed all belong to closed functional classes, so the þ- and hw- morphemes do not distribute as morphemes in productive, open classes. This observation is undoubtedly the reason that the þ- and hw- based forms have not been treated as morphologically complex in traditional grammar. It is interesting to note that the morphological approach to the forms listed is perfectly uncontroversial in historical linguistics and comparative Germanics. Nevertheless, in traditional contemporary grammars, the morphemes are not recognized, only the full forms are recognized. And apparently outside circles of historical linguists, only very few linguists seem to operate with Germanic þ- and hw- roots. We need to add some further comments to (2) and (3). We note that the three modern languages employ different variants of the old pan-Germanic þ- and hw- roots. The original shared Germanic initial þ-, i.e. the voiceless spirant /θ/, has been retained in Modern Icelandic, but in English it became weakened into the voiced spirant /ð/ in the transition from Old English (OE) to later Middle English (MidE). In German and Danish it eventually developed from /θ/ in the old languages into /d/, most likely after an intermediate stage with /ð/ (Braune 1987, Prins 1972). So the graphemes þ, th and d as well as their corresponding phonemes /θ/, /ð/ and /d/ are variants of the same morpheme frozen at various stages of phonetic development with the oldest form surviving in Icelandic, the middle form surviving in English, and the – so far – youngest variant in the continental Germanic languages, including German and Danish. Interestingly enough, the only place in the Germanic languages where the voiced spirant /ð/ survives as an initial consonant is in the ModE forms listed in (2) and in the complementizer that, in the complementizer though and in the complementizer/preposition than. The complementizers are derived from lexemes which contained the same morpheme, but the forms have become recategorized as non-complex complementizer lexemes. The

 Alex Klinge

three graphemes wh-, w- and hv-, representing the two phonemes /w/ in English and /v/ in German and Danish, also developed as variants of the older Germanic form hw-. A couple of odd forms call for special mention: ModE who, i.e. /hu:/, or weak /hu/ and /u/, and how, i.e. /ha~/, do not follow the regular English pattern of initial /w/. It should also be noted that the forms listed in (2) and (3) are classified in traditional descriptive grammars such as Quirk et al. (1985) as ’articles’ and ’pronouns’ of various ilks. So, for instance, distributional facts have led to a strong tradition of distinguishing a category of articles, comprising the and an in English, and a category of demonstratives, comprising this and that. Distributional facts have also led to a strong tradition of distinguishing a set of ‘relative pronouns’, including which and who in English, and a set of ‘interrogative pronouns’, including which and who. The fact that these forms have been relegated to different syntactic (sub-)categories in the grammar has meant that they have also been treated as inherently different, and little effort has gone into considering what they in fact share. An obvious question is whether it is the case that there is only one functional item which, which occurs in a distributional range where it interacts with other morpho-syntactic phenomena to produce different semantic results? A null-hypothesis would take which to be the same functional item across its range of distributions, which would make the so-called ‘relative pronoun’ which and the so-called interrogative pronoun which one and the same functional item in two different syntactic environments. Similarly, German would only have one functional item der, which is found in three different distributions: (a) as a transitive or intransitive determiner head, i.e. a so-called ‘article’ or ‘demonstrative’; (b) as an intransitive determiner head in the anaphoric nominal in relative clauses, i.e. a so-called ‘relative pronoun’, and (c) as an intransitive determiner head, i.e. a so-called ‘pronoun’. Does German have one functional lexeme der with a range of different distributions, or does it have three different lexemes with the exact same phonetic shape? The proposal I intend to make in the following is that the functional items in (2) and (3) need to be considered together in terms of their striking similarities and in terms of their paradigmatic complementarity. A nominal phrase structure in terms of the general assumptions of Abney’s DP analysis makes it possible to consider the forms in (2) and (3) in terms of unitary formal features, unitary distributional potential and in terms of a level of unitary semantics. As a pre-condition of treating the, and unstressed das and det together with the other þ- based forms, it is necessary first to disentangle them from their alleged relation with the so-called ‘indefinite articles’ of a traditional two-member category of articles. This can only be convincingly accomplished by first presenting the argument that there is no such two-member category in the Germanic languages.

Stating the case for þ- root and hw- root determiners 



2. The fallacy of the category of ‘articles’ One of the most deeply entrenched grammatical categories of traditional grammar is the category of articles. It is widely assumed that a category exists in which two members are in paradigmatic opposition such that one encodes indefiniteness, i.e. an ‘indefinite article’, and the other encodes definiteness, i.e. a ‘definite article’. So, the argument goes, in English an is the ‘indefinite article’ and the is the ‘definite article’; in German, ein is the (masucline) ‘indefinite article’ and der is the (masculine) ‘definite article’; and in Danish en is the (common gender) ‘indefinite article’ and den is the (common gender) ‘definite article’. According to Quirk et al. (1985: 253): ”The most common and typical central determiners are the definite and indefinite articles, the and a/an, respectively.” The two articles may be illustrated by the cognate examples in (4) and (5). (4) a. She has read a[indef] new book. b. She has read the[def] new book. c. *She has read the a new book. (5) a. Sie hat ein[indef] neues Buch gelesen. b. Sie hat das[def] neueBuch gelesen. c. *Sie hat das ein neue Buch gelesen. The traditional analysis which suggests that English and German both have two-member categories of articles may be supported by the observations in (6).

(6) a. The two articles appear to have the complementary functions of marking an NP as either ‘indefinite’ or ‘definite’, i.e. their semantic raison d’être is to divide up the semantic notion of definiteness into contradictory opposites; b. the two articles are similarly prosodically weak; c. unlike for example this and dieser, the two articles have no independent distribution without a following noun (and so cannot be classified in the traditional sense as ‘pronouns’) – not counting of course constructions such as the more, the merrier; d. the two articles cannot co-occur in a DP, so they appear to be in complementary distribution, which, given that their semantics is in contradictory opposition, is a defining feature of paradigmatically contrasting members of the same grammatical category. The case for a category of articles which is distinct from a category of demonstratives would appear to be fairly good given the observations in (6). However, if we take a slightly different approach, it soon becomes clear that the category of articles is probably a fallacy following from unwarranted conclusions on the basis of the observations in (6). It is widely accepted that historically the so-called indefinite article developed from the cardinal quantifier through a process of grammaticalization at some point in

 Alex Klinge

the transition from the old Germanic languages to their modern versions (e.g. Heine 1997). The exception is Icelandic, which never developed an indefinite article. It is also widely assumed that the cardinal quantifier and the indefinite article parted ways and went on to do separate jobs as a quantifier and as an indefinite article. And given the observations in (6), the form that developed into an article apparently found a partner in a definite article, which itself had developed through a process of grammaticalization from the old Germanic demonstrative paradigms. Exactly why, how and when the two otherwise completely unrelated functional items should meet to form their own separate grammatical category remain murky points. The question we need to address here is whether perhaps they never really left their points of origin to come together as a coherent two-member category. First, there is the question of form. It is quite remarkable that in German and Danish in a syntactic constellation of ModG ein and ModD en and an N-head, the only formal difference between the cardinal quantifier and the alleged indefinite article is prosodic – the cardinal quantifiers receive independent stress, whereas the indefinite articles receive weak, dependent stress. So depending on prosodic contour, Danish common gender en and neuter et may both either be the cardinal quantifier or the indefinite article. This level of formal identity may be taken to be especially remarkable in the case of German, which has a three-way gender distinction and a four-way case distinction – so in other words the two lexemes would have 12 different forms which might have developed slight differences. If the indefinite article parted ways with the cardinal quantifier a thousand years ago, why should it have remained formally identical with the cardinal quantifier? If we turn to English, we do seem to have two formally distinct lexemes, viz. the cardinal quantifier one and the indefinite article an, and we would have to stretch credulity to bursting point if we tried to connect them phonologically (although just such a rule was formulated by Perlmutter 1970) as two distinct forms of the same lexeme. So while it is tempting on formal grounds to conflate the two prosodic forms into one lexeme in German and Danish, the English forms an and one resist such economy. In his comprehensive study of definiteness, C. Lyons (1999: 34–35) discusses the viability of conflating the English indefinite article and the English cardinal quantifier, but concludes that conflation is not an option, because “the differences are more substantial in English”. As C. Lyons correctly points out, an analysis which links the indefinite article and the cardinal quantifier as weak and strong variants of one lexeme would predict that if we assign contrastive stress to a or an, we should get one, which we do not (ibid). The historical facts of the two English forms are that the OE cardinal quantifier was ān and that the distinct ModE cardinal quantifier one was a late OE and early MidE development from southwestern and western English dialect (Barnhart and Steinmetz 1988). This suggests that one OE form developed into two distinct, and distinguishable lexemes, viz. ModE an and one. One important question is why the dialectical variation should only target one function, i.e. the quantifying function, and leave the other untouched. Is it a fact which might suggest that two distinct versions of

Stating the case for þ- root and hw- root determiners 



ān had already developed, one which was an article and another which was the cardinal quantifier? Or may it simply be the case that for independent phonetic reasons the change only targeted the stressed variant of ān? That is of course a question which we will have to leave open in this context. The important observation is that an and one had the same point of origin, viz. OE ān, which may already have had a weak and strong variant in exactly the same manner as Danish en. The outstanding question is whether we want to say that at some point an left the general paradigmatic relation with the cardinal quantifiers and other quantifiers to become a marker of indefiniteness and enter into a completely different paradigmatic relationship with a definite article. As C. Lyons (1999: 33–34) observes, mass nouns and plural nouns do not require an article to be understood as semantically indefinite. Consider his examples (ibid): (7) a. John has gone out to buy milk. b. I have already put spoons on the table. Strictly speaking, the bare N-heads of the NPs milk and spoons should be described as ‘non-definite’ rather than ‘indefinite’ because if we assume that a bare N-head is nondefinite, it can be overridden both by an inherently definite determiner, such as my milk and these spoons, or by an inherently indefinite quantifier such as some milk and enough spoons without contradiction. But the question is why, unlike Icelandic, English, German and Danish should require an indefinite article to handle distinctions of definiteness. Why should (8) not be as perfectly perspicuous as (7b)?

(8) *I have just put spoon on the table.

Another question which follows from (7) is that if an went its own way as an indefinite article contrasting with a definite article, then why has it failed after almost a thousand years to generalize to mass nouns and plurals? At this point, a traditional article analyst would have to point to the origin of an as a singular cardinal quantifier. So one argument counting against full generalization is that its origin as a numeral continues to make it incompatible with mass nouns and plurals. But why should its origin continue to affect it if it changed status from a quantifier to an article? A much better analysis would be simply to say that an is still to be analysed as a quantifier – not as an article. If we assume that an is still to be analysed as a quantifier, we can also explain why it occurs naturally where the syntactico-semantic context indicates that it is the head of a QP. Consider the following attested example:

(9) Actually, the Pedersen Zoo, thanks to Sarah and Kelly, consists of three horses, three dogs, two cats, a snake and a bird.

Here the so-called indefinite article would appear to have much the same distribution – though not exactly the same semantics – as the cardinal quantifier one, cf. (10).

 Alex Klinge

(10) Actually, the Pedersen Zoo, thanks to Sarah and Kelly, consists of three horses, three dogs, two cats, one snake and one bird. Quite clearly the first three conjoints of the coordination structures in (9) and (10) are QPs headed by cardinal quantifiers. Is there any reason for assuming that a snake and a bird should be simple (indefinite) NPs rather than QPs? If so, how do we explain that clearly simple NPs do not coordinate well with the three first QP conjoints, cf. (11)? (11) Actually, the Pedersen Zoo, thanks to Sarah and Kelly, consists of three horses, three dogs, two cats, and {?birds,?fish, fifteen birds, lots of fish, many lizards}. It would seem that (11) suggests that while an coordinates smoothly with the clear Qheaded conjoints, coordination is less felicitous – though probably not ungrammatical – with bare NPs. So an sides with the cardinal quantifiers. Moreover, if we assume that an in ModE is still to be analysed as a quantifier, then we are also in a much better position to say why it cannot co-occur with a whole range of other quantifiers, cf. (13). (12) We have a {*many, *some, *three, *several, *enough, … } birds. If we analyse an as a Q-head, we can simply say that an cannot co-occur with many, some, three, several and enough because they are in paradigmatic contrast as quantifiers – there is no reason to call it an article which happens to have retained traces of its origin as a quantifier. As already suggested in various ways by other writers, e.g. C. Lyons (1999), Vangsness (2001) and Wood (2003), the purpose of the weak forms of ein and en in German and Danish and of an in English is to mark singular grammatical or singular semantic number, whereas the purpose of the strong forms ein and en and of one is to mark numerical singular in contrast to two, three, etc. If we consider the facts of Germanic syntax, such an analysis makes a lot of sense. It would not be unreasonable to assume that determinate number is an obligatory feature of Germanic nominals because nominals have to incur number agreement relations with the head of IP as well as with any co-referential intransitive DPs, i.e. ‘pronouns’ in traditional grammar. As convincingly argued by Vangsness (1999 and 2001), an analysis of the weak forms of ein and en as well as an in terms of grammatical number marking tallies quite well with the hypothesis that OE, OD and OG did not need a quantifier to mark singular number because the inflectional paradigms for nouns in the old languages contained singular and plural suffixes which marked singular and plural number. According to such an analysis, it was the loss of the singular suffixes which meant that the cardinal quantifier was drafted in in a weak form to do the job of marking singular grammatical number, i.e. ‘one’ without the counting contrast. Vangsness (1999 and 2001) argues that such an analysis allows us to explain why Icelandic never developed an ‘indefinite article’, i.e. a weak non-counting singular form. There was no need because Icelandic never lost the singular suffixes.

Stating the case for þ- root and hw- root determiners 



One of the strongest traditional arguments that supports the analysis that there is a grammatical category of articles with an indefinite member and a definite member is that since an and the cannot co-occur, they must be in paradigmatic opposition within a grammatical category. The observation of complementary distribution is indisputably correct, but the conclusion is wrong. It is correct that the definiteness associated with the blocks its co-occurrence with an. However, this is not the result of a paradigmatic opposition within a category but the result of a syntagmatic opposition in phrasal projection. In other words, the reason that the and an cannot co-occur is that the semantic features ‘definite’ and ‘indefinite’ are contradictory, and a projection with the ‘definite’ feature cannot be merged with a projection with the ‘indefinite’ feature. The contradiction of features that blocks the and an similarly blocks co-occurrence of my and an, of this and an, of the and enough, etc., cf. (13a-d). (13)

a. b. c. d.

*She has read the[def] a[indef] book[non-def]. *She has read my[def] a[indef] book[non-def]. *She has read this[def] a[indef] book[non-def]. *She has bought the[def] enough[indef] pasta[non-def].

Note that given the unacceptability of (13b)-(13d), (13a) in itself does not necessarily constitute evidence for the conclusion that an and the form a two-member category. No-one I know of has so far claimed that blocked co-occurrence of the and enough should be taken as evidence for a two-member category. The correct conclusion on the basis of the observations in (13a-d) is that the, my and this are determiners which share the feature ‘definite’, whereas an and enough are quantifiers which share the feature ‘indefinite’. The simple generalisation is that a Dhead marked for the feature ‘definite’ cannot be merged with a QP complement with a Q-head marked for the feature ‘indefinite’ because the two features are in contradiction. A definite D-head can of course take a non-definite QP or a non-definite NP as its complement, cf. (14a-b). (14) a. She has read {my[def], the[def], these[def] } many[non-def] books[non-def]. b. She has read {my[def], the[def], these[def] } books[non-def]. In other words, the contrast between an and the is syntagmatic, not paradigmatic. However, the contrast between the quantifiers an and many, some, three, several and enough, which is evident in (13) above, is in fact paradigmatic.

 Alex Klinge

The syntagmatic and paradigmatic relations in a multiheaded nominal projection can now be summed up for Germanic on the basis of ModE as the example: (15) Basic architecture of Germanic nominals exemplified by ModE: DP D

QP Q

NP2 NP1

AP N

XP

A

the[def] *the [def] *that [def] *the [def] the[def] my[def] it[def]

new a[indef] new a[indef] new a[indef] new enough[indef] new many[non-def] new new

book [non-def] book [non-def] book [non-def] book [non-def] books [non-def] books[non-def] book [non-def]

I can now return an whence it came: an belongs in the paradigmatic class of quantifiers. There is no coherent two-member category of articles in English, nor is there in German or Danish. As a result, we also no longer need to analyse the, das and det in terms of a paradigmatic contrast with an, ein and et in a category of articles. The next step is to return the, das and det whence they came.

3. The þ- root determiners It is generally assumed that the weak variants, the so-called ‘definite articles’, of ModG das, ModD det, and ModE the had their origin in the demonstrative paradigms of the old ancestors of the modern languages. Als sicher weiß man nicht viel mehr, als dass der Artikel aus dem Demonstrativpronomen hervorgegangen ist. Über den Weg und die Etappen des Weges bis zum Artikel herrscht ziemliches Dunkel. (Hodler 1954: 13)

Stating the case for þ- root and hw- root determiners 



[Little else is known for sure than the article developed out of the demonstrative pronoun. The path and the stages leading to the article remain murky]

The assumption underlying Hodler’s conclusion is that the modern prosodically weak forms of the Germanic languages at some point became distinct lexical items, i.e. ‘definite articles’, parting ways with the strong, demonstrative forms. As we saw above, such an assumption may be problematical because apart from the difference in prosodic prominence, they are formally indistinguishable. But more importantly, we also saw that it is highly doubtful whether a coherent category of articles exists, so if the weak variant can no longer be called an article, how do we accommodate it? It might be counter-argued that the evidence for separate distinct forms is readily available in ModE. ModE has three distinct forms, viz. the, a definite article, that, a distal demonstrative, and this, a proximal demonstrative. The is clearly not merely a weak variant of the same lexeme as strong that, or strong this, for that matter. And since the three distinct forms this, that and the correspond semantically with ModG dieser/ModD denne and ModG strong der/ModD strong den, and, finally, ModG weak der/ModD weak den, then surely the strong and weak forms in ModG and ModD must be distinct forms as well. Though such an argument might seem persuasive, it is most likely wrong. In order to appreciate the situation in ModE, we first need to see how the paradigmatic contrasts work in ModG, OD, and ModD. We will find that these languages are remarkably similar in the way their so-called ‘demonstrative paradigms’ are constituted formally and semantically in terms of a formal two-way distinction which is interpretable in terms of a three-way semantic interpretation. There is every indication that OE followed the same two-way form distinction but three-way semantic distinction, so the outstanding explanation is why there is a formal three-way split in ModE, a question to which we will return later. ModG has a paradigm of forms which are clearly marked for the semantic feature proximal: (16) ModG proximal paradigm:

Nominative Accusative Genitive Dative

Masculine

Neuter

Feminine

Plural

dieser diesen dieses diesem

dieses dieses dieses diesem

diese diese dieser dieser

diese diese dieser diesen

In traditional grammar the forms are listed as morphologically non-complex. However, it should be obvious that the forms lend themselves to decomposition into the stem dies and some bound suffixes, given in bold, which mark gender, case and number. The paradigm suggests that the stem might be diese, but since neuter nominative and

 Alex Klinge

accusative are often reduced to dies in spoken German, we assume that the shorter form is the stem. ModG has a contrasting non-proximal paradigm. (17) ModG non-proximal paradigm:

Nominative Accusative Genitive Dative

Masculine

Neuter

Feminine

Plural

der den des/dessen dem

das das des/dessen dem

die die der der

die die der den/denen

If we accept that the forms in (16) are complexes of a stem and a set of suffixes marking gender, case and number, then, similarly, we can identify a set of almost identical suffixes marking gender, case and number in (17). The masculine set is identical, the neuter nominative and accusative differ minimally in the epenthetic vowels of the suffix. However, the most obvious difference lies in the phonemes which realize the feminine and plural nominative and accusative, i.e. proximal /∂/ vs. non-proximal /i:/. However, the variation is still fully within the bounds of what we might reasonably consider to be variants of the same morpheme – possibly weakened in the second nonstressed syllable of the proximal paradigm. We concluded that the forms of the proximal paradigm are composed of a stem, viz. dies and a set of gender, case and number suffixes. Since we have the same set of suffixes in the non-proximal paradigm, we also have a stem to which they attach, viz. the morpheme d-. Given this observation, we can go back and further decompose the stem of the proximal paradigm into a root, viz. d-, and the stem-forming bound morpheme -ies. Now we can assume that the two paradigms share the root morpheme d-, and that the bound morpheme -ies encodes the marked proximal feature of the proximal paradigm. The root morpheme d- is a consonant, so it requires a supporting epenthetic vowel to form a stem to which the suffixes can attach. By default we might set the epenthetic vowel as the minimal form /∂/, which may then be overridden by stronger vowels drawn from the suffixes. The epenthetic vowel is fixed as /i/ in the proximal stem dies. In the non-proximal forms, the vowel is drawn from the various suffixes – feminine and plural nominative and accusative are marked only by the vowel. If we abandon the idea that the two paradigms are constituted by simplex forms and adopt the complex morphological analysis, we can re-write the two paradigms as compositional forms consisting of a d- + vowel root, a d- + vowel stem, a dies-stem and some inflectional suffixes, cf. (18). Note that the schema indicates that in the non-

Stating the case for þ- root and hw- root determiners 



proximal ‘demonstrative’, the epenthetic vowel is drawn from the suffixes whereas it is fixed in the proximal ‘demonstrative’. (18) The composition of ModG ‘demonstratives’: Stem Root

Inflex.

Non-prox. dem.

d-

epenthetic vowel -{...}-

Proximal dem.

d-

-ie-

Prox.

-s

Gender/case/ number -{er, en, as, es, em, er, en, ie} -{er, en, es, es, em, er, en, e}

A compositional approach to ModG ‘demonstratives’ along the lines proposed is by no means new or original.  The etymological complexity of dies and its cognates in the other Germanic languages is well-known in historical linguistics and can be traced back to the early forms of the Germanic languages. A detailed and highly convincing compositional analysis of the ModG forms based on the d- morpheme was recently proposed by Wiltschko (1998: 149), who also refers to similar compositional analyses by Koopman (1993), Olsen (1991) and Tappe (1990) (cf. interestingly enough, while Eisenberg (2004: 170) is happy to treat dies- as a stem, he leaves the status of d- open: “Ob man d den Status einer morphologischen Einheit grunzätzlich absprechen kann, lassen wir an dieser Stelle offen” 2004: 176 [whether it is possible to deny d the staus of a morphological unit will be left open here][my translation]). Various compositional analyses have also been proposed for the cognate English th- forms, e.g. (Cotte 1993), (Langacker 2001), (Wiltschko 1998) and (Déchaine and Wiltschko 2002). We will return to some of those when we discuss ModE below. Consider now the cognate forms as they are rendered in paradigms for OE and OD. (19) OE proximal ‘demonstrative’ (Mitchell and Robinson 2001: 18):

Nominative Accusative Genitive Dative

Masculine

Neuter

Feminine

Plural

þes þisne þisses þissum

þis þis þisses þissum

þēos þās þisse, þisre þisse

þās þās þissa, þisra þissum

 Alex Klinge

(20) OE non-proximal ‘demonstrative’ (Mitchell and Robinson 2001: 18): Masculine

Neuter

Feminine

Plural

Nominative Accusative Genitive

se þone þæs

þæt þæt þæs

sēo, sīo þā þ5re

þā þā þāra, þ5ra

Dative

þ5m, þām

þ5m, þām

þ5re

þ5m, þām

While the paradigms are not quite as regular as ModG, we can nevertheless discern a þ- root and a þ + vowel + s stem, although unlike in ModG there are variants of the epenthetic vowel in the OE proximal paradigm. The two most striking odd forms are masculine nominative se and feminine nominative sēo/sīo. According to Lass (1992: 112), generalized þe is found as a variant of masculine, nominative se as early as the tenth century, so se and sēo/sīo may be seen as a vestige of earlier Indo-European forms (Lass 1994: 143). Turning now to OD, we find the following forms: (21) OD proximal ‘demonstrative’ (Andersen 1936: 63):

Nominative Accusative Genitive Dative

Masculine

Neuter

Feminine

Plural masc. neut. fem

þessi þenna þessa þessum

þetta þetta þessa þessu

þessi þessa þessar þessi

þessir, þessi, þessar þessa, þessi, þessar þessum þessa

We note here the irregularity of masculine accusative singular and neuter nominative and accusative singular, which do not have the expected vowel + s morpheme marking proximity. (22) OD non-proximal ‘demonstrative’(Andersen 1936: 63):

Nominative Accusative Genitive Dative

Masculine

Neuter

Feminine

Plural masc. neut. fem

sā þann þess þeim

þat þat þess þvi

sū þā þeirar þeiri

þeir, þau, þær þā, þau, þær þeim þeirar

The three-way gender distinction of OD was reduced to a two-way distinction between common gender and neuter, and except for the genitive the gender distinctions were lost in all but a few determiners. The OD dative survives as an objective case

Stating the case for þ- root and hw- root determiners 



merging accusative and dative in the demonstrative non-proximal determiner when it is used intransitively without a following NP, cf. (23). So the two contrasting paradigms in ModD are given in (23) and (24): (23) ModD proximal ‘demonstrative’:

Non-genitive Genitive

Common

Neuter

Plural

denne dennes

dette dettes

disse disses

With a few exceptions, OD generally followed the established Germanic proximal stem in having the stem þes, but it only survives in the ModD plural stem since the surviving forms in ModD are the OD masculine accusative form þenna and the neuter nominative/accusative þetta. (24) ModD non-proximal ‘demonstrative’: Common Subj./obj. Genitive

den dens

Neuter det dets

Plural de/dem (intrans.) deres

Again we see that it was the OD accusative forms þann and þat which survived. This also meant that Danish rid itself of OD masculine sā and feminine sū, the older irregular forms not based on þ- roots. While it is not possible here to uncover all the details, we can draw the following conclusions about the basic Germanic contrast on the basis of the picture which is emerging: (25) The basic Germanic contrasts: a. the Germanic languages we have discussed so far have ‘demonstrative determiners’ based on the three variants /θ/, /ð/ and /d/ + epenthetic vowel of the ancestral Germanic þ- root; b. if we assume that by definition a morpheme is recognized as a stable, predictable correlation between form and meaning, it should be possible to identify the semantics which is associated with the Germanic þ- root; c. if we disregard the gender, case and number suffixes, throughout the Germanic languages discussed so far, the persistent contrast is one between an unmarked þ- root + epenthetic vowel and a marked þ- root + epenthetic vowel + -s morpheme, e.g. OE masculine accusative singular nonproximal þ-o-(ne) as against proximal þ-i-s-(ne);

 Alex Klinge

d. if the formal contrast is binary, our point of departure is also to assume that the basic semantic contrast is binary and follows from the presence vs. absence of the –s morpheme. We have now seen how ModG, ModD, OE and OD all have a basic formal contrast between a þ- root serving with an epenthetic vowel as a stem for gender, case and number suffixes, and a þ-vowel-s stem serving with an epenthetic vowel as a stem for gender, case and number suffixes. We have focussed on forms in the nominal domain, but as already pointed out in the introduction, it would of course be possible to identify further forms based on the same morpheme outside the nominal domain, such as the so-called demonstrative adverbs ModG da (there) and ModG dann (then), ModD der (there) and ModD da (then), etc. As we have seen, these may all be given demonstrative paraphrases, e.g. ModG da (there) = at that place, ModG dann (then) = at that time. Moreover, so far the question of the status of the so-called 2nd person ‘personal pronoun’ ModG/ModD du (you) has not been addressed, but, as we shall see, it may well be based on the same morpheme. And up to this point, little has been said about ModE and about the finer points of the semantic contrasts. We will come back to those concerns after we have considered determiners based on the hw- root.

4. The hw- root determiners As we have seen in the previous sections, there is no coherent category of articles. We have also seen that there is plenty of circumstantial etymological and cross-linguistic evidence that in fact we need to operate with a Germanic þ- root. In a similar way we can identify a Germanic wh- root. If we consider the ModG paradigm for so-called ‘interrogative pronouns’, it is quite clear that they contain the same suffixes we found in the German ostensive determiner d- root, which in the case of the ‘interrogative pronouns’ leaves a w- root. The only difference is that the paradigm based on the w- root does not have plural forms and it is not based on a three-way gender distinction, but on a distinction between ‘human’ and ‘non-human’ referents. ‘Human’ referents are signalled by masculine suffixes and ‘non-human’ referents are signalled by neuter suffixes. (26) ModG ‘interrogative’ paradigm

Nominative Accusative Genitive Dative

‘human’

‘non-human’

wer wen wes/wessen wem

was was wes/wessen wem

Stating the case for þ- root and hw- root determiners 



We see that the suffixes are totally identical to those found in the þ- based paradigm, even down to intransitive dessen and wessen as against transitive des and wes. So we can indisputably identify a w- morpheme in a paradigm contrasting with the paradigm based on the d- morpheme. The two morphemes are clearly roots which form stems to which inflexional suffixes attach. Similarly, if we turn to OE, we find the same suffixes as in the ostensive þ- based paradigm, but here of course they attach to a stem formed by the OE hw- root, cf. (27). (27) OE ‘interrogative’ paradigm:

Nominative Accusative Genitive Dative

‘human’

‘non-human’

hwā hwone hwæs hw5m/hwām

hwæt hwæt hwæs hw5m/hwām

The situation in contemporary Danish is that hvem (whom) is the generalized non-genitive form, and hvis (whose) is the genitive form. Hvem is derived from the OD dative form hveim, which perfectly matched the OD masculine dative ostensive determiner þeim. A set of modern Germanic variants of the old Germanic hw- morpheme has now been established in the nominal domain. Again the focus has been on forms in the nominal domain, but it would of course be possible to identify further forms outside the nominal domain, such as the so-called interrogative adverbs ModG wo (where) and ModG wann (when), ModD hvor (where), etc. In ModE there is still a perfectly transparent morphological relationship between there-where, then-when, thencewhence and thither-whither. We will now turn our attention to the semantic and pragmatic aspects of the þand hw- roots. It is by bringing the formal observations together with the semantic description that we will be able to appreciate the full implications of operating with the two morphemes. More particularly, once we understand the semantics which cuts across the Germanic languages, we can also begin to understand the three English forms the, this and that.

5. The basic contrasts of the nominal þ- based forms and the case of Modern English Since we have identified a þ- root and a þ-vowel-s stem, our concern is now to identify the basic semantics which cuts across the forms based on the morpheme. The clearest semantics is found in the marked proximal forms because we can readily identify the basic semantics of the þ-vowel-s stem, including its variants in Danish denne and dette.

 Alex Klinge

(28) a. b.

ModG: Ich ziehe dieses alte Buch von Kurt Vonnegut vor. (I prefer this old book by Kurt Vonnegut.) ModD: Jeg foretrækker denne gamle bog af Kurt Vonnegut. (I prefer this old book by Kurt Vonnegut.)

In possible utterances of the examples given in (28) the speaker does two things: she points out an entity she has in mind, properly denoted by the noun lexemes Buch (book) and bog (book), and she provides the search information that the entity is proximate relative to her own location either in the physical world or in discourse space. We can now consider minimally contrasting examples with the þ- + vowel stem. (29) a. b.

ModG: Ich ziehe das alte Buch von Kurt Vonnegut vor. (I prefer that/the old book by Kurt Vonnegut.) ModD: Jeg foretrækker den gamle bog af Kurt Vonnegut. (I prefer that/the old book by Kurt Vonnegut.)

Note that in both cases there are two possible interpretations of das and den. The two interpretations are brought out in the two alternative English translations. One interpretation is that the entity pointed out by means of Buch and bog is distal relative to the speaker in space or discourse, the other interpretation is that the entity is pointed out but not located relative to the speaker, i.e. it is deictically neutral. The two interpretations share the feature that they are not specified for proximity. In a written version of the sentences in (29), the hearer’s assignment of a distal or deictically neutral interpretation depends on the wider context of interpretation. In spoken discourse, however, the speaker can guide the interpretation through prosodic contour. Prosodic demotion of the determiner through weak stress correlates with the neutral, non-deictic interpretation (see Fretheim and Amfo, this volume, for some details of Norwegian). In fact, in modern spoken Danish the proximal form denne is becoming obsolete and is being superseded by den her, with den doing the pointing out, and her giving the proximal deicitic location, which reveals the semantic composition of denne. According to the argument presented so far, what we have in ModG and ModD is a binary formal and semantic contrast, i.e. between proximal and non-proximal, but a ternary interpretive contrast, i.e. between proximal, non-proximal distal and nonproximal neutral. The full set of meaning contrasts, semantic and pragmatic, can be rendered in terms of the tree in (30). (30) Ostension and meaning contrasts: ostension proximal

non-proximal neutral

distal

Stating the case for þ- root and hw- root determiners 



Since the bare þ- + vowel and the þ-vowel-s stems share the semantics of ‘pointing out’ an entity, i.e. an ostensive act, I will assume that the semantics encoded in the þ- root itself is ostensive, which is of course also an obvious semantics for þ- based forms outside the nominal domain, e.g. ModG da (there) and dann (then). The full implications of an ostensive semantics in the nominal domain will be further elaborated below. As a point of terminological clarification, we need to distinguish a simple ostensive act from an ostensive act + a deictic instruction. The traditional term ‘demonstrative’ covers the complex semantic notion of an ostensive act + a deictic instruction. As a consequence of the morphological analysis, ‘demonstrative determiners’ should be renamed ‘ostensive determiners’ because what they share is the ostensive þ- root. So, using ModG and OE as examples, we can represent the full set of contrasts encoded in the bare þ- root and the þ-vowel-s stem as the following three levels: (31) The basic semantic contrasts of ModG and OE ostensive determiners: Semantic field: Semantic contrast: = coded level Interpretable as:

ostension proximal ModG: dieses OE: þis proximal

non-proximal ModG: das OE: þæt neutral

distal

The representation in (31) can be spelled out as a chain of claims. The general semantic field within nominal determination which is encoded in the þ- morpheme itself is the field of ostension, i.e. pointing out a discourse entity. This, as we shall see below, is the level at which the þ- morpheme contrasts with the hw- morpheme. The vowel + s morpheme adds ‘proximity’ as the marked semantic feature of the complex stem þ + vowel + s, which means that the interpretation of the complex stem is locked as proximal, providing the basic contrast between proximal and non-proximal. It follows from the general principles of markedness that the marked semantic feature of a feature opposition carries the specific meaning which is inalienable, so the meaning of dieses is locked as proximal. In contrast, the unmarked ostensive root which forms the non-proximal stem þ + vowel may be interpreted in context as either deictically distal or deictically neutral, as illustrated also in (30) above. We cannot know for sure whether this in fact was a clear-cut set of contrasts in the old languages. Lass (1992: 113) warns us that “the semantics of the Old English opposition between the proto-definite article sē/sēo/þæt and the ‘emphatic demonstrative’ þes/þis/þēos is not entirely clear [ … ].” Indeed the lack of native-speaker informants makes it extremely difficult to know for sure exactly what the contrasts were. However, given the observation that the distinctions are fairly solid in ModG and ModD, it is certainly not unreasonable to expect that a similar set of contrasts were observed in the old languages, including OE. I will return below to a possible semantic

 Alex Klinge

explanation why this should be the basic set of contrasts. The outstanding question is, however, how the observations made, and the conclusions drawn, fit ModE. If we leave the status of thou and they to one side for a moment, ModE differs from its modern Germanic siblings in having a ternary formal contrast in its þ- based nominal determiners. Given the basic Germanic contrasts in (31) above, we can easily see that the ternary distinction of ModE simply conflates the formal level with the interpretation level of the other Germanic languages. The interpretation level of the other Germanic languages, including OE, is simply formally coded in ModE. (32) The basic semantic contrasts of ModE ostensive determiners Semantic field: Semantic contrast: Interpretable as: = coded level

ostension proximal proximal ModE: this

non-proximal neutral ModE: the

distal ModE: that

How do we explain the difference between ModE and the other Germanic languages considered here? If we adopt the morphological analysis proposed, we can postulate a fairly straightforward explanation for the ModE state of affairs. The two paradigms of OE ostensive determiners are provided in (19) and (20), above. While conceding that it is based on informed conjecture, I have assumed that just like its ModG and ModD counterparts, the non-proximal OE paradigm in (20) was unmarked ‘non-proximal’ and did dual service as a distal ostensive determiner and a neutral ostensive determiner. The context of utterance determined the interpretation. And, again by analogy, we might assume that the speaker could guide the hearer’s interpretation through prosodic contour, i.e. the assignment of weak stress signalled neutral interpretation and the assignment of strong stress signalled the distal interpretation. Given the fact that the gender and case distinctions of OE were gradually lost through late OE and MidE, the morphological analysis also gives us a basis for arguing that the gender and case distinctions of the ostensive determiner of OE were lost. Such a loss can either manifest itself in one form surviving as generalized and other forms disappearing, or as a levelling of all forms towards a generalized reduced form. The reduction process from OE to ModE is totally transparent in the proximal paradigm, where the loss of gender and case distinctions led to a reduction to the proximal stem with a generalized vowel, viz. þ + vowel + s = ModE this (with the subsequent /ð/ variant of the þ- morpheme). A similar analogical reduction affecting the non-proximal paradigm is in fact also totally transparent because, ex hypotesi, it should simply led to a reduction towards the stem, i.e. þ + epenthetic vowel. If we assume a maximal reduction and similarly generalize the epenthetic vowel, we arrive at ModE the, phonetically represented as /ð∂/ or

Stating the case for þ- root and hw- root determiners 



/ði/. However, we might further assume that this drastic phonetic reduction towards the stem took place only in the prosodically weak distribution of the non-proximal ostensive determiner, which, as we have seen, is confined to its occurrence in DPs with a deictically neutral interpretation. In prosodically prominent distribution, the full singular neuter form þat was generalized and became ModE that (again with the subsequent /ð/ variant of the þ- morpheme). The prosodically prominent distribution was of course with a non-proximal, distal interpretation in transitive D-NP sequences, or as a non-proximal intransitive D-head, or ‘pronoun’ in traditional terms. If this is a viable interpretation of events, we can also explain why that became a marker of nonproximal distal deixis while the became a marker of non-proximal, neutral ostension. They are the results of two different effects in different prosodic and semantic environments of the loss of gender and case distinctions from OE to ModE. Obviously, the proposed development towards ModE this, the and that is based on correlations with what we can observe in modern Germanic languages, and it clearly calls for more detailed studies of texts reflecting the process through some four hundred years – a task far beyond what can be done here. What is argued here is that the fact that English has the three forms the, this and that, is a consequence of the basic Germanic system of formal and semantic contrasts and of the special development of English, not the yardstick by which we should analyse the other Germanic languages. Just like ModG dieses and das, ModE the, this and that are based on the þ- root and they compete for the same syntactic slot as D-heads. They contrast paradigmatically in terms of the ostensive instructions they convey. Now I have provided a plausible account of the formal inventory and the basic semantic distinctions. It is now necessary to spell out what is meant by ‘ostension’ and ‘ostensive instruction’. Once we understand the semantics of the þ- root, we can also see how it contrasts paradigmatically as a D-head with the hw- root.

6. Some previous proposals The point I have made in the previous sections is that there are very good formal and etymological grounds for operating with a Germanic þ- root and a Germanic hw- root. If we accept the formal evidence in favour of a morphological analysis, our next step has to be to spell out the semantic content which is associated with the two morphological forms, i.e. what is their dimension of contrast? So far the focus of interest has been the distribution of the morphemes as D-heads (with or without gender, case and number suffixes) in the nominal domain. We will now consider the semantic basis of the two roots, which we must assume is apparent throughout their distribution, also outside the nominal domain. Various proposals have been made for the semantics of forms based on the þ- and hw- roots. I will briefly consider some of the possibilities already on offer.

 Alex Klinge

In their compositional approach to pronouns, Déchaine and Wiltschko (2002: 410, 422) recognize ModE th- as a morpheme, and they argue that it is a bound Dmorpheme associated with definiteness. It should be noted that their claim is that the entire class of pronouns qualifying as pro-D heads are inherently definite, so their claim equally covers I, we and th-, and is not made specifically for the th- morpheme. Below I will argue that definiteness is a semantic feature derivatively associated with the th- morpheme, not an inherent feature. Note that the paradigmatic contrast between þ- and hw- roots would also not be explainable in terms of a contrast of definiteness – the contrast between ModG der and wer cannot be accounted for as a contrast of definiteness. Moreover, as argued in detail by Bernstein (this volume), the ModE th- morpheme may be used without being associated with definiteness, which we would not expect if definiteness were the raison d’être of the root morpheme, cf. (33). Moreover, although Déchaine and Wiltschko do not discuss wh- as a morpheme contrasting with th-, it is noteworthy that the ModE wh- morpheme seems to distribute equally well in definite and indefinite contexts. This is apparent in existential thereclauses, which allow only indefinite interpretations. (33) a. There was this boy who studied linguistics … b. There was what boy in the garden? c. *There was which boy in the garden? Bernstein (this volume) analyses the ModE th- morpheme as a 3rd person marker unspecified for gender and number. Such an analysis seems to be supported by the compositional analyses of the Germanic þ- based paradigms above, e.g. ModD den = d[3rd pers] + en[masc, sing]. Moreover, there is little doubt that the locus of person distinctions tends to be the D-head (see also Longobardi, this volume). However, in the light of the argument which is being developed here, there are certain problems which follow from making 3rd person marking the raison d’être of the th- morpheme. First of all, as Bernstein herself notes, “a common noun must always be 3rd person even though there is no morpheme that can be identified as such.” This means that unless a nominal contains a 1st or 2nd person marker, the nominal will be associated with 3rd person by default. It follows that the NP individuals, the DP somebody, the QP many individuals and the DP which individuals are all as much 3rd person as those individuals and the individuals. How do we disentangle the semantic person feature of the th- forms if it is there by default? Moreover, if it is there by deafault, how do we account for the redundant 3rd person feature introduced by th-? We also saw above how ModG has a totally transparent morphological correlation between the d- forms der, den, dessen, dem and the w- forms wer, wen, wessen, wem. Both sets of forms are 3rd person, but they are certainly not synonymous, so some other meaning contrast must inhere in d- and w-. In other words if d- is taken to be a carrier of the 3rd person feature, then what is w-? A further problem arises with the þ- based forms outside nominal distribution. If we assume that the th- morpheme encodes 3rd person, we would either have to extend



Stating the case for þ- root and hw- root determiners 

the concept of person to the adverbial domain in there, then, thence, thither and thus, or argue that the non-nominal forms do not contain the morpheme. Moreover, we would find it hard to provide a coherent compositional account of the contrasting forms therewhere, then-when, and cf. ModG dann-wann. A final problem is that, subject to further scrutiny, we may want to include ModE thou and ModG/ModD du in the forms based on the þ- root. So like definiteness, 3rd person might well be associated with th- derivatively, or by default, but there must be more to the þ- root than person. In his proposal for wh- as a morpheme, Langacker (2001: 137) observes that the formal similarities of ModE interrogative and relative wh- forms is no mere coincidence and that they must be related through some notional conceptual basis. He also argues that wh- and th- forms are related in a non-trivial way. He characterizes the shared conceptual basis as “some general notion of selecting from a range of alternatives.” (2001: 139). He goes on to discuss interrogative and relative distribution of the wh- morpheme in terms of the notion of choosing from a range of alternative potential referents, or, in the case of whether from a range of alternative propositions. Langacker’s notion of choice from a range of alternative referents is primarily intended as a potential candidate for the semantics of the hw- morpheme. However, it is precisely the kind of semantics that would fit the picture which is being developed here. If we take the traditional perspective of historical linguistics, the þ- based forms are usually summed up as ‘demonstrative’, which is a fairly general characterization of their shared kinship, e.g. Shipley (1984). In the introduction to this article, we also saw how most of the forms lend themselves to a demonstrative paraphrase. But it has also been argued here that ‘demonstrative’ may be broken down further into the primitive semantic notions of ostension, i.e. pointing out, and deixis, i.e. providing a location relative to the speaker. This is what we might explore further to arrive at the semantic field in terms of which the þ- and hw- morphemes contrast. We will approach the two morphemes as two complementary ways referent selection may be guided by the speaker.

7. Towards a procedural ostensive semantics In an article about deixis as a source of reference, Lyons (1975: 61) proposed that we can understand the ontogenesis of reference “on the basis of a prior understanding of the deictic function of demonstrative pronouns and adverbs in what might be loosely described as concrete or practical situations.” Lyons draws a parallel between ostension in non-linguistic communication with the grammar of Quasi-English. Lyons’ approach suggests that many of the features which have been associated with the, this and that follow from the basic processes of non-linguistic ostension. Following this insight by Lyons, I have suggested elsewhere (Klinge, 2006) that simple ostension is the key to understanding the unitary procedural semantics encoded in the shared th- morpheme of the, this and that. If we extend that to the full range of English forms which are based on the þ- morpheme, ostension will involve

 Alex Klinge

linguistically pointing out an ‘entity’ (the, this and that), a ‘place’ (there, thence, thither), a ‘time’ (then) or a ‘manner’ (thus) (and of course ‘person’ if we include thou and they, see further below). We will begin by outlining the basic features of a non-linguistic act of ostension. A communicator who engages in a non-linguistic act of ostension in the physical hereand-now of the instance of communication, for instance by pointing to an entity, a place or a person, does so in order to direct the addressee’s attention to a specific entity, place or person which the gesturer has in mind and which she wishes to raise to the addressee’s attention because she believes that the addressee will draw some intended inferences from having the specific entity, place, manner or person raised to attention. For ease of reference, in the following I will subsume entity, place, manner and person under the term ‘entity’. Using the general framework and terminology of relevance theory, we can say that in directing the addressee’s attention to the specific entity she has in mind, the speaker intends the addressee to derive some cognitive effects beyond the mere recognition of the entity. The addressee combines already existing assumptions with the assumptions that the act of communication gives rise to in order to derive ‘contextual effects’ in the form of new, strengthened or cancelled assumptions. For an act of ostensive communication to succeed, certain presuppositions must be met (cf. Klinge, 2006). (34) A successful ostensive act in a physical here-and-now environment presupposes that: a. the entity pointed out by the gesturer exists; b. the entity may be identified by the addressee; c. the gesturer has a communicative intention to produce contextual effects in the hearer by raising the entity to attention by individuating it relative to other referents potentially available to the addressee. Since it is an act of communication, it will only be successful if it gives rise to certain contextual effects. (35) A successful ostensive act in a physical here-and-now environment entails that: a. the addressee recognizes the gesturer’s intention to communicate through the ostensive act; b. the addressee accesses the set of assumptions available in his cognitive context to identify the particular entity the speaker wishes him to raise to attention; c. the addressee accesses the identified entity as a constraint on the contextual effects that may be derived from the ostensive act, and derives the intended effects. Note that, as I have already argued, in the unmarked case, pointing out is away from the speaker, i.e. non-proximal – statistically, the potentially interesting discourse referents



Stating the case for þ- root and hw- root determiners 

which are away from the speaker far outnumber the discourse referents which are proximal to the speaker. The basic system of the Germanic languages reflects this iconically: proximity is both semantically and formally marked, non-proximity is unmarked. We can now transfer these observations from a physical here-and-now environment to communication through a universe of discourse created by linguistic communication. I will assume that the þ- morpheme is a linguistic marker of an ostensive instruction in discourse. Its semantic purpose is to indicate that the speaker has a specific entity in mind which she instructs the hearer to identify. This means that it encodes a procedure rather than a concept, i.e. its semantics is procedural rather than conceptual (in the sense of Blakemore 1987 and 2002). It follows from (34) and (35) that the act of ostension itself carries an existential presupposition, which has often wrongfully been associated with the ‘definite article’. The existential presupposition of ostension holds across all the forms based on the þ- morpheme in contexts which do not explicitly cancel it because they encode ostensive acts in various domains. The act of ostension itself also presupposes the accessibility or identifiability often associated with the ‘definite article’ and the ‘demonstrative pronouns’ (I assume that identifiability is a special instance of accessibility where the specific discourse referent is given in the discourse, whereas accessibility allows the hearer to construct the specific discourse referent, see further below). The accessibility or identifiability presupposition holds across all the forms based on the þ- morpheme. The ostensive instruction itself instructs the hearer to identify the specific entity the speaker has in mind in order to derive the contextual effects intended by the speaker. All the forms based on the þ- morpheme communicate to the hearer that he is to access the specific entity the speaker has in mind. So in all their basic interpretations and in their basic distribution, all forms based on the þ- morpheme share these procedural semantic features. I cannot hope to cover the full distribution of all the forms based on the þ- morpheme in all the Germanic languages, but I can provide a fairly basic sample which illustrates the ostensive instruction encoded by the þ- morpheme, cf. (36). (36) a. She found {the book, them}. b. She found {this book, that book, this, that}. c. She found me {there, then, thus}. In (36a) the speaker instructs the hearer to identify the specific entity or entities she has in mind because it is from the specific entity or entities the intended contextual effects are to be drawn. Unmarked ostension is away from the speaker, so both neutral the and neutral them will be assigned the unmarked non-proximal reading. Note that it is the ostensive act itself which presupposes specificity and accessibility of the referents, not any semantic features of the or them. So the and them only imply specificity and accessibility indirectly via the basic ostensive semantics encoded in the th- morpheme. It is here that thou and du might actually fit the picture. We note that again we have a root forming a stem to which case can attach, cf. ModG du[nominative], d[ichaccusative], d[eingenitive], d[irdative] as against irregular ich[nominative], m[ichaccusative], m[eingenitive], m[irdative].

 Alex Klinge

The ostensive þ- morpheme would actually be ideal for pointing out the only nonproximal discourse participant whom the speaker can systematically specifically point to in the discourse context. The 2nd person plural may involve one or more referents who are not locatable relative to the speaker, so the þ- morpheme is not used in the 2nd person plural forms. Given the fact that as a point of departure the so-called ‘personal pronouns’ in Germanic languages are in fact a motley collection of forms, we should not be deterred from pursuing the idea that thou and du are in fact based on the ostensive þ- morpheme. I will, however, leave the status of thou and du unresolved and postpone any conclusions until such a time as good etymological evidence can be found. In (36b) this and that add a deictic feature to the ostensive instruction which locates the entity relative to the speaker’s location at the time of utterance. So this and that are used to identify entities which are identifiable relative to the speaker at the time of utterance, which is what Hawkins (1978: 154) called the ‘matching constraint’ (see further in Klinge, 2006). It follows that an ostensive act involving this and that presupposes identifiability, because if the hearer cannot identify the referent, he also cannot locate it relative to the speaker. In contrast, the merely requires accessibility. In (36c) the speaker instructs the hearer to identify a specific place, a specific time or a specific manner which she has in mind from which the intended contextual effects are derivable. While ‘familiarity’ has been one of the most persistent semantic features associated with the definite article, it should be clear that the semantics of the th- forms provided in (36) is not ‘familiarity’ per se. A speaker who instructs a hearer to identify the specific entity, place or time she has in mind as a prerequisite to deriving the intended contextual effects, must also assume that the hearer is capable of actually identifying the entity, place or time – otherwise the intended contextual effects cannot be derived and the intended communication will fail. It should also be quite clear that the analysis proposed here for the th- morpheme does not allow pride of place to any discourse organization effects of the and an (contrary for instance to Heim’s ‘file change semantics’ 1988) – any such effects are derived, which is corroborated by the finding in Fraurud (1990) that the correlation between definite DPs and first mention was much higher than we should expect if the primary purpose of the was to pick up already textually activated discourse referents. We now need to contrast the semantics proposed for the forms based on the þmorpheme with forms based on the hw- morpheme. If we construct sentences with wh- forms in situ, we can actually compare sets of minimal pairs. (37) a. She found them. b. She found {what, who}? (38) a. She found me there. b. She found me where? By uttering the (a) sentences, the speaker instructs the hearer to identify the specific entities or place that the speaker has in mind and insert them into the propositional



Stating the case for þ- root and hw- root determiners 

form the hearer develops on the basis of the utterance of the sentences. In other words, the speaker predetermines the entities or place included in the completed propositional form. In the (b) examples, on the other hand, the wh- forms introduce the entities and the place as variables which the hearer has to fill on the basis of a range of appropriate specific entities or place which will make the propositional forms contextually relevant. Thus, the hearer determines the entities or place included in the completed propositional forms. It is in this contrast between the speaker and the hearer determining the appropriate referents of the propositional form that we find both the semantic raison d’être and the paradigmatic contrast of the þ- and hw- morphemes. But why not simply follow the tradition which says that the þ- based forms are ‘demonstrative’ and the hw- based forms are ‘interrogative’? While I have argued that the þ- based forms are indeed ‘ostensive’ (rather than ‘demonstrative’), it is not the case that ‘interrogative’ will account for the full distribution of the hw- based forms. It is important to note that hw- based forms also distribute as bound and free nominal or adverbial ‘interrogatives’ and ‘relatives’, cf. (39) and (40). (39) a. She found the persons who were responsible. (bound ‘relative’) b. She found me at the square where the ancient Romans used to stage chariot races. (bound ‘relative’) (40) a. She knew who was responsible. (free ‘interrogative’) b. I will call you when I arrive in Wellington. (free ‘adverbial’) c. Peter liked what I had bought. (free ‘relative’) It is not the case that the sentences provided in (39) and (40) are ‘interrogative’ in the illocutionary sense of contributing to a question, so ‘interrogative’ is not a good general characterization of the wh- morpheme. It is, however, the case that all the subclauses containing a wh- based determiners give rise to incomplete representations with free variables which the hearer has to fill in (to the degree of specificity which is required to make sense of the utterances in the context). In the case of free relatives the variable has to be recovered from contextually accessible assumptions. We can contrast the wh- morpheme and the th- morpheme in (41a) and (41b). (41) a. Peter liked what I had bought b. I bought the shoes. In (41a) the internal argument of bought is only semantically constrained by the inanimacy of what and by the buy (x, y) predicate frame, and pragmatically constrained by the wider context in which the sentence is parsed. In (41a) the speaker does not instruct the hearer to identify any specific entity or entities which she has in mind – in fact, whether the hearer is capable of inserting an argument beyond ‘inanimate object of a buying action’ is semantically left open. In contradistinction, in (41b) the speaker semantically instructs the hearer to identify the specific entities she has in mind. The

 Alex Klinge

introduction of variables by the hw- based determiners may be made explicitly limited to a contextually given set, as we see in ModE which, or explicitly unlimited, as we see in compound determiners such as whichever, whoever, etc. In the case of bound relatives the variable is simply recovered from the entity already activated by the antecedent N-head. This also explains why bound relatives may in principle either be realized as þ- based or hw- based determiners. Standard ModG has þ- based determiners in bound relative distribution, ModE and ModD have hwbased determiners in bound relative distribution. A þ- determiner in relative distribution simply instructs the hearer to identify the specific entity the speaker has in mind, which happens to be co-extensive with the entity introduced by the antecedent Nhead. Since determiners in free relatives cannot rely on an entity already introduced by an antecedent N-head, they always have hw- based determiners, also in ModG. Note also that for the same reason clausal relatives always have hw- based determiners. This appears from the cognate examples of clausal relatives in (42). (42) a. ModE: The number of cars on the roads is growing, which is bad for the environment. b. ModG: Die Anzahl der Autos auf den Strassen nimmt zu, was für die Umgebung schädlich ist. c. ModD: Antallet af biler på vejene stiger, hvilket er skidt for miljøt. There is, however, one distribution which will not be explained easily on the basis of the semantics postulated here, viz. the use of hw- based determiners in exclamations in ModE and ModG. (43) a. ModE: What a tricky question! b. ModG: Was für eine verzwickte Frage!.

8. Conclusion In this article I have argued that in the Germanic languages investigated we can identify variants of a pan-Germanic þ- morpheme and variants of a pan-Germanic hwmorpheme in forms which distribute as determiners in the nominal and the adverbial domain. This means that it is possible to argue that the, this, that, there, then, etc are related, not through some obscure shared ‘demonstrative’ ancestor, but through a shared morpheme. This means that a level of common semantic description is called for. That common level of description is ‘ostension’, i.e. the speaker drawing the hearer’s attention to the specific entity the speaker has in mind. Similarly, a level of common description is required for who, where, when, etc. That common level of description is the introduction by the speaker of a ‘variable’ referent. However, the complementary distribution of the þ- based and the hw- based forms as D-heads raises the question whether they should not also be described together in terms of a paradigmatic contrast

Stating the case for þ- root and hw- root determiners 



between two complementary types of procedural semantics. I argued that the two morphemes encode two basic ways that reference assignment can be guided by the speaker. The two complementary ways may be summarized as in (44) and (45). (44) The speaker determines the referent which the hearer is supposed to insert in the propositional form. (45) The speaker leaves it for the hearer to determine the referent which the hearer inserts in the propositional form. Since the two morphemes distribute as D-heads, it does not come as a surprise that their central semantic purpose is to guide the procedure of reference assignment. Notions such as definiteness, familiarity, and accessibility are probably derived from the procedure of reference assignment. If the analysis of the formal and semantic aspects of the þ- and hw- morphemes proposed here survives further meticulous scrutiny, the standard analyses of the categories involved in nominal determination will have to be reformulated. According to what has been proposed here, þ- determiners and hw- determiners form sub-categories of a larger reference-assignment category. In particular, if the analysis proposed here eventually turns out also to include thou and du, the traditional category of ‘personal pronouns’ is neither formally nor semantically coherent and will have to be reconsidered. According to the analysis proposed here, the distribution of þ- determiners and hw- determiners in DP architecture reflects their complementary semantics and complementary distribution, as in (46). (46) þ- determiners and hw- determiners in DP architecture DP D

QP Q

NP 2 AP

NP 1 N

A

the this which the my it what who

a many

new new new new new new

book book book book books book

XP

 Alex Klinge

References Abney, S.P. 1987. The English Noun Phrase in its Sentential Aspect. PhD dissertation. MIT. Andersen, H. 1936. Oldnordisk Grammatik. Copenhagen: J.H. Schultz Forlag. Mitchell, B. & Robinson, F.C. 2001. A Guide to Old English. 6th edn. Oxford: Blackwell. Barnhart, R.K. & Steinmetz, S. (eds). 1988. Chambers Dictionary of Etymology. Edinburgh: Chambers Harrap Publishers. Bernstein, J. 2001. The DP hypothesis: Identifying clausal properties in the nominal domain. In The Handbook of Contemporary Syntactic Theory, M. Baltin & C. Collins, 536–561. Oxford: Blackwell. Blakemore, D. 1987. Semantic Constraints on Relevance. Oxford: Blackwell. Blakemore, D. 2002. Linguistic Meaning and Relevance. The Semantics and Pragmatics of Discourse Markers. Cambridge: CUP. Braune, W. 2004. Althochdeutsche Grammatik I. Laut- und Formenlehre. 15th edn. By I. Reiffenstein. Tübingen: Max Niemeyer. Cotte, P. 1993. De l’etymologie à l’énonciation; deixis, anaphore abstraite, syntaxe génétique dans quelques mots en th de l’anglais contemporain. Travaux de Linguistique et de Phiolologie 4. Déchaine, R-M. & Wiltschko, M. 2002. Decomposing Pronouns. Linguistic Inquiry 33(3): 409–442. Fraurud, K. 1990. Definiteness and the processing of noun phrases in natural discourse. Journal of Semantics 7(4): 395–433. Hawkins, J. A. 1978. Definiteness and Indefiniteness: A Study in Reference and Grammaticality Prediction. London: Croom Helm. Heim, I.R. 1988. The Semantics of Definite and Indefinite Noun Phrases. New York NY: Garland. Heine, B. 1997. Cognitive Foundations of Grammar. Oxford: OUP. Hodler, W. 1954. Grundzüge einer germanischen Artikellehre. Heidelberg: Carl Winter. Klinge, A. 2006. The unitary procedural semantics of the, this and that. Acta Linguistica Hafniensia 38: 54–77. Koopman, H. 1993. The Internal and External Distribution of Pronominal DPs, Ms, UCLA. Langacker, R. W. 2001. What wh- means. In Conceptual and Discourse Factors in Linguistic Structure, A. Cienki, B.J. Luka & M.B. Smith (eds), 137–151. Stanford CA: SCLI Publications. Lass, R. 1992. Phonology and morphology. In The Cambridge History of the English Language I, 1066–1476, N. F. Blake (ed.). Cambridge: CUP. Lass, R. 1994. Old English. A Historical Linguistic Companion. Cambridge: CUP. Lyons, C. 1999. Definiteness. Cambridge: CUP. Lyons, J. 1975. Deixis as the source of reference. In Formal Semantics of Natural Language. Papers from a Colloqium Sponsored by the King’s College Research Centre, Cambridge, E. L. Keenan (ed.). Cambridge: CUP. Olsen, S. 1991. AGR(eement) und Flexion in der deutschen Nominalphrase. In Syntactic Phrase Phenomena in Noun Phrases and Sentences, C. Bhatt, E. Löbel & C. Schmidt (eds), 51–69. Amsterdam: John Benjamins. Payne, J. & Huddleston, R. 2002. Nouns and noun phrases. In The Cambridge Grammar of the English Language, R. Huddleston & G.K. Pullum (eds), 323–523. Cambridge: CUP.



Stating the case for þ- root and hw- root determiners  Perlmutter, D.M. 1970. On the article in English. In Progress in Linguistics, M. Bierwisch & K.E. Heidolph (eds). The Hague: Mouton. Prins, A.A. 1972. A History of English Phonemes. Leiden: Leiden University Press. Quirk, R., Greenbaum, S. Leech, G. & Svartvik, J. 1985. A Comprehensive Grammar of the English Language. London: Longman. Shipley, J.T. 1984. The Origins of English Words. A Discursive Dictionary of Indo-european Roots. Baltimore MD: John Hopkins University Press. Tappe, H.T. 1990. Determiner Phrases and Agreement in German, Ms. Vangsness, Ø. 1999. The Identification of Functional Architecture. PhD dissertation, University of Bergen. Vangsness, Ø. 2001. On noun phrase architecture, referentiality, and article systems. Studia Linguistica 55(3): 249–299. Wiltschko, M. 1998. On the syntax and semantics of (relative) pronouns and determiners. Journal of Comparative Germanic Linguistics 2: 143–181. Wood, J. L. Definiteness and Number: Determiner Phrase and Number Phrase in the History of English. PhD dissertation, Arizona State University. Zamparelli, R. 2000. Layers in the Determiner Phrase. New York NY: Garland.

On certain differences between noun phrases and clauses Naoki Fukui and Mihoko Zushi

Sophia University/Kanagawa University Along with the noted similarities that have prompted much important study in the development of the theory of phrase structure and movement, nouns and verbs (and their projections) exhibit remarkable differences. These differences have often been neglected or otherwise discredited in favor of the similarities that have been a target for novel proposals regarding the internal structures of noun phrases and clauses. This paper tries to pin down the locus of the differences, and identifies the fundamental difference between noun phrases and clauses as the fact that noun phrases (nominal expressions) have a singlelayered internal structure having a single phase and are completed (or “closed”) in terms of licensing of internal elements, whereas clauses have a double-layered internal structure with two internal phases one of which (vP) is not completed (or “open”) in the sense that outside probes (namely, C and C-T) play a role in determining the inner workings of vP. The paper argues in a preliminary form that from this fundamental difference (which itself seems to be rooted in considerations of the C-I interface, i.e., clauses are propositions while nominal expressions are typically arguments), various syntactic differences between the two classes of categories, particularly those with respect to A-movement and A’movement, naturally follow.

1. Introduction The properties of nominal structures have attracted considerable attention in the literature ever since the earliest years of modern generative grammar, particularly after Chomsky (1970). The striking similarities observed between nominal structures and clausal structures have led researchers to construct a general theory of phrase structure to capture the cross-categorial parallelism. X’-theory, as developed since Chomsky (1970), has yielded the desired cross-categorial parallelism, allowing the internal structure of nominals to be sufficiently similar to that of clauses. This line of approach attained its full development in the 1980’s when two hypotheses emerged, the predicate-

 Naoki Fukui and Mihoko Zushi

internal subject hypothesis and the DP hypothesis. These hypotheses make it possible to treat in a completely uniform fashion nominal structures and clausal structures: Arguments of a noun/verb are projected within its projection (NP for noun phrases; VP for verb phrases/clauses), and the subject of a noun/verb is raised to the specifier position of a functional category that selects the lexical category (DP for noun phrases; T/IP for clauses). Further development in the study of nominal structures was made in the 1990’s, refining the internal structure of noun phrases, with various functional categories being postulated within noun phrases (see Cinque 1994, 2002 and references therein). Although these lines of research have been highly successful in capturing the parallelism between a noun phrase and a clause, too much emphasis seems to have been placed on the similarities between the two classes of categories, leaving important differences between noun phrases and clauses unexplored. There are in fact several asymmetries between them that should not be dismissed. Thus, a noun/nominal differs from a verb/clause in terms of A-movement within it: A noun (such as belief), unlike a verb (such as believe), does not allow the subject of its clausal complement to move to the matrix subject position, despite the fact that both nouns and verbs allow for a passive-like operation (Chomsky 1970; Kayne 1984). Also, differences between a noun/nominal and a verb/clause have been reported with respect to extraction: A nominal, unlike a clause, generally prohibits extraction out of it (Bach and Horn 1976: Chomsky 1973, 1977; among many others). In addition, it is well-known that a verb phrase as a whole can be fronted, but a noun phrase cannot (see Saito and Murasugi 1999). In this paper, we reconsider nominal structures in light of the recent development of minimalism, with special attention to the differences between nominals and clauses in terms of movement from and within them. This paper is organized as follows. In section 2, we first examine the internal structure of noun phrases and argue that noun phrases are represented as nP, which constitutes a “phase” with a single-layered structure, in contrast to clauses that have a twolayered structure with both CP and vP being phases. Then, we discuss A-movement properties within noun phrases with special focus on the difference between noun phrases and clauses in terms of subject raising. The fundamental difference between nouns and verbs in terms of φ-feature specifications plays a major role in accounting for the presence/absence of subject raising. Section 3 deals with another difference between nouns/nominals and verbs/clauses with respect to a preposing operation: the presence of VP-preposing and the absence of NP-preposing. The proposed nominal structure, together with the notion of phases, gives a straightforward account for the absence of NP-preposing. Section 4 is devoted to a discussion of further differences between nominals and clauses in terms of extraction. We argue that an idea suggested by Chomsky (2005) opens up a novel approach to capturing some of the effects of the Left Branch Condition, the Specificity Condition, and the Complex NP Constraint observed in extraction from noun phrases, in contradistinction with corresponding cases of extraction from verb phrases/clauses. Section 5 concludes the paper.



On certain differences between noun phrases and clauses 

2. Movement within noun phrases This section deals with the differences between nouns/nominals and verbs/clauses in terms of operations of A-movement. It has been argued in the literature that while nouns/nominals are parallel to verbs/clauses in that a passive-like operation is possible in both types of structures, the parallelism breaks down in another type of A-movement, namely, subject raising. That is, a verb allows the subject of its clausal complement to move to the matrix subject position, but a noun does not. The question to be resolved is what derives the similarities and differences between nouns/nominals and verbs/clauses. Before going directly into these issues, we first consider the structure of noun phrases and their Case properties.

2.1

The structure of noun phrases

Certain derived nominals exhibit the following two types of structures with essentially the same meaning: (1) a. the enemy’s destruction of the city b. the city’s destruction by the enemy Since Chomsky (1970), the nominal pair in (1) has been treated in parallel to the sentential active/passive pair in (2), i.e., (1b) is the passive counterpart of (1a), just like (2b) is the passive counterpart of (2a). (2) a. The enemy destroyed the city b. The city was destroyed by the enemy Under the DP hypothesis (Abney 1987; Brame 1981, 1982; Fukui and Speas 1986), the structural parallelism between nominals and clauses can be fully captured. Thus, in the nominal case, the subject of a noun is generated within the noun’s projection (NP) and moves to the specifier position of a higher head (Spec of DP, henceforth designated as Spec(D) and the like), completely in parallel to the clausal case, in which the subject of a verb is generated within the verb’s projection (VP) and moves to Spec(I) (or Spec(T), depending on one’s analysis of inflectional elements in clauses), as illustrated by (3). (3) a. [DP the enemyi D [NP ti [N’ destruction (of) the city ]]] b. [IP the enemyi I [VP ti [V’ destroyed the city ]]] In the cases of corresponding passives, the object, instead of the subject, moves to a higher Spec (Spec(D) for nominals; Spec(I) for clauses).1 1. The nature of the movement operation involved in passive nominals has been a matter of debate, since several differences between passive nominals and passive sentences have been observed (see Anderson 1979; Fiengo 1979; Giorgi and Longobardi 1991: Jaeggli 1986; Longobardi 2001). Thus, Longobardi (2001) argues, based on the differences between the two, that passive

 Naoki Fukui and Mihoko Zushi

Although this analysis allows us to capture certain similarities between nouns/ nominals and verbs/clauses, it deserves careful reconsideration in light of the recent development of the theory of phrase structure. Chomsky (1995) argues that the projection of a verb (VP) is headed by a functional head, a light verb v, producing the structure of a clause represented roughly in (4a).2 Also, there have been occasional suggestions in the literature (see among others Valois 1991, Sportiche 1998, Radford 2000) that noun phrases have a similar “shell” structure to that of verbs in that they comprise an outer nP headed by a “light noun” n and an inner core NP headed by a lexical noun N. Drawing on these suggestions, let us pursue further the idea that the internal structure of nominals ought to be completely parallel to that of clauses, by analyzing the structure of nominals as in (4b), where the projection of a noun (NP) is taken as a complement of a functional head n that corresponds to v in clauses. (4) a. [CP [TP [vP SUBJ v [VP V OBJ ]]]] b. [DP [nP SUBJ n [NP N OBJ]]] In a clausal structure like (4a), vP forms a domain in which all the arguments of a verb are projected, and a higher domain of CP expresses specifications of Force in the sense of Chomsky (1995) (or clausal-type in Cheng 1991). The two phrases, vP and CP, have been taken as syntactic domains called “phases” since Chomsky (2000), in which Case/ agreement features must be fully determined. In the minimalist model which has developed since around 1990, recursive Merge (the fundamental structure-building operation) incorporates the effects of three compositional cycles postulated in the Extended Standard Theory (i.e., the sensori-motor system, the conceptual-intentional system, and Logical Form (LF)), while eliminating the internal levels of D-structure and S-structure. It is argued in the recent literature (cf. Nissenbaum 2000) that the final internal level LF can also be abandoned, if at various stages of computation there are “Transfer” operations, which on the one hand transfers the syntactic object already constructed to the phonological component (“Spell-Out”), and which on the other hands over the syntactic object to the semantic component, which in turn maps it to the conceptual-intentional system. These syntactic objects (i.e., the targets for Transfer operations) are called phases. See Chomsky’s series of works, particularly Chomsky 2005, for detailed discussion on the nature of phases. nominals are more like middles than passives. If the analysis of unaccusatives proposed by Chierchia (2004) can be extended to middles, as he briefly suggests, the movement operation in middles/passive nominals can be regarded on a par with the one taking place in unaccusatives, which are derived, in Chierchia’s analysis, from their corresponding transitives by identifying the subject and object and externalizing the remaining argument. 2. In Chomsky’s series of recent works (Chomsky 2000, 2001, 2004, 2005), the verbal projection of a transitive is represented as v*P, while the projection of a passive and an unaccusative is notated as vP. Since this distinction is not particularly relevant for the present discussion, we use vP in the sense of v*P of Chomsky (2005) for ease of exposition.



On certain differences between noun phrases and clauses 

In the vP domain, the Case/agreement features of the object are determined, while in the higher CP domain, the Case/agreement features of the subject are determined. Chomsky (2005) proposes that it is in fact C, not T as customarily assumed, that is responsible for determining the Case/agreement features of the subject. In particular, C transmits its feature inducing agreement (the Agree-feature in Chomsky 2005) to T, turning T into a head that triggers raising of the subject to its Spec. In other words, the Agree-feature of C transmitted to T serves as a “probe” that takes the subject bearing φ-features and a Case feature, as a “goal.” In this system, it is solely the head of a phase, v or C, which determines the Case/agreement features of the argument of the verb and triggers raising of the argument to its Spec. Given this system, it is necessary to revise the structure of the noun phrase given in (4b), in order to fully capture the parallelism between nominals and clauses. There seem to be two possibilities regarding the required revision, as illustrated in (5). In (5a), DP is taken to be analogous to TP with a subject being raised to its Spec, and is selected by another head, whose projection is represented here as XP; XP corresponds to CP in the clausal case. In (5b), DP is taken to be analogous to CP in the clausal case and selects another category, XP, which behaves like TP hosting a subject in its Spec. (5) a. [XP [DP SUBJi [nP ti [NP N OBJ ]]]] b. [DP [XP SUBJi [nP ti [NP N OBJ ]]]] Neither of these structures seems to be promising, however. Although it is plausible to assume that nominals (at least derived nominals) contain a domain in which all arguments of the head noun are projected, just as vPs constitute a propositional domain, it is extremely unlikely that nominals have a higher domain corresponding to CP in clauses. Unlike clauses, nominals do not express specifications of force (declarative, interrogative, etc.) and of mood (indicative, subjunctive, conditional, etc.) and do not generally host operators (that is subject to syntactic movement) at the edge.3 Expressions related to discourse, such as topic and focus phrases, can appear in the left periphery of clauses, but not in the comparable position of nominals. Furthermore, nominals differ from clauses in terms of their Case properties: In clauses two Case features are licensed, nominative in the C-T domain and accusative in the v-domain, whereas only genitive Case is typically licensed in nominals. It is then natural to conjecture that nominals do actually lack a higher phase corresponding to CP, and that only nP, which is analogous to vP, is present in the nominal domain.4 3. Definiteness is an implausible candidate for “mood indication” of a nominal expression, particularly given the fact that some languages (such as Japanese) do not have definiteness effect on noun phrases but nonetheless exhibit mood distinctions. 4. Alternatively, we might assume, in an attempt to maintain the insights of the DP-analysis, that the relevant head in nominals is D, rather than n as assumed in the text. Another possibility is that n and D (and possibly other kinds of functional heads associated with a noun) alternate as the relevant head, either within a single language or cross-linguistically (or both), possi-

 Naoki Fukui and Mihoko Zushi

Thus, we propose that a noun phrase has the following internal structure with n being a phase head analogous to v in clauses5:

(6) [nP n [NP SUBJ N OBJ]]

The question immediately arises under this analysis as to how a determiner fits in this structure. Given that n is a phase head, having an edge-feature (see Chomsky 2005), it is natural to assume that the edge-feature is satisfied by merger of a determiner and the projection of n, forming an nP, as illustrated in (7a). (7) a. [nP D n [NP SUBJ N OBJ]] b. [nP SUBJi n [NP ti N OBJ ]] When a determiner is absent, the edge-feature of n can be satisfied by raising of an argument of the noun, say the subject, to its Spec, as illustrated in (7b).6 On the assumption that Merge alone is not sufficient to induce agreement (which requires Agree and Merge), the subject should come from within NP. Assuming that n also has a set of φ-features (an Agree-feature), it serves as a probe taking the subject with an uninterpretable Case feature as a goal, causing the latter to raise to its Spec, through which the Case feature of the subject is valued genitive. As for the object, no functional head is available to establish an agreement relation with it to value its uninterpretable Case feature, and thus a last resort operation of of-insertion takes place to salvage the otherwise Caseless noun phrase. When a passive-like operation takes place, the sole argument of a noun, the object, raises to Spec(n), where its Case feature is valued genitive. This issue is taken up in the following subsection.

2.2

Passive vs. Subject raising

As we briefly discussed above, although nominals are parallel to clauses with respect to the applicability of passive, the parallelism breaks down in the case of subject raising. bly affecting the phase status of a nominal. The choice between these alternatives awaits future research. In any case, the crucial ingredient of our analysis is the idea that nominals have a onelayered internal structure with a single phase head (n or D), while clauses have a double-layered structure with two phase heads (v and C). 5. See Fukui and Kasai (2004) and the references cited there, for discussion of the possibility that DP (nP in our terms) counts as a phase. 6. Our analysis would not by itself preclude multiple occurrences of elements in Spec(n), e.g., *the John’s destruction of the city. It is not obvious, however, whether such constructions, while clearly excluded in English, should be ruled out by principles of Universal Grammar. Thus, multiple satisfactions of the edge-feature of n seem to be available in many languages including Hungarian, in which both subject and D appear at the edge of an nP (see Szabolcsi 1983). Crosslinguistic comparison in this respect goes far beyond the scope of this paper, however. We leave this issue for future research.



On certain differences between noun phrases and clauses 

The subject of an infinitival clause selected by a verb can be raised to the matrix subject position, as shown in (8a), while the corresponding movement operation is not allowed in nominals, as illustrated by (8b). (8) a. Johni appears to ti have left b. *Johni’s appearance to ti have left The questions that need to be addressed here include the following. First, Why do nouns/nominals and verbs/clauses differ in terms of the possibility of subject raising? And, Why is a passive-like operation allowed in nominals but subject raising is not (see Chomsky 1986 for a Case theoretic approach and Kayne 1984 for an ECP-based approach toward this problem). Let us consider the first question raised by the contrast in (8). It is generally assumed that in the case of raising constructions like (8a), whose structure is roughly represented as in (9), the C-T complex serves as a probe seeking the goal, John, having a set of φ-features and an uninterpretable Case feature.7

(9) [ C [ T [VP V [TP John [ have left ]]]]]

The Agree relation is established between T and John, resulting in the raising of John to the higher Spec(T). Now let us consider subject raising within a noun phrase (8b). We assume that N, like V, takes a TP as its complement, as illustrated in (10). Following the analysis of genitive Case licensing presented above, it is the Agree-feature of n that serves as a probe seeking a goal in terms of A-movement.8 (10) [ n [NP N [TP John [ have left ]]]] The impossibility of raising of John to Spec(n) in (10), in contrast to the possible raising in (9), suggests that the association of the probe n and the goal John in (10) is blocked for some reason. The most likely reason is that the locality condition on the operation of Agree is not fulfilled here. That is, there is another element with the relevant features 7. As discussed above, Chomsky (2005) argues that it is in fact C transmitting the relevant features to T that is responsible for raising subjects. See section 4 for further discussion. 8. To supplement our discussion that follows from a slightly different angle, let us recall that Chomsky (2005) claims that a phase head generally transmits its Agree feature to the head it selects, which turns the latter into a phase head, making it possible to raise a noun phrase to its Spec. If n indeed transmitted its Agree feature (uninterpretable φ-features) to N, an argument of the noun would raise to Spec(N). This is not a welcome result, given that the subject of a noun may appear in Spec(N). We assume here that the transmission does not occur in the phase nP. The most plausible reason for this is that N itself bears a set of φ-features, which makes transmission of the φ-features of n to N rather redundant (hence impossible). Note also that the uninterpretable Case feature of N can and must be inherited by n, once genitive Case assignment (caused by n) is carried out. The Case feature of n-N serves as a goal for the probe v with respect to accusative Case licensing. See section 4 for more discussion.

 Naoki Fukui and Mihoko Zushi

that are matched with the features of n which is closer to n than is John. Given that the probe-goal association in (9) is not blocked, the offending element that blocks the probe-goal relation in (10) must be N, as N has a set of φ-features, it c-commands John, and is c-commanded by n, thereby being closer to n than John. The presence of the φ-features of N breaks the required Agree relation between n and John in (10), rendering the movement of John impossible.9 By contrast, V does not block the Agree relation between T and John in this way, because it bears no φ-features, allowing John to move to the matrix Spec(T). Thus, the difference between nouns/nominals and verbs/clauses with respect to subject raising comes down to the fundamental and general asymmetry between nouns and verbs: Nouns have a set of φ-features, but verbs do not. The proposed analysis also offers a simple and straightforward account for the second question raised above, i.e., the difference between passive and subject raising in nominals. Consider (11). In the case of passive, the object of a noun, the city, is to be moved to Spec(n)10: (11) [ n [NP N the city ]] This is possible, because in (11), N and the object are equidistant to the probe n. Thus, N does not block the Agree relation between n and the object, allowing the latter to move to Spec(n). Notice that on the proposed analysis, it is only an element within the minimal domain of a head noun that can be raised to Spec(n). Otherwise, the head noun blocks the search by the probe n in the first place. This analysis gives a straightforward account not only for the impossibility of subject raising in nominals as just discussed, but also for the impossibility of subject raising in the following examples (see Chomsky 1986b and Kayne 1984 for relevant discussion): (12) a. Mary is believed to have left by John b. *Mary’s belief to have left by John (13) a. Mary was believed a genius by John b. *Mary’s belief a genius by John The examples in (12) involve an ECM structure and those in (13) a small clause structure. In either case, raising of the embedded subject over a nominal head to Spec(n) is banned due to the presence of the φ-features of the head noun. The noun that c-commands the embedded subject in either an ECM or a small clause structure blocks the

9. The intervention effect here may fall under the scope of the feature-based A-over-A principle proposed by Fukui (1999) on the assumption that N’s φ-features and Case feature percolate to its projection, making the set of those features dominate in the tree structure the set of features of the embedded subject. 10. We assume that the subject of a noun is not syntactically projected as an argument due to the process of middle formation as suggested in note 1.



On certain differences between noun phrases and clauses 

Agree relation between n and the embedded subject. Therefore, subject raising is impossible in these examples.

2.3

Raising to object

We have argued that the lack of subject raising within a noun phrase is due to the intervening φ-features of the head noun that blocks the Agree relation between n and the embedded subject. Let us next consider the active counterparts to (12a) and (12b): (14) a. Mary believed John to have left b. *Mary’s belief John to have left In the ECM structure (14a), Case/agreement properties of the embedded subject, John, are determined by the matrix light verb v. Agree holds between v and John, by which the uninterpretable Case feature of John is valued accusative. This sort of Agree relation cannot be established for John in (14b), however. As we argued above, the Agree relation between n and John cannot be formed, because the φ-features of the head noun block it. Given the internal structure of a noun phrase proposed in 2.1, in which only nP forms a phase with no intermediate phase within it, it follows that there is no functional head other than n within a noun phrase that could determine the Case properties of John. There is no way to eliminate the uninterpretable Case feature of John, yielding the ill-formedness of (14b). The remaining problem is to explain why the otherwise Caseless noun phrase John in (14b) cannot be saved by the rule of of-insertion as in (15a), which applies to the object in examples like (1a), repeated here as (15b): (15) a. *Mary’s belief of John to have left b. the enemy’s destruction of the city Although the exact mechanism of the of-insertion rule needs to be made precise, it is sufficient for the present purposes to observe that the rule of of-insertion applies to an element generated as a sister to N. Thus, it applies to the city in (15b), whereas it does not apply to John in (15a) (or to John in (8b)). It remains to be seen why the rule of ofinsertion is designed exactly in this way, which we leave open here for future research.

3. NP-preposing vs. VP-preposing In connection with their analysis of NP-deletion (commonly known as “N’-deletion”), Saito and Murasugi (1990, 1999) discuss the differences between NP and VP with respect to preposing operations (see also Lobeck 1990; Zagona 1982; among many

 Naoki Fukui and Mihoko Zushi

others). It is generally the case that VP-preposing is allowed, whereas NP-preposing is not, as the contrast between (16) and (17) shows.11 (16) a. *Book, I read John’s. b. *Destruction of the city, I witnessed the barbarian’s. (17) a. John said he would leave early, and leave early he will. b. I believe that Mary would win the race, and win the race, she (actually) did. As pointed out by Saito and Murasugi, the contrast between NP and VP with respect to preposing operations appears to be unexpected under the standard hypothesis that the internal structure of DP is parallel to that of T/IP. In our analysis, however, the observed difference is rather expected, given that the parallelism between noun phrases and clauses regarding their internal structures breaks down in certain respects. Before illustrating how the absence of NP-preposing (as opposed to the presence of VPpreposing) is straightforwardly accounted for under our analysis, let us have a somewhat closer look at the nature of VP/NP-preposing. Huang (1993) argues, based on reconstructions facts, that a preposed VP in the examples like (17) contains a trace of the subject that has moved to Spec(I/T). This analysis captures, among others things, the important differences between VP-preposing and operations such as wh-movement and topicalization with respect to reconstruction effects. Consider the following examples. (18) a. Which pictures of himselfi/j does Johni think Billj likes t? b. Those pictures of himselfi/j, Johni thinks Billj will buy t. c. Criticize himself*i/j, Johni said Billj never will t. The moved wh-phrase in (18a), as well as the topicalized phrase in (18b), has two reconstruction possibilities. The anaphor (himself) within the moved phrase can be bound by either John or Bill. On the other hand, the preposed VP in (18c) has only one reconstruction possibility. An anaphor himself can be bound only by Bill. Huang accounts for this difference between VP-preposing and wh-movement/topicalization by attributing the narrower range of reconstruction options of VP-preposing to the presence of a subject trace within a preposed VP. Under the predicate-internal subject hypothesis, the preposed VP contains a trace of the subject, while no such trace is present within a preposed wh-phrase/topic. Now the preposed VP (but not the 11. We do not discuss in this paper the phenomenon of NP-deletion that Saito and Murasugi (1990, 1999) extensively discuss. In view of the fact that NP-deletion, as well as VP-deletion, does not obey island constraints, in clear contrast to VP-preposing which obeys island constraints, and given the fact that NP-deletion applies across utterance boundaries, it seems plausible to assume that the process of deletion of this kind takes place in the phonological component, rather than in the (narrow) syntactic component (see Lobeck 1995 and the references therein for relevant discussion).



On certain differences between noun phrases and clauses 

preposed wh-phrase/topic), which contains a subject trace, constitutes the minimal domain relevant for the version of the binding theory proposed by Chomsky (1986). Therefore, an anaphor within the preposed VP must be bound within it, i.e., by the subject trace contained in the preposed VP. This is why himself in (18c) can only be bound by Bill. Recasting Huang’s idea in current theory of argument alignment, we can say that what is preposed in the process of VP-preposing is actually a vP rather than VP, since a subject is merged to vP (becoming Spec(v)) in the theory of clause architecture we’re adopting. Thus, the operation in question should be analyzed as “vP-preposing” rather than VP-preposing. Let us now turn to NP-preposing in light of our analysis of nominal structures presented in section 2. According to our analysis, the structure of the noun phrase in (19a) is roughly represented as (19b). (19) a. the barbarian’s destruction of the city b. [nP the barbariani n [NP ti destruction (of) the city ]] Under this analysis of noun phrases, the preposed elements in the examples in (16) are an NP that contains a trace of the subject which had been raised to Spec(n). The situation is exactly parallel to the vP-preposing cases just discussed. Then, where does the contrast between (16) and (17) come from? The crucial difference can be actually stated in terms of the category that is subject to the operations (vP-preposing and NPpreposing). In the vP-preposing cases, what is preposed is a phase, while in the NPpreposing cases, the preposed phrase is not a phase. Based on this independently established distinction, it seems possible to give a unified account of the asymmetry exhibited by these operations. In current theory of grammar, the central role that a phase plays is to define the domain to which an operation applies in a derivation. Once a phase is passed in a derivation with all the uninterpretable features within it being deleted, no further operation applies to that phase, thereby excluding, among other things, counter-cyclic derivations. In this way, derivations proceed phase by phase (see Chomsky 2001 and much subsequent work, particularly Chomsky 2005). As we have discussed, the (desirably universal) set of phases include the following: CP, vP, nP. Other categories such as NP, VP, TP, etc. are not phases. If we turn our attention to the set of categories that are subject to Agree/Move, we come up with basically the same set, namely, CP, vP, and nP. Unless this is a mere accident (not a desirable result), a natural hypothesis that suggests itself is that the notion of phase plays a role not only in defining the computational domain, but also in characterizing a possible target for computations.12 Thus, we have the following tentative hypothesis. (20) Only phases are subject to Move. 12. Actually, the inability of, say, TP to be moved is sometimes taken to be a piece of evidence that TP is not a phase. See Chomsky (2005) for discussion.

 Naoki Fukui and Mihoko Zushi

To the extent that this hypothesis is tenable, we have a principled explanation as to why vP-preposing is possible but NP-preposing is not. Other cases of movement, including, in particular, wh-movement, “NP-movement” (Passive and Raising), topicalization, etc. are to be analyzed as nP-movement under this hypothesis. There is of course much to discuss regarding the hypothesis (20), including the very nature of this principle. One might also raise the question as to why this principle does not seem to hold for Merge, the fundamental operation in human language. Other questions remain, but they are outside the scope of this preliminary study. Thus, we leave fuller discussion on the nature of the principle for future work.

4. Extraction out of noun phrases We have so far argued that a noun phrase is represented as an nP that constitutes a phase. Given this hypothesis, it seems possible to claim that n allows an element at its edge to be extracted to a higher phase in accordance with the Phase Impenetrability Condition (PIC), in exactly the same way as the strong phases CP and vP allow extraction to their outer phase. The work on the Romance and other languages which found that extraction out of a noun phrase is possible only through its Spec position (see Cinque 1980; Giorgi and Longobardi 1991; Torrego 1986, among others) offers some support for this hypothesis. But the hypothesis is not in accord with the traditional observation that movement across an NP node yields a partial decrement in acceptability and that the possibility of extraction from noun phrases is restricted much more severely than that of extraction from clauses. Extraction from noun phrases exhibits the effects of the Left Branch Condition, the Specificity Condition and the Complex NP Constraint, but none of these conditions/constraints is placed on extraction from clauses. We argue in this section that the idea suggested by Chomsky (2005) for deriving the effects of the Subject Condition opens up a new way of looking at the differences between noun phrases and clauses with respect to extraction.

4.1

Extraction of a subject: effects of the left branch condition

As is well known, the subject of a noun cannot be singled out and extracted in English (a part of the effects of the Left Brach Condition), while the subject of a verb can be freely extracted (modulo the that-trace effect). (21) a. *Whose did you see t book? b. Who do you think t saw John? The impossibility of extraction of a subject in noun phrases/nPs, as opposed to the possibility of extraction of a subject in clauses/vPs, suggests that nP, unlike vP, forms an opaque domain for further computations. Put differently, Spec(n), unlike Spec(v),



On certain differences between noun phrases and clauses 

cannot be seen by a higher probe, rendering extraction from that position impossible. What, then, brings about this asymmetry between nP and vP? To resolve this question, it is instructive to see how the subject of a verb is moved via wh-movement in the system of Chomsky (2005). The relevant portion of example (21b) has a representation like (22a) as a “base structure,” then converted to (22b) via raising operations. (22) a. C [T [who [v [see John]]]] b. whoi [C [whoj [T [whok v [see John]]]]] Chomsky (2005) argues that the phase head C has two probes: C’s edge-feature and an Agree-feature (φ-features) that is transmitted from C to T. Both the edge-feature and the Agree-feature of C (the latter inherited by T) seek the goal who in Spec(v). The Agree-feature of C-T seeks the goal who in Spec(v), moving it to Spec(T) and the edgefeature of C also seeks the goal who, moving it to Spec(C), resulting in the structure (22b) with three copies of who: whoi, whoj, and whok. Crucially, the derivation in which the Agree-feature of C-T raises who to Spec(T), from where who again is raised to Spec(C) (i.e., successive-cyclic raising of the wh-phrase to Spec(T) and then on to Spec(C)) is excluded. The failure of whoj in Spec(T) serving as a goal for the edge-feature of C is accounted for in Chomsky’s system by the principle that a head of A-chain is invisible to further computation when its uninterpretable features are valued. This forces the edge-feature of C to access who in its base position, i.e., in Spec(v). This is possible because both the search of the Agree-feature of C-T and that of the edge-feature of C proceed in parallel. The search of the edge-feature of C cannot be followed by the search of the Agree-feature of C-T. If it were possible, the uninterpretable Agreefeature of C-T would remain without being eliminated. On this analysis, the vP-phase in (22a) is not, in a sense, “completed” since the uninterpretable Case feature of the subject (who) at its edge is not yet valued. The presence of the uninterpretable Case feature in Spec(v) allows computations to extend the vP-phase domain to the next higher phase CP. The wh-phrase in Spec(v) with the uninterpretable Case feature can be raised by wh-movement to Spec(C), if both Agree and wh-movement are simultaneously driven by the same head C. With this in mind, let us consider extraction of subjects out of nP in examples like (21a) with the schematic representation in (23). (23) v [V [nP whoi [n [NP ti book ]]]] The structure (23) is basically the same as (22a), except that the relevant categories differ in the two cases: nP in (23), like vP in (22a), is a phase and the v in (23), like C in (22a), is a probe. The crucial difference between the two configurations, however, is that as opposed to the subject in Spec(v) in (22a), the subject, who, in (23) has no uninterpretable Case feature, since its Case feature has already been eliminated by the Agree relation with n. The nP in (23), then, is “completed” in the sense that its Spec has no uninterpretable Case feature.

 Naoki Fukui and Mihoko Zushi

Following Chomsky (2005), let us assume that the head of an A-chain is immune to further computation once its uninterpretable Case feature is valued. Thus, who in Spec(n) in (23) cannot be seen by the probe v, since it heads an A-chain with no unvalued uninterpretable features. Then, v needs to seek another goal. It cannot access who in its base position, however, because who in that position is too deeply embedded within NP which is not a phase (a PIC violation). Therefore, who cannot move out of nP in (23). In this way, the effect of the Left Branch Condition is successfully accounted for under the present approach.

4.2

Extraction of an object: specificity condition effects

Let us next consider the cases of extraction of an object out of a noun phrase. It has been argued that the effects of the Specificity Condition (Fiengo and Higginbotham 1980), along with the definite effects, which we set aside here, show up in extraction from a noun phrase: Extraction out of a noun phrase is blocked by the presence of a subject, as the contrast in (24) shows. On the other hand, no such effect is observed in extraction from a clause, as illustrated by (25): (24) a. Who did you see a picture of t? b. *Who did you see Bill’s picture of t? (25) a. Who did he expect Bill to see t? b. Who did you tell me that Bill saw t? Note that (24b) has the following partial structure at the relevant point of its derivation: (26) v [ V [nP who [nP Bill n [ N who ]]]]] In (26), the subject of the noun (Bill) is moved to Spec(n), as we discussed above. Suppose that the edge-feature of n also allows who, the object of the noun, to be moved to its (outer) Spec. Then, the presence of who in Spec(n) creates a problem for the Agree relation between v and n.13 The head of nP has an uninterpretable Case feature inherited from N (see note 7) that must be eliminated through the Agree relation with v and valued as accusative. The φ-features of who in (26) block the Agree relation to be established between v and n. This prevents who from being moved to Spec(n). If who stays at its base position, however, the probe v cannot access it, as we have seen above in the case of subject extraction. Thus, there is no way to relate v (the probe) and who (a potential goal). Hence the impossibility of the extraction of who.

13. Note that the presence of a subject of the noun cannot block the Agree relation between v and n, as the subject heads an A-chain which is invisible to a higher probe.



On certain differences between noun phrases and clauses 

A blocking effect of this sort does not hold in the cases of wh-movement of an object of a verb. Consider the following structure: (27) C [T [whoi [Bill [v [V whoj ]]]]] In (27), the object, who (its base position is indicated by whoj), moves to the outer Spec(v) (i.e., the position indicated by whoi). As Chomsky (2005) points out, raising of the subject, Bill, to Spec(T) crosses whoi in the outer Spec(v), but subject raising is not blocked by the presence of whoi. This is because both subject raising and wh-movement of who are driven by the probe C in parallel. Thus, no blocking effect is induced, and wh-movement of an object over the subject is possible.

4.3

The complex NP constraint

There is still another difference between noun phrases and clauses with regard to extraction: The clausal complement of a noun resists extraction out of it (the ComplexNP Constraint, CNPC), while no such effect is observed with extraction out of the clausal complement of a verb. Observe the following contrast. (28) a. *Who do you believe the claim that Mary likes t? b. Who do you believe that (Bill claims that) Mary likes t? Our analysis for the lack of subject raising within nP proposed in section 2 again enables us to account for the effect of CNPC. Recall that we argued there that the subject of a clausal complement of a noun cannot move to Spec(n), since the φ-features of the head noun cause an intervention effect. In contrast, there are no φ-features associated with a verb, and thus no intervention effect is induced when the subject of a clausal complement of a verb moves to Spec(v). The same intervention effect also shows up in the extraction of a wh-phrase out of the clausal complement of a noun. Consider the following structure. (29) [nP who n [NP N [CP who [ Mary likes who ]]]]

*

In (29), who cannot move to Spec(n) due to the intervening φ-features of the head noun. And if the wh-phrase is stuck at the embedded CP level, the probe v (which is located outside of nP) cannot have an access to it, as the embedded CP phase has already been passed in the derivation. Thus, the wh-phrase in the clausal complement of a noun cannot be moved out of it, showing the effect of the CNPC. In contrast to the extraction from a complex NP, no such intervention effect is obtained in the cases of extraction from a complex VP as in (28b). Recall that we demonstrated in section 2 that subject raising is allowed in clauses, because the matrix verb has no φ-features, and thus does not block the movement of the embedded subject to the matrix Spec(T). Exactly the same reasoning applies to the lack of intervention

 Naoki Fukui and Mihoko Zushi

effect in extraction from a complex VP. The lack of φ-features of the verb selecting a complex VP straightforwardly accounts for the nonexistence of an intervention effect with the movement of a wh-phrase from Spec(C) in the embedded clause to Spec(v) in the matrix clause. The analysis of CNPC just laid out also gives a simple account for the observation made by Chomsky (1973) that extraction from a noun phrase embedded within a higher noun phrase is ruled out. (30) a. *Who did you hear stories about a picture of t? b. *What do you receive requests for articles about t? The wh-phrase originated in the embedded nP may move to Spec(n), but it cannot move from that position to the higher Spec(n), moving over the head noun. As already discussed, this is because the φ-features of the noun induce an intervention effect, barring the movement over the noun.

4.4

Remaining problems

We have argued so far that noun phrases, as opposed to clauses, resist extraction in general. While general patterns of extraction support this position, there are two sets of facts that appear to run counter to our analysis. First, it is a well-known fact about English that in certain cases, extraction from a noun phrase indeed yields grammatical sequences. Second, in the Romance languages and others, extraction from a noun phrase seems to be permitted under certain conditions, i.e., extraction is possible only through the Spec position of a noun phrase. In this subsection, we briefly address these problems. Let us consider the first problem posed by, for example, the following well-formed examples of extraction from a noun phrase: (31) a. Who did you see a picture of t? b. Who did John write a book about t? It has been observed in the literature that several factors are involved in determining the grammaticality of these examples, including semantic properties of the verb that allow for extraction from the noun phrase that it selects (see Bach and Horn 1976, Chomsky 1973, 1976, Erteschik-Shir 1981, Diesing 1992, among many others). Thus, the grammaticality of the examples in (31) contrasts with the ungrammaticality of the following example: (32) *Who did John destroy a book about t? This situation has led many researchers to the conclusion that extraction in such examples as (31) is not genuine extraction from within a noun phrase, but rather, the permitted extraction cases actually involve extraction from a phrase that is located



On certain differences between noun phrases and clauses 

outside of a noun phrase. Thus, Bach and Horn (1976) claim that the wh-phrase in such cases is generated inside a PP that is directly dominated by the verb, from where the movement operation proceeds without violating any locality condition. Chomsky (1977) argues that a readjustment rule, triggered by certain type of verbs, is at work, which extraposes the PP complement of a noun, locating it outside the noun phrase and thereby rendering extraction possible (in accordance with the subjacency condition then assumed). An analysis along these lines gets further support from cross-linguistic considerations, as argued by Haraguchi and Washio (1988). Japanese is a language with a system of overt case particles, one of which is no “GEN(itive)”. This particle has to be attached to any noun phrase/postpositional phrase appearing in a noun phrase, and can never be attached to a PP outside of a noun phrase. Thus, if the postposition ni tuite “about/ of ” appears inside a noun phrase, no has to be attached, whereas if it appears outside of a noun phrase, no is not attached to it. (33) a. John ni tuite *(no) hon about -gen book ‘a book about John’ b. Boku-wa John ni tuite (*no) hanasita I -top about -gen talked ‘I talked about John’ In this way, the distribution of no gives us a definitive procedure for determining the positioning of the PP in question, i.e., whether it is located inside the noun phrase or not. Now consider the following paradigm. (34) English: whoi did you {write, read, buy, *destroy, *burn, *lose} a book about ti Japanese: John-wa sono -zinbutu ni tuite hon-o {kaita, yonda, -TOP that person about book-ACC wrote, read, ??katta, *yabuita, *moyasita, *nakusita } bought destroyed/tore burned, lost ‘John wrote/read/bought/destroyed/lost/a book about that person’ Note that no is not attached to the PP (sono-zinbutu ni tuite “about that/the person”) in the above example. The ungrammaticality of the Japanese examples with such verbs as yabuita “destroyed/tore,” moyasita “burned,” and nakusita “lost” shows that with these verbs, sono-zinbutu ni tuite “about that/the person” must be inside a noun phrase, while it is not located inside the noun phrase with verbs like kaita “wrote” and yonda “read.” (Notice that if no is attached to the PP, then the example becomes grammatical with verbs like yabuita, moyasita, etc. Cf. John-wa sono-zinbutu ni tuite-no hon-o yabuita. ‘John destroyed/tore a book about that/the person.’) And there is an unmistakable correspondence between the class of verbs (in Japanese) that require an associated PP to be located outside of a complement noun phrase, and the class of verbs (in English) that allow extraction from an associated PP. To the extent that this corre-

 Naoki Fukui and Mihoko Zushi

spondence is not accidental, the Japanese fact just reported offers indirect support for the analyses explored by Bach and Horn (1976) and Chomsky (1977). One way of capturing the insights of these analyses is to assume that a certain type of creation verb (i.e., the class of verbs that require the PP to be outside of the complement noun phrase) allow for the structure given in (35), in which the PP complement of a noun is in fact base generated outside of nP, from where it binds a null element in nP. (35) v [VP write [nP a book ei] [PP about who] i] Thus, extraction of who takes place directly from VP, rather than from within nP. It is clear that this mode of extraction is free from all the blocking effects discussed so far, and thus should be allowed. Of course, the legitimacy of this particular analysis remains open to further scrutiny (and within a much wider context), and we leave for future research the important tasks of justifying and refining the kind of analysis suggested above. Let us next consider the second problem, i.e., the fact that the subject of a noun can be extracted through Spec. It has been observed in the literature that in the Romance languages and others, only the subject of a noun can be extracted, as illustrated by (36) (taken from Cinque 1980; see also Giorgi and Longobardi 1991, Torrego 1986). (36) a. b. c.

Una persona di cui apprezziamo [la grande generosità t] (è Giorgio) A person of whom we appreciate the great generosity (is Giorgio) *Il paese a cui ricordiamo [un/l’attacco t] (è la Polonia) A country on which we remember an/the attack (is Poland) *Non è posto da cui possano minacciarci [il licenziamento t] It is not a position from which they can threaten us the dismissal

It seems that the analysis we have proposed for the impossibility of subject extraction out of a noun phrase is, at least in its present form, too strong, as it erroneously excludes the possible cases of subject extraction out of a noun phrase observed in the Romance languages. Here, we follow Rizzi (1990) and Cinque (1990), and assume that extraction from a noun phrase is generally prohibited (the conclusion we have independently arrived at in this paper), but a special device is at work in Romance which allows for subject extraction from a noun phrase under certain specific circumstances. Rizzi (1990) suggests that a noun, a [-V] head, is generally insufficient for licensing traces (see also Cinque 1990), thus disallowing for extraction out of its projection. However, an agreement between the subject of a noun and the head noun turns the latter into a sufficient and legitimate head governor, so that the trace left by movement of the subject can be governed (and licensed) by the noun, satisfying the ECP. Recasting Rizzi’s analysis in the present framework, we can deal with the relevant phenomena in the Romance languages in the following way. The subject wh-phrase of a noun is raised to Spec(n), which then agrees with the n-N complex. Thanks to this agreement, the presence of a wh-phrase in Spec(n) fails to block the Agree relation between v and n-N. Rather, the wh-phrase also agrees with v by transitivity, which makes



On certain differences between noun phrases and clauses 

it possible for the wh-phrase to be associated with v, freeing its movement to Spec(v). Details need to be worked out, but we assume an analysis along these lines will turn out to be effective in dealing with the exceptional subject extraction cases in Romance.

5. Summary and conclusion In this paper, we investigated a theory of nominal structure with a particular emphasis on several remarkable differences between nouns/nominals and verbs/clauses with respect to A-movement and A’-movement. More specifically, we argued (i) that a noun phrase forms an nP in which a functional head n, which corresponds to v in a clausal structure, selects a lexical category NP; and (ii) that an nP constitutes a single-layered structure with only nP being a phase, in contrast to a clausal architecture that involves a double-layered structure with both CP and vP being phases. With regard to A-movement, a noun/nominal differs from a verb/clause in that the former does not allow the subject of its clausal complement to be raised to the matrix subject position (section 2). We attributed this difference between the two classes of categories to the fundamental asymmetry having to do with their φ-feature specifications: Nouns bear φ-features, but verbs do not. The φ-features of a head noun block raising of the embedded subject, while the lack of such features associated with a head verb results in no intervention effect for subject raising. Although a head noun (or its φ-features) blocks subject raising, it does not block raising of its object in a passive-like operation, since this movement satisfies the required locality condition on movement. The intervention effect induced by a head noun also plays a crucial role in excluding extraction out of a complex NP, as discussed in section 4. The second difference between a noun/nominal and a verb/clause is concerned with the presence of VP-preposing vs. the absence of NP-preposing, as detailed in section 3. Having first established that “VP-movement” is in fact “vP-movement,” the proposed analysis of nominal structures nicely accounts for this asymmetry, by attributing it to the fact that vP is a phase but NP is not. Given the general hypothesis that only phases are visible to Move, the absence of NP-preposing (as opposed to the presence of vP-preposing) is explained on principled grounds. The differences between nouns/nominals and verbs/clauses with respect to extraction were discussed in section 4. We argued there that the general ban on extraction of the subject of a noun (part of the effects of the Left Branch Condition) is due to the “completeness” property of noun phrases; noun phrases are complete in the sense that the subject of the noun is licensed within nP itself, not requiring any licenser outside the nP. Since the subject raised to Spec(n) heads an A-chain being immune to a higher probe, it cannot be seen (detected) by a probe, v, which is located outside of nP. Nor can the subject in its base position (Spec(N)) be seen (detected) by the probe, because it is too deeply embedded within NP. This situation contrasts sharply with extraction of the subject of a verb, where C is responsible both for A-movement of the subject to

 Naoki Fukui and Mihoko Zushi

Spec(T) and for A’-movement to Spec(C) (Chomsky 2005). As for the extraction of an object wh-phrase, we attributed the impossibility of extraction of the object from a noun phrase to the fact that nouns, unlike verbs, need to receive accusative Case (via the Agree relation with v). However, once a wh-phrase is raised to Spec(n), the required Agree relation between v and n is blocked by the raised object wh-phrase (or its φ-features), rendering accusative Case-marking of the nP impossible. This accounts for the impossibility of extracting an object wh-phrase out of the nP. Along with the noted similarities that have prompted much important study in the development of the theory of phrase structure and movement, nouns and verbs (and their projections) exhibit remarkable differences. These differences have often been neglected or otherwise discredited in favor of the similarities that have been a target for novel proposals regarding the internal structures of noun phrases and clauses. In this paper, we have concentrated on these differences and have tried to put forth a theory that explains the differences on principled grounds. We have pinned down and identified the fundamental difference between noun phrases and clauses as the fact that noun phrases (nominal expressions) have a single-layered internal structure and are completed (or “closed”) in terms of licensing of internal elements, whereas clauses have a double-layered internal structure and one of these phases (vP) is not completed (or “open”) in the sense that outside probes (namely, C and C-T) play a role in determining the inner workings of vP. From this fundamental difference (which itself seems to be rooted in considerations of the C-I interface, i.e., clauses are propositions while nominal expressions are typically arguments), various syntactic differences between the two classes of categories follow. This paper should be taken as the first preliminary step toward a more comprehensive and well-balanced theory of noun phrases and clauses.

Acknowledgments We are grateful to the editors of this volume (Henrik Høeg Müller and Alex Klinge), Teresa Griffith, Hironobu Kasai, Chunyan Ning, and Sze-Wing Tang for their invaluable comments on an earlier version of this paper. The research reported in this article was supported in part by a research fund from the Sophia Linguistic Institute for International Communication (SOLIFIC), and also by a research grant from Kozo Keikaku Engineering, Inc.

References Abney, S. 1987. The English Noun Phrase in its Sentential Aspect. PhD dissertation, MIT. Anderson, M. 1979. Noun Phrase Structure. PhD dissertation, University of Connecticut. Bach, E. & Horn, G. 1976. Remarks on conditions on transformations. Linguistic Inquiry 7: 265–300.



On certain differences between noun phrases and clauses  Brame, M. 1981. The general theory of binding and fusion. Linguistic Analysis 7(3): 277–325. Brame, M. 1982. The head-selector theory of lexical specifications and the nonexistence of coarse categories. Linguistic Analysis 10(4): 321–325. Cheng, L. 1999. On the Typology of Wh Questions. PhD dissertation, MIT. Chierchia, G. 2004. A semantics for unaccusatives and its syntactic consequences. In The Unaccusativity Puzzle, A. Alexiadou, E. Anagnostopoulou & M. Everaert (eds), 22–59. Oxford: OUP. Chomsky, N. 1970. Remarks on nominalization. In Readings in English Transformational Grammar, R. Jacobs & P. Rosenbaum (eds), 184–221. Waltham MA: Ginn. Chomsky, N. 1973. Conditions on transformations. In A Festschrift for Morris Halle, S. Anderson & P. Kiparsky (eds), 232–386. New York NY: Holt, Rinehart and Winston. Chomsky, N. 1977. On wh-movement. In Formal Syntax, P. Culicover, T. Wasow & A. Akmajian (eds), 71–132. New York NY: Academic Press. Chomsky, N. 1986. Knowledge of Language: Its Nature, Origin and Use. New York NY: Praeger. Chomsky, N. 1995. The Minimalist Program. Cambridge MA: The MIT Press. Chomsky, N. 2000. Minimalist inquiries: The framework. In Step by Step: Essays on Minimalist Syntax in Honor of Howard Lasnik, R. Martin, D. Michaels, and J. Uriagereka (eds), 89–155. Cambridge MA: The MIT Press. Chomsky, N. 2001. Derivation by phase. In Ken Hale: A Life in Language, M. Kenstowicz (ed.), 1–52. Cambridge MA: The MIT Press. Chomsky, N. 2004. Beyond explanatory adequacy. In Structures and Beyond, A. Belletti (ed.), 104–131. Oxford: OUP. Chomsky, N. 2005. On phases. Ms, MIT. Cinque, G. 1980. On extraction from NP in Italian. Journal of Italian Linguistics 5: 47–99. (Reprinted in Cinque 1995, Italian Syntax and Universal Grammar, 7–53) Cinque, G. 1990. Types of A’-Dependencies. Cambridge MA: The MIT Press. Cinque, G. 1994. Evidence for partial N-movement in the Romance DP. In Path towards Universal Grammar: Studies in Honor of Richard Kayne, G. Cinque, J. Koster, J-Y. Pollock, L. Rizzi & R. Zanuttini (eds), 85–110. Washington DC: Georgetown University Press. Cinque, G. (ed). 2002. Functional Structure in DP and IP: The Cartography of Syntactic Structures, Vol. 1. Oxford: OUP. Diesing, M. 1992. Indefinites. Cambridge MA: The MIT Press. Erteschik-Shir, N. 1981. On extraction from noun phrases (picture noun phrases). In Theory of Markedness in Generative Grammar: Proceedings of the 1979 GLOW Conference, A. Belletti, L. Brandi & L. Rizzi (eds), 147–169. Pisa: Scuola Normale Superiore di Pisa. Fiengo, R. 1979. Surface Structure: The Interface of Autonomous Components. Cambridge MA: Harvard University Press. Fiengo, R. & Higginbotham, J. 1980. Opacity and NP. Linguistic Analysis 7: 395–422. Fukui, N. & Speas, M. 1986. Specifiers and projection. In MIT Working Papers in Linguistics 8, 128–172. (Reproduced in Fukui, N. 2006. Theoretical Comparative Syntax: Studies in Macroparameters. London: Routledge, pp. 9–37.). Fukui, N. 1999. An A-over-A perspective on locality. In Linguistics: In Search of the Human Mind, E. Iwamoto & M. Muraki (eds), 109–129. Tokyo: Kaitaikusha. (Reproduced in Fukui, N. 2006. Theoretical Comparative Syntax: Studies in Macroparameters. London: Routledge, pp. 209–223.). Fukui, N. & Kasai, H. 2004. Spelling-out scrambling. In Linguistic Variation Yearbook 4: 109–141.

 Naoki Fukui and Mihoko Zushi Haraguchi, S. & Washio, R. 1988. Henkei (Transformations). Tokyo: Kenkyusha. Jaeggli, O. 1986. Passive. Linguistic Inquiry 17: 587–633. Giorgi, A. & Longobardi, G. 1991. The Syntax of Noun Phrases: Configuration, Parameters and Empty Categories. Cambridge: CUP. Huang, J. 1993. Reconstruction and the structure of VP: Some theoretical consequences. Linguistic Inquiry 24: 103–138. Kayne, R. 1984. Connectedness and Binary Branching. Dordrecht: Foris. Lobeck, A. 1995. Ellipsis: Functional heads, Licensing, and Identification. Oxford: OUP. Longobardi, G. 2001. The structure of DPs: Some principles, parameters, and problems. In The Handbook of Contemporary Syntactic Theory, M. Baltin & C. Collins (eds), 562–603. Oxford: Blackwell. Nissenbaum, J. 2000. Investigations of Covert Phrase Movement. PhD dissertation, MIT. Radford, A. 2000. NP shells. Essex Research Reports in Linguistics 33: 2–20. Rizzi, L. 1990. Relativized Minimality. Cambridge MA: The MIT Press. Saito, M. & Murasugi, K. 1990. N’-deletion in Japanese. UConn Working Papers in Linguistics 3: 87–107. Saito, M. & Murasugi, K. 1999. Subject predication within IP and DP. In Beyond Principles and Parameters: Essays in Memory of Osvaldo Jaeggli, K. Johnson & I. Roberts (eds), 167–188. Dordrecht: Kluwer. Sportiche, D. 1998. Partitions and Atoms of Clause Structure. London: Routledge. Szabolcsi, A. 1983. The possessor that ran away from home. The Linguistic Review 3: 89–102. Torrego, E. 1986. On empty categories in nominals. Ms. University of Massachusetts, Boston. Valois, D. 1991. The Internal Syntax of DP. PhD dissertation, UCLA. Zagona, K. 1982. Government and Proper Government of Verbal Projections. PhD dissertation, University of Washington.

Determination, nominalisation and conceptual processing Helle Dam-Jensen

Aarhus School of Business The purpose of this article is to account for the interpretive process initiated by nominalisations in Spanish. I shall start from the assumption that both morphological nominalisations and syntactic nominalisations, in which a determiner merges with either an infinitive phrase or a complementizer phrase, generate a complex interpretive process due to a clash between an interpretation in terms of ‘entity’, on the one hand, and an interpretation as ‘situation’, on the other. I shall substantiate this claim by analysing nominalised infinitive phrases, nominalised complementizer phrases and morphological nominalisations.

1. Introduction The object of study of this article is Spanish nominalisations. The nominalisations in question are those where the nominalisation is predicative in nature. Such noun phrases are nominalisations. There are different operations of nominalisation. One of those is a syntactic operation which merges a determiner with other categories than nouns. This applies for example to noun phrases headed by an infinitive phrase as illustrated in (1) or noun phrases headed by complementizer phrases as exemplified in (2). A different kind of nominalisation is the morphological nominalisation shown in (3).

(1) ¿Puede ser peligroso (el) beber mucha agua? ‘Can it be dangerous (the) drink a lot of water?’



(2) Era verdad (el) que yo le debía un favor ‘It was right (the) that I owed him/her a favour’



(3) Convención Internacional sobre la Eliminación de todas Formas de Discriminación Racial ‘International Convention of the Elimination of all Types of Racial Discrimination’

It is the purpose of this article to account for the interpretive process initiated by noun phrases such as those exemplified in (1) – (3). It is assumed that nominalisations

 Helle Dam-Jensen

generate a complex interpretive process due to a clash between information provided by the determiner and information stemming from the semantic head; determiners carry instructions for the creation of entities, whereas the semantic heads create situations. This claim will be substantiated below by an account of the interplay between determiners and infinitive phrases, complementizer phrases and morphological nominalisations in the interpretation of nominalisations. For this purpose I will rely on the notions of ‘entity’ and ‘situation’. However, before that the theoretical assumptions on which the analysis rests will be set out. This comprises, on the one hand, an outline of the structure of noun phrases, a description of the layered analysis of noun phrases and a discussion of the notions of ‘entity’ and ‘situation’ which is explored in section 2 below. But firstly, it is necessary to stress that the analysis presented in this article only concerns Spanish, and that the analysis proposed does not necessarily extend to other languages.

2. The structure of determiner phrases In the following, I will adopt the basic determiner phrase (DP) analysis as it was introduced by Abney (1986). In other words, I will assume that a DP is headed by a determiner, which may be realised as an empty category, and that it may take an NP complement headed by a member of the category of nouns. The determiner analysis has been discussed and largely accepted in the analysis of different languages within different theoretical persuasions. Longobardi (1994) argues for the DP-analysis in Romance and Germanic within a generative framework, Hewson (1991) defends the analysis in a cognitive framework for various Indo-European languages, and Eguren (1989), also working within a generative field, offers syntactic evidence for assuming the DP-analysis in Spanish.1 It is a major advantage of the DP-analysis that it allows us to generalise over structures in which the determiner is followed not only by a noun, as in example (3), but also by other categories, as exemplified in (1) and (2). It is the determiner which guides the interpretation of noun phrases. In more specific terms, a determiner carries instructions for the interpretation of the category which it takes as its complement. It is the complement of the determiner which carries the representational semantics which makes it the semantic head of the nominal as a whole. The point of departure is the assumption that determiners as a grammatical class have a constant meaning which instructs the hearer to set up an entity. Determiners instruct the hearer to establish an entity in a universe of discourse. The semantic head, in turn, provides descriptive information about this entity. The interpretive interplay between determiners and semantic heads is then a function of the type of determiner used and the type of category in the scope of the determiner. The simplest 1. It is worth noting that Radford (1993) suggests an analysis of noun phrases which presupposes double-headed structures. On this assumption, noun phrases have an immediate head and an ultimate head.



Determination, nominalisation and conceptual processing 

case is a determiner followed by a noun which denotes what Lyons (1977: 442–443) calls first order entities. However, the case becomes complicated if the semantic head is realised by an infinitive phrase, a complementizer phrase or by a morphological nominalisation, which denote entities of a higher order. In terms of this work, these categories create situations. As a consequence, when these categories form part of a determiner phrase, the information provided by them and the information stemming from the determiner are merged resulting in a recategorisation of the referent as a situation construed as an entity. This may be regarded as a cognitive reformulation of the Kantian conception of hypostatization. However, the way this recategorisation is performed differs according to the type of linguistic expression in the scope of the determiner. With respect to determiners, it is important to underline that the analysis presented in this article pertains only to the definite and the null article. Other scholars have described the semantic effect of nominalisation. An example is Azpiazu (2004: 66) who argues as follows: la consecuencia directa de la abstracción es la condensación del contenido proposicional, ya que, en definitiva, nominalizar un predicado consiste en sintetizar en un sólo concepto la información que cabe en un juicio ‘the direct consequence of the abstraction is the condensation of the propositional content, since, after all, to nominalise a predicate consists in synthesizing the information of a judgment in a single concept’. To this we can add Langacker’s (1991: 34–35) characterisation which goes as follows: “The effect of reification (marked by that) is to “step back” from the situation – including both the event and its relation to the ground – and construe it as an abstract object or proposition capable of being manipulated, evaluated, and commented on. Instead of being asserted, this proposition is taken as one of the participants in a higher-order relationship of belief, denial, evaluation, etc., whence its role as a clausal subject or object”. Although these characterisations to a large extent correspond to my conception of the effect of nominalisation, it would seem that they lack an explanation of how the different parts of nominalised structures contribute to the interpretation and how different types of nominalisation lead to different interpretations. The next section sets up the theoretical guidelines for the analysis of the three types of nominalisations.

3. Conceptual processing and layered analysis of NPs On the view that “the only way for finite agents such as ourselves to master a language is by having knowledge (implicit or explicit) of the relations between the meaning and structure of complex expressions and the meaning and structure of simpler expressions, in addition to knowledge of the meanings of a stock of basic items (…)” (Barwise and Echemendy 1993: 208), one of the major tasks of a semantic study is to account for how the meaning of smaller expressions contributes to the meaning of larger expressions of which they form part. The way in which such relations are accounted for differs, though, depending on the framework applied. As we have seen, it is fundamental

 Helle Dam-Jensen

to the assumptions made above that linguistic expressions initiate a process of interpretation. The performance of this may be called ‘conceptual processing’. This view has at least three implications: firstly, meaning is constructed mentally, secondly, meaning construction is dynamic, and thirdly, meaning construction is a process in which data, linguistic expressions, are converted into something else. In other words, linguistic expressions provide information for the dynamic construction of mental representations. Conceptual processing in general refers to our mental elaboration of any kind of sensory input, including language. Language differs, though, from other kinds of input in being representational (Thrane 2004: 39); that is, it is designed so as to inform, not about itself, but about something else. This again implies that it is designed in a way which allows us to decode it in accordance with a set of more or less fixed conventions. The decoding, or conceptual processing, combines information from language with contextual information, viz. other types of sensory input (Thrane 2004: 44). However, if we want a more substantial understanding of how language contributes to this process, we must be able to account for the relationship between the different units making up complex linguistic expressions, as noted above. Assuming a cognitive view of language implies that the compositionality of linguistic expressions is not to be view as the mere addition of the meaning of the parts of which they are composed. Rather, it should be seen as the cognitive result of combing the semantic effect of the parts of complex expressions with the structural rules of which they are composed (Thrane 1999: 23). In the analysis of nominalised expressions presented in this paper, the focus point is therefore the interpretive effect of each of the parts making up nominalised expressions and the interpretive result of this process. In other words, complex linguistic expressions are assumed to consist of different layers each contributing to the interpretive work to be performed by the hearer. Different suggestions of layered analysis of the meaning of complex expressions have been made in studies of linguistics. Outstanding exponents are Dik (1997) and Langacker (1987; 1991). They differ though in that Langacker’s analysis is conceptual, whereas Dik’s analysis is purely functional in the sense that it focuses on communicative interaction (Harder 1998: 37). Langacker (1987; 1991) offers a cognitive analysis of word classes, grammatical functions and the way in which they combine into larger units. However, in the theoretical foundations of cognitive grammar, Langacker (1987: 3) claims that basing a linguistic analysis on a distinction between lexical and grammatical units is not productive. A major reason for opposing this view, I shall argue, is that the picture of how the different linguistic expressions interact in meaning construction gets muddled. The application of the traditional distinction between grammar and lexicon (or developments of it) facilitates a clear account of the relative categorial status of the constituents of complex expressions. In this line of argument, both Harder (1996; 1998) and Thrane (1983) take as a starting point the relative contribution of the constituents of complex expressions. In Thrane’s (1983) work, a division of labour is made between functional and contentive categories. The former operate on the latter in the sense that functional categories map contentive categories into more



Determination, nominalisation and conceptual processing 

complex expressions. In the analysis of determiner phrases this means that when a determiner is applied to a linguistic expression, the output is a determiner phrase. Harder (1998: 558–59) uses the terms operator and operand for the terms ‘functional and descriptive terms’ respectively2 and specifies that the two expression types are communicatively incomplete. According to this view, functional categories are conceptually incomplete because they require a different element to operate on. Contentive categories, on the contrary, are functionally incomplete in the sense that they lack specifications with regard to function. In the case of determiner phrases, this means that determiners are incomplete as they require a lexical category to operate on3; determiners “do not mean by themselves”. A noun, on the other hand, lacks specifications with regard to function. Such specifications are provided by determiners, which turn nouns into determiner phrases. The case, however, is somewhat more complicated if a determiner has a phrase in its scope. In the sense of Thrane (1983), phrases and lexemes are contentive. Phrases and lexemes form two types of contentive categories located at two different levels in the structure of clauses. Phrases are the result of an operation of grammatical inflectional elements in which a lexeme is mapped into a different category, a phrase, by a functional category. When a determiner has a complementizer phrase in its scope, the contentive category is therefore the result of a different grammatical operation. This, however, does not prevent us from giving a uniform analysis of the composition of determiner phrases. If we accept ‘nounhood’ as a functional category, Fc, which may take a descriptive category, Dc, in its scope, in accordance with the proposal of Thrane (1983), we may generalise over phrase structures in the following way:

A. X(Fc(Dc)) → XP

Applying this general formula to the analysis of the three structures, we get the following three specifications:

B. D(n(Dc)) → DP C. D(Inf(Dc) → InfP) → DP D. D(que(IP) → CP) → DP

What these specifications make explicit is that nominalised infinitive and complementizer phrases involve more complex syntactic structures than the mere combination of a determiner and a noun. Formulated in cognitive terms, I shall suggest that the processing depth required of the language user differs depending on the expression in the scope of the determiner. Determiner phrases formed by a determiner followed by a noun denoting a so-called first order entity initiate a simple process of interpretation, 2. Operators are functional in the sense that they relate content to situation, whereas operands have conceptual content (Harder 1998: 58). 3. This generalisation does not apply to all determiner types in the sense of Abney (1986) as he also includes pronouns.

 Helle Dam-Jensen

whereas determiner phrases in which the determiner is followed by a noun denoting what is known as second or third order entities, such as a morphological nominalisation, an infinitive phrase or a complementizer phrase, require more processing depth. The fact that functional categories determine the reading of the categories by which they are followed, has led to the view that some practitioners within cognitive semantics take them to be instructional (see for example Thrane 1999, Harder 1996, Dam 2001, Bache & Davidsen-Nielsen 1996). On this view, determiners instruct the hearer to create an entity, regardless of whether they are followed by a noun phrase, a complementizer phrase or an infinitive phrase. It was suggested above that nominalisation leads to a recategorisation of the referent. It is important to note that recategorisation does not imply that the grammatical meaning applicable to the semantic head is neutralised; because the determiner is the syntactic head, we are instructed to set up an entity, but due to the internal structure of the semantic head, we are at the same time instructed to establish a situation. As argued above, conceptualisations created on the basis of nominalisation are taken to be a complex of ‘entity’, at a primary level, and ‘situation’, at a secondary level. However, the properties of ‘situation’ may be more or less focalised depending on the semantic head. This assumption will be supported by consideration of how nominalisations behave with respect to grammatical categories applicable to nominal structures. The following parameters will be used: I. Determination II. Modification III. Agreement Within each type of nominalised expression, it will be explained how the interpretation of ‘situation’ is established, and how the operation of nominalisation contributes to the interpretive process. Before going into the concrete constructions, I shall start by discussing the notions of ‘entity’ and ‘situation’.

4. Entities and situations In studies on language, different classifications of entities and situations have been suggested starting from different theoretical angles. Exponents of these are type theory (see fx Cann 1990) in the logical semantic tradition, Lyons’ (1977) ontology of semantic types in semantics and what appears to be a development of this ontology in a functional framework (Dik 1989; Dik & Hengeveld 1991). A common point of these classifications, I shall claim, is a lack of ability to explain the semantics of nominalisation. On a traditional semantic view, linguistic expressions have semantic types. There are two primitive types: t which represents declarative sentences, and e which represents first order entities. This distinction, however, does not reflect the semantics of



Determination, nominalisation and conceptual processing 

nominalised expressions such as those exemplified in (1) – (3). A different typology of semantic types is Lyons’ ontology of entities which distinguishes three types of entities (Lyons 1977): first-, second- and third- order entities. These entities are directly related to linguistic expressions. Like the logical typology of semantic types, Lyons’ ontology is not fine-grained enough to capture the complex semantics attached to nominalised expressions; that is, due to the fact that it presupposes a one-to-one relationship between linguistic expressions and “things in the world”, it cannot account for the semantics of nominal expressions which, because they are nominal expressions, create entities, but which, due to their internal structure, create situations. The same applies to the classification proposed by Dik (1989) and Dik & Hengeveld (1991). This classification is a four-way distinction between individuals, stateof-affairs, propositional content and speech acts. Although the distinction of these entities to a large degree corresponds to Lyons’, the way they are interrelated differs. Where Lyons’ classification is an attempt to account for the relation between language and the world, on the basis of what he himself calls an ontological assumption of “naïve realism” (Lyons: 1977: 442), Dik & Hengeveld’s classification is an attempt to explain the construction of clauses; that is, each layer of clause formation designates an entity of a different order (Dik & Hengeveld 1991: 233). It is a basic premise of this article that language in general carries instructions for the creation of either entities or situations. The former is static, whereas situations are either dynamic or static, although prototypically dynamic. Entities and situations are mental representations formed by language users on the basis of different linguistic expressions. Semantically, entities are characterised as being locatable in situations. As opposed to situations, they are atemporal; they do not establish a dimension in time. Furthermore, entities may be assigned properties. In contrast to entities, situations are created by verbs, finite or nonfinite, combined with the necessary arguments. The way we understand ‘situation’ differs, though, depending on the way in which it is established. According to Lyons (1977: 483), a situation may be a state, an event, a process or an action. A situation is understood here as a conceptualisation of a scene, or a space in the sense of Fauconnier (1994), in which entities may be located. Situations carry information about modality and time either concretely or potentially. A distinction is made between two types of situations: concrete situations and abstract situations. Linguistic expressions which carry information about time and modality create concrete situations and correspond to the semantic type t. Complementizer phrases create concrete situations. Situations of type t are complete because they carry the necessary information about time and modality. Verbal forms which do not carry this type of information, on the other hand, create abstract situations in which time and modality are only potential. This applies to infinitives, gerunds and participles. We shall call this situation type s. Abstract and concrete

 Helle Dam-Jensen

situations have as a common feature that they may be recategorised by means of nominalisation.4 To sum up, nominalisations carry instructions for creating entities, either syntactically or morphologically. By abstracting away from the properties ascribable to ‘situation’, the situation is frozen, so to speak. According to Thrane (1998: 59), nominalisation is a way of speaking about situations as if they were entities. The consequence of this is that the entity thus established may be located in a different situation; the situation established by the sentence of which the nominalisation forms part. It is important to note, though, that nominalisations, at a secondary level, set up situations, in virtue of their internal structure. This account differs from Lyons’ in not assuming three distinct types of entities. Instead it takes into account the fact that language enables us to talk about situations as if they were entities. Below it will be illustrated how focus may vary between ‘entity’ and ‘situation’ in each of the expressions mentioned above depending on how the nominalisation reacts to the parameters mentioned above.

5. The infinitive Being a nonfinite form, the infinitive is unmarked for mood and tense. It is therefore not capable of indicating the speaker’s attitude towards the utterance, nor is it capable of locating the situation described in time deictically in relation to the time of utterance. Temporal orientation is only implied indirectly in virtue of what Guillaume (1965: 115–120) calls le temps in posse, the inner time. This means that it does not imply a concrete dimension in time. Conrad (1982: 118–119) claims with respect to the infinitive in English that it does not refer to individuated occurrences, but that it is compatible with locatable occurrences depending on the context. In other words, the infinitive designates an abstract situation, s, and the suggested features creating its instantiated impact are following: + verbal form -tense -mood

Situation: s

Figure 1.

This is the starting point of the analysis of the interpretation of nominalised infinitives. 4. It should be emphasised that the only clauses which allow nominalisation are relative clauses and complementizer phrases.



Determination, nominalisation and conceptual processing 

5.1

Nominalised infinitive phrases

Infinitives may be nominalised by any member of the determiner paradigm, but they are most frequently nominalised by either the null article, as in (4) below, or by the definite article, as in (5). As the definite article always appears in the masculine, gender, and thus agreement (Skydsgaard 1977: 1924), is irrelevant. (4) Ø(Inf(Lc) → InfP) → dinf → NP Comer es una fiesta ‘eat is a party’ (5) el(Inf(Lc) → InfP) → dinf → NP El comer es una fiesta ‘the eat is a party’ I shall argue that infinitives which allow the appearance of the definite article are nominalised, as in (4) and (5). Other occurrences of infinitives are heads of verbal periphrases, as shown in (6a) – (6b) and (7a) – (7b) below. (6) a. Quiero ser profesor ‘I would like be a teacher’ b. *Quiero el ser profesor ‘I would like the be a teacher’ (7) a. Debe saber que el odio genera odio ‘You ought know that hate generates hate’ b. *Debe el saber que el odio genera odio ‘You ought the know that hate generates hate’

Whenever infinitives which are postposed a finite verb cannot be introduced by an article, as shown in (6b) and (7b), the two verb forms are closely integrated both syntactically and semantically. The word forms which do not allow nominalisation seem to pattern with the verbs which according to Jensen (1990: 101) may form verbal periphrases. Jensen (1990: 101) applies object shift as the main criterion for displaying the close unit between two verbs, a finite and a nonfinite form, forming a verbal periphrasis. This is shown in (8a) – (8b). (8) a. Debe saberlo ‘You ought know it’ b. Lo debe saber ‘It you ought know’ Semantically, the finite form is characterised as having as its primary task to indicate mood, tense and aspect, its “independent lexical meaning” is weakened, so to speak. Contrary to (6) and (7) the two verb forms of (9) below may be separated by an article. They are therefore more loosely related both syntactically and semantically. (9) a. Los médicos aceptan recetar cannabis ‘The doctors accept prescribe cannabis’ b. Los médicos aceptan el recetar cannabis ‘The doctors accept the prescribe cannabis’

 Helle Dam-Jensen

Unlike (8), (9) does not allow object shift, as shown in (10) below. This supports the assumed connection between nominalisation and formation of verbal periphrases. (10) *Los médicos lo aceptan recetar ‘The doctors it accept prescribe’ Furthermore, there seems to be a correlation between superordinate structures with a factive meaning and infinitives in the object position which may be introduced by the definite article, as noted by Hernanz (1999: 2279). However, this does not explain that one and the same context allows the appearance of the definite article as well as the null article without the meaning of the sentence being altered, as evidenced by (9a) – (9b).5 The fact that the constellation of aceptan and recetar in (9b) allows the appearance of the definite article, in front of the infinitive form, indicates that the verb forms are neither syntactically nor semantically integrated. According to Herslund & Korzen (1999: 169), a loose degree of integration between the finite form and the infinitive form patterns with finite forms with a precise and independent semantics [aceptar], whereas a close integration between the two forms patterns with finite forms with a less precise semantics [querer]. It should be noted, though, that not all verbs which may enter into verbal periphrases are equally imprecise in meaning. A verb such as [lograr] for example (‘succeed’) has a more precise semantics than [querer]. I shall propose that structures such as (9a) and (9b) create situations conceptualised as an entity, whereas infinitives which merge with an auxiliary forming verbal periphrases, as shown in (6a), create situations described by the infinitive and modalised and located in time by the auxiliary. In more specific terms, nominalised infinitives, whether nominalisation is created by a null or a definite article, instruct the hearer to create an entity which may be located in a situation. Infinitives which form part of a verbal periphrasis, on the other hand, establish situations. As it is the nonfinite form which is semantically weighty, it is this form which provides the descriptive content of the situation. The finite verb, in contrast, has as its primary task to relate this situation to the utterance situation in virtue of being marked for tense, mood and aspect. Having thus described the syntactic process leading to nominalisation and its general semantic effect, I shall explain what is implied by conceptualising a situation designated by an infinitive as an entity. To this end, it will be discussed how nominalised infinitives behave with respect to the parameters described above. I will start by approaching agreement on the basis of the following example. (11) No es el estudiar lo que no me gusta ‘It is not the study [NEUTER ARTICLE] which I do not like’ This example is a copular construction in which the subject is realised by the nominalised infinitive, el estudiar, and the subject predicative is realised by a determiner phrase, lo que no me gusta, semantically headed by a relative clause and mapped into a determiner phrase by the neuter article, lo. Agreement is not formed by grammatical gender 5.

See pages 16-17 for elaboration of this point.



Determination, nominalisation and conceptual processing 

since the infinitive is introduced by the definite article in the masculine, but in virtue of the semantic content. I shall claim that the neuter article carries the information that the entity created is to be conceived as inanimate and uncountable.6 As agreement is not formed grammatically between the two articles, but semantically between the information provided by the neuter article and the lexical content of the infinitive, it will be assumed that it is ‘situation’ which is emphasised, rather than ‘entity’. Next we shall see how nominalised infinitives behave with respect to determination. It is a basic assumption of a cognitive framework that every formal difference between constructions results in a conceptual difference. Accordingly, determiner phrases semantically headed by nouns differ in meaning depending on whether they are introduced by the definite article or by the null article: in general, an existential presupposition holds in determiner phrases introduced by the definite article, but not in those introduced by the null article. However, in the case of determiner phrases headed by the infinitive, it is remarkable that this difference does not seem to hold. Examples (12) and (13) below only differ in type of determination, but not in presupposition. (12) Lamento el ser a veces tan espontánea ‘I regret the be sometimes so spontaneous’ (13) Lamento ser a veces tan espontánea ‘I regret be sometimes so spontaneous’ [Lamentar] belongs to the group of verbs which Kiparsky & Kiparsky (1971: 347) call ‘factive predicates’. Syntactically, this group of verbs shares a number of characteristics. One is that they allow the appearance of ‘the fact’ introducing a ‘that’-clause (see section 6.1 on equivalent structures in Spanish). Semantically, factive verbs carry presupposition. In this line of argument, the lexical content of the verb [lamentar] indicates that presupposition holds in both (12) and (13); that is, there is a situation describable by ‘ser a veces tan espontánea’. In both (12) and (13), the infinitive is nominalised. But, although nominalisation is realised by the null article in (13), the presupposition holds due to the lexical content of [lamentar] although this contrasts with the general meaning of the null article. It seems that the general semantic opposition between the definite article and the null article is weakened, if not neutralised. The difference must then reside in something else. According to Alarcos Llorach (1994: 143), infinitives introduced by the definite article may have an emphatic effect. The difference is therefore a matter of style. Seeing that the semantic opposition between the definite article and the null article does not hold when determiner phrases are semantically headed by infinitives, it is argued that focus is on ‘situation’, rather than on ‘entity’. This supports the conclusion drawn above. However, in favour of the entity reading is the fact that infinitives, just like nouns, may be introduced by any member of the determiner paradigm, 6. Leonetti (1999b: 834) puts it in the following terms: la capacidad de denotar únicamente entidades inanimadas o no humanas, o quizá como la capacidad de denotar sólo lo no contable o no discreto ‘the capacity for denoting only inanimate or nonhumans entities, or perhaps as the capacity for denoting only the uncountable or nondiscrete’.

 Helle Dam-Jensen

and that other types of determiners than the definite article and the null article have the meaning which they are generally assigned. In conclusion, the parameter of determination does not clarify the vacillation in reading between ‘entity’ and ‘situation’. Nor does the parameter of modification provide a univocal picture of interpretation. As illustrated in (14) – (17) below, the infinitive may be subject to different kinds of modification which emphasise either ‘entity’ or ‘situation’. (14) De este modo, los futuros esposos se ahorran el buscar desesperadamente restaurantes y más restaurantes ‘In this way the future married couple save the look desperately for restaurants and more restaurants’ (15) Los del otro lado, se contentaban con ver el ir y venir de los compradores que recorrían la Avenida Central ‘Those from the other side made do with watching the go and come of the shoppers who went up and down Avenida Central’ (16) ¿Es correcto el comer animales? ‘Is it correct the eat animals?’ (17) Podía ir cualquier noche a vivir la vida loca de los discos, (...), el agitado ir y venir de las calles (...) ‘He could any night live the crazy life of the discos, (...), the agitated go and come of the streets (…)’ When the infinitive is modified by an adverb phrase, as shown in (14), a situation reading is emphasised. This also applies to infinitives combining with a prepositional phrase semantically headed by a noun which indicates an agent, as in (15), and to infinitives with a determiner phrase complement, as in (16), in which animales is the logical direct object of the infinitive. In (15), however, it should be emphasised that the article is obligatory when nominalised infinitives are modified. This seems to weaken the situation reading. In (17), the infinitive is modified by an adjectival phrase which generates emphasis on ‘entity’. Since infinitives most frequently appear to be modified by adverb phrases, also the parameter of modification emphasises an interpretation as ‘situation’. In conclusion, the interpretation of nominalised infinitives is a merger of the information from the infinitive form and from the operation of nominalisation. The lack of the dimensions of tense and mood in the infinitive form is in accordance with the characteristics of ‘entity’, but due to the fact that “they are potentially there”, the interpretation of ‘situation’ is also manifest. This vacillation between ‘situation’ and ‘entity’ is made explicit in the analyses of the way in which nominalised infinitives behave with respect to the parameters just described. The effect of nominalisation may then be described as an operation which emphasises the characteristics of the infinitive form which are in accordance with the ‘entity’ reading, but the “verbal nature” of the infinitive generates an interpretation as ‘situation’ thus restricting the entity reading. The parameters applied above show, however, that ‘situation’ is predominant compared to ‘entity’.



Determination, nominalisation and conceptual processing 

6. Complementizer phrases The complementizer, que7, heading complementizer phrases carries the information that complementizer phrases are dependency structures. This implies that situations established by complementizer phrases are situations which are dependent on, or embedded in, a different situation. Every piece of information provided by complementizer phrases is then to be evaluated against the information provided by the superordinate structure. Complementizer phrases are furthermore characterised as being indirectly marked for tense and mood in the verb phrase. Therefore they both locate the situation described in time and are capable of indicating the speaker’s attitude towards the situation described. As opposed to infinitives, complementizer phrases then establish concrete situations, t. The features which constitute the instantiated impact of complementizer phrases may be summarised as follows: + verbal form combined with necessary arguments + dependency

Situation: t

+ tense + mood

Figure 2.

6.1

Nominalised complementizer phrases

After having established the semantic features of complementizer phrases, the effect of nominalisation will be considered. Nominalisation of complementizer phrases is established either by the null article, as in (18a), by the definite article, as in (18b), or by the definite article extended by hecho de (“fact” [prep]), as in (18c). (18) a. Ø (que(IP) → InfP) → dcp → NP Me extrañaba que no hubiera habido especulaciones sobre quién sería encargado de servir en la boda real ‘It seemed strange to me that there had been no speculations as to who should be in charge of serving at the royal wedding’ b. El (que(IP) → InfP) → dcp → NP Me extrañaba el que no hubiera habido especulaciones sobre quién sería encargado de servir en la boda real ‘It seemed strange to me the that 7. I shall ignore other types of complementizer phrases than those which are introduced by que.

 Helle Dam-Jensen

there had been no speculations as to who should be in charge of serving at the royal wedding’ c. El(n(Lc) (PREP)(que(IP)) → InfP) → dcp → NP Me extrañaba el hecho de que no hubiera habido especulaciones sobre quién sería encargado de servir en la boda real ‘It seemed strange to me the fact [prep] that there had been no speculations as to who should be in charge of serving at the royal wedding’ Whether a complementizer phrase allows nominalisation or not, depends on the type of situation created by the matrix clause, but it appears to be complicated to unveil a precise pattern of phrases that do allow (or even require) this extension of the complementizer phrase and those which do not. The fact that the meaning of hecho implies factuality, and that the meaning of the definite article creates presupposition, may give support to the idea that the meaning of the complementizer phrase establishes presupposition. This is in fact the main point of Delbeque & Lamiroy’s (1999) analysis of the use of el (hecho de) as introducing complementizer phrases. According to Delbeque & Lamiroy (1999: 1969), complementizer phrases introduced by the definite article (+ hecho de) have a presupposed meaning, thus corresponding to the general meaning of the definite article. Along the same lines, Leonetti (1999a: 2101) claims that it is mainly predicates of superordinate structures with a factive meaning, those which create second order entities, which allow nominalisation. As opposed to this, verbs which do not imply factuality and verbs with a “pure propositional content” (such as verbs which designate thoughts and hypothesis) do not as easily allow nominalisation of the complementizer phrase. To substantiate this claim, he provides the following examples8. (19) Nos dijeron (*el hecho de) que ellos se conocían. ‘They told us the fact [prep] that they knew each other’ (20) Necesitas (*el hecho de) que alguien te ayude. ‘You need the fact [prep] that someone helps you’ (21) Prometo (*el hecho de) que lo tendré listo mañana. ‘I promise the fact [prep] that I will have it ready for tomorrow’ Contrary to this, nominalisation is allowed in example (18) above because the superordinate predicate does create presupposition. Saying: Me extrañaba (el hecho de) que no hubiera habido especulaciones sobre quién sería encargado de servir en la boda real presupposes that ‘there had been no speculations as to who should be in charge of serving at the royal wedding’. However, the fact that factive predicates in general seem to allow nominalisation of complementizer phrases does not explain that some predicates in which

8. The emphasis is mine.



Determination, nominalisation and conceptual processing 

presupposition does not hold allow nominalisation all the same. This is evidenced by the following three examples. (22) Temo (el hecho de) que pueda haber otras preguntas más difíciles ‘I fear (the fact [prep]) that there may be other more difficult questions’ (23) Seguro que muchos preferimos (el hecho de) que no fuese aceptado ‘It is certain that we are many that preferred (the fact [prep]) that it would not get accepted’ (24) Incluso es posible (el hecho de) que usted mantenga su página en otro servidor ‘Moreover it is possible (the fact [prep]) that you keep your web site in a different server’ Neither temer, preferir nor ser posible create a factual situation. To this it should be added that some predicates even require the appearance of the definite article. Delbecque (2000: 64) gives the following example. (25) La grúa no resuelve {el hecho de / el problema de / *ø} que no haya aparcamiento ‘The crane9 does not solve {the fact [prep] / the problem [prep] / *ø} that there is no parking’ According to native informants, it is the meaning of presupposition of the superordinate structure which makes the definite article compulsory in this sentence. This, however, does not appear to be the general case as the majority of predicates does not seem to require the appearance of the definite article. In example (26), it is not the predicate of the superordinate structure which requires the definite article, but the syntactic composition of the sentence. (26) Pero aseguró que el hecho de que haya sido emitido por el tribunal de Cuentas, en el que son mayoría los miembros nombrados por el PP, le confiere “un extramado grado de neutralidad” ‘But he assured that the fact [prep] that it had been emitted by the court of auditors, of which the majority has been appointed by Partido Popular, gives him an “extremely high degree of neutrality”’ The fact that the complementizer phrase introduced by el hecho de is the subject of a different complementizer phrase makes the definite article (+hecho de) compulsory in order to avoid a clash between two different instances of que. In conclusion, a precise hypothesis cannot be formulated on the basis of factuality which explains both that predicates which create presupposition allow nominalisation of the complementizer phrase, and that some predicates which do not create presupposition also allow nominalisation. With respect to interpretation, this means that not all situations established by complementizer phrases allow recategorisation. Complementizer phrases of which the 9. Towing cars away.

 Helle Dam-Jensen

superordinate structure does not allow nominalisation create a situation which is not to be conceptualised as an entity, whereas complementizer phrases of which the superordinate structure does allow nominalisation, makes the entity reading available. With regard to choices of article (null article vs. definite article), as we saw with respect to infinitives, it does not have substantial effect on interpretation whether nominalisation is a product of a null article or of the definite article (+hecho de). In the words of Leonetti (1999a: 2100–2101): La secuencia el hecho de, (...) refleja y resume el contenido semántico de la subordinada a la que precede, sin imponer restricciones gramaticales (...). A la carga semántica reducida de esta secuencia nominal se debe el hecho de que pueda ser elidida sin dar lugar a incorreciones (...) (“The sequence el hecho de, (...) reflects and resumes the semantic content of the subordinate phrase which it precedes, without imposing grammatical restrictions (…) That this nominal sequence has reduced semantic importance is due to the fact that it may be deleted without making the sentence incorrect”).

The difference between complementizer phrases introduced by the definite article and those introduced by the null article may, as we saw with respect to infinitives, in the majority of cases be ascribed to style. One may argue that the effect of nominalisation is a matter of degree in the sense that complementizer phrases introduced by the definite article emphasise the entity reading compared to those introduced by null articles, whereas complementizer phrases introduced by el hecho de are even more entity-like as they are dominated by a noun, hecho. In the same line of argument as proposed with respect to infinitives, I shall propose that the parameter of determination creates a greater emphasis on ‘situation’ compared to ‘entity’. This proposal is supported by the fact that complementizer phrases do not allow other syntactic variations within the determiner paradigm than between the null article and the definite article. Further support to the claim that focus is on ‘situation’ rather than on ‘entity’, is that complementizer phrases do not allow modification. Moreover, the example below shows that the parameter of agreement is in accordance with this conclusion. (27) Incluso es posible (el) que usted mantenga su página en otro servidor, lo que le ahorrará de gastos ‘It is furthermore possible (the) that you keep your site on a different server, [neuter article] which may save you expenses’ In this example, agreement is formed between the complementizer phrase, (el) que usted mantenga su página en otro servidor, and the neuter article, lo, which introduces the relative clause, que le ahorrará de gastos. As we have already seen, the neuter article is characterised as carrying the information that the entity set up is to be conceived as inanimate and uncountable. This entity is capable of subcategorising the entity established by the nominalisation because it is in accordance with its meaning. Therefore agreement is formed semantically between the neuter article and the content of the complementizer phrase, rather than grammatically by gender.



Determination, nominalisation and conceptual processing 

In conclusion, complementizer phrases generate concrete situations which are dependent on a superordinate situation. When they merge with a determiner, the situation is recategorised as an entity; the situation is frozen and located as an entity in a superordinate situation. Due to the fact that situations established by complementizer phrases are concrete situations, they do not acquire the properties which belong to ‘entity’ − it is the mere act of “freezing” or condensation of the situation which makes the entity reading explicit − and the interpretation of complementizer phrases is therefore closer to ‘situation’ than to ‘entity’. Lastly, it should be mentioned that complementizer phrases may be topicalised, which, naturally, contributes to the way the situation is understood. Information structure could therefore be employed as a further parameter in the evaluation of the meaning of nominalised complementizer phrases. However, using this as a parameter would require the implication of a quite different set of theoretical tools, which is why I shall leave this aspect unaccounted for.

7. Morphological nominalisations Nominalisation refers both to the morphological operation of derivation in which a suffix is added to a lexical root and to the output of this process. Supposing that the composition of nominalisation is a hierarchy of morpho-syntactic operations, we need an explanation of how we get from the root to the category of nominalisation. This process may be explained as a mapping relation between levels. The basic level is the root. It is characteristic of nominalisations that they may share root with a verb.10 The lexical structure of roots may be defined by their selectional properties. In the case of nominalisation, the selectional properties are defined as carrying the same argument structure as the related verb. Depending on the suffix added to the root, the output is either a noun or a verb as made explicit below:

A. -rir (ADQUI) → adquirir ‘acquire’ B. -sición (ADQUI) → adquisición ‘acquisition’

The suffix is then a functional category which maps the lexical root into a category at a different level: a lexical type. In the first formula, the suffix –rir, maps ADQUI into a verb, in the second formula, –sición maps ADQUI into a noun, a nominalisation. In Spanish, nominalisation may be formed by adding different kinds of suffixes to the root. I shall only comment on the type which is formed by the suffix –sición. The lexeme thus formed may then be mapped into a determiner phrase by a determiner:

C. La (adquisición) → la adquisición ‘the acquisition’

10. I shall ignore the fact that nominalisations also may be lexically related to other word classes.

 Helle Dam-Jensen

The layered structure of nominalisation provides us with different clues for the interpretive process. The root carries information for establishing a set of relations in virtue of its selectional properties. It is then these relations combined with the necessary constituents which create a situation. However, the situation is manifested differently depending on whether it is mapped into a verb or into a noun. If it is mapped into a verb, the lexical structure is manifested in accordance with the characteristics of situations; we are instructed to set up a situation in which entities may be introduced. If it is mapped into a noun, on the other hand, we are instructed to set up an entity in a situation. However, due to the internal structure of morphological nominalisations, this entity itself creates a situation in which entities may be located. Being nouns, morphological nominalisations designate abstract situations as shown below: + suffix:- ción -tense -mood

Situation: s

Figure 3.

Situations established by morphological nominalisations may be perceived as either static or dynamic. The reading is a function of the context as evidenced by the following two examples. (28) Pues esa es mi nueva adquisición, mejor dicho, la de mi madre ‘Well this is my new acquisition, in other words, the of my mother’ (29) No es la adquisición de la tecnología lo que diferencia una empresa de otra ‘It is not the acquisition of technology [neuter article] which differentiates one company from another’ In (28), the subject and the verb, esa es, give us information for understanding the situation set up by la adquisición as static; the entity set up by the nominalisation is understood as a first order entity. In (29), on the other hand, la adquisición establishes a dynamic reading due to the referential link established with the determiner phrase, lo que diferencia una empresa de otra. This example is a copular construction in which the referent of the subject predicative subcategorises the referent of the subject. The determiner phrase in the subject position, la adquisición de la tecnología, is semantically headed by a nominalisation. The determiner phrase in the subject predicative position, in turn, lo que diferencia una empresa de otra, is realised by a relative clause headed by the neuter article, lo. The subject predicative does not agree in grammatical gender with the subject, which is in the feminine, but in virtue of the lexical content. With respect to the parameter of agreement, this leads to the conclusion that emphasis



Determination, nominalisation and conceptual processing 

is on situation in (29) as agreement is formed between la adquisición and the neuter pronoun, lo, which instructs us to perceive the referent of la adquisición as uncountable. In (28), in which the entity set up is a first order entity, agreement is established grammatically by gender with the determiner phrase, la de mi madre. Because of the internal structure of morphological nominalisation, emphasis is still on situation; only in this case, the situation is conceived as being static as stated above. I claimed above that the definite article when introducing determiner phrases semantically headed by nouns always creates the same meaning: presupposition. This, consequently, also applies to determiner phrases headed by nominalisations. Moreover, being nouns, morphological nominalisations allow variation within the determiner paradigm. Consequently, according to this parameter, ‘entity’ is emphasised. This also applies to the parameter of modification as morphological nominalisations may be modified by adjectival phrases. As only the parameter of agreement underlines the situation reading (and this does not even apply to all types of nominalisation), it is inarguable that the interpretation established by morphological nominalisations emphasises the properties belonging to ‘entity’, rather than those belonging to ‘situation’.

8. Concluding remarks It is a common feature of the three structures analysed in this article that they create a mapping relation between ‘entity’ and ‘situation’. This mapping implies that we are instructed to create situations as if they were entities. By abstracting away from time and modality, the situation is frozen. The entity created is located in a situation, but due to the internal structure of the semantic head of the determiner phrase used, information is provided for creating a different situation, at a “secondary level”. The linguistic information for creating a mapping between ‘entity’ and ‘situation’ is provided at different levels in the three structures. In the case of morphological nominalisation, it occurs at the morphological level, whereas it is established at the phrasal level in infinitive and complementizer phrases. A consequence of this appears to be that the definite article differs functionally depending on the type of structure in the scope of the determiner. Determiners are functional elements which create determiner phrases. Semantically, this results in the establishment of an entity in a universe of discourse. In the case of morphological nominalisations, the definite article complies with this function, but with regard to infinitive and complementizer phrases, the definite article both recategorises the situation created by the semantic head as an entity and introduces this entity into a universe of discourse. This difference between morphological nominalisation on the one hand and infinitive and complementizer phrases on the other, is due to the linguistic level at which nominalisation takes place. It may be assumed that the effect of this is that the general meaning of the determiner paradigm is weakened in the case of syntactic nominalisation; “full variation” within

 Helle Dam-Jensen

the paradigm is not allowed, and the general meaning of the definite article is suspended. It may furthermore be argued that the fact that the definite article is in the masculine, and not in the feminine, is due to it being unmarked; it has a greater functional potential. Summing up, according to the parameter of determination, morphological nominalisation has greater emphasis on ‘entity’ in comparison with infinitive and complementizer phrases. This also applies to the parameter of modification. In contrast to complementizer phrases, infinitives do allow modification, but as adjectival modification is rare, it is ‘situation’ which is underlined according to this parameter. Naturally, this applies even more so to complementizer phrases. The parameter of agreement shows that the three structures have as a common point that they all agree with the neuter article, thus favourising an interpretation in terms of ‘situation’. In conclusion, although all three types create situations, they each represent a specific manifestation of this semantic type. Complementizer phrases create concrete situations, which per se focalise ‘situation’ rather than ‘entity’. Furthermore, the parameters applied also favour the situation interpretation. Infinitives and morphological nominalisations, on the other hand, create abstract situations. According to the analyses above, they differ, though, in that morphological nominalisations emphasise ‘entity’, whereas infinitives emphasise ‘situation’. The determiner paradigm has the function at the syntactic level of turning categories into phrases, and at the semantic level they introduce entities into a universe of discourse and specify the way in which we understand such entities. However, a further dimension of this is that they may result in recagorising certain grammatical categories, thus providing speakers with the possibility of talking about situations as if they were entities.

References Abney, S. 1986. Functional elements and licensing. Paper presented to GLOW 1986. Gerona, Spain. Alarcos L.E. 1994. Gramática de la lengua española. Madrid: Editorial Espasa Calpe. Azpiazu, S. 2004. Las estrategias de nominalización [Studien zur Romanischen Sprachwissenschaft und interkulturellen Kommunikation]. Frankfurt: Peter Lang. Bache, Carl and Niels Davidsen-Nielsen. 1996. Tempus og aspekt i engelsk. Ny forskning i grammatik [Fællespublikation 4], L.F. Jakobsen & G. Skytte (eds), 26–48. Odense: Odense Universitet. Barwise, J. & Etchemendy, J. 1993. Model theoretic semantics, Foundations of Cognitive Science, Vol. 1, M. I. Posner (ed.), 207–243. Cambridge MA: The MIT Press. Cann, R. 1990. Formal Semantics. Cambridge: CUP. Conrad, Bent. 1982. Referring and Non-referring Phrases (Publications of the Department of English 11). Copenhagen: Akademisk forlag.



Determination, nominalisation and conceptual processing  Dam, H. 2001. A Referential Analysis of Definite Noun Phrases in Spanish. PhD dissertation, The Aarhus School of Business. Delbecque, N. & Lamiroy, B. 1999. La subordinación sustantiva: Las subordinadas enunciativas en los complementos verbales. In Gramática Descriptiva de la Lengua Española, Vol. II, I. Bosque & V. Demonte, 1965–2104. Madrid: Editorial Espasa Calpe. Delbecque, N. 2000. La estructura [El Nabstracto de que + completiva]: Variación formal y functional. En torno al sustantivo y adjetivo en el español actual, G. Wotjak (ed.). Frankfurt:Vervuert. Dik, S. & Hengeveld, K. 1991. The hierarchical structure of the clause and the typology of perception-verb complements. Linguistics 29: 231–259. Dik, S. 1989. The Theory of Functional Grammar [Functional Grammar Series 9]. Dordrecht: Foris. Dik, S. 1997. The Theory of Functional Grammar. Berlin: Walter de Gruyter. Eguren, L. 1989. Algunos datos del español en favor de la hipótesis de la frase determinante. Revista Argentina de Lingüística 5(1–2): 163–203. Fauconnier, G. 1994. Mental Spaces. Cambridge: CUP. Guillaume, G. 1965. Temps et verbe. Paris: Librairie Honoré Champion. Harder, P. 1996. Functional Semantics: A Theory of Meaning, Structure and Tense in English. Berlin: Mouton de Gruyter. Harder, P. 1998. Function, cognition, and layered clause structure. In Cognitive Semantics, J. S. Allwood & P. Gärdenfors, 37–66. Amsterdam: John Benjamins. Hernanz, M. L. 1999. El infinitivo. In Gramática descriptiva de la lengua española, Vol. II, I. Bosque & V. Demonte, 2197–2352. Madrid: Editorial Espasa Calpe. Herslund, M. & Korzen, H. (eds.) 1999. Det franske sprog: Kapitel VIII, 1. Den komplekse prædikation 1. Hewson, J. 1991. Determiners as heads. Cognitive Linguistics 2(4): 317–337. Jensen, K. 1990. Spansk basisgrammatik. Århus: Akademisk Forlag. Kiparski, P. & Kiparsky, C. 1971. Fact. In Semantics. An Interdisciplinary Reader in Philosophy, Linguistics and Psychology, D. Steinberg & L.A. Jakobovits, 345–369. Cambridge: CUP. Langacker, R.W. 1991. Foundations of Cognitive Grammar, Vol. II. Descriptive Application. Stanford CA: Stanford University Press. Langacker, R.W. 1987. Foundations of Cognitive Grammar, Vol. 1.Theoretical Prerequisites. Stanford CA: Stanford University Press. Leonetti, M. 1999a. La subordinación sustantiva: Las subordinadas enunciativas en los complementos nominales. In Gramática descriptiva de la lengua española, Vol. I. I. Bosque & V. Demonte, 2083–2104. Madrid: Editorial Espasa Calpe. Leonetti, M. 1999b. El artículo. In Gramática descriptiva de la lengua española, Vol. I. I. Bosque & V. Demonte, 787–890. Madrid: Editorial Espasa Calpe. Longobardi, G. 1994. Reference and proper names: A theory of N-movement in syntax and logical form. Linguistic Inquiry 25(4): 609–665. Lyons, J. 1977. Semantics. Vol. I-II. Cambridge: CUP. Radford, A. 1993. Head-hunting: on the trail of the nominal Janus. In Heads in Grammatical Theory, G.G. Corbett, N. Fraser & S. McGlashan (eds). Cambridge: CUP. Skydsgaard, S. 1977. La combinatoria sintáctica del infinitivo español, I-II. Copenhague: Estudios románicos. Szabolcsi, A. 1983. The possessor that ran away from home. The Linguistic Review 3: 89–102. Thrane, T. 1983. On the Universality of AUX. Acta Linguistica Hafniensia. Vol.18(2): 154–200.

 Helle Dam-Jensen Thrane, T. 1998. Nominaler, nominaliseringer og semantisk kompleksitet. Hermes 21: 39–66. Thrane, T. 1999. Understanding functionality. Ms. Thrane, T. 2004. Hvorfor er sproget så svært at forstå når det er så let at forstå? Tidsskrift for Sprogforskning 2(2): 37–70.

The semantics and pragmatics of the possessive determiner Georges Kleiber

Université Marc Bloch “This article deals with the semantics and pragmatics of the possessive adjective in French. As a point of departure we do not base this article, as it is traditionally done, on the general semantic category of ‘possession’, but instead we take our point of departure in the competition of the possessive determiner and the definite article in the context of associative anaphors. This article shows firstly, that the possessive adjective requires an a priori dependant asymmetry between the two entities E1 and E2 which are implied by a possessive description, secondly, that this asymmetry is satisfied either by the ontological status of the entities involved; by the semantico-lexical relation between the Ns of the entities involved which indicate an a priori subordination; or by a discursive relation between two specific groups or classes of referents established by the linguistic or situational context.”

1. Introduction1 Generally, research on the possessive has either focused on its syntactic restrictions and the (re)-presentation of its pronominal origins (Godard, 1986, Gross, 1986, ZribiHertz, 1999, etc.), or has examined possession as a semantic category, especially in typological work where the approach has been more or less openly universalist and cognitivist (Seiler, 1983, Barker, 1995, Taylor, 1996, Heine, 1997, Baron, Herslund and Sørensen, eds, 2001, etc.). There has also been much careful description and analysis of the semantic and pragmatic relations which can hold between the ‘possessor’ and the ‘possessed’ in particular languages (Baron et Herslund, eds, 1997, Tasmowski, ed., 2000, Heinz, 2003b, etc.). It seems to me however that it would be interesting to adopt a different approach, and to look at the textual properties of the possessive, in particular 1. This paper summarises, develops and complements, with appropriate modifications and corrections, previous work on this subject by Kleiber (2003 and 2004). The author wishes to thank Christopher Gledhill for translating this article.

 Georges Kleiber

when it coincides with the corresponding use of an associative definite article, as in the following examples: (1) a. Il s’abrita sous un vieux tilleul. Le tronc était tout craquelé. (He sheltered under an old limetree. The trunk was full of cracks.) b. Il s’abrita sous un vieux tilleul. Son tronc était tout craquelé. (He sheltered under an old limetree. Its trunk was full of cracks.) We shall not be dealing here with the questions that are routinely asked about the possessive, namely: What is possession? Is there a general semantic relation which accounts for all cases of the possessive? or What are the basic relations created by the possessive and how do we explain their implications? These are clearly legitimate and relevant issues, and indeed they turn out to be essential in the argument developed by Baron, Herslund and Sørensen (2001), when they talk of “dimensions of possession”.2 But unfortunately the level of generality and the terminological debate which these questions raise do not leave much room for the equally important issue of the discursive role of the possessive determiner. In fact, in the literature on textual reference the possessive determiner is hardly mentioned. It seems to me that it would be extremely fruitful therefore to take this indirect textual approach rather than attempt a frontal attack on the possessive. The approach taken here only seems to have been taken previously by Fradin (1984).3 That is, this paper starts from the discursive context established by an associative anaphor (Kleiber, 2003a) and the paper examines how the possessive determiner operates on the basis of this. As we will see below, this method makes it possible to shed new light on the semantics and pragmatics of the possessive, and perhaps even makes it possible to relate the definite article to the overall approach. Although this is a preliminary study, it may in fact lead to intuitive and original observations about the definition and role of this referential marker. The possessive adjective and the definite article can be directly compared in situations where both are markers of anaphoric reference corresponding to a predictable set of discourse criteria. These criteria correspond to two conditions: (I) There are two entities, the first E1, normally realised by a nominal antecedent (thus N1), and a second entity (E2), which follows E1 and is also nominal (N2) and which is considered to be an anaphoric expression. The phrases NP1 and NP2 which denote E1 and E2 cannot be coreferential, since they are two different entities (E1 ≠ E2).

2. The introduction to the book by Herslund and Baron has the same title (2001: 1-25). 3. As well as Mitsuru Ohki in articles published in Japanese in 1991 and 1993, which unfortunately it has only been possible to take account of superficially on the basis of abstracts published in French.



The semantics and pragmatics of the possessive determiner 

(II) The entities E1 and E2 can be related by use of a preposition such as de in French or of in English as in le N2 de NP1or the N2 of NP1.4 These conditions allow for both an indirect associative anaphor with the definite article, or a possessive anaphor, where the possessive adjective indicates coreference with the referent E1 introduced by NP1. In our example tilleul / limetree (E1) and tronc / trunk (E2) can be linked either by the associative use of the definite article (un tilleul —> le tronc / a limetree —> the trunk ) or by a possessive adjective (un tilleul —> son tronc / a limetree —> its trunk). In both cases, the referent of NP2 can be reformulated as a prepositional phrase: le N2 de NP1 (le tronc du tilleul) / the N1 of NP1 (the trunk of the limetree). Not all the different E1 – E2 combinations corresponding to conditions (I) and (II) allow for the use of the associative definite article and the possessive determiner. There are three possibilities: A) the possessive adjective can do the job, but the definite article cannot, B) the definite article works, but the possessive does not, or C) both can be used. The source of this variation, as we will see when we look at each case below, is the kind of entities encountered as E1 – E2. In the following sections each of these cases will be examined and explained in turn. In the final section, the main differences between possessive adjective and definite article are explained.

2. The possessive does the job, but not the definite article There are a number of cases where the definite article cannot be used.5 The first involves instances where E2 is a property of an animate or inanimate E1, as in the following two examples: (2) a. b.

? L’homme enleva sa casquette. La calvitie plut à tout le monde. (‘lit:’ The man took off his cap. Everyone was delighted by the baldness.) ? Paul a acheté une Clio, parce qu’il a été séduit par la sobriété. (‘lit:’ Paul bought a Clio, because he was seduced by the simplicity.)

While there is nothing from a cognitive point of view to prevent us from understanding the baldness and the simplicity as in the baldness of a man taking off his cap and the simplicity

4. Since the main focus is on French examples, I do not dwell here on the use of the genitive in English. 5.

A detailed analysis of this problem is set out in Kleiber (1999a and 2001a).

 Georges Kleiber

of the Clio, it appears that the definite article does not allow us to make this anaphoric connection. On the other hand, the possessive6 appears to fit the bill perfectly: (3) a. b.

L’homme enleva sa casquette. Sa calvitie plut à tout le monde. (The man took off his cap. Everyone was delighted by his baldness.) Paul a acheté une Clio, parce qu’il a été séduit par sa sobriété. (Paul bought a Clio, because he was seduced by its simplicity.)

A second instance involves an awkward use of the definite article when the E2 refers to a process (an event, an activity, etc.). Thus in the following examples, the article cannot be used to emphasise that the words and gestures are those of Paul: (4) a. b.

? Paul entra. Les paroles réveillèrent l’assemblée. (‘lit:’ In came Paul. The words stirred the audience.) ? Paul entra. Les gestes étaient brusques. (‘lit:’In came Paul. The gestures were brusque.)

There is no a priori cognitive reason why the relevant associative inference cannot be made. But again, the possessive allows us to make the necessary link: (5) a. b.

Paul entra. Ses paroles réveillèrent l’assemblée. (In came Paul. His words stirred the audience.) Paul entra. Ses gestes étaient brusques. (In came Paul. His gestures were brusque.)

Looking now at E1, a rather surprising constraint occurs when E1 is an animate entity, as has often been noted in work on part-whole relations.7 Contrary to parts of inanimate E1’s, as in: (6) a b.

Il s’abrita sous un vieux tilleul. Le tronc était tout craquelé. (He sheltered under an old limetree. The trunk was full of cracks.) La voiture fit une embardée. Les freins avaient lâché. (The car swerved. The brakes had failed.)

E2s referring to parts of animate E1s do not usually allow for the definite article: (7) a. b.

? Une femme rêvait. Les yeux étaient fermés. (‘lit:’ A woman was dreaming. The eyes were closed.) ? Le chien eut peur. Les oreilles se dressèrent. (‘lit:’ The dog became afraid. The ears pricked up.)

6. In French, the pronoun en can sometimes be used anaphorically:  cf. ?Ils habitent un quartier central. J’apprécie beaucoup le calme / ?’They live in a central neighbourhood. I like the calm’. vs Ils habitent un quartier central. J’en apprécie beaucoup le calme / ‘They live in a central neighbourhood. I like the calm of it.’ (Fradin, 1984). 7.

For an overview, see Spanoghe (1995) and Salles (1995a) inter alia.



The semantics and pragmatics of the possessive determiner 

Although less clear-cut, this also occurs in references to the clothes of E1, as Fradin (1984) has noted: (8) ? Les enfants sont rentrés. Les souliers étaient pleins de boue. (‘lit:’ The children came home. The walking shoes were covered in mud.) It has also been occasionally remarked that the ‘intentional’ parts of an animate E1 share the same constraint: (9) ? Paul pouvait enfin se reposer. L’esprit était libre de tout souci. (‘lit:’ Paul was finally able to rest. The mind was free of all worries.) Once more, the possessive is needed in all these cases to repair the anaphoric link of identity between E1 and E2: Une femme rêvait. Ses yeux étaient fermés. (A woman was dreaming. Her eyes were closed.) Le chien eut peur. Ses oreilles se dressèrent. (The dog became afraid. His ears pricked up. Les enfants sont rentrés. Leurs souliers étaient pleins de boue. (Fradin, 1984). (The children came home. Their walking shoes were covered in mud.) d. Paul pouvait enfin se reposer. Son esprit était libre de tout souci. (Paul was finally able to rest. His mind was free of all worries.) (10) a. b. c.

This distribution is rather surprising, because it is not intrinsically obvious just what is blocking the use of the article. We thus have an intriguing problem which cries out for an explanation. One solution I have proposed (Kleiber, 1999a and 2003a) in order to prevent the associative definite article referring to NP2, is to posit a constraint that I have termed the alienation condition: “Le référent d’une anaphore associative doit être présenté ou donné comme aliéné par rapport au référent de l’antécédent.” (Kleiber, 1999a and 2003a) [The referent of an associative anaphor must be alienable or be interpreted as such in relation to the referent of the antecedent.]

The origin of this condition is iconic. The aim is to account for the difference between an NP with an associative definite article, such as the trunk, and a post-modified NP with the definite article, such as the trunk of the limetree, which gives us the identity of the tree whose NP2 referent is the trunk. In contrast to the complex descriptive NP, the simple version, the trunk, even though it refers to the same referent as the complex descriptive NP, refers by way of a single noun without any semantic subordination to another individual. It is thus semantically independent or autonomous. Although on a pragmatic level the simple NP remains dependent on a prior mention of another individual, as far as its referential interpretation is concerned, the form the N gives it an

 Georges Kleiber

iconic freedom, and thus it can be said to be ‘alienated’.8 Of course it should be pointed out that the trunk is not materially separated from the limetree: it is merely understood as an autonomous individual. To explain this, I have previously used the analogy of a camera focussing on separate things. The fact that a camera can take a picture of a whole limetree or a close-up of its trunk does not mean that the trunk becomes detached from the rest of the tree (Kleiber, 1999a and 2003a). The alienation condition also allows us explain in a straightforward way why the definite article cannot perform an associative link between a property and a process. I have termed these entities syncategoremic (Kleiber, 1981: 40), because in contrast to categoremic entities, whose particular occurrences are autonomous, syncategoremic items are not ontologically independent, but depend on the existence of other items. A particular occurrence of a property or an event is not as autonomous as a categoremic noun. In contrast to chimpanzee, for example, a property such as bald (or baldness) or simplicity, or an event such as speaking / speech or gesturing presupposes an individual who is bald or simple, who speaks or uses gestures. The best test for this is still the deletion test: if you delete the individual on which they depend, these occurrences simultaneously disappear. If you do the opposite, that is you try to ‘alienate’ the syncategoremic item visually or in thought, it seems impossible to carry out such an operation. I cannot detach the baldness, simplicity, speech and gestures from an individual who is bald, simple, uses gestures, and so on. This takes us back to Husserl’s distinction between independent contents and dependent contents, i.e. between contents which can be represented separately and contents which cannot9: “on peut assurément se représenter une tête séparée de l’homme auquel elle appartient, on ne peut se représenter de cette manière une couleur, une forme, etc., elles ont besoin d’un substrat, dans lequel on les remarque sans doute exclusivement, mais dont elles ne pourront être séparées.” (Husserl, 1962: 24). [one can surely imagine a head separated from a man’s body, but one cannot similarly represent a colour, a shape, etc. These require a substratum in which one clearly notices them on their own, but from which they cannot be separated.]

We can now understand why anaphoric associative reference is not possible with properties and events. These do not satisfy the alienation interpretation required by the form le N / the N, since it is not possible to detach them from the individual on which they depend. We can also clarify the role of the possessive adjective in this case, which appears to operate at one remove from the notion of possession.10 The possessive adjective, by an anaphoric link of coreference with an individual, helps to express 8. According to Azoulay (1978: 29), the part is considered to be «considérée dans son existence propre» [seen from the point of view of its own existence]. 9. See Salles (1995a and b) on this point. 10. Although of course possession is still relevant!



The semantics and pragmatics of the possessive determiner 

anaphorically a discourse continuity between the individual E1 and a property or event by respecting the ontological relationship of inalienability which unites E2 and E1. How does this relate to parts of animate bodies? The definite article, as we have seen, is awkward in this context. Yet as Husserl’s example of the body and its head shows us, imaginary alienation is possible and thus, even if we are dealing with syncategoremic nouns, where the part depends on the whole, the alienation condition is satisfied. An associative definite article is therefore possible, as we saw with parts of inanimates: (11) a. b.

Il s’abrita sous un vieux tilleul. Le tronc était tout craquelé. (He sheltered under an old limetree. The trunk was full of cracks.) La voiture fit une embardée. Les freins avaient lâché. (The car swerved. The brakes had failed.)

There must therefore be another reason for those instances where the article cannot be used. The model of perceptual alienation (whether real or imaginary) assumes that basic referential features are used to identify different entities. In order to be detached, a part must have the same features as the whole from which it is being isolated. Thus if we take a concrete object, an object with the features ‘substance’ and ‘shape’11, we can only visually alienate an entity with the same basic ontological features. This explains why it is impossible to alienate properties and events relating to a concrete object: they do not have the same ontological ingredients as their wholes. They have no substance or shape of their own, and cannot as a consequence be isolated from the substance and shape of the object to which they relate. This notion also explains how parts of animates are different from parts of inanimates. Only the latter are alienable because they possess the same ontological features as the wholes from which they have been separated. I can isolate the trunk from a limetree because, just like a limetree, the trunk possesses the same substance and shape which allow us to represent it separately from the tree. Parts of animates do not obey this principle, which I have called the principle of ontological congruence (Kleiber, 1999a et 2003a). This is because together with substance and shape, an animate noun also has intentionality (or the feature ‘animacy’). A part only carries the features substance + shape if it is part of a concrete object, and may only have the feature ‘animate’ when it refers to an intentional object (see the above example the mind of Paul / Paul’s mind), in which case it contravenes the principle of ontological congruence. Thus parts of bodies and the other components associated with the notion of animacy are not just parts of an animate referent. They are both parts of the body, which is an animate referent (and which indeed has both substance and shape) and also parts of its intentional 11. A concrete object, at least according to one of its meanings, is denoted by a noun whose referents have substance and shape, as I have shown elsewhere in collaboration with Michel Galmiche (Kleiber, 1994a, Ch. 3). Other dimensions can be added, such as a temporal dimension (see Kleiber, 1994a for more details).

 Georges Kleiber

component. We can note in support of this analysis that it is difficult to have expressions of the type ? a part of John,?a part of our dog, etc.12, whereas it is quite possible to talk about a part of a chair, a part of a car etc. This is a particular difficulty of using the article in an associative context with parts of animate bodies. The possessive adjective is on the other hand well suited to this kind of link, in that it maintains an explicit connection with the animate referent, avoiding the ontological break that would be implied by the article. Thus in contrast to the definite NP the N, the possessive NP Poss Adj. + N keeps its ‘animate’ character by virtue of the possessive anaphor. This brings us to a curious feature of French, where in certain appositive contexts, the definite article can find its rightful place again, as has been noted in various studies (Vergnaud et Zubizaretta, 1992): (12) a. b.

Une femme rêvait, les yeux fermés. (‘lit:’ A woman was dreaming, the eyes closed.) Paul pouvait enfin se reposer, l’esprit libre de tout souci. (‘lit:’ Paul was finally able to rest, the mind free of all worry.)

The explanation for this lies in the absolute construction.13 In French, the absolutive use of the article describes or expresses the attitude of an animate referent, functioning like a circumstantial of manner (Hanon, 1989 and Choi, 1991), and thus allowing for alienation of a part of the animate referent, either its body (substance + shape), or its ‘intentionality’. From a discourse perspective, alienation is also possible in intersentential contexts, as long as the alienation from the body is in some way justified by the context, as in the following examples mentioned by Julien (1983: 137), Fradin (1984: 362) and Salles (1995a/b): (13) a. Le malade est livide. Les yeux sont hors de leurs orbites. (‘lit:’ The sick man is livid. The eyes are poking out of their sockets) (Julien, 1983)

12. The problem is more complex than this. We would need to study the potential referents of part and what conditions determine how a part is identified. It is interesting to note that, from a very different line of argument, Tamba (1994) comes to a similar conclusion concerning the notion of part. 13. Other kinds of alienation are also at work in French, as in Il lève les yeux (?He raises the eyes) or in expressions with a direct object complement (Sylvie a les yeux bleus / ?Sylvie has the blue eyes). A ‘syntactic’ explanation for this is given in Vergnaud and Zubizaretta (1992). In constructions such as Il lui prend le bras (?He took from him the arm = He took his arm) or Il se pince la peau (?He pinches himself on the skin = He pinched his skin), it is noticeable that dependency is marked by a personal or reflexive pronoun: lui / se (to him, from him / himself).



The semantics and pragmatics of the possessive determiner 

b. Autour de la table les joueurs s’épiaient. Les mains étaient crispées sur les revolvers. (‘lit:’ The players stared at each other across the table. The hands clutching the revolvers.) c. Les coureurs redoublent d’effort. On voit les muscles saillir sous les maillots. (‘lit:’ The racers are redoubling their effort. You can see the muscles bulge under the jerseys) (Fradin, 1984) d. Hughes recommençait chaque soir le même itinéraire, suivant la ligne des quais, d’une démarche indécise, un peu voûté déjà, quoiqu’il eût seulement quarante ans. Mais le veuvage avait été pour lui un automne précoce. Les tempes étaient dégarnies, les cheveux pleins de cendre grise. (‘lit:’ Every night Hughes took the same route, following the quayside, unsteady and already rather stooped, although he was only forty. The widowerhood had been an early autumn for him. The temples were drawn, the hair speckled ash grey. (G. Rodenbach, Bruges-la-morte, Babel, 1989, 25) e. Il la dévisagea; elle était pâle; la bouche était serrée, les yeux pluvieux battaient. (‘lit:’ He stared at her, she was pale, the mouth tight shut, the tearful eyelids trembling) (J.-K. Huysmans, Là-Bas, Gallimard, Folio, 1991, 292) f. Le Christ (...). L’aisselle éclamée craquait, les mains grandes ouvertes brandissaient des doigts hagards (...); les pectoraux tremblaient (...) (‘lit:’ Christ…. The disjointed armpit cracked, gaunt fingers dangled from the wide open hands,… the pectorals twitched…) (J.-K. Huysmans, Làbas, Gallimard, Folio, 1991, 33) (these last three examples cited by Salles). In all of these referential links, a part of the body is isolated by a kind of alienation which is indicated by verbs of perception relating to vision (as in the two examples from Fradin: s’épier / stare at each other, voir / see). Salles analyses Rodenbach’s example and Huysmans’ first example as follows: “Et, quelquefois, c’est simplement le regard qu’on porte sur un personnage qui permet une énumération descriptive des parties au moyen d’une anaphore associative.” [Sometimes the portrayal of a character allows for a descriptive enumeration of its parts by way of an associative anaphor.] (Salles, 1995 b: 54).

It is worth emphasising that this perceptual alienation does not operate directly on the animate referent, but only on the body.14 In Julien’s example, this is brought to the fore by the fact that the animate referent is sick, and in the other examples by the fact that

14. I believe this kind of explanation could also apply to ‘residuals’ in English cited by Ebert (1982), Vergnaud et Zubizaretta (1992): if, contrary to expectations, the definite article appears instead of the possessive, this is likely to be because the part has been removed from the body.

 Georges Kleiber

a zone of the body has been prominently placed in view, as illustrated strikingly in the following example taken from a cheap paperback novel: (14) a. Il s’assit sur le lit et la regarda. Les paupières étaient boursouflées et les po­ ches sous les yeux étaient striées de veinules bleues. (‘lit:’ He sat on the bed and looked at her. The eyelids were swollen and the bags under the eyes were lined with blue veins.) Since the ontological congruence has been respected, the definite article can be used in these examples, at least in French. Having said this, it is now necessary to account for the fact that the possessive adjective can also be used in the same contexts. Indeed, it is obligatory in English in all of these cases, as we can see in the following possessive versions: (15) a. Le malade est livide. Ses yeux sont hors de leurs orbites. (The sick man is livid. His eyes are poking out of their sockets) (Julien, 1983) b. Autour de la table les joueurs s’épiaient. Leurs mains étaient crispées sur les revolvers. (The players stared at each other across the table. Their hands clutching their revolvers.) c. Les coureurs redoublent d’effort. On voit leurs muscles saillir sous les maillots. (The racers are redoubling their effort. You can see their muscles bulge under their jerseys) (Fradin, 1984) d. Hughes recommençait chaque soir le même itinéraire, suivant la ligne des quais, d’une démarche indécise, un peu voûté déjà, quoiqu’il eût seulement quarante ans. Mais le veuvage avait été pour un automne précoce. Ses tempes étaient dégarnies, ses cheveux pleins de cendre grise. (Every night Hughes took the same route, following the quayside, unsteady and already rather stooped, although he was only forty. His widowerhood had been an early autumn for him. His temples were drawn, his hair speckled ash grey.) (G. Rodenbach, Bruges-la-morte, Babel, 1989, 25) e. Il la dévisagea; elle était pâle; sa bouche était serrée, ses yeux pluvieux battaient. (He stared at her, she was pale, her mouth tight shut, her tearful eyelids trembling.) (J.-K. Huysmans, Là-Bas, Gallimard, Folio, 1991, 292) f. Le Christ (...). Son aisselle éclamée craquait, ses mains grandes ouvertes brandissaient des doigts hagards (...); ses pectoraux tremblaient. (Christ … His disjointed armpit cracked, gaunt fingers dangled from his wide open hands,… his pectorals twitched…) (J.-K. Huysmans, Là-bas, Gallimard, Folio, 1991, 33) (these last three examples cited by Salles). We set out below in the conclusion how the parallel uses of the article and possessive differ in these examples. A partial explanation here would be to suggest that the two referential markers express a difference of scope. The possessive adjective presents a



The semantics and pragmatics of the possessive determiner 

part of the body in terms of individual identification, underlined by a dependency relation with the animate individual. In contrast, the definite article implies a more generic notion relating to body parts. The individual is no longer in the foreground, and instead the part is taken to be a stereotypical component of the body, detached iconically from the individual in question. The possessive contributes to the individualisation of the body part, while the article contributes to the opposite process. Even if les yeux are in fact the eyes of the sick man in the first example, they are not presented as such, a function associated with the possessive, but rather they are a generic part of the man’s body, as opposed to any other part. It should be pointed out here that that we are not dealing with a generic NP. The claim here is that that the isolation of a part is expressed at the generic level in terms of: un homme a des yeux / a man has eyes.15 The associative link, as shown at length elsewhere (Kleiber, 2001a), is based on an a priori relation, which is part of our shared knowledge of how objects are related in reality or in terms of more stereotypical relationships. The main evidence for this is that the addition of a specifier which runs contrary to our usual lexico-stereotypical expectations, prohibits the definite article and makes the possessive a necessary part of the expression (see especially Julien, 1982): (16) a. b.

? Le malade est livide. Les yeux bleus sont hors de leurs orbites. (‘lit:’ The sick man is livid. The blue eyes are poking out of their sockets.) Le malade est livide. Ses yeux bleus sont hors de leurs orbites. (The man is livid. His blue eyes are poking out of their sockets.)

It should be added that as far as the enumeration of parts is concerned, as can be seen in Rodenbach and Huysmans’ examples quoted above, the definite article is in theory more appropriate than the possessive determiner in French simply because it does not mark, as the possessive does, a reference back to E1 each time. It is thus more suited, at least in French, to the close determination of individual parts of E1 than the possessive.

3. Cases B) and C) We now turn to the other two possibilities, where either the possessive is not possible or where both determiners can be used.

3.1

Cases where the article can be used, but not the possessive

There are few examples where the definite article can be used, but the possessive cannot. As can be seen in the context of actancial associative anaphors (Kleiber, 2001a and 2003a), the possessive can never be used in place of the article: 15. This also explains the definite article in the attributive complement structure Sylvie a les yeux bleus (?Sylvie has the blue eyes -> Sylvie has blue eyes) (see Riegel, 1988, 1989 et 1997).

 Georges Kleiber

(17) a. Paul s’est pendu. Sa corde s’est cassée. (sa = Paul, *sa = that of the hanging) (Paul hanged himself. His rope snapped.) (his = Paul, *its = that of the hanging) b. Paul s’est pendu. La corde s’est cassée. (Paul hanged himself. The rope snapped.) c. Il y a eu un assassinat hier soir à Souffelweyersheim. *Son assassin a pris la fuite. (*son = that of the murder). (There was a murder last night in Souffelweyersheim. *Its murderer ran off.) d. Il y a eu un assassinat hier soir à Souffelweyersheim. L’assassin a pris la fuite. (There was a murder last night in Souffelweyersheim. The murderer ran off.) These examples can be partly explained by invoking our initial constraint on forming a prepositional phrase of the type le N1 de NP2 / the N1 of NP2. Clearly not all actancial anaphors fulfil this condition. Thus we can have a prepositional construction with of referring to the rope or the hanging, even if the phrase is rather awkward, such as: (18) La corde de la pendaison. (The rope of the hanging.) But we cannot have in the case of murder a NP with dc: (19) ? L’assassin de l’assassinat. (‘lit:’ The murderer of the murder) The same goes for collective anaphors (Kleiber, 2001a and 2003a), which appear to resist the possessive and do not allow for replacement by the definite article, even though we can use a corresponding nominal phrase with de / of. A similar case involves kinship relations. Although we have in the case of mother and family, couple and husband post-modified NPs such as the mother of the family and the husband of the couple, a possessive determiner could not be used to mark such a relation: (20) a. Dans les familles d’origine immigrée notamment, *leur mère (= des familles) est en porte à faux entre sa culture d’origine et sa volonté d’intégration, elle est complètement larguée au niveau scolaire et *leurs enfants (= des familles) en profitent. (In immigrant families especially, *their mother (= of the families) is caught between her original culture and her desire to integrate; she is out of her depth in the educational system and *their children (= of the families) take advantage of this.) b. Dans les familles d’origine immigrée notamment, la mère est en porte à faux entre sa culture d’origine et sa volonté d’intégration, elle est complètement larguée au niveau scolaire et les enfants en profitent.



The semantics and pragmatics of the possessive determiner 



(In immigrant families especially, the mother is caught between her original culture and her desire to integrate; she is out of her depth in the educational system and the children take advantage of this.) c. Un couple s’installa à la terrasse. *Son mari (= du couple) commanda une 1664. (A couple sat down outside a café. *Its husband (= of the couple) ordered a glass of 1664.) d. Un couple s’installa à la terrasse. Le mari commanda une 1664. (A couple sat down outside a café. The husband ordered a glass of 1664.)

3.2

Cases where both determiners can be used

In contrast, there are plenty of situations where both referential markers are allowed. In the context of a locative anaphor (Kleiber, 2001a), both markers can appear and are thus in competition: (21) a. b.

Le village était situé sur une butte. Son église dominait toute la region. (The village was on top of a hillock. Its church dominated the whole area). Le village était situé sur une butte. L’église dominait toute la region. (The village was on top of a hillock. The church dominated the whole area).

The same can be said of meronymic anaphors (Kleiber, 2001a), which express a partwhole relation with inanimates.16 As we saw in our previous examples, the article can be easily substituted for the possessive: (22) a. b. c. d.

Il s’abrita sous un vieux tilleul. Le tronc était tout craquelé. (He sheltered under an old limetree. The trunk was full of cracks.) Il s’abrita sous un vieux tilleul. Son tronc était tout craquelé. (He sheltered under an old limetree. Its trunk was full of cracks.) La voiture fit une embardée. Les freins avaient lâché. (The car swerved. The brakes had failed.) La voiture fit une embardée. Ses freins avaient lâché. (The car swerved. Its brakes had failed.)

The same goes for functional associative anaphors (Kleiber, 2001a): (23) a. b. c.

Une voiture s’est renversée hier dans le fossé. Le conducteur s’était assoupi. (Yesterday a car crashed into the ditch. The driver had fallen asleep.) Une voiture s’est renversée hier dans le fossé. Son conducteur s’était assoupi. (Yesterday a car crashed into the ditch. Its driver had fallen asleep.) Le village de Pfaffenheim est de plus en plus fleuri. Les habitants / Le maire raffole(nt) des géraniums.

16. It is necessary to bear in mind here the specific case of parts of animates, discussed above.

 Georges Kleiber



(The village of Pfaffenheim is getting more and more flowery. The villagers are / The mayor is wild about geraniums.) d. Le village de Pfaffenheim est de plus en plus fleuri. Ses habitants / Son maire raffole(nt) des geraniums. (The village of Pfaffenheim is getting more and more flowery. Its villagers are / Its mayor is wild about geraniums.) Finally, a member-collection relationship also allows for both referential markers, at least for a certain category of N: (24) a. Nous entrâmes dans une forêt magnifique. Les arbres resplendissaient de lumière verte. (We went into a magnificent forest. The trees were ablaze with green light.) b. Nous entrâmes dans une forêt magnifique. Ses arbres resplendissaient de lumière verte. (We went into a magnificent forest. Its trees were ablaze with green light.) c. Le régiment a été défait. Les soldats n’ont pas eu le temps de combattre. (The regiment was defeated. The soldiers did not have time to fight back.) d. Le régiment a été défait. Ses soldats n’ont pas eu le temps de combattre. (The regiment was defeated. Its soldiers did not have time to fight back.) Situations B) and C) thus raise two specific questions: Q1) Why is the possessive suitable in the C) examples but not in the B) examples? Q2) When the possessive is in competition with the definite article, is there a discourse explanation which can help to distinguish between the two types of occurrence?

3.3

Question 1: Ontology and Lexical Relations

3.3.1 The a priori dependency asymmetry We should recall here that the second contextual condition which was initially proposed (condition II) was that it should be possible to have for any two entities E1 and E2 an NP of the type le N2 de NP1 / the N1 of NP2. This was based on the observation, commonly noted in the literature on the possessive adjective, that there is a close relationship between the NP Poss. Adj. + N2 and prepositional NPs of the type le N2 du (d’un) N1 / the N2 of the (of a) N1. The possessive determiner, as noted by Godard (1986: 102): (25) “…possède une propriété spécifique: il alterne avec un complément de Nom de la forme de GN”.



The semantics and pragmatics of the possessive determiner 

[…has a specific property: it alternates with a post-modifier of the form of + Noun Group]17 Whichever syntactic explanation may be adopted18, this combination, which is not necessarily an ‘assimilation’19, can be summarised in the interpretive equation Poss. Adj. + N1 = the N1 of NP. This implies that an NP such as his book can be interpreted as more or less the same as the book of Paul (see Bartning, 1989 et 1996). However, whether it is possible to have the possessive in cases B) or C) does not depend on this, since condition II has already been satisfied. However, it does contribute greatly to the interpretation of N1 and N2. So how does it operate? In the first instance, for E1 a ‘human’ or ‘animate’ feature is generally considered to be relevant in the motivation of a possessive NP. The ‘human’ or ‘animate’ feature also figures, as pointed out by Bartning (1996), at the top of Hawkins’ possessive hierarchy (1981), whereas inanimates are placed at the bottom (see also Seiler, 1983: 81). However, as all the grammarians who have commented on this have pointed out, this feature cannot be treated as a formal constraint since the possessive determiner can be easily used with inanimate antecedents, or even inanimate N1s in the structure le N2 du (d’un) N1 / the N2 of a (of the) N1, as the following example from Bartning (1996) shows: (26) La gravité de la situation. Sa gravité. (The seriousness of the situation. Its seriousness.) We have already seen this in the following examples: (27) a. Le tronc du tilleul / son tronc (The trunk of the limetree / its trunk.) b. Les freins de la voiture / ses freins (The brakes of the car / its brakes.) 17. Most authors point out that the NP1 is a personal pronoun and they therefore relate the possessive phrase Poss. Det + N to the structure Le N de + Pro / The N of + Pro. This approach is adopted by Wilmet (1986: 108, 1997: 241), who talks about quantifiants-caractérisants personnels and Riegel et al. (1994: 158), who argue that “le déterminant possessif est l’équivalent de le (...) de moi, le (...) de toi, etc. […et] représente la synthèse de deux éléments généralement disjoints du GN : l’article défini et un complément du nom introduit par de (en l’occurrence un pronom personnel)” [the possessive determiner is the equivalent of the … of me, the … of you etc. and this represents the synthesis of two elements which are generally separate in the noun group: the definite article and the noun complement introduced by of (in the case of a personal pronoun)]. 18. See Godard (1986). 19. See Gross (1986) for a presentation of the two principal theoretical positions on the possessive adjective, i.e. (i) the possessive is an irreducible expression, (ii) it is the result of a transformation. Option number (i) is preferred. Bartning (1989: 196-197) summarises the syntactic debate of the 1970s and 80s. In Godard (1986) a generative solution is worked out according to Chomsky’s modular grammar of 1981, a solution which «combines hypotheses concerning lexis, syntax and interpretative semantics» (Godard, 1986: 103). Zribi-Hertz (1999) proposes a generative analysis in terms of Chomsky’s Principes and Parameters adopting an ‘autonomist’ conception of inflectional morphology which also goes by the name of Distributional Morphology.

 Georges Kleiber

c. L’église du village / son église (The church of the village / its church.) d. Les arbres de la forêt / ses arbres (The trees of the forest / its trees.) We cannot ignore the characteristics ‘animate’ or ‘human’ and the more general ontological dimension in our explanation. But these features nevertheless need to be associated with two further factors, namely the lexico-semantic relations between N1 and N2 and the discourse context. To start with, it has often been noted that the entity E1, which is referred to by the possessive determiner, is itself used to determine, localise or identify the entity E2, which is referred to by the possessive description and which belongs to the category N2 representing the N from the possessive NP. E2 and E1 can be termed respectively the target and the site, or using Culioli’s terms repère (marker) and repéré (marked) or Langacker’s (1993, 1995 and 2004) target and domain. The main point, as shown by the complementarity of a prepositional NP and the possessive, is that an NP of the type Poss. Det + N marks a referential dependency, an asymmetry between the two entities involved in the NP.20 This asymmetry has a consequence that is often acknowledged, but rarely explained. It is required that E1 should identify or localise E2 inside the class N2 in which, as has been pointed out, E2 has been categorised. In other words, the referential dependency of E2 on E1 and the categorisation of E2 as an N2 both lead to a specific distinctiveness condition: entity E1 must be such that it can be distinguished or isolated from E2 within the class N2, i.e. by particularising it in terms of other N2 entities. A second, less acknowledged, feature can be added to the notion of referential asymmetry and to its corollary, the distinctiveness between N2 entities. This relates to the anaphoric pronominal character of the possessive determiner. It has already been noted that several authors draw a parallel between the possessive determiner and a prepositional structure involving an article and personal pronoun (my N = the N of me, his N = the N of him). And as pointed out, we are not obliged to assimilate these structures, since the possessive determiner can be considered irreducible even though it can at the same time be reformulated as a definite article structure with a preposition and personal pronoun. The more important point is that the possessive determiner still has a personal pronoun function, which is partly obscured by the term ‘possessive adjective’. Thus by relating the pronominal features of the possessive determiner with the distinctiveness condition, which is a consequence of the referential asymmetry 20. For Zribi-Hertz (1999), this is explained by predication: the predicative character of the relation YP-XP is a necessary (but not sufficient) condition for the use of the possessive. For her, (1999: 15), «l’équation Possession = Prédication éclaire les commentaires souvent confus ou ad hoc faits par les grammairiens sur la relation qu’ils nomment ‘possessive’. En réalité, la Possession est une relation juridique et la Prédication une relation grammaticale indépendante de la première.» […the equation Possession = Predication sheds light on the often confused or ad hoc comments made by grammarians on the relationship they call ‘the possessive’. In reality, Possession is a legal relationship, and Predication is a grammatical one independent of the former.]



The semantics and pragmatics of the possessive determiner 

between E1 and E2-N2, and which is fulfilled by the possessive determiner denoting entity E1, the implication is that E1 can be understood as one of a group of entities of the same type. If we take my / your / her hat, the possessive determiner only has a specifying function if it assumes that other entities of the type that it refers to could also identify or localise the E1 entity, hat. If such a paradigm were not available, the possessive would not be able to distinguish the specific E1 entity from the class of hats. That this comes from the pronominal nature of the possessive determiner can be easily understood for 1st or 2nd person possessives which retain referential identity because of their marked roles in the speech context. But for 3rd person possessives, things are less clear-cut, because like the 3rd person pronoun they can be applied to virtually any named or pre-classified entity (Kleiber, 1994b), and thus to a group of entities which have little in common. However, this difficulty is less real than it seems, since this heterogeneity vanishes in actual usage, where only one specific category can be understood. Incidentally, the plural pronoun ils / they cannot be applied to a group of different entities precisely because of this well-known requirement for prior naming or classification. It could be objected that the same argument applies to the prepositional NP in de or of as in le chapeau de Jean / the hat of Jean. This is certainly valid, but the explanation is down to the properties of the Ns in the phrase, i.e. Jean, chapeau, hat, as argued below, and not to the properties of the prepositions de / of, and consequently not to the intrinsic properties of the post-modified NP itself. This is possible even if the entity E1 is not understood as a natural localiser of E2, as is the case with the possessive determiner in a variety of structures. We only need to examine Bartning’s examples (1996, 1998 et 2001), which (s)he terms ‘discursive’ to distinguish them from ‘prototypical’21 occurrences of the type le chapeau de Jean. We can very well have an NP of the type le chapeau de la voiture / the hat of the car where the E1 entity la voiture localises entity E2 le chapeau without assuming that the distinction implies the possibility that cars can localise these types of E2 entities. We can indeed create a context which could justify this post-modified NP, as the following sequence shows: (28) Sur la banquette arrière de la voiture, le commissaire remarqua un curieux chapeau vert, orné d’une fleur rouge. Ce n’est qu’au bout d’un moment qu’il fit le rapprochement: le chapeau de la voiture était celui de l’avocate assassinée il y a trois mois. (On the back seat of the car, the officer noticed a curious green hat, decorated with a red flower. It took a while for him to make the connection: the hat of the car was that of the lawyer murdered three months ago.)

21. In the case of ‘prototypical’ interpretation, the meaning of the determinant relation is given by the micro-structure of the phrase, in the case of ‘discursive’ interpretations, the meaning follows from the information provided in the discourse context.

 Georges Kleiber

The possessive determiner is prohibited in this context, since we cannot have the NP son chapeau / its hat as a replacement of the post-modified NP: (29) Sur la banquette arrière de la voiture, le commissaire remarqua un curieux chapeau vert, orné d’une fleur rouge. Ce n’est qu’au bout d’un moment qu’il fit le rapprochement: *son chapeau était celui de l’avocate assassinée il y a trois mois. (On the back seat of the car, the officer noticed a curious green hat, decorated with a red flower. It took a while for him to make the connection: *its hat was that of the lawyer murdered three months ago.) The constraint on paradigm or on prior category for E1 that has just been set out means that referential dependency cannot be established between the specific or particular entity E1 and the entity E2 expressed by N2, but must already exist in the homogenous class in which E1 and the class of N2s to which E2 belongs can be recognised. This relation is expressed by saying that the possessive determiner requires an a priori dependent asymmetry. 3.3.2 The three motivations for the possessive22 This a priori asymmetry is motivated by three criteria, of which two correspond to an intrinsic subordination related to the Ns in context. The third corresponds to the subordination of an E2 class to an E1 class established by context:

(1) the ontological status of the entities involved,



(2) a lexico-semantic relation between the Ns of the entities involved indicating an a priori dependent asymmetry,



(3) a relation which is not lexico-semantic as such but rather discursive, between two specific groups or classes of N1 and N2 established by the linguistic or situational context.

Criterion (1) posits that if general categories or ontological types to which the different entities belong are in a dependency relation, a possessive connection is possible with the most prominent entity as antecedent of the possessive, i.e. the entity which serves as determiner. The scale of ontological dependency on which these categories can be placed is more or less as follows23: Humans > Animals > Concrete objects > Events > Properties The second and third categories account for uses of the possessive which appear to contravene the first category. If we have two entities of the same type, or two entities of which the lower one in the hierarchy appears as the antecedent of a higher entity, there 22. In Kleiber (2003a and 2004), only the first two criteria are emphasised. 23. This hierarchy may require more refined analysis. It is perhaps just worth pointing out here that it corresponds approximately to the one which is often used for referential salience.



The semantics and pragmatics of the possessive determiner 

is either a lexical relation between the Ns involved which can provide a necessary a priori determinative asymmetry, or a discursive situation exists which makes possible a dependency relation similar to that of the second criterion, but at a contextual level, and which is thus necessarily construed before the use of a possessive NP. This approach posits that the absence of a prior intrinsic referential asymmetry24 is ontological or lexico-semantic, and would indicate which of the two Ns present is dependent on the other and thus blocking a possessive connection. Thus to give a basic example, if we have two concrete objects which are at the same ontological level and whose nominal expressions do not have any a priori semantic relation between them, the possessive determiner is impossible, even if the discourse context permits a contingent and entirely specific link between the two entities which can be expressed by a post-modified NP with de / of. We have already seen this in the example le chapeau de la voiture / the hat of the car which cannot allow a possessive NP in the context set out above. Criterion (2) cannot be applied here in order to obtain a possessive, since there is no lexico-semantic relation between voiture / car and chapeau / hat which would establish the necessary a priori referential asymmetry for a possessive determiner. But criterion (3) is available, even though it appears rather contrived. For this we just need to create a context in which a series of cars each have a hat inside, allowing us to justify the possessive expression son chapeau / its hat. Imagine an antique car exhibition where each car has a hat on the bonnet with the make of the car written on it. We would then be able to freely accept sequences of the type: (30) Lorsqu’il arriva au niveau de la deuxième voiture, Paul ne put savoir quelle était sa marque. Son chapeau (= le chapeau de la voiture) avait été volé par un visiteur collectionneur. (When he got to the second car, Paul couldn’t work out what make it was. Its hat (= the car’s hat) had been stolen by a marauding visitor). The possibility that the possessive determiner can localise or distinguish the hat by reference to the category E1 (car) is due to the fact that its hat can be contrasted with the hats of other cars. Here we have a construction involving a class which is entirely contingent (the class of cars with hats on) in a part-whole relation (the hats in this particular reading become honorary parts of cars25). The situation changes when we have entities of a different ontological order. Let us examine two cases. First, a concrete object with a property, such as a car with a colour. Here we have an a priori referential dependency, i.e. criterion (1), since the occurrence 24. See Strawson (1973) for this. Zribi-Hertz (1999: 23) comes to the same conclusion positing that «l’acceptabilité du syntagme possessivisé semble (...) solidaire de l’interprétation prédicative de la relation YP-XP dans le XP de YP”. [“the acceptability of the possessive phrase seems …to support the predicative interpretation of the YP-XP relation in the XP of YP.” ] 25. In other words, we can refer to a prior statement such as the cars have a hat. It happens that we can also say: chaque voiture a son chapeau / every car has its hat.

 Georges Kleiber

of colour depends, by its ontological nature as a property, on an occurrence such as car. The reverse is not true: the occurrence car is not subordinate to colour, because its properties (including that of colour) depend ontologically on the object whose property they are, and objects do not depend on their properties. This was mentioned above in terms of the alienation condition and syncategoremic entities. So, the possessive determiner is possible in this case since there is an a priori referential dependency which is not limited to the two particular entities at stake: (31) a. sa couleur (its colour = the colour of the car) b. *sa voiture (*its car = *the car of the colour) Now let us take a second example, in which we have a human being and a concrete object, as in Jean and book. The feature ‘human’ means that there is a prior asymmetry between the two entities, i.e. criterion (2), an asymmetry which is a permanent difference between these entities and not just contingent. Clearly, the animate or human controls the concrete object rather than the other way round. In Strawsonian terms of subject-object asymmetry, it is clear that only one order is allowed: the concrete object can be predicated on the animate human being, but the animate or human being cannot be predicated on the inanimate.26 The consequence is that we can freely form NPs of the type son livre / his book for le livre de Jean / the book of Jean, whereas the converse *son Jean / *its Jean27 is not possible. However, it is possible to dream up contexts which might temporarily justify the reverse, and may for example allow for such constructions as the post-modified NP le Jean du livre / the Jean of the book. There is in fact one example of this in Alexandre Dumas, quoted by Bartning (2001: 148, following Eriksson 1980: 399), which illustrates a discursive construction using the preposition de: (32) Debout devant la cheminée était un homme de moyenne taille, à la mine haute et fière; (…). De temps en temps, l’homme de la cheminée levait les yeux de dessus les écritures. (A. Dumas, Les Mousquetaires, 183–184) (Standing in front of the fireplace was a man of medium build, his face held high and haughty; … From time to time, the man of the chimney looked up from the writings.) It is clearly not possible to have a possessive determiner in this example. But even though it may appear somewhat contrived, perhaps even more so than in our first example, criterion (3) in which the prior establishment of a context in which two entities E1-E2 share referential dependency, can once again allow the possessive determiner to reappear. This can be justified by relying again on the relational mode introduced by the second criterion. In the example le Jean du livre / the Jean of the book, if we can imagine 26. As pointed out before, this position has also been argued by Zribi-Hertz (1999) 27. This obviously does not mean that we can never have a possessive NP of the type son Jean / its Jean. Such NPs are in fact quite common (cf. ton Zidane / your Zidane..., ma chère Françoise / my dear Françoise , etc.). But the possessive is not referring in these instances to a concrete object.



The semantics and pragmatics of the possessive determiner 

that several books are talking about Jean, we can talk about a particular book in a series as son Jean / its Jean in order to talk about each individual instance of Jean. 2.3.3 How the three criteria apply to situations B) et C) Our criteria (1) and (2) are enough to deal with situations B) et C), as discussed above, because the associative anaphor, which is our basic starting point, is ‘fuelled’ by a stereotype relation. This is not a contingent relation since it is intrinsically linked to the content of the Ns present (Kleiber, 2001a). On the other hand, our ‘third way’, criterion (3) involving the prior contextual construction of dependency relations, is hardly rele­ vant to situations B) and C). The conjunction of criteria (1) and (2) allows us therefore to explain the presence of the possessive in the context of meronymic and locative anaphors, as in our previous examples: (33) a. b.

Il s’abrita sous un vieux tilleul. Son tronc était tout craquelé. (He sheltered under an old limetree. Its trunk was full of cracks.) Le village était situé sur une butte. Son église dominait toute la region. (The village was on top of a hillock. Its church dominated the whole area).

The appearance of the possessive here can not be explained by ontological factors (the two entities are at the same ontological level), but by the lexical relationship between N1 and N2, which creates a necessary asymmetry or dependency hierarchy. There is thus a meronymic relationship between tilleul – tronc / limetree – trunk (Kleiber, 2001a) and a stereotypical locative relationship between village – église / village – church (Kleiber, 2001 a). The point here is that this relation is asymmetrical, i.e. one N is defined by the other N. In the case of a meronymic relation, the part N(i) is defined in terms of the whole expression, and not the reverse. It is interesting to note that if a noun in the anaphoric expression is in fact intrinsically marked as being a part, the antecedent noun which represents the whole does not qualify as a ‘holonym’, as Tamba has argued (1994). The antecedent cannot possess the equivalent semantic trait of wholeness which a meronym would imply when used to express ‘a part of ’. Thus although trunk and wheel are defined semantically as parts of a tree or a car, tree and car are not intrinsically defined as being ‘wholes’ (Tamba, 1994 and Kleiber, 2001a). In the case of our locative relations, a church is not defined in terms of a village, whereas a village is a place where we can usually find a church. The locative noun28 village – N(j) – is thus defined as an entity where the entity church – N(i) – can be found and not the reverse. In support of this observation, we can note that the possessive is not allowed in the converse expressions: son N2 / its N2 even though we could have a prepositional phrase of the type le N2 de NP1 / the N2 of the NP2. Thus 28. Bartning (1998) refers to ‘locative nouns’ and categorises them as N2s whose lexical meaning contributes to the relational meanings associated with complex NPs post-modified by de / of.

 Georges Kleiber

we could not have its car (where its = of the steering wheel) or its village (where its = of the church), but we could have instances such as the following (although this may be more awkward in the English version): (34) La voiture du volant qu’on a ramassé dans le fossé n’a jamais été retrouvée. (?The car of the steering wheel that was recovered from the ditch has never been found.) It is noticeable that the meronymic and locative contexts share a common characteristic in that they represent a general form of inclusion, the meronym possibly being considered as a particular case of internal localisation. In both cases, at an intrinsic level the included item is E2 and the inclusive item is E1. A second point is that the possessive determiner can also be explained in what I have termed have termed functional contexts: (35) a. Une voiture s’est renversée hier dans le fossé. Son conducteur s’était assoupi. (A car crashed into a ditch yesterday. Its driver had fallen asleep.) b. Le village de Pfaffenheim est de plus en plus fleuri. Ses habitants / Son maire raffole(nt) des geraniums. (The village of Pfaffenheim is getting more and more flowery. Its villagers are / Its mayor is wild about geraniums.) While the ontological hierarchy implies that the E1 antecedent is less prominent than E2, since it is a non-human, the existence of a functional semantic relation such as X is N2 of Y in fact allows us to invert this relation. But now there is no longer a relationship between a human and the E1 (car, village), but between a non-human car or village and a facet of a human corresponding to a functional predicate noun such as, driver, mayor, villager etc. Evidence for this comes from the impossibility of relating the N1 voiture / car, to the possessive son N2 / its N2 when we have an N2 such as automobiliste / car user. This is because these are not functional Ns. On the other hand there is no problem with a predicate N such as conducteur / driver (Kleiber, 2001c): (36) a. son conducteur (its driver) = le conducteur de la voiture (the driver of the car) b. *son automobiliste (*its car user) =? l’automobiliste de la voiture (? the car user of the car) Thirdly, the impossibility of having a possessive determiner in an agentive context can also be explained by the ontological hierarchy, since the concrete object corde / rope is much higher up on the scale than the event pendaison / hanging29: (37) Paul s’est pendu, mais *sa corde (= la corde de la pendaison) s’est cassée. (Paul hanged himself, but *its rope (= the rope of the hanging) snapped.) 29. The reverse is not possible either (cf. ? sa pendaison / its hanging), but here things are already awkward because of the level of dependency of the prepositional phrase (*La pendaison de la corde / *The hanging of the rope) (cf. above).



The semantics and pragmatics of the possessive determiner 

Our final case involves collective expressions, where the situation is less clear, since, as we have seen, the possessive is possible in certain cases: (38) a. Nous entrâmes dans une forêt magnifique. Ses arbres resplendissaient de lumière verte. (We went into a magnificent forest. Its trees were ablaze with green light.) b. Le régiment a été défait. Ses soldats n’ont pas eu le temps de combattre. (The regiment was defeated. Its soldiers did not have time to fight back.) though not in others: (39) Un couple s’installa à la terrasse. *Son mari (= du couple) commanda une 1664 (A couple sat down outside a café. *Its husband (= of the couple) ordered a glass of 1664.) This is rather intriguing, since it is difficult to see why the member – collection lexical relation, which can be used to explain the use of the possessive in the case of forest – tree and regiment – soldiers30, cannot be used in the same way when dealing with kinship relations. Presumably, if the NP *son mari / *its husband is not possible, this is because the kinship items do not functionally refer to a collective term (i.e. mari / husband to couple or mère / mother to famille / family) but instead to their complementary kinship terms: mari / husband to femme / wife and mère / mother to enfants / children (Kleiber, 1999b et 2003b).

4. Conclusion: Two markers in competition We now have to address the issue of ‘discourse’ which was raised at the beginning. If the associative definite article and the possessive can at times occur in the same contexts, what is the difference between the two items in such contexts? A reply has already been sketched out to this in the discussion above about parts of animate bodies. In this section this point is developed again and applied to the wider set of contexts where both items are possible. Since the identity of the referent is the same for both the article and the possessive in these cases, the distinction can only come from the way in which both markers view the identical referent from differing semantic perspectives. This semantic contribution is entirely down to a distinction of scope and topical continuity. The associative definite article, as pointed out above, downplays the role of the antecedent, while the possessive determiner actively underlines a topical link with the antecedent by creating a coreferential anaphor. In the first instance, a referent is identified as part of an antecedent whole by means of an associative anaphor (Kleiber, 2001a) which functions as a kind of referential space containing entities which can also be potentially part of this associative anaphor. 30. A forest has trees, a regiment has soldiers, but not the reverse.

 Georges Kleiber

The associative referent is thus understood in virtual opposition to these other entities which could equally have been taken up associatively. The referent of le tronc (du tilleul) / the trunk (of the limetree), as in: (40) Il s’abrita sous un vieux tilleul. Le tronc était tout craquelé. (He sheltered under an old limetree. The trunk was full of cracks.) is considered to be the unique object of the antecedent whole tilleul / limetree and is understood in terms of the paradigm of other entities (or Ns) which also have an associative link with the antecedent whole (leaves, branches, etc.). The particular or specific relation which links le tronc / the trunk and le vieux tilleul / the old limetree is based, as have argued above, on a generic relation between limetrees and trunks (un tilleul a un tronc / a limetree has a trunk). This leads us to express the referent, albeit particular and specific, in terms of a generic part-whole relation, thus making explicit its generic identity with other trunks. In contrast, the possessive creates a different topical chain in which the antecedent remains salient at the level of the new referent of a possessive expression. The identification of this referent is not based on a whole unit containing other entities at the same level, but instead comes from a direct link with the antecedent. This has two consequences. First, whereas the associative article implies an internal contrast between the antecedent and different Ns capable of being in the same associative relation, the possessive implies a contrast with different antecedents. The possessive son tronc / its trunk opens up the paradigm of trunks belonging to other individuals: the trunk of tree X, Y, or Z. As has also been pointed out above, the referent is thus oriented towards the identification of an individual rather than of a stereotypical or generic relation, as is the case with the definite article. Secondly, the pronominal nature of the possessive adjective requires the antecedent to be in a position of semantic salience (Kleiber, 1994b) and to have an individualising function (De Mulder et Tasmowski, 2000). This feature is not shared by the antecedent of the associative article and, while it leads to an interpretation of particularity, also builds on a rather transitional kind of discourse continuity. The possessive NP maintains a thematic link with its antecedent and at the same time creates a new topic capable of being the subject itself of a new continuity. It comes as a no surprise to see that in the following examples both determiners can be used with a referent which is identical but open to different readings: (41) a. Si tout se passe comme prévu, l’arbre sera enlevé le lundi 12; d’après le scénario, le / son tronc devrait être ébranché, coupé en trois morceaux et évacué avant midi. (DNA, 27/12/97) (Provided everything happens on time, the tree will be removed on Monday 12th. According to the schedule the / its trunk should be lopped, cut into three pieces and taken away by midday.)



The semantics and pragmatics of the possessive determiner 

b. La ville, à sept heures du matin, n’avait pas perdu cet air de vieille maison que lui donne la nuit. Les / Ses rues étaient comme de grands vestibules, les / ses places comme des cours. (J.L. Borgès, Fictions) (At seven in the morning the town had not lost the look of an old house at night time. The / Its streets were like great hallways, the / its squares like courtyards.) c. Dessiné par Gérard Ecklé, le grand bâtiment comprendra un rez-de-chaussée et 6 étages, les deux niveaux se trouvant en retrait. Les / Ses façades seront rythmées par des terrasses et des loggias. (DNA, 05/06/96) (Designed by Gérard Ecklé, the tall building will comprise a ground and 6 floors, with two levels being set in relief. The / Its façades will be broken by terraces and loggias.) d. Cette entreprise mène une action intéressante à signaler: les / ses salariés fabriquent bénévolement des fenêtres — 750 jusqu’à nos jours — au profit de la fondation. (DNA, 30/1/1998) (This firm is making a very significant gesture: the / its employees have been making windows free of charge – 750 of them up to now – as a gift to the foundation.) The discourse context lends itself here to both a possessive interpretation, individualising a particular referent in relation to the antecedent, as well as an interpretation based on the generic definite article linking the class of N1 antecedents to the class N2. Just taking the last example, with the article les salariés / the employees, the referent is understood by means of the functional relation X is an employee of (firm) Y or a firm has employees. Here no comment is made whatsoever about the particularity of the firm and its employees. Such an interpretation creates the possibility of making a virtual contrast with other entities in the same relationship, such as le patron / the boss. The employees are only particular occurrences of employees (working for the firm), when we consider the generic relationship which unites salariés / employees to entreprise / firm. On the other hand, the possessive (ses salariés / its employees) has the same referent, which is understood immediately as a particular one with specific employees (the employees of this firm). This individualisation is marked by a pronominal anaphor and reinforced by the fact that we are dealing with a particular firm whose employees are specifically working for charity. Our analysis predicts that if the context does not include elements which motivate an individualising specification of this kind, then the possessive becomes superfluous as we see in the following: (42) a. Le jour déclinait lorsqu’il arriva, avec son troupeau, devant une vieille église abandonnée. Le /?Son toit s’était écroulé depuis longtemps, et un énorme sycomore avait grandi à l’emplacement où se trouvait autrefois la /?sa sacristie. (Paolo Coelo, L’alchimiste)

 Georges Kleiber



(The day was waning when he arrived with his flock in front of an old abandoned church. The /?Its roof had collapsed long ago, and an enormous sycamore tree had grown at the spot where before there had been the /?its sacristy.) b. Fleischmann arriva enfin dans la rue de banlieue où il habitait chez ses parents dans une petite villa entourée d’un jardin. Il ouvrit la /?sa grille, sans aller jusqu’à la /?sa porte d’entrée. (J.L. Borgès, Fictions) (Fleischmann arrived at last in the suburban street where he had lived with his parents in a little detached house with a garden. He opened the /?its gate, without going up to the /?its door.) c. La voiture dérapa et alla s’écraser contre un platane. Le /?Son chauffeur avait été pris d’un malaise. (DNA, 15/02/92) (The car skidded and went on to crash into a plane tree. The /?Its driver had been taken ill.) Let us look at the last example, since the first two are straightforward enough. If son chauffeur / its driver seems less natural than le chauffeur / the driver, this is because the possessive focuses on a singularity and not a virtual opposition between other drivers which the context does not justify. The context, given that it deals with an accident, foregrounds the functional relation X is driver of Y and a coherent reading requires the associative definite article rather than the possessive. We would need to place more emphasis on the individuality of the antecedent than is required by the higher level pronominal definition implied by the possessive, and we would require more detail about how and at what level the different ingredients of this context fit together before justifying it. Such a task is clearly beyond the objectives of this paper. But at least we may have begun to set out a clearer path for future analysis!

References Azoulay, A. 1978. Article défini et relations anaphoriques en français. Recherches linguistiques françaises 7: 5–46. Barker, C. 1995. Possessive Descriptions. Stanford CA: CSLI Publications. Baron, I. & Herslund, M. (eds). 1997. Possessive Structures in Danish [KLIMT 3]. Copenhagen: Copenhagen Business School. Baron, I., Herslund, M. & Sørensen, F. (eds). 2001. Dimensions of Possession. Amsterdam: John Benjamins. Bartning, I. 1989. Le déterminant possessif et les compléments adnominaux en de. Revue Romane 24(2): 163–203. Bartning, I. 1996. Eléments pour une typologie des SN complexes en de en français. Langue française 109: 29–43, Bartning, I. 1998. Modèle intégré des syntagmes nominaux complexes en de. Romanische Forschungen 110(2): 165–184.



The semantics and pragmatics of the possessive determiner  Bartning, I. 2001. Towards a typology of French NP de NP structures or how much possession is there in complex noun phrases with de in French? In I. Baron & M. Herslund (eds), 147–167. Choi, I. 1991. Etude des compléments de manière non prépositionnels du type les yeux fermés, Thèse de doctorat de l’Université de Provence, Aix-en-Provence. Chomsky, N. 1981. Lectures on Government and Binding. Dordrecht: Foris. De Mulder, W. & Tasmowski, L. 2000. Déterminant possessif et anaphore associative. In L. Tasmowski (eds), 69–82. Ebert, K. 1982. Definite articles with inalienables in English and German. In W. Lohnes & E. Hopkins (eds), The Contrastive Grammar of English and German. Ann Arbor MI: Karoma. Eriksson, O. 1980. L’étiquette (fixée) sur la valise — l’étiquette de la valise — l’étiquette de valise. Studia Neophilologica 52: 389–414. Fradin, B. 1984. Anaphorisation et stéréotypes nominaux. Lingua 64: 325–369. Godard, D. 1986. Les déterminants possessifs et les compléments de nom. Langue française 72: 102–122  Gross, G. 1986. Syntaxe du déterminant possessif. In Déterminants: Syntaxe et sémantique, J. David & G. Kleiber (eds), 87–111. Paris: Klincksieck. Hanon, S. 1989. Les constructions absolues en français moderne. Louvain-la-Neuve: Editions Peeters. Hawkins, R. 1981. Towards an account of the possessive constructions: NP’s Nù and the N of NP. Journal of Linguistics 17: 247–269. Heine, B. 1997. Possession. Cognitive Sources, Forces and Grammaticalization. Cambridge: CUP. Heinz, M. 2003. Le possessif en français. Aspects sémantiques et pragmatiques. Bruxellles: De Boeck-Duculot. Herslund, M. & Baron, I. 2001. Introduction: Dimensions of possession. In Dimensions of Possession, I. Baron, M. Herslund & F. Sørensen (eds), 1–25. Amsterdam: John Benjamins. Husserl, E. 1962. Recherches logiques. Vol. 2. Paris: PUF. Julien, J. 1983. Sur une règle de blocage de l’article défini. Le français moderne 51(2): 135–156. Kleiber, G. 1981. Problèmes de référence. Descriptions définies et noms propres. Paris: Klincksieck. Kleiber, G. 1994a. Nominales. Paris: Armand Colin. Kleiber, G. 1994b. Anaphores et pronoms. Bruxelles: Duculot. Kleiber, G. 1999a. Anaphore associative et relation partie-tout: Condition d’aliénation et principe de congruence ontologique. Langue française 122: 70–100. Kleiber, G. 1999b. Les noms relationnels en anaphore associative: Le cas des noms de parenté. Studi Italiani di Linguistica Teorica e Applicata XXVIII(2): 282–300. Kleiber, G. 2001a. L’anaphore associative. Paris: PUF. Kleiber, G. 2001b. Anaphore associative, lexique et référence ou Un automobiliste peut-il rouler en anaphore associative. In Anaphores pronominales et nominales, W. de Mulder & C. Vet (eds), 27–42. Amsterdam: Rodopi. Kleiber, G. 2003a. The possessive via associative anaphora. In From NP to DP. Vol.II: The Expression of Possession in Noun Phrases, M. Coene & Y. D’Hulst (eds), 43–71. Amsterdam: John Benjamins. Kleiber, G. 2003b. Un « puzzle » référentiel en anaphore associative. In Lingua portuguesa: estruturas, usos e contrastes, F.I. Fonseca & A.M. Brito (eds), 97–110. Porto: Centro Linguistica da Universidade do Porto.

 Georges Kleiber Kleiber, G. 2004. Défini associatif et possessif en concurrence textuelle. In J.M. Adam, M. Ali Bouacha & J.B. Grize (eds), Texte et discours: Catégories pour l’analyse, 51–61. Dijon: Editions Universitaires de Dijon. Langacker, R.W. 1993. Reference-point constructions. Cognitive Linguistics 4(1): 1–38. Langacker, R.W. 1997. Possession and possessive constructions. In Language and the Cognitive Construal of the World, J.R. Taylor & R.E. Mac Laury (eds), 51–79. Berlin: Mouton de Gruyter. Langacker, R.W. 2004. Remarks on nominal grounding. Functions of Language 11(1): 77–113. Riegel, M. 1988. L’adjectif attribut de l’objet du verbe avoir: Amalgame et prédication complexe. Travaux de linguistique 17: 69–87. Riegel, M. 1989. Avoir + attribut de l’objet: Construction syntaxique et paradigme idiomatique. Recherches germaniques 2: 337–347. Riegel, M. 1997. De Il a les yeux bleus à Je n’ai pas le cinq à sept facile: Les avatars de la constrution attributive de avoir. Travaux et Recherches en Linguistique Appliquée, série B, 1: 99–108. Riegel M., Pellat J.C. & Rioul, R. 1994. Grammaire méthodique du français. Paris: PUF. Salles, M. 1995a. La relation lexicale « partie-de ». Thèse de Doctorat de l’Université de Caen. Salles, M. 1995b. Anaphore, partie-de et stéréotypes. Scolia 3: 47–58. Seiler, H.-J. 1983. Possession as an Operational Operator of Language. Tübingen: Narr. Spanoghe, A.-M. 1995. La syntaxe de l’appartenance inaliénable en français, espagnol et en portugais. Berne: Peter Lang. Strawson, P.F. 1973. Les individus. Paris: Seuil. Tamba, I. 1994. Un puzzle sémantique: Le couplage des relations de tout à partie et de partie à tout. Le gré des langues 7: 64–85. Tasmowski, L. (ed.). 2000. The Expression of Possession in Romance and Germanic Languages. Cluj-Napoca: Editura Clusium. Taylor, J.R. 1996. Possessives in English. An Exploration in Cognitive Grammar. Oxford: Clarendon Press. Vergnaud, J.R. & Zubizaretta, M.L. 1992. The definite determiner and the inalienable constructions in French and in English. Linguistic Inquiry 23(4): 595–652. Wilmet, M. 1986. La détermination nominale. Paris: PUF. Wilmet, M. 1997. 2003. Grammaire critique du français. 3rd edn. Bruxelles: Duculot. Zribi-Hertz, A. 1999. Le système des possessifs en français moderne. Langue française 122: 7–30.

Reference, determiners and descriptive content Thorstein Fretheim and Nana Aba Appiah Amfo

Norwegian University of Science and Technology, University of Ghana This paper starts out arguing that Gundel et al.’s claim that whatever a demonstrative can do, a definite article can do equally well is in need of revision. Then, against the tenor of Gundel et al.’s Givenness Hierarchy model, we postulate a univocal lexical meaning for determiners and corresponding pronouns in Norwegian, but we also show that what appears to be a conflation of definite article and distal demonstrative determiner in certain syntactic environments in Norwegian is two distinct linguistic phenomena in spoken Norwegian, and finally we argue that segmentally identical determiners and pronouns in the Niger-Congo language Akan are semantically distinct lexemes.

1. Introduction Some definite descriptions with the definite article as determiner have a descriptive content that helps the addressee to recognize the referent as intended by the communicator; others have a descriptive content that is of much less help in the task of reference resolution, because the addressee cannot immediately link the description to his current discourse model. In the latter case the substitution of a demonstrative determiner for the definite article will provide the addressee with information that is likely to lead to successful identification of the referent, a point argued for by Francis Cornish (1999: 53–59). While Cornish pleads for recognition of a qualitative difference between the semantic role of demonstrative descriptions with a deictic item like a demonstrative determiner and definite descriptions with a non-deictic definite article, Jeanette Gundel and her co-authors (Gundel, Hedberg and Zacharski 1993, 2001; Gundel and Mulkern 1998) define the semantic difference between demonstratives and the definite article exclusively in terms of their model of how pronouns, demonstrative determiners and articles all “serve as processing signals that assist the addressee in restricting possible referents” (Gundel, Hedberg and Zacharski 2001: 275). They postulate a universal implicational scale – the Givenness Hierarchy – of six distinct but implicationally related

 Thorstein Fretheim and Nana Aba Appiah Amfo

cognitively important memory and attention states, so-called ‘cognitive statuses’, which are matched in various ways with types of referring expressions in the world’s languages – not necessarily in such a way that each of the six cognitive statuses is correlated with a specific type of referring expression conventionally associated with it, although that happens to be the English situation as illustrated in (1) below. Each cognitive status, from the most restrictive one on top of the implicational scale (‘in focus’) down to the least restrictive (‘type identifiable’), is encoded by specific types of referring expression containing a determiner or a pronoun (exemplified to the right of each status1). Because any status on the hierarchical scale unidirectionally entails (>) all lower statuses, the actual cognitive status of a discourse referent may be higher than the status encoded by the determiner or pronoun appearing in the referring expression. For example, legitimate use of the definite article the in English requires a referent that is uniquely identifiable for hearer as well as speaker but an expression of the type the N can be used to refer even to a discourse entity whose cognitive status is manifestly higher than uniquely identifiable. Conversely, it is impossible to use a determiner or pronoun that encodes a higher status if the cognitive status of the referent is manifestly lower than the status encoded by the referring form, so a personal pronoun or a proximal demonstrative may not be used to refer to a non-activated discourse entity. (1)

Gundel et al.’s Givenness Hierarchy (with relevant English forms) in focus it > activated this; this N; that > familiar that N > uniquely identifiable the N > referential indefinite this N > type identifiable aN

Gundel et al. (1993) emphasize that use of the definite article does not require the addressee’s prior familiarity with the referent, because all that is needed is that the referent is uniquely identifiable and the definite article guarantees that that condition is met. When the referent is not familiar to the addressee, the conceptual content of the definite description should enable him to pick out a unique referent. In § 2 we argue that interpretation of descriptions with the definite article are just as context-dependent as interpretation of complex demonstratives, against Gundel et al.’s claim that a non-demonstrative definite description alone may be sufficient to secure unique identifiability. § 3 addresses the question raised in the opening lines of this paper, i.e. whether the definite article is the most relevant choice of determiner when the information needed 1. Gundel et al. customarily write the cognitive statuses and their associated referring devices from left (most restrictive) to right (least restrictive). We are here presenting the hierarchy vertically so as to enhance, by iconic means, the important point that lower statuses are implied by higher ones.



Reference, determiners and descriptive content 

to form a mental representation of the intended referent must be retrieved by means of a search among candidate referents stored in working memory. We argue that the descriptive content of a definite NP is occasionally quite unsuitable as a clue to reference resolution but that does not mean that there is anything wrong with the description, because all languages have demonstratives, a type of indexical expression that constrains the search space in which the referent is to be found. Consequently the appropriateness, and relevance, of a definite NP may depend on the fact that the speaker uses a demonstrative rather than the definite article. In § 4 we contrast proximal and distal demonstratives with reference to German, English and Norwegian. It is found that the German distal demonstrative jener has lexical properties that may pose a problem for the Givenness Hierarchy model, because the referent of a phrase modified by jener should be familiar to the addressee, yet it will not be a referent in focus. The Givenness Hierarchy approach to the relationship between referring expressions and the cognitive status of discourse referents has nothing to say about determiners and pronouns that set an upper bound on the implicational scale in addition to the encoded lower bound. The Scandinavian languages use nominal inflectional suffixes to express definiteness. However, when a head noun is qualified by a pre-nominal adjective, a pre-adjectival determiner appears as definite article, mostly in addition to the later definiteness suffix in Swedish and Norwegian but invariably instead of the suffix in Danish. Norwegian sequences of the type the Attribute N are indistinguishable from that Attribute N in writing but they are systematically kept apart in speech, at least in spoken East Norwegian as described in § 5. The Givenness Hierarchy in (1) shows that Gundel et al. consider that as demonstrative pronoun and as determiner to encode two different cognitive statuses, which implies that they are different lexical items. We are forwarding an alternative analysis. § 6 provides evidence that demonstrative determiners and corresponding pronouns, and even unstressed pronouns with the same form, are integrated in a single lexical item in Norwegian. In earlier work (Amfo and Fretheim 2005) we also postulated a univocal lexical analysis of the function word no in the Akan language spoken in Ghana (the Kwa branch of the Niger-Congo family), which serves the functions of definite article, distal demonstrative determiner, marker of certain dependent clauses, and personal pronoun. In Norwegian, the fact that a pronoun signals that the intended referent is at least activated is not a lexical property of the pronoun as such but a consequence of the fact that the referring expression contains no descriptive content that would be needed to refer successfully to a non-familiar or non-activated discourse entity. In Akan, however, the pronoun no is invariably realized with a high tone (nó) and all other functions of no are realized with a low tone (nò), which is detrimental to a unified analysis that subsumes the pronominal use of no and its other uses under a single lexical meaning.

 Thorstein Fretheim and Nana Aba Appiah Amfo

2. Definites, identifiability and familiarity Definite descriptions have an encoded descriptive content that regularly underdetermines the intended reference of the phrase, so the linguistic semantics of the referring expression needs to be supplemented by pragmatic inference, so-called enrichment of the encoded logical form (cf. Carston 1988, 2002, 2004). In his monograph on ‘complex demonstratives’, King (2001: 69) asserts that “the semantics of descriptions doesn’t allow supplementation by properties determined by speaker intentions”, and then he says that “the semantics of ‘that’ phrases does allow for such supplementation” (ibid). The former claim, which applies to definite descriptions with the as determiner, is truly remarkable for its total disregard of empirical facts. A quick look at any page in King’s own monograph should convince him that the reference of definite descriptions is indeed determined by the author’s intentions. For instance, when he writes, “This occurs in (15), and so the use of the ‘that’ phrase is felicitous” (ibid), his own intention is that readers should not understand the definite description “the ‘that’ phrase” within the larger definite description “the use of the ‘that’ phrase” to refer generically to ‘that’ phrases and the larger phrase generically to their use; rather, we infer that the intended reference is to a unique token of the kind of NP which appeared in his example (15) earlier on the same page 69 of his book. Any other interpretation would amount to a misunderstanding of King’s intention, we believe. A producer of linguistic stimuli relies on the addressee’s ability to draw context-dependent inference of the type made familiar in pragmatic research through the concept of enrichment (e.g. Carston 2002), and no matter how much descriptive content is included in a given definite description, the communicator will expect the addressee to add extra-linguistic inference to decoded information whenever needed to resolve the reference of the definite phrase. Gundel et al. (2001) do recognize the fact that pragmatic (context-based) inference permeates not only resolution of the reference of NPs with demonstratives but also resolution of the reference of NPs with the definite article, though we have previously accused Gundel et al. of inadvertently playing down the importance of the addressee’s pragmatic enrichment of the meaning of definite descriptions and its consequences for successful reference resolution (Amfo and Fretheim 2005). Gundel et al. address, and confront, the traditional view that the function of the definite article in English is to signal that the referent is familiar to the addressee, rejecting it in favour of their own view that the article signals no more than that the communicator assumes the addressee will be able to form a unique representation of the intended referent in his mind. Important cues intended to help the addressee identify the referent of a phrase of the type the N are obviously found in the descriptive content of the phrase, in whatever information is encoded in the complement of the. Other information that may be needed to establish a representation of a unique referent must be retrieved through pragmatic inference based on representations of first-order or higher-order entities stored either in the long-term memory or the working memory of the addressee. In a footnote, Gundel et al. (2001: 280) warn the reader of any possible misconception of



Reference, determiners and descriptive content 

their position by saying, “Note that the Givenness Hierarchy framework, or any other theory of definites that we know of, does not itself address the question of how the language processor determines the means for assigning a unique representation, e.g. whether by constructing a bridging inference or by searching for an existing representation in long-term or working memory. Such questions, we maintain, are outside the domain of a theory of definites or other referring forms, and can only be adequately addressed within a general pragmatic theory of language understanding”, and then they refer to Relevance Theory as an example of such a theory. Still, our impression is that their distinction between uniquely identifiable and familiar as two distinct cognitive statuses on their implicational scale is justified principally by the distinction between linguistically encoded information that is needed to identify a unique referent and the combination of linguistically encoded information and retrieval of information stored in memory needed to identify a so-called familiar referent. In the spirit of Relevance Theory, we (Amfo and Fretheim 2005) have criticized Gundel et al. for not granting more room for context-based retrieval of pertinent information even in situations where the referent is not familiar to the addressee, just uniquely identifiable. Let us look at a small part of an extended illustration that they have used (Gundel et al. 2001: 278). The reference here is to the (dropped) criminal case against Linda Tripp, the Pentagon employee whose secretly recorded conversations with Monica Lewinsky exposed an affair that led to President Clinton’s impeachment.

(2) The decision concludes the only criminal charge against a major figure in the most serious Oval Office scandal since Watergate.

Gundel et al. note that the intended referent of the descriptively rather meagre phrase the decision is uniquely identifiable by virtue of the fact that it is familiar. One sentence in the preceding discourse starts like this, “The chief state prosecutor of Maryland decided today to drop criminal wiretap charges against Linda R. Tripp, …”, so resolution of the reference of the later definite description the decision in (2) depends on whatever information stored in the reader’s short-term memory could be described as a decision to do something; or, if that information is already lost on the reader, then the relevance of his mental representation of the decision will depend on his ability to go back in the written text and find the required antecedent information. This latter procedure involves a bit of added processing effort triggered by the thought that it must be possible to identify the abstract-entity referent of the decision by tracing it via an examination of the preceding written lines, but this is not gratuitous processing, because the alternative would be to just continue reading without having anchored the definite description the decision contextually, that is, without having identified the explicature of (2). Whatever the reader will do to enrich the truth-conditionally underspecified phrase the decision so as to identify its intended referent, the enrichment is based on information that is accessible due to the structure of the discourse. According to Gundel et al., the longer definite description at the end of (2) contains “enough descriptive content encoded in the phrase itself to allow the reader to

 Thorstein Fretheim and Nana Aba Appiah Amfo

construct a new unique representation” (Gundel et al. 2001: 279), that is, one which is not based on familiarity with the scandal referred to there. The referent is claimed to be uniquely identifiable even if this scandal is not familiar to the reader. Now, what exactly does it mean to say that the addressee forms a unique representation of the referent in the absence of already known information? Surely, the descriptive content ‘most serious Oval Office scandal since Watergate’ was chosen with the kind of reader in mind who is at least familiar with the Oval Office in the White House in Washington D.C. and with the concept of Watergate and the ramifications of the clandestine operation known by that name. If none of that background information is familiar to the addressee, can we still maintain that the addressee has a unique representation of the intended referent upon hearing and processing (2)? It would seem that the relevance of (2) depends on retrieval of familiar information, even if the addressee is a person who fails to equate (what is here described as) the most serious Oval Office scandal since Watergate with the Clinton – Lewinsky scandal. We would claim that the linguistic form of the definite description is not relevant to those readers who are unable to associate anything with the Oval Office or Watergate. Readers who fail to do that may well arrive at some sort of mental representation of the referent but this would fall short of the unique representation which the writer intended them to activate. No amount of encoded information will normally be sufficient to dispense with inference based on information stored in the addressee’s memory. In fact, the more concepts encoded in a definite description, the more potential links can be made to referents that are stored in memory as part of the addressee’s encyclopaedic knowledge.

3. Demonstrative determiners vs. the definite article We mentioned above that it is not always the case that the description encoded in an NP is primarily geared to the addressee’s need for information that facilitates identification of the intended referent (Cornish 1999; Amfo and Fretheim 2005). When the description appears to be of an entity that is not readily retrievable in an accessible context, the definite article will be a bad choice of determiner. On the other hand, if the definite article is replaced by a demonstrative determiner, resolution of the reference of the phrase may depend entirely on the presence of the demonstrative, not on the descriptive content of the referring phrase, which may even contain information that is totally new to the addressee. Consider the difference between (3) and (4).

(3) We went past the town of Collymore. The mountainous road takes longer but it is definitely more scenic.



(4) We went past the town of Collymore. That mountainous road takes longer but it is definitely more scenic.



Reference, determiners and descriptive content 

An addressee who is not familiar with Collymore and the region in which that town is situated may understand the second utterance in (3) to be a predication about an alternative route, a stretch of road which does not include Collymore but which the addressee is able to form a mental representation of based on general knowledge of the region.2 We are not saying that this is the only possible referential interpretation of the mountainous road in (3) but it is a readily accessible interpretation. Contrasting (3) and (4), we find that substitution of demonstrative that in (4) for the in (3) gives the addressee the requisite procedural information that causes him to infer that ‘mountainous road’ is the speaker’s description of the road past Collymore, a discourse entity activated in the preceding utterance. In (4) the hearer does not have to know in advance that Collymore is located in a mountainous region, because correct reference assignment depends not on the descriptive content but on the deictic nature of the distal demonstrative determiner that. Gundel et al. would have no problem with what we are saying about the pair of (3)-(4), because their point is that a definite article and a demonstrative determiner have inherent semantic properties that force the addressee to associate a referring phrase of the type the N with a discourse entity that is at least uniquely identifiable and a referring phrase of the type that N with a discourse entity that is at least familiar. Nevertheless, there is nothing in their theory that accounts for the fact that the second utterance in (3) is less relevant than the second utterance in (4), if the speaker’s intention is to state that the road past Collymore takes longer and is more spectacular than some alternative route. The indexical nature of that in (4) eliminates the referential indeterminacy of the definite description in (3). According to the Givenness Hierarchy and its implications for English usage, the definite article in (3) should be sufficient to do the job that the demonstrative determiner that is doing in (4), because a determiner which encodes the information that the referent must be uniquely identifiable is appropriate when the referent is not just uniquely identifiable but even familiar or activated as well. Gundel et al. might respond that the referring phrase the mountainous road in (3) is an ‘inferrable’ (Prince, 19813; Gundel et al. 1993; Gundel 1996; Amfo and Fretheim 2005), which in the absence of a proper antecedent in the preceding utterance must be assigned reference via the bridging inference ‘The road past Collymore is a/the mountainous road’, or something similar. That kind of reaction would be fair enough, but Gundel et al. also predict that with ‘inferrables’, it should be relatively speaking easier to use a phrase which signals that the referent may be no more than uniquely identifiable than to use one which, like the determiner that in English, signals that the referent is not only uniquely identifiable but even familiar (see Gundel et al. 1993: 280–283, their section on ‘Inferrables’). Their 2. The relevance of (3) could depend on activation of the implicated assumption that they chose to go via Collymore because it is a shorter route. 3. Prince used a double ‘r’ rather than the orthographically correct form inferable, possibly to highlight the status of inferrable as a technical term different from the regular adjective inferable.

 Thorstein Fretheim and Nana Aba Appiah Amfo

prediction seems to be valid for those English examples that they themselves offered, some of which are attested data from fictional texts, but it is not true more generally. Both the mountainous road in (3) and that mountainous road in (4) must be pragmatically enriched so as to match the speaker’s intention to communicate an explicature, but only the latter expression directs the addressee’s attention to antecedent information in the preceding discourse, to be used in the enrichment process. We conclude that the special function of demonstrative determiners is that they direct the addressee’s attention to a familiar and, with proximals, activated referent regardless of the amount of conceptual meaning encoded in the referring phrase. The ability of demonstratives, whether determiners or pronouns, to facilitate the reference resolution process for an addressee rests partly on the procedural meaning (Blakemore 1987; Wilson and Sperber 1993) encoded by demonstratives, i.e. on the role they play as constraints on how the intended referent is to be identified. Referring expressions with demonstrative determiners are sometimes produced in a discourse situation where the corresponding phrase with the definite article in lieu of the demonstrative would have done a poorer job. Cornish (1999: 57) says that a referring expression with a demonstrative determiner like French ce(t)/cette or English this can introduce a predicate with a ‘non-presupposed status’, like the French head noun Américain in his attested example (5), where the use of italics is the author’s. (5) [Beginning of a radio news broadcast, France Culture, 13 August 1993] On apprend la mort hier de John Sowers. Cet Américain, âgé de 55 ans, était à l’origine de la photocopieuse... ‘We hear of the death yesterday of John Sowers. This American, aged 55, was the inventor of the photocopier …’ Cornish makes the point that the French public could not be expected to know who John Sowers is, so the information that he is an American is certain to be new to the listeners to the radio programme in which (5) appeared, and he goes on to say, “Based on this assessment of the common ground existing between newsreader and listeners, she [the newsreader; TF & NAAA] chose a demonstrative expression rather than a definite description, which would have presupposed the information regarding the referent’s nationality; such an expression type (i.e. L’Américain ‘The American’) would indeed have led to a degree of incoherence in this context, requiring the listener to set in motion an interpretative ‘repair’ strategy by ‘accommodating’ this presupposition into the common ground set up by his/her processing of the text up to this point. No such accommodation was needed, however, in the case of the original utterance, since this information was presented as being a new, ‘classifying’ item of knowledge.” (Cornish 1999: 57). A singular definite description with the definite article will normally enable the addressee to pick out a unique individual that fits the description but with demonstrative determiners there is no uniqueness presupposition associated with the descriptive content, which may describe many individuals correctly (Cornish op cit: 58). This is nicely illustrated by the following example which is due to Lita Lundquist (2005).



Reference, determiners and descriptive content 



(6) Après deux années passées auprès du premier ministre, don’t il était le très fidèle conseiller en communication, l’ancien producteur de télévision Dominique Ambiel teste bien malgré lui la “télé-réalité”. Soupçonné d’avoir eu recours à la prostitution d’une mineure, ce fidèle de Jean-Pierre Raffarin est obligé d’interrompre immédiatement sa carrière dans les coulisses du pouvoir. Après les téléfilms historiques [...] et enfin le pouvoir, cet infatigable homme de l’ombre est soudainement placé sous les projecteur des faits divers... ‘After two years with the Prime Minister whose loyal counsellor he was, the former TV-producer Dominique Ambiel is now himself testing a «reality show». Accused of having used a minor prostitute, this faithful friend of JeanPierre Raffarin is forced to immediately interrupt his career behind the political scene. After historical TV-films (…) and finally power, this untiring man from the shadow is all of a sudden placed in the limelight …’ Substituting le fidèle de Jean-Pierre Raffarin (‘the faithful friend of J-P R’) for ce fidèle de Jean-Pierre Raffarin (‘this faithful friend of J-P R’) in (6), we somehow get the feeling that the description ‘faithful friend of Jean-Pierre Raffarin’ is meant to fit only the individual Dominique Ambiel, but that is hardly the communicator’s intention. Similarly, the later phrase cet infatigable homme de l’ombre (‘this untiring man from the shadow’) introduces a new subordinate predication about M. Ambiel, which is in order because the description is preceded by an indexical, cet, which invites a bottom-up referential processing, as opposed to the top-down processing required to access the referent of the definite description l’infatigable homme de l’ombre (‘the untiring man from the shadow’) – if we pretend that that was the phrase employed by the newsreader.4 Amfo and Fretheim (2005) cite several attested examples from Norwegian of demonstrative determiners that are followed by a description which introduces information that is not contextually retrievable. Here we would like to add an authentic example from Akan (the Akuapem dialect), which illustrates the same point. Only the second sentence is rendered in its original Akan form, the preceding frame sentence is only given in English.

(7) ‘She got to an extremely old lady who was very scary indeed. This child (= she) stayed with this algae-infested old lady.’ … Abofra yi ne aberewa a ne ho nkonkonene yi tena-e. child pd conj old.lady rel poss skin algae pd stay-compl There are two tokens of the proximal demonstrative determiner yi (‘this’) in this Akan sentence. We are mainly interested in the second token at the end of the phrase aberewa a ne ho nkonkonene yi (‘this algae-infested old lady’), because there is no overt 4. Lundquist’s point was a different one, as she presented (6) with its two italicized definites in order to illustrate that what is actually new information may be deliberately presented by the communicator as information already known to the addressee. Our judgement is that such a position is not supported by empirical or theoretical evidence.

 Thorstein Fretheim and Nana Aba Appiah Amfo

information in the preceding sentence that explains why she looked so frightening. That she was infested with algae is new information to the reader. Even though the descriptive noun aberewa (‘old lady’) appears in both sentences, substitution of the definite article no for demonstrative yi, as in (7’), may confuse the Akan reader, who is forced to adopt a top-down referential processing strategy rather than a bottom-up strategy that exploits the fact that yi, being a proximal demonstrative determiner, guarantees easy access to a local discourse antecedent for the descriptive phrase aberewa a ne ho nkonkonene. (7’) … Abofra yi ne aberewa a ne ho nkonkonene no tena-e.5 child pd conj old.lady rel poss skin algae def stay-compl ‘This child stayed with the algae-infested old lady.’ Gundel et al. (1993: 294) make the following observation about the proper use of English determiners, ”Since each of the cognitive statuses in the Givenness Hierarchy entails all lower statuses, a particular form can often be replaced by forms which require a lower status. For example, the proximal demonstrative these, which requires that its referent be activated, could be replaced by those, which requires only familiarity; by the, which requires only unique identifiability; and sometimes even by an indefinite article (or zero if it is plural), which requires only type identifiability.”. They offer (8) as illustration.

(8) These incredibly small magnetic bubbles are the vanguard of a new generation of ultradense memory-storage systems. a. These systems are extremely rugged; … b. Those systems are extremely rugged; c. The systems are extremely rugged; d. New generation ultradense memory-storage systems are extremely rugged; The formulation “can often be replaced by forms which require a lower status” implicates that it is not always natural to replace a form encoding a high cognitive status with a form encoding a lower cognitive status. However, they do not mention that there are restrictions on the conceptual information entered in a phrase of type the N in contexts where the cognitive status of the referent is high enough to allow use of a proximal demonstrative. The definite article is a good choice of determiner if the descriptive content of its complement is chosen specifically for the purpose of directing the addressee to the intended referent, but as our examples (5), (6) and (7) show, not all descriptions help to direct the hearer to a representation of a unique referent, and the description may even have been offered for a different purpose altogether, being a novel characterization of the referent instead of offering indexical information. Their example (8), 5. PD stands for Proximal Demonstrative Determiner, REL for relative clause marker, and COMPL for completive aspect.



Reference, determiners and descriptive content 

in which (8c) is identical to (8a) and (8b) except for the choice of determiner, and where (8c) is felt to be as natural as the other two, may be an idealized case. Gundel et al. (1993) account for our tendency to understand pronominal demonstratives like that and this to refer to entities that are not in focus by invoking Grice’s First Maxim of Quantity (Q1), which demands that communicators make their contributions as informative as required (Grice, 1989). They say (Gundel et al. 1993: 297), “… use of a demonstrative pronoun, which requires only activation, often implicates by Q1 that the referent is not currently in focus; …”. While they make no explicit statement about similar scalar implicatures evoked by the use of demonstrative determiners, it is reasonable to interpret their generalization as one that extends to that N and this N as well. They furthermore contend that “the definite article is used to code infocus referents more frequently than demonstratives are” (Gundel et al. 2001: 277) and explain this with reference to the interaction of the two parts of Grice’s Maxim of Quantity, Q1 and Q2, the latter advising communicators to say no more than what is required in order to bring the intended message across to the hearer. For Gundel et al., using a demonstrative pronoun or determiner is to give more information about the referent than what is necessary. What the interlocutor has to know is that the referent is a uniquely identifiable entity, but the information encoded in a definite description is sufficient to achieve that, they say. Using a demonstrative determiner instead of the definite article would be a violation of Grice’s sub-maxim known as Q2. We judge their account to give a distorted picture of the very important difference between using an indexical like a demonstrative, which attracts the hearer’s attention to a discourse-local antecedent6 and a non-demonstrative device like the definite article, which implies that the hearer has to rely on top-down inferential processing to get at the intended referent.

4. Proximals and distals German usage shows a striking tendency for communicators to employ a demonstrative determiner at the expense of the definite article when the antecedent structure that must be located in order for the addressee to identify the referent is located in the immediately preceding part of the discourse. The German proximal dies- series contrasts with the distal series of demonstratives whose stem is jen-, but there are heavier constraints on the use of jen- forms in German than on English that and those. While the distal determiner those in the English source text of (9), from the Oslo Multilingual Corpus (OMC)7, corresponds to jener in the German target text, distal that of that smell in (10), a translation of 6. An ‘antecedent trigger’ in the sense of Cornish (1999). 7. OMC is a bi-directional translation corpus (http://www.hf.uio.no/german/sprik/english/ corpus.shtml). It includes source and target texts from English, German, French and Norwegian.

 Thorstein Fretheim and Nana Aba Appiah Amfo

Norwegian den (‘that’), corresponds to German proximal diesen and could not have been translated as jenen there. (9) His school, their school, in fact, had been one of those decrepit but expensive establishments created and run by an ex-army officer and his harassed wife that proliferated in England in the years before the Second World War. →8 Seine Schule, ihre Schule eigentlich, war eines jener baufälligen, aber teuren Institute, …

(10) Besides, Stine secreted a special smell when she exerted herself. That smell had blessed Benjamin ever since he was a baby. ← … Den lukten [= that smell] hadde gjort Benjamin velsignet helt fra han var spebarn. Lit. that smell had made Benjamin blessed all from he was infant → … Diesen [#jenen] Geruch hatte Benjamin schon als Kleinkind außerordentlich gemocht. Lit. this-obj smell had Benjamin already as small child extraordinarily liked German distal demonstrative forms in the jen- series not only signal conventionally that the minimum requirement is a familiar referent; they also signal that there is an upper bound as far as the cognitive status of the referent goes. By convention, as long as there is no more than one accessible candidate antecedent, the referent of a phrase modified by the distal determiner jen- cannot be activated. Gundel et al. (1993) consistently refer to minimum cognitive status requirements and do not allow for the fact that certain demonstratives might be associated with a maximum cognitive status requirement as well. Distal that of that smell in (10) is a perfect choice of determiner, because, as Gundel et al. say, activated implies familiar, so that smell will be in order even if the cognitive status of the referent is high enough for appropriate use of this smell, and possibly even high enough to make the personal pronoun it acceptable as a substitute for the distal demonstrative description appearing in (10). Norwegian is just like English in this respect. Den lukten (‘that smell’) is as acceptable in (10) as the proximal demonstrative alternative denne lukten (‘this smell’) would have been. However, German jenen in a phrase like jenen Geruch (‘that smell’) is a different kind of distal determiner than English that and Norwegian den, as it is reserved for referents whose cognitive status is either non-activated or not the most salient (in-focus) one of two referents that are both activated.

8. → means that what precedes that symbol is a source text and what follows it is a target text; ← means that the order of presentation of the original text and the target text is the opposite one.



Reference, determiners and descriptive content 

A proximal demonstrative determiner cannot be replaced by either a definite article or a distal demonstrative determiner when the reference is to something in the written text (metadiscursive reference). (11) is a part of a letter cited in a Norwegian OMC source text that is translated into English and German. (11) I felt like some wretched lizard, some twitchy creature that should have stayed on the ground, a conviction linked to the fact that it really was an old, blasé and half-way cocky descendant of a lizard that was flying the plane. You realise, because you’re reading these lines [#those lines; #the lines] and because you met me in Salamanca some months later, that the plane came down in one piece. ← … Ettersom du leser disse linjene [#de linjene (‘those lines’); #linjene (‘the lines’)] og dessuten møtte meg i Salamanca noen måneder senere, … ‘As you’re reading these lines and moreover met me in Salamanca some months later …’ → … Da du diese Zeilen [#jene Zeilen; #die Zeilen] lesen kannst und wir uns außerdem einige Monate nach diesem Flug in Salamanca begegnet sind, … ‘As you can read these lines and we moreover met with one another in Salamanca some months after this flight, …’ The German proximal demonstrative determiner also appears in referring expressions that have no overt antecedent in the preceding discourse and whose reference must therefore be established via a so-called bridging inference (Clark 1977). The English OMC source text in (12) contains a phrase with a distal demonstrative which was translated into German as a proximal demonstrative. There has been no prior mention of high walls but any reader will know that a cathedral has high walls and that knowledge is apparently sufficient to allow English distal those to be translated as German proximal diesen. (12) One could imagine them swarming into the cathedral for refuge in times of war or pestilence; how they would have scuttled through the squalor of their little hovels and tunnels to find sanctuary among those high walls against whatever threatened from outside. → Man konnte sich vorstellen, wie sie in Kriegs- oder Pestzeiten zufluchtsuchend in die Kathedrale geströmt, wie sie durch den Schmutz ihrer Häuschen und engen Gassen gehuscht und gekrochen waren, um zwischen diesen hohen Wänden eine Freistatt zu finden, in der sie sicher waren vor allem, was sie von draußen bedrohte. In the corresponding Norwegian target text rendered in (13) below, the original’s phrase those high walls is translated as de høye murene (the high-PLUR walls-DEFPLUR) where the determiner de [di], whose segmental form is the same as for the

 Thorstein Fretheim and Nana Aba Appiah Amfo

personal pronoun meaning ‘they’ and for demonstrative ‘those’, will automatically be read as the definite article because the referring phrase is an inferrable and above all because we have, as readers of a written text, a very strong tendency to read a determiner like de as unstressed by default. That processing works well in (13). As will be shown in § 5, it would in fact have been acceptable to pronounce the determiner de in such a way that it could be said to correspond to English those, but no reader could conceivably impose that marked prosody on the word in a read text as long as an interpretation of it as definite article makes the utterance optimally relevant. (13) Man kunne se for seg hvordan de strømmet inn i katedralen for å søke tilflukt der i tider med krig eller pest; hvordan de pilte gjennom elendigheten i sine egne små rønner og tunneler for å finne et tilfluktssted mellom de høye murene mot alt som truet utenfra. ‘One could envisage how they streamed into the cathedral to seek refuge there in times with war or pestilence; how they scuttled through the squalor in their own small shacks and tunnels in order to find a place of refuge between the high walls against all that threatened from outside.’ In § 5 it will be argued that in spoken (East) Norwegian, the pre-nominal definite article de in a phrase like de høye murene in (13) is a different lexical item than the distal demonstrative that the article is identical to in written Norwegian. Contrariwise, the 3rd person plural pronoun de (‘they’) turns out to be the same lexical item as the distal demonstrative determiner de (‘those’), and that is a point we are going to argue in § 6.

5. The Norwegian pre-nominal article vs. the distal demonstrative determiner Numerous languages inside and outside Europe use the same segmental forms for pronouns and determiners of the distal demonstrative type, such as English that and those, or German jener with its gender-, case- and number-dependent variants. Another common phenomenon is the formal identity of the definite article and 3rd person pronominal forms, as in Latin and its Romance descendants (cf. Harris 1980). The Scandinavian languages deviate from the rest of the Germanic languages in their apparent failure to distinguish phonologically not only between demonstrative pronouns, demonstrative determiners and unstressed pronouns, but also between the definite article and distal demonstrative determiner in definite NPs with a pre-nominal adjective. The Norwegian anaphor den in (15) is a demonstrative pronoun that could represent the same referent as the definite NP døra (‘the door’) does in (14). (14) Kontoret mitt er bak døra. office-def mine is behind door-def ‘My office is behind the door.’



Reference, determiners and descriptive content 

(15) Kontoret mitt er bak den. office-def mine is behind that ‘My office is behind that.’ When an adjective, or more generally an attribute, is added, as in (16), where the door is described as being red, the definite article is not just the suffix -a as in (14); encoding the definite article now includes a pre-nominal determiner, den, with the same morphological properties of non-human, common gender, singular as the prepositional object pronoun in (15), yielding what could be described as a discontinuous definite article den … -a. (16) Kontoret mitt er bak den røde døra. office-def mine is behind def red-weak door-def ‘My office is behind the red door.’ However, if we preserve the determiner den in the absence of an attribute, as shown in (17), the determiner is no longer the definite article, it is a distal demonstrative determiner. The fact that (14) and (17) have different meanings implies that den in (17) does not encode the same as den in (16) where the form is required for expression of definiteness. (17) Kontoret mitt er bak den døra. office-def mine is behind that door-def ‘My office is behind that door.’ If den is a definite article in (16) and a distal demonstrative determiner in (17), the obvious question to ask is how Norwegians differentiate formally between the definite article and the distal demonstrative when the noun phrase includes a pre-nominal adjective – if they ever make that distinction. Den døra in (17) refers to a door that must be familiar to the hearer, not just uniquely identifiable, but the phrase den røde døra appears to underdetermine the difference between ‘the red door’, which refers to a door that is uniquely identifiable but not necessarily familiar to the hearer, and ‘that red door’, which encodes the information that the door is familiar. Is that just a context-driven pragmatic difference between (16) and (16’), the latter differing from (16) only in our glossing of den as ‘that’? (16’) Kontoret mitt er bak den røde9 døra. office-def mine is behind that red-weak door-def ‘My office is behind that red door.’ As there is apparently no formal distinction between den røde døra glossed as ‘the red door’ in (16) and den røde døra glossed as ‘that red door’ in (16’), we can conclude one 9. This adjective is composed of the stem rød- (‘red’) and the so-called weak form of the adjective, an invariant suffix -e regardless of the gender and number of the head noun. The weak adjectival form is a marker of agree-ment with the definiteness of the NP.

 Thorstein Fretheim and Nana Aba Appiah Amfo

of two things: (a) Norwegian lacks the ability to express a distal demonstrative when the noun phrase includes a description in the form of an attributive adjective, or (b) the definite phrase den røde døra is inherently ambiguous between a definite article reading of den … -a and a distal demonstrative reading of den, the latter implying that the definite article is encoded solely by the suffix -a. Option (a) would seem to be supported by the observation that there is no formal difference between (16) and (16’). One could hypothesize that the pre-adjectival determiner is the definite article in (16)/(16’). It would then be semantically vague and could be glossed as ‘the’ or ‘that’, depending on context. However, there is a systematic formal difference between encoding with the definite article and encoding with a distal demonstrative (in addition to the suffixed article) when no attributive adjective is present. Den døra in (17) means ‘that door’, and contrasts with døra (‘the door’) in (14). This fact suggests that option (b) involving an ambiguity analysis may have the upper hand. It will now be shown that the ambiguity analysis is the correct one for spoken Norwegian. While it is not impossible to pronounce a Norwegian demonstrative determiner with an accent for a contrastive purpose, such words are more frequently unstressed in Norwegian utterances. Word-accent is not normally used to keep demonstrative determiner and definite article apart. Still, spoken Norwegian does use a prosodic contrast to distinguish systematically between demonstrative and article, not only in cases like (16) where a linguistic contrast might be communicatively important but also in descriptive phrases like den døra (‘that door’) in (17), where the syntactic form and the prosodic handling of den separately mark it as a demonstrative. In a given discourse, an utterance of (16) produced with the intonation contour of Figure 1 would normally be an acceptable alternative to the contour of Figure 2, but not vice versa. 240 F0 [Hz] 160 80 0

0

Time [sec]

2.31838

[[kon[1toret – mitt – er]AU [ 1bak – den]AU røde [1DØRA]AU ]FP ]IU office-def mine is behind def red-weak door-def ‘My office is behind the red door.’ Figure 1.



Reference, determiners and descriptive content  240 F0 [Hz] 160 80 0

0

Time [sec]

2.31838

[[kon[1toret – mitt – er]AU [ 1 bak ]AU den røde [1DØRA]AU ]FP ]IU office-def mine is behind that red-weak door-def ‘My office is behind that red door.’ Figure 2.

The stretch of syllables bak den forms an Accent Unit (AU) in Figure 1, [1bak – den]AU, which is here composed of the accented preposition bak (‘behind’) and the unaccented pre-adjectival determiner den (‘the’). The word-tone on the syllable bak is Low and there is a pitch step up to the unaccented syllable den, across the unvoiced segment [k] of bak, which discloses the AU-internal nature of the unaccented syllable. The demonstrative den in Figure 2, on the other hand, is extrametrical, in the sense that it belongs to no AU. This implies that the pitch movement from Low to High on bak den observable in Figure 1 is now replaced by a similar Low-High pitch movement completed in the course of the single syllable bak that exhausts the AU in Figure 2. The two syllables constituting the adjectival form røde is AU-external in both F0 trackings, and since the next tone in the global contour is the Low tone on the accented syllable of the noun døra, it is natural that the pitch decreases as it approaches that target tone. It is the extrametrical nature of the determiner den in Figure 2, which encodes the information that this is a demonstrative. As opposed to the pre-adjectival definite article, a demonstrative determiner is invariably extrametrical. The fact that the demonstrative determiner den in (17) cannot be incorporated in the same AU as bak proves that extrametricality is a conventional prosodic property of the demonstrative determiner den, while the (pre-adjectival) definite article den is invariably included in the same AU as the preceding accented item. A spoken utterance of the sentence in (17) must be pronounced with an AU-external form den (‘that’) as indicated in (17’a). The intonational phrasing indicated in (17’b) where the determiner is AU-internal is illformed, because it is only as a definite article that den can be AU-internal. This data suggests that the definite article in an adjective-free sentence like this one is just the suffix -a of døra, as in (14).10 10. The Danish phrase corresponding to Norwegian den røde døra is den røde dør (‘the red door’). If the suffix -a is the only marker of [+definite] when den in the Norwegian phrase is extrametrical, then what item encodes the definite article in the Danish phrase that contains no such suffix? Alex Klinge reminded us that one should have very good arguments for rejecting an analysis of the determiner den as article in Norwegian, as Norwegian and Danish are so closely related. However,

 Thorstein Fretheim and Nana Aba Appiah Amfo

(17’) a. [[kon[1toret – mitt – er]AU [ 1 bak ]AU den [1 DØRA]AU ]FP ]IU office-def mine is behind that door-def ‘My office is behind that door.’ b. *[[kon[1toret – mitt – er]AU [ 1 bak – den ]AU [1 DØRA]AU ]FP ]IU The extrametrical determiner den in Figure 2 is a distal demonstrative requiring a referent that is familiar to speaker and hearer, despite the fact that the word is unstressed and that its segmental form is identical to that of the definite article. As the form den in the adjective-free sentence (17) is an unequivocal demonstrative, it would not be necessary to make that word form AU-external in order to signal its category membership. Den is nevertheless obligatorily AU-external in a spoken utterance of (17). That is what we would predict, given our claim that the special AU-external handling of den in (17’a) is a conventional way of expressing the Norwegian demonstrative determiner. By making the determiner IU-internal, as in Figure 1, the speaker signals that the determiner is not a demonstrative but the definite article, which it cannot be in (17) for syntactic reasons, and (17’b) is therefore ill-formed. The pair of utterance types whose intonation structure is indicated by the labelled bracketing notation of (18) is of a similar sort. A potential pragmatic difference between them is that (18a) with its AU-internal definite article den would be in order in a context in which the definite description den neste filmen hans (the next film-DEF his; ‘his next film’) is used attributively (Donnellan 1966) to refer to whatever movie may be his next one, while an utterance of (18b) with the AU-external den implicates that the speaker is familiar with what has been announced to be the film director’s next project. (18) a. b.

[[jeg ser [1FRAM-til-den]AU ]FP [[2NESTE]AU ]FP [1filmen-hans]AU ]IU I look forward to def next film-def his ‘I’m looking forward to his next film.’ [[jeg ser [1FRAM-til]AU ]FP [ den [2NESTE]AU ]FP [1filmen-hans]AU ]IU I look forward to that next film-def his ‘I’m looking forward to that next film of his.’

We have argued in the present section that the determiner is a definite article in Figure 1 and a distal demonstrative in Figure 2, and that the two are lexically distinct. It was shown that the way you express the definite article as a separate word in Norwegian and the way you express the distal demonstrative determiner follow mutually exclusive prosodic patterns, even though the segmental phonological forms of the article

there is a bit of variation between the Scandinavian languages in this area of grammar. Swedish allows definite NPs with a pre-nominal adjective but without a pre-nominal determiner, just the suffix, as in en bild på stora rummet och öppna spisen (a picture of big room-DEF and open fireplace-DEF, i.e. ‘a picture of the big room and the open fire-place’). See Dahl (2004).



Reference, determiners and descriptive content 

and the demonstrative are identical. The demonstrative is AU-external and unstressed and the definite article is AU-internal and unstressed.11

6. The dual category membership of demonstratives If a hearer is able to form a mental representation of a unique referent, it is of minor interest to reflect upon the referent’s actual cognitive status at the time of utterance. We agree with Gundel et al. on that point. An interesting upshot of this realization is that we should not be surprised to find that one and the same lexeme helps us pick out a referent that is non-activated when the item is a determiner in a descriptive phrase but activated when it is a pronoun used without the support of an encoded conceptual structure. There is a universal tendency for series of pronouns to be phonologically identical to corresponding series of determiners (if one disregards possible affixes attached to the word stem). And it is the distal determiners that typically pattern with pronouns. Let us turn to a consideration of the relationship between the Norwegian distal determiner forms and a pronoun like den in (15), repeated here. (15) Kontoret mitt er bak den. office-def mine is behind that ’My office is behind that.’ We would like to look into the possibility that the pronoun den in (15) and the AUexternal determiner den represent a single lexeme with a dual category membership. Their different pragmatic roles could be a function of the fact that the indexical den in (15) contains no descriptive content, while the demonstrative determiner occurs in referring expressions with more or less descriptive content encoded by a head noun and possible modifiers. The former situation requires an activated referent, while the latter situation only requires one whose cognitive status is familiar. A pronoun like den in (15) may be either unstressed or stressed, which has an impact on our ideas about the cognitive status that the referent has for the speaker but which should not tempt us to conclude that unstressed and stressed den are distinct lexical items like it and that. Languages generally distribute stressed and unstressed word forms according to very well known patterns. For instance, an unstressed proper name signals that the referent is activated, an accented one does not. The complementary syntactic distribution of pronouns and determiners arguably supports an analysis that subsumes segmentally identical forms under a single lexeme. The analysis propounded by Gundel et al. (1993) is that unstressed personal pronouns encode the information that the referent is in focus and demonstrative pronouns 11. It should be noted, though, that an AU-external word form is perceptually closer to an accented form than an AU-internal form is, so the extrametricality probably does imply a heightening of the prominence of the word.

 Thorstein Fretheim and Nana Aba Appiah Amfo

that it is activated but not in focus; a distal determiner encodes the information that the referent is familiar but not necessarily activated, and so on. We are offering a radically different proposal that emphasizes rather more the presence or absence of encoded conceptual content in a definite NP. When a linguistic form is used as a pronoun, it is redundant to specify as part of its lexical meaning that the entity it refers to is activated. What definiteness in NPs and pronouns really amounts to is that the referent can be assumed to be uniquely identifiable. Higher cognitive statuses, as defined within the Givenness Hierarchy framework, can be encoded linguistically, but our recognition that the cognitive status of a referent is at least activated can also follow from the absence of descriptive content that would help to direct the hearer to the intended referent. What may be a single syntactic pattern, and what is definitely the same propositional content, in a pair like (18a) and (18b), was argued to be two semantically distinct utterance types, defined by Norwegian grammar, because an utterance of (18b) encodes the assumption that the speaker presents the film referred to as familiar, while (18a) just encodes the assumption that the referent is uniquely identifiable. This difference has communicative consequences, as only (18b) allows the interlocutor to derive the conclusion that the speaker knows the director has concrete plans to make a new film. It is also evident that (14) with just a suffixed article and (17) with an added determiner encode different information about the cognitive status of the referent, uniquely identifiable in the former, familiar in the latter – though the actual cognitive status of the door referred to may well be activated for the conversational parties in a real discourse that includes (14) or (17). One potential problem that faces an attempt to equate pronouns and determiners lexically in Norwegian is the fact that the singular pronoun den cannot normally be used to refer to a human referent, but the determiner den is fine even if the reference is to a human being. The prepositional object argument in (19a) refers to a non-human entity, like the determiner in (19b), but den in (20b) is the correct determiner whether the referent is human or non-human. (19) a. b.

Se på den. look at that/it-utr ‘Look at that (one).’ / ‘Look at it.’ Se på den båten. look at that-utr boat-def-masc ‘Look at that boat.’

(20) a. b.

Se på han. ‘Look at him.’ Se på den mannen. look at that-utr man-def-masc ‘Look at that man.’



Reference, determiners and descriptive content 

When the NP is plural, however, there is no encoded human/non-human opposition. Pronouns and distal demonstrative determiners have the same form, subject to person distinctions only. Consider (21)-(25). (21) De er mine. they are mine-plur ‘They are mine.’ (22) De avisene er mine. they papers-def are mine-plur ‘Those newspapers are mine.’ (23) De barna er våre. they children-def are ours-plur ‘Those children are ours.’ (24) Dere må vente. you-plur must wait ‘You have to wait.’ (25) Dere andre må vente. you-plur others must wait ‘The rest of you have to wait.’ Use of a 1st or 2nd plural pronoun as determiner should not be confused with use of such pronouns in generic NPs, where the pronoun appears before a plural noun that is not definite. Compare the determiner use of vi (‘we’) in (26) with (27) where vi is a pronominal head and where the indefinite plural nordmenn (‘Norwegians’) is used generically. The subject of (26) does not refer generically to the Norwegian population but to a designated set of individuals. (26) Vi nordmennene sa ingenting. we Norwegians-def said nothing ‘Those of us who were Norwegians said nothing.’ (27) Vi nordmenn er dårlige tapere. we Norwegians-indef are bad-plur losers ‘We Norwegians are bad losers.’ Lyons (1999) introduces the term ‘personal determiner’, by which is meant “forms related to personal pronouns occurring within full phrases, of the type we teachers, you students.” (Lyons 1999: 141). There is no formal differentiation in English, however, that corresponds to the difference between Norwegian (26) and (27). Certain patterns of usage that are typical of colloquial Norwegian but hardly ever found in written form suggest that Norwegian pronouns and determiners are felt to belong together lexically. A syntactic form like (28) has become a popular alternative to (20b) in urban and rural dialects alike.

 Thorstein Fretheim and Nana Aba Appiah Amfo

(28) Se på han mannen. look at he man-def-masc ‘Look at that man.’ If we expand (28) by adding an attributive adjective between determiner and head N, the AU-external prosodic handling of the determiner is linguistically mandated, as shown in (30). (29) Se på han lange mannen. look at he long-weak man-def ‘Look at that tall man.’ (30) a. [[[1se-på]AU han [2lange]AU [1mannen]AU ]FP ]IU b. *[[[1se-på-han]AU [2lange]AU [1mannen]AU ]FP ]IU A personal pronoun like han is a demonstrative determiner in (29), which is why it cannot be AU-internal as in (30b). The prosodic pattern with the determiner han contrasts with the options available when han is replaced by inanimate den, shown in (31)-(32). Because den is either a distal demonstrative or (part of) a definite article, (32a) and (32b) are both acceptable intonations. (31) Se på den lange mannen. ‘Look at that tall man.’ / ‘Look at the tall man.’ (32) a. b.

[[[1se-på]AU den [2lange]AU [1mannen]AU ]FP ]IU ‘Look at that tall man.’ [[[1se-på-den]AU [2lange]AU [1mannen]AU ]FP ]IU ‘Look at the tall man.’

The structural difference between (30a-b) compared to (32a-b) makes us still more reluctant to subsume the Norwegian pre-adjectival definite article forms under the same lexeme as the corresponding pronouns/demonstrative determiners, but a conclusive judgement on that issue would be premature. Norwegian determiners that are formally identical to personal pronouns are also frequently found to appear in an expanded form consisting of the pronoun and the place adverb der as illustrated in (33). (33) Har du sett de der fakturaene? have you seen they there invoices ‘Have you seen those invoices? (I can’t find them.) In a large number of Norwegian dialects the proximal demonstratives too have a basic form that does not differ from personal pronouns and pronominal anaphora. You simply add the proximal place adverb her (‘here’) to the determiner, as in (34), instead of the distal place adverb der (‘there’) of (33) and get a distal/proximal distinction reminiscent of French ce N-là vs. ce N-ci.



Reference, determiners and descriptive content 

(34) Har du sett de her fakturaene?12 have you seen they here invoices ‘Have you seen these invoices?’ There are many dialects in which specific proximal forms like disse (‘these’), denne (‘this’, non-neuter) or dette (‘this’, neuter) are non-existent, being replaced by a demonstrative plus the proximal locative index her (‘here’). Although there is no sign that the pronouns han (‘he’) and hun (‘she’) encoding [+human] are going to supplant den and det as determiners when the reference is to a human discourse entity, the tendency to let the same form function both as pronoun and distal determiner is seen to be quite strong in Norwegian. In the present section we have tried to demonstrate that personal pronouns and distal demonstrative determiners pattern together in Norwegian. Both are indexicals, but while the determiners are accompanied by a phrase that offers a description of the referent, the pronouns are not. The fact that pronouns must be saturated by association with an antecedent in the immediately preceding discourse is due to the lack of conceptual meaning encoded by an unaccompanied pronoun, as opposed to a pronoun that is the head of a relative clause, as in de som har spist middag (they that have eaten dinner, i.e. ‘those who have had dinner’). Thus a pronoun like de does not encode the information that the cognitive status of the intended referent is activated. In contrast, a proximal demonstrative determiner encodes the procedural information that the referent is activated, no matter whether its lexical form deviates from the distal counterpart or whether the distal/proximal distinction is taken care of by means of the distal and proximal place adverbs der and her, respectively, which are integral parts of the demonstratives. Diessel (1999) makes the point that many languages do not distinguish formally between demonstrative pronouns and demonstrative determiners, and he notes that there may even be languages with words whose function resembles that of demonstrative determiners but which are grammatically pronouns. Reporting on Mithun’s analysis of demonstratives in Tuscarora (1987), he says (Diessel 1999: 4) “Unlike adnominal demonstratives in French, adnominal demonstratives in Tuscarora do not function as determiners; rather, they are used as independent pronouns that are only loosely adjoined to a coreferential noun in apposition.” Tuscarora seems to represent an extreme case where the grammatical category of determiners is wiped out. There may be a gamut of relations, however, between demonstratives as independent pronouns and demonstratives as determiners in an NP with minimally a head noun in addition to the determiner. One apparently rather frequent situation is the one we have postulated for Norwegian. Distal demonstrative pronouns and the corresponding determiners

12. The distribution of nominative vs. accusative forms of the determiners in non-subject phrases also depends on the dialect spoken. Many dialects use accusative case forms, e.g. dem der/ dem her instead of the nominative forms de der/de her that appear in (33)–(34).

 Thorstein Fretheim and Nana Aba Appiah Amfo

are not lexically distinct but the lexical members of the set of demonstratives behave syntactically either as pronoun or as determiner. In the next section we will consider to what extent arguments similar to the ones brought forth in the present section are valid in an account of the semantics of determiners and pronouns in Akan.

7. Akan no: a pronoun and a determiner? The word no in Akan has a number of grammatical functions. First, no is an anaphoric non-subject pronoun referring to a 3rd person singular animate entity.13 Second, no is sometimes referred to as a definite article contrasting with the indefiniteness marker bi. However, the no – bi contrast is not an opposition on a par with the grammatical [± definite] distinction in English. Bi and no can co-occur, in the sequence bi no, and bi encodes an instruction to the addressee to look for a referent with Gundel et al.’s status referential, as opposed to just type identifiable. An Akan speaker can introduce a female discourse referent by referring to her as maame bi (‘a certain woman’). The next time around she can be referred to as maame bi no (‘the same woman’), where the part maame bi echoes the previously employed form. An entity whose cognitive status is just type identifiable will be referred to by means of a bare noun, without bi. Third, no is a so-called dependent clause marker (DCM) attached to the right of an embedded clause whose propositional content is presented as given information. A dependent clause followed by no can (a) impose a temporal constraint on the main clause proposition, as in (35), (b) mark a relative clause in a syntactic construction where the overt head noun is outside the scope of no, or (c) mark one of the propositions in a substitutive (‘instead of ’) construction exemplified in (36), where the future tense prefix bε- of bεtf indicates that the state of affairs described in the no-marked clause is unfulfilled. (35) Adjoa ba-e no, me-a-n-hu no. Adjoa come-compl dcm I-compl-neg-see her ‘When Adjoa came, I didn’t see her.’ (‘I didn’t see Adjoa when she arrived.’) (36) Yε-de bε-tf atadeε no, yε-n-fa n-tf bukuu. we-take fut-buy dress dcm we-imp-take imp-buy book ‘Instead of buying a dress, let’s buy a book.’ An utterance of (35) presupposes the hearer’s familiarity with the fact that Adjoa came and an utterance of (36) presupposes that someone brought up the possibility of buying a dress. In a recent paper (Amfo and Fretheim 2005) we argued that no has the dual grammatical category membership postulated for Norwegian determiners/pronouns in § 6. 13. An inanimate non-subject is often represented by zero.



Reference, determiners and descriptive content 

True, no is both a pronoun and a determiner and in the capacity of determiner its role is not only that of a ‘definite article’ and an unmarked (i.e. distal) demonstrative, because it can even take scope over a clause, as illustrated in (35) and (36). In our glossing of the Akan data in (35) we used DCM and her, respectively, and we are using the gloss DEF for the ‘definite article’ function of no in (37) below. While this glossing reflects what we take to be three different grammatical functions of Akan no, we postulated a single lexeme with a univocal procedural semantics, to cover the pronominal, determiner and DCM functions. While the post-nominal determiner no encodes the information that the referent is (at least) uniquely identifiable, no may co-occur with the pre-nominal demonstrative determiner saa, which presupposes a later appearance of no, or with the proximal demonstrative determiner yi of N yi (‘this N’), whose pronominal counterpart is eyi, both encoding that the referent is activated. (37) Mε-kyere wo saa mfoni no a e-firi Egypt no. I.fut-show you dem pictures def rel they-come.from Egypt dcm ‘I’ll show you those pictures from Egypt.’ As saa requires a familiar referent for the referring phrase saa mfoni no (‘those pictures’), there is a potential difference in communicated meaning between (37) and (38). The definite phrase in (37) refers to certain photos that the hearer is assumed to be familiar with, while mfoni no in (38) can refer to photos that the speaker believes the hearer to have no prior knowledge of. (38) Mε-kyere wo mfoni no a e-firi Egypt no. I.fut-show you pictures def rel they-come.from Egypt dcm ‘I’ll show you the pictures from Egypt.’ The demonstrative determiner saa has a contrastive function, which explains why it can appear with proximal yi as well as with no. The existence of a determiner saa demanding a familiar referent causes no problem for a univocal lexical semantics for no similar to what we postulated for Norwegian pronouns and determiners, but we now see that there are valid grounds for rejecting an analysis that does not differentiate lexically between the pronominal and the determiner use of no. This word occurs either with a high (H) or a low (L) tone – nó vs. nò – and the distribution of H and L is such that H is an inherent lexical property of the pronoun, while the other functions require L. While nó and nò were presumably lexically undifferentiated at an earlier historical stage, the systematic distribution of H and L in present-day Akan is indicative of a lexical split. There is a formal and functional difference between nó, which refers to an animate non-subject referent, and f-nó and -nó, which are emphatic pronouns whose respective prefixes represent animate and inanimate. The latter forms are typically not used with reference to in-focus entities. When the prefixes f- (animate) and ε- (inanimate) appear with verbs, they are encoders of highly accessible (in-focus) subject referents.

 Thorstein Fretheim and Nana Aba Appiah Amfo

(39) is an approximation to what a Givenness Hierarchy representation of the more salient determiners and pronouns in Akan might look like. There is an undeniable semantic difference between the forms aligned with ‘in focus’ and the forms aligned with ‘activated’ here, which differs from the Norwegian situation displayed in (40), and no appears twice, as nó on top of the hierarchy and as nò further down.14 By comparison, the Norwegian pronoun/distal demonstrative determiner den (‘that’, ‘it’) occurs just once in (40), where it encodes the status familiar. (39)

The Givenness Hierarchy (for Akan) in focus f- ‘he/she’; ε- ‘it’; nó ‘him/her’ > activated f nó ‘S/HE’; εnó ‘IT/THAT’; N yi ‘this N’; eyi ‘this’ > familiar saa N nò ‘that N’ > uniquely identifiable N nò ‘the N’ > referential N bi ‘a certain N’ > type identifiable NØ

(40)

The Givenness Hierarchy (for Norwegian) activated denne N-en ‘this N’; denne ‘this’ familiar den (N-en) ‘that N-DEF’, ‘that’, ‘it’ uniquely identifiable (den Adj) N-en (‘the Adj) N-DEF’ type identifiable en N

> > >

No personal pronoun in Norwegian encodes the highest cognitive statuses in focus or activated, but [±stress] in pronouns15 is a universal non-lexical procedural device that distinguishes between reference to the most salient discourse entities and reference to those that are activated but not maximally salient.

8. Conclusion This paper has addressed a number of predictions that follow from Gundel and al’s Givenness Hierarchy, a theoretical framework which accounts for the way pronouns and determiners constrain an addressee’s search for the referent of NPs. We have warned against Gundel et al.’s expression of the view that the definite article plays a role not significantly different from that of demonstratives, their reason being 14. It looks as if we are listing two different nò’s as well, in saa N nò, but this is the same nò that encodes the status uniquely identifiable. It occurs twice in (39) because saa presupposes a later nò in the same NP. 15. This is true not only of dialects in which linguistic structures such as (28) and (29) are fine, it is also true of a pronoun like han (‘he’) in Standard Norwegian phrases where han is the head of a complex nominal with a restrictive relative clause, like han som jeg tenker på (lit. he that I think on), meaning ‘the one [male] that I’m thinking of ’, whose referent may be unfamiliar to the addressee at the time of utterance.



Reference, determiners and descriptive content 

that all a hearer needs to know to identify a referent is that it can be picked out uniquely and a definite article does exactly that. We pointed out certain consequences of the fact that demonstratives, unlike articles, are indexicals that direct the hearer to an accessible referent independently of the descriptive content of the referring expression. The Givenness Hierarchy model postulates different meanings for the determiner that of that N and for the pronoun that. We have argued that pronouns and corresponding determiners encode the same information about the cognitive status of the discourse entity referred to. The reference of a pronoun is resolved on the basis of information represented in the working memory of the addressee. That their referents must be activated follows from the fact that these words are anaphora. Since it is predictable, this information should not be specified in the lexicon. Gundel et al. consider use of a demonstrative determiner or pronoun to be potentially misleading when the reference is to an in-focus entity, but at the same time there is no conventional meaning in demonstratives which prohibits reference to such an entity. More generally, there is no rule against referring to an in-focus entity with the help of a referring expression associated with a less restrictive cognitive status. We have shown, however, that German jener is a demonstrative which requires not only that the referent is at least familiar but also that it is at most activated. The determiner in Norwegian NPs with an adjective is the same form for the definite article and the distal demonstrative in written Norwegian. Spoken Norwegian reveals that this formal identity is not due to a neutralization of the opposition between definite article and demonstrative determiner. Intonational evidence was provided, which proves the two to be lexically distinct. On the other hand, Norwegian pronouns and segmentally identical demonstratives were found to belong to a single lexical item, different from the pre-nominal definite article. In contrast to the Norwegian picture, what may look like a single multifunctional word no in Akan was shown to be two distinct lexemes, because the pronoun no requires H tone and the determiner L tone.

Acknowledgements We would like to express our gratitude to Jeanette Gundel, to Walter De Mulder, to Francis Cornish, to Jorunn Hetland, to our colleagues Wim van Dommelen and Kweku Osam, and to the editors Alex Klinge and Henrik Høeg Müller whose inspiring comments on the first draft made the revision a very enjoyable task. We are solely responsible for any infelicities or graver mistakes.

 Thorstein Fretheim and Nana Aba Appiah Amfo

References Amfo, N.A.A. & Fretheim, T. 2005. A relevance-theoretic approach to linguistic signaling of the cognitive status of discourse referents. In Grammar and Beyond: Essays in Honour of Lars Hellan, M. Vulchanova & T. A. Åfarli (eds), 131–156. Oslo: Novus. Blakemore, D. 1987. Semantic Constraints on Relevance. Oxford: Blackwell. Carston, R. 1988. Implicature, explicature and truth-theoretic semantics. In Mental Representations: The Interface between Language and Reality, R. Kempson (ed.), 155–181. Cambridge: CUP. Carston, R. 2002. Thoughts and Utterances: The Pragmatics of Explicit Communication. Oxford: Blackwell. Carston, R. 2004. Explicature and semantics. In Semantics: A Reader, S. Davis & B.S. Gillon (eds), 817–845. Oxford: OUP. Clark, H.H. 1977. Bridging. In Thinking: Readings in Cognitive Science, P. Johnson-Laird & P. Wason, 411–420. Cambridge: CUP. Cornish, F. 1999. Anaphora, Discourse, and Understanding: Evidence from English and French. Oxford: Clarendon Press. Dahl, Ö. 2004. Definite articles in Scandinavian: Competing grammaticalization processes in standard and non-standard varieties. In Dialectology meets Typology: Dialect Grammar from a Cross-Linguistic Perspective, B. Kortmann, 147–180. Berlin: de Gruyter. Diessel, H. 1999. Demonstratives. Form, Function and Grammaticalization. Amsterdam: John Benjamins. Donnellan, K. 1966. Reference and definite descriptions. Philosophical Review LXXV: 281–304. Grice, P. 1989. Studies in the Way of Words. Cambridge MA: Harvard University Press. Gundel, J.K. 1996. Relevance theory meets the givenness hierarchy. An account of inferrables. In Reference and Referent Accessibility, T. Fretheim & J.K. Gundel, 141–154. Amsterdam: John Benjamins. Gundel, J.K., Hedberg, N. & Zacharski, R. 1993. Cognitive status and the form of referring expressions in discourse. Language 69(2): 274–307. Gundel, J.K., Hedberg, N. & Zacharski, R. 2001. Definite descriptions and cognitive status in English: Why accommodation is unnecessary. English Language and Linguistics 5(2): 273–295. Gundel, J.K. & Mulkern, A.E. 1998. Quantity implicatures in reference understanding. Pragmatics & Cognition 6(1/2): 21–45. Harris, M. 1980. The marking of definiteness in Romance. In Historical Morphology, J. Fisiak (ed.), 141–156. The Hague: Mouton. Heim, I. 1982. The Semantics of Definite and Indefinite Noun Phrases. PhDdissertation, MIT. King, J.C. 2001. Complex Demonstratives. Cambridge MA: The MIT Press. Lundquist, L. 2005. Contrastive text linguistics: A study of lexical anaphora in Danish and French. Plenary lecture at the 1st Scandinavian PhD conference in linguistics and philology. Bergen. Lyons, C. 1999. Definiteness. Cambridge: CUP. Mithun, M. 1987. The grammatical nature and discourse power of demonstratives. Berkeley Linguistic Society 13: 184–194. Prince, E.F. 1981. Toward a taxonomy of given-new information. In Radical Pragmatics, P. Cole (ed.), 223–256. New York NY: Academic Press. Wilson, D. & Sperber, D. 1993. Relevance and linguistic form. Lingua 90: 1–25.

Index A A priori  309, 312, 319, 322, 326–328 Abstract  2–3, 17–21, 46–47, 50–51, 53–54, 57, 59–60, 76, 113–114, 116–119, 186–187, 208, 293–294, 304, 306, 341 Accent unit  353 Accessibility  257–258, 261 Accusative  4, 214–215, 217, 219–220, 224–225, 228, 243–249, 257, 269, 271, 273, 278, 284, 359 Action verb  176–177, 185 Activity verb  184 Agent  85, 125, 289, 298 Agentive  184, 330 Agreement  103, 116–119, 203–206, 214, 218, 222, 225, 228–230, 240, 268–270, 273, 282, 292, 295–297, 302, 304–306 Akan  337, 339, 345–346, 360–363 Alienation  313–317, 328 All-quantifier  181, 186 Alternation  31, 33, 166, 170, 173, 175, 177, 179, 220 A-movement  265–267, 271, 283 Anaphor  204–206, 217, 220, 274–275, 309–311, 313, 316–317, 319–321, 329, 331, 333, 350, 358, 363, Anaphoric  143–144, 150, 153–154, 156, 217–218, 229, 236, 310, 312–315, 324, 329, 360 Animate  52, 54, 57–59, 87, 184, 311–313, 315–317, 319, 321, 323–324, 328, 331, 360–361 Antecedent  29, 53, 204, 260, 310, 313, 323, 326, 329–334, 341, 343–344, 346–349, 359 Apposition  104, 118–119, 129, 224, 316, 359

Argument  45–46, 48–49, 83, 88, 92, 95, 104, 112–113, 122–123, 184–185, 189–195, 197–199, 201–202, 206–208, 226–227, 268–270 Asymmetry  27, 142, 202, 272, 275, 277, 283, 309, 322, 324, 326–329 Autonomous  4, 13, 22, 58–60, 156, 313–314 B Bare noun  35, 46, 48–49, 57, 105, 111, 120, 123, 132, 164, 192–196, 206–207, 360 Bottom-up referential processing  345 C Cardinal  14, 22, 68, 172, 175, 237–240 Case  49, 51, 56, 88, 151–152, 200, 225, 235, 238, 243–248, 252–253, 257, 267–273, 277–278, 281, 284 Categoremic  314–315, 328 Category  8, 12, 14, 17, 19–20, 22–23, 88, 179, 181–185, 189–193, 195–196, 198–201, 208, 214–215, 217, 221, 236–238, 241–243, 248, 261, 265–266, 275, 283–284, 287–292, 303, 306, 324–327, 354–355, 359–360 c-command  272 Classification  45, 49–55, 59–60, 122, 129, 131, 146, 155, 292–293, 325 Classifier  27, 35, 38, 40, 42, 55 Clause  7–9, 109, 114, 118, 139–142, 145–146, 156, 158–159, 254, 259, 265–272, 274–276, 278–280, 283–284, 293, 360–362

Coda  131–132, 134, 139–148, 151–159 Cognitive  45, 124, 256, 288–292, 297, 311–312, 338–339, 341, 346, 348, 355–356, 359–360, 362–363 Collective  35–37, 50–54, 76, 116, 320, 331 Complement  92, 95, 105, 123–124, 126, 241, 266–268, 279, 281–283, 288 Complementary distribution  114, 237, 241, 260–261 Complementizer  215, 235, 287–289, 291–293, 299–303, 305–306 Completeness  283 Complex Np  164, 266, 276, 279, 283, 329 Composition  6–7, 21, 47, 159, 179, 244–245, 250, 254–255, 268, 290–291, 301, 303 Compound  81–82, 89, 102, 105, 108, 119, 163–164, 166–170, 187, 260 Concept  5, 45, 51, 69, 85, 163, 165–166, 168, 171, 173, 177–179, 187, 203, 206, 255, 257, 289, 340, 342 Conceptual  8, 10–11, 13, 170–171, 173, 190, 207–208, 255, 257, 268, 287, 289–293, 338, 344, 346, 355–356, 359 Concrete  52–53, 87, 90, 125, 166, 171, 185, 255, 293–294, 299, 303, 306, 315, 326–328, 330 Condensation  289, 303 Condition  133, 135, 139, 144, 153–156, 158–159, 266, 271, 276, 278, 281, 283, 310–311, 313–315, 322–324, 328 Congruence  315, 318 Constitutive  173

 Essays on Nominal Determination Constraint  131–132, 134–136, 138–139, 141–145, 147, 150, 152, 154–156, 158–159, 195–197, 258, 266, 276, 279, 312–313, 320, 326 Content syntax  5 Contentive category  290–291 Contextual effect  256–258 Coreferential  240, 310, 331, 359 Count  38–39, 47–49, 51, 54–55, 76–77, 85–87, 89–90, 93, 95, 101–104, 113–115, 170, 173, 176, 198–200, 208 Cp  113, 215–216, 266, 268–269, 275–277, 279, 283, 291 Cross-categorial  265 D Danish  5, 14–15, 20, 29, 31–34, 41, 79–84, 87–88, 90, 92–94, 96, 167–168, 233–240, 242, 247, 249–250, 339 Decoding  290 Decompose  127, 155, 214, 243–244 Definiteness  14, 16, 131–136, 138–139, 142, 146–148, 151–156, 158–159, 198–200, 213–215, 217–218, 221, 223, 228–230, 237–239, 254–255, 339, Deictic  10–11, 85, 169–170, 213–218, 220, 224, 229–230, 250–251, 253, 255, 258, 262, 294, 337, 343 Demonstrative  27–31, 33, 213–216, 219, 224, 236, 238, 242–243, 245–248, 251, 255, 259–260, 337–339, 342–344, 346–355, 358–359, 361–363 Dependency  15, 28, 33–34, 196, 207, 299, 316, 319, 322, 324, 326–330 Derivation  115, 123, 154, 190, 193, 199, 213–214, 221, 228–230, 275, 277–279, 303 Descriptive  1–3, 16, 20–22, 58, 60, 181, 186, 291, 296, 337, 339–344, 346, 352, 355–356, 363 Determination  45, 49, 51, 54–55, 59–60, 84–89, 92, 94–97, 163, 165–166, 175–177, 306 Deverbal  165, 185–186

Differential object marking  45, 49, 55–56, 58, 60 Differentiation  1, 3, 6–11, 13–14, 21–22, 357 Discourse  10, 16, 69–70, 88–89, 95–97, 256–258, 315–316, 322, 331–333, 337–339, 343–344, 346–347, 359–361 Discursive relation  309 Distal demonstrative  243, 337, 339, 343, 348–352, 354, 357–359, 361–363 Domain  118–120, 194, 200, 204, 208–209, 233–234, 248–249, 251, 253, 255, 260, 268–269, 271–272, 275–277 Double-layered  265, 270, 283–284 Dp  108–109, 111–120, 123–124, 126–129, 132–148, 150–159, 201- 202, 204, 218, 221, 227–230, 233, 236–237, 261, 266–270, 288 D-structure  268 Dual category membership  355 E E1  309–313, 315, 319, 322–328, 330 E2  309–313, 315, 322, 324–326, 328, 330 Ecm  272–273 Empty category  193, 195–196, 288 Endocentric language  79, 82–83, 88, 90, 93, 95 English  6, 17, 20, 42, 80–82, 103, 111, 114–116, 120, 132–138, 142–146, 156–159, 194–198, 200–204, 206–207, 213–222, 224–231, 233–240, 249–251, 253, 255, 338–340, 343–350 Event  15–16, 96, 112, 121, 141–142, 144–146, 156, 158–159, 183, 185–186, 314–315, 326 Existential  46, 105, 131–159, 163, 166, 192, 196, 213, 215, 219, 228, 230, 254, 257, 297 Exocentric language  79–80, 82–83, 87–89, 95 Extension  67–70, 73, 75, 182, 194 Extraction  108, 266, 276–284

Extrametrical  353–355 F Familiarity  69–71, 168, 258, 261, 338, 340, 342, 346, 360 Filler  3, 5, 13–15, 22–23 Focus  132, 134, 143, 145–146, 148, 150–159, 171, 338–339, 347–348, 355–356, 361–363 Functional  1–15, 17–22, 113, 115–116, 169, 198–199, 208, 233–236, 266, 268–270, 290–292, 305–306, 330, 333–334 Function-based structure  4, 20–21 G Gender  49–54, 60, 103, 116–121, 125, 128, 200–202, 220–222, 230, 235, 237–238, 243–248, 252–254 Generic  10, 81, 85–86, 89, 91–94, 124–125, 163, 165–166, 168, 179–183, 185–187, 196, 319, 332–333, 357 German  20, 28, 41, 51, 102, 105, 116, 119, 123, 198, 200, 205, 225, 229, 233–240, 242, 244, 248, 339, 347–350, 363 Germanic  87–88, 95, 116, 195–198, 204, 207, 233–236, 238, 240, 242–243, 245, 247–249, 252–254, 257–258, 260 Givenness Hierarchy  337–339, 341, 343, 346, 356, 362–363 Grammatical  83, 115–117, 134–136, 170, 202–203, 217, 219–224, 227–228, 237–238, 240–241, 290–292, 302, 359–361 Grammaticalization  13, 54, 60, 205–208, 237–238, 335, 364 Grounding  1–3, 5, 8–11, 13, 15–18, 21–22 H Hearer  84–85, 216, 218, 250, 252, 256–261, 288, 290, 346–347, 351, 354–356, 360–361, 363 Heterogeneous  1, 38, 52–53, 163, 234



Index  Hierarchy  31, 85, 87–88, 303, 323, 326, 329–330, 337–339, 341, 343, 346, 356, 362–363 Homogeneous  35, 38–39, 163, 170, 217 Human  105, 116–117, 122–123, 125, 128, 190, 248–249, 323–324, 328, 330, 356–357, 359 hw- root determiner  233, 248 Hyperonym  82, 88, 90–91, 93, 182–183 Hyponym  81, 87–88, 90, 93–94, 182–183 Hypostatization  289 I Icelandic  195, 205, 235, 238–240 Identifiable  31, 59, 79, 83–85, 88, 95, 97, 133, 257–258, 338, 340–343, 346–347, 351, 356, 360–362 Identification  10–11, 29, 59, 65, 75, 111, 184, 215, 220, 263, 286, 319, 332, 337, 342 Inalienable  251, 315 Inanimate  53–54, 58, 87, 259, 297, 302, 311–312, 323, 328, 358, 360–361 Inclusion  330 Incorporation  89–92, 94–95, 97, 133, 153–154, 164, 169, 175–177, 184–185 Indefinite article  19–20, 27, 34–35, 38, 40–42, 47–48, 54, 59, 66, 87–88, 114, 216, 236–240 Indefiniteness  14, 159, 169, 215, 237, 239, 360 Individual  51, 54, 58–59, 68–71, 109–110, 121–122, 164–166, 170–173, 176–179, 182–185, 189–193, 199, 202–206, 227, 313–315, 319, 332–334, 344–345 Inergative  184 Inference  16, 22, 256, 312, 340–343, 349 Infinitival  108, 271 Infinitive  84, 229, 287–289, 291–299, 302, 305–306 Information  13, 15–16, 79–80, 115, 118–119, 131–132, 139, 152–156, 159, 230, 288–290, 293, 297–299, 302–305, 340–347, 359–361, 363

Instruction  10, 16, 22, 59, 85, 133–134, 251, 253, 257–258, 288, 292–294, 360 Intensional  35, 57, 67, 75, 85–86, 89, 113, 125, 182 Intention  190, 256, 268, 313, 315–316, 340, 343–345 Interface  127, 190, 199, 265, 284 Internal structure  2, 163–164, 170, 175, 195, 223, 265–266, 268, 270, 273–274, 284, 292–294, 304–305 Interrogative  15, 235–236, 248–249, 255, 259, 269 Intrinsic  94–95, 198, 200, 202–203, 313, 325–327, 329–330 Isomorphic  1, 5, 214, 219–220 K Kind  112, 124–128, 137, 156, 183–186, 190–194, 196–198, 206, Kinship  255, 320, 331 L Layer  2, 9, 12, 18–19, 111, 113, 265–266, 270, 283–284, 288–290, 293, 304 Left branch  266, 276, 278, 283 Lexeme  171, 175–177, 182–183, 235–238, 243, 291, 303, 355, 358, 361, 363 Lexicon  83, 88–89, 97, 170, 208–209, 290, 363 Lf  18, 115, 123, 196–197, 226, 268 Locative  134–144, 146, 153, 156–157, 184, 215, 321, 329–330, 359 M Mapping  2, 19, 112, 190, 192, 196–197, 207–209, 303, 305 Mapping parameter  207–208 Mapping theory 190, 192, 196 Mass  38, 46–51, 54, 57–60, 76–77, 96, 103–104, 113–117, 165–166, 170–179, 186–187, 239 Matrix subject position  266–267, 271, 283, 300 Meaning construction  290 Meronym  321, 329–330 Minimalism  208, 266 Modality  293, 305 Mood  83, 88, 95, 269, 294–299, 304

Morphology  49, 53–54, 60, 67, 88, 108, 115, 152, 200, 205, 323, 364 Movement  127–128, 191–193, 196–197, 201, 203, 226, 234, 266–284 N N1  164–173, 176, 179, 185–187, 310–311, 320–326, 329–330, 333 N2  164–179, 184–187, 310–311, 322–326, 329–330, 333 N2 modifier  168 Nominal denotation  190 Nominal expression  9–13, 15, 18, 20–22, 133, 193, 227, 269, 284, 293, 327 Nominal level  3, 9, 166, 184–187 Nominalisation  83, 108, 121, 176, 184–185, 288–289, 292–306 Nominalization, see nominalisation Nominative  88, 151, 215, 224–225, 235, 243–249, 257, 269, 359 Non-argument  192, 198–199, 201–202, 226 Non-human  116, 248, 330, 351, 356–357 Non-referential  133, 150, 164–165, 216, 230 Non-specific  35, 46–47, 54–58, 85–87, 159, 165–168, 173, 176–179, 216, 230 Norwegian  102, 104, 116, 121, 195, 226, 250, 339, 345–363 Np-preposing  266, 273–276, 283 N-raising  191, 196, 198 N-to-D  21, 191–192, 197, 203, 207 Number  48–54, 60, 70, 83, 88, 95, 107, 112, 200–208, 218–225, 229–230, 243–244, 247, 253, 350–351 Numeral  14, 34–35, 41–42, 47, 54, 66–68, 89, 115, 118, 239 O Object  37, 46–49, 55–60, 66, 73, 85, 89–92, 104, 151, 183–184, 193, 259, 267–273, 278, 283, 289, 295–298, 351, 356 Ontological status  326 Oslo Multilingual Corpus (Omc)  347

 Essays on Nominal Determination Ostension  191, 250–257, 260 Ostensive act  251, 256–258 P Paradigm  11–14, 22, 42, 202, 224, 243–249, 252–254, 281, 295–297, 302–306, 325–326, 332 Parallelism  150, 266–270, 274 Parameter  19, 111, 135–136, 194, 197, 204–209, 254, 298, 302–306 Parametric minimalism  20 Partial autonomy  3–4 Partitive  35, 39, 47–49, 54–59, 66, 70–77, 87, 108, 151–152, 171 Part-whole  312, 321, 327, 332 Passive  147–149, 266–272, 276, 283 Perception  76, 166, 222, 317 Person  29–30, 83, 95, 200–214, 217–232, 248, 254–255, 258, 325, 350, 357, 360 Person head  201–203 Person marker  214, 217–220, 223, 229, 254 Personal pronoun  11, 29, 53, 58, 138, 200, 205–206, 215, 221–226, 248, 261, 323–324, 338, 348, 355–359 Phrasal  164, 169, 241, 305 Phrase structure  2, 28, 197, 236, 268, 291 Plural  14, 35–39, 48–55, 69–73, 85, 114–117, 128, 187, 198, 203, 218–229, 244, 247, 258, 325, 346, 357 Possessed  49, 309 Possession  14, 310, 314, Possessive  14, 169, 184, 204, 215, 219, 228–229, 310–334 Possessive adjective  310–311, 314–318, 322, 332 Possessive anaphor  311, 316 Possessive determiner  310–311, 319–331 Possessive pronoun  169, 215, 219 Possessor  118, 123, 229, 309 Pragmatics  10, 310 Predicate type  183–184 Prefix  360–361

Presupposition  120, 123–124, 133–134, 256–257, 297, 300–301, 305, 344 Probe  269–272, 277–279, 283 Procedural meaning  344 Procedural semantics  255, 261, 361 Processing  7, 13, 289–292, 341–344, 346–347, 350 Projection  2, 108, 111–116, 120, 125, 201, 208, 230, 234, 241–242, 266–272, 282, 284 Pronominal  29, 34, 40, 53–54, 203–204, 214, 218–221, 225–226, 230, 324–325, 332–339, 347, 357–358 Proper name  11, 20–22, 58, 69, 116, 124–128, 134, 143, 190–207, 226–227, 355 Properties  4–10, 17–22, 47, 55–56, 110–114, 124–126, 133–138, 156, 183–185, 190–209, 266–269, 273, 280, 292–294, 303–305, 314–315, 325–328, 343 Proposition  18, 84, 113, 139, 255, 284, 289, 360 Prototype  8, 81, 90, 165–166, 170–187 Prototypicality  179, 186 Proximal demonstrative  243, 338, 345–349, 358–361 Q Quantification  35, 74, 110, 121, 144, 175, 190 Quantifier  34–35, 40–41, 68–74, 181, 191, 198, 206, 216, 238–240 R Recategorisation  176, 289, 292, 301 Reference  1, 2, 10–13, 20–22, 51, 110–114, 154–156, 179, 190–197, 202, 234, 255–256, 310, 327, 339–344, 349, 356–363 Relative pronoun  215, 236 Relevance Theory  256, 341 Representational semantics  288 Rigid designator  11 Romance  29–31, 35, 46–55, 59–60, 80–83, 92–96, 111, 147–152, 191–207, 228, 276, 280–283, 350

Root  50, 54, 81, 116, 234–235, 244–249, 251–255, 257, 303–304 S Salience  124, 332 Scope  5, 17–18, 24, 67, 87, 137–138, 181, 276, 288–291, 305, 318, 331, 360–361 Second person  30, 200, 203, 214, 218, 221–231, 248, 254, 258, 325 Sentence level  89, 166–169, 173, 180, 184–186 Single-layered  266, 283–284 Singular  14, 28, 37–38, 46–54, 69–74, 85, 91–92, 102–104, 111–115, 120, 165–166, 171–179, 198–203, 219–226, 239–240, 351, 360 Singular individual  165–166, 171–179 Site  324 Situation  3–4, 15, 88, 96, 176, 185–186, 288–289, 292–306, 323, 344, 355 Slot  1, 11–16, 19, 22–23, 31–32, 41, 143, 253 Small clause  109, 114, 118, 139, 141–142, 145–146, 156–159, 272 Spanish  29–36, 46, 49, 52–60, 102–104, 132–138, 144–159, 165–168, 175, 180–184, 204, 229, 288, 303 Spec  109, 124–128, 267–284 Specific  1, 10, 16–20, 46, 56–59, 85, 121, 159, 164–168, 173–179, 185, 216, 251, 256–260, 323–326, 332–333 Specificity  56–58, 80–91, 94–95, 132, 159, 168, 257, 266, 276–278 Specifier  2, 266–267, 319 S-structure  18, 268 Stage level predicate  145, 183–186 State verb  184–185 Stereotype  126, 168, 176, 183, 329 Subject raising  266–267, 270–273, 279, 283 Subordinate  31, 302, 328, 345 Suffix  5, 14, 52, 116–118, 120, 128, 198, 204, 235, 240, 243–249, 253, 303, 339, 351–353 Superordinate  87, 296, 299–303



Index  Symbolic-conceptual meaning  8, 11, 15, 17, 21 Syncategoremic  314–315, 328 T Target  193, 268, 275, 324, 353 Textual  34, 87–88, 310 Th-, 214–231, 245, 254–259 þ- and hw- morphemes  235, 255, 259, 261 þ- root determiner  242 Third person  29, 200–204, 214–231, 254–255, 325, 350, 360

Top-down processing  345 Topic  8, 83–85, 136, 140, 145–146, 153–159, 269, 274–275, 332 Topical  58–59, 331–332 Truth value  19, 35, 56–58, 72, 106, 138–139, 181–184, 196–198, 218, 229, 234, 328, 344, 361 Type identifiable  338, 360–362 U Unaccusative predicate  185 Underlying  2, 5, 15–20, 29, 116, 135, 155, 210, 243

Unique reference  174, 177 Uniqueness  10, 69, 120, 123–124, 133–134, 156, 168, 344 V Vernacular  214, 224 Vocatives  199, 226–227, 230 vP  268–269, 275–277, 283–284 Vp-preposing  266, 273–275, 283 Z Zero determiner  93, 110, 165, 168, 171

Studies in Language Companion Series A complete list of titles in this series can be found on the publishers’ website, www.benjamins.com 103 Josephson, Folke and Ingmar Söhrman (eds.): Interdependence of Diachronic and Synchronic Analyses. viii, 344 pp. + index. Expected August 2008 102 Goddard, Cliff (ed.): Cross-Linguistic Semantics. 2008. xvi, 356 pp. 101 Stolz, Thomas, Sonja Kettler, Cornelia Stroh and Aina Urdze: Split Possession. An areallinguistic study of the alienability correlation and related phenomena in the languages of Europe. 2008. x, 546 pp. 100 Ameka, Felix K. and M.E. Kropp Dakubu (eds.): Aspect and Modality in Kwa Languages. 2008. ix, 335 pp. 99 Høeg Müller, Henrik and Alex Klinge (eds.): Essays on Nominal Determination. From morphology to discourse management. 2008. xviii, 369 pp. 98 Fabricius-Hansen, Cathrine and Wiebke Ramm (eds.): 'Subordination' versus 'Coordination' in Sentence and Text. A cross-linguistic perspective. 2008. vi, 359 pp. 97 Dollinger, Stefan: New-Dialect Formation in Canada. Evidence from the English modal auxiliaries. 2008. xxii, 355 pp. 96 Romeo, Nicoletta: Aspect in Burmese. Meaning and function. 2008. xv, 289 pp. 95 O’Connor, Loretta: Motion, Transfer and Transformation. The grammar of change in Lowland Chontal. 2007. xiv, 251 pp. 94 Miestamo, Matti, Kaius Sinnemäki and Fred Karlsson (eds.): Language Complexity. Typology, contact, change. 2008. xiv, 356 pp. 93 Schalley, Andrea C. and Drew Khlentzos (eds.): Mental States. Volume 2: Language and cognitive structure. 2007. x, 362 pp. 92 Schalley, Andrea C. and Drew Khlentzos (eds.): Mental States. Volume 1: Evolution, function, nature. 2007. xii, 304 pp. 91 Filipović, Luna: Talking about Motion. A crosslinguistic investigation of lexicalization patterns. 2007. x, 182 pp. 90 Muysken, Pieter (ed.): From Linguistic Areas to Areal Linguistics. 2008. vii, 293 pp. 89 Stark, Elisabeth, Elisabeth Leiss and Werner Abraham (eds.): Nominal Determination. Typology, context constraints, and historical emergence. 2007. viii, 370 pp. 88 Ramat, Paolo and Elisa Roma (eds.): Europe and the Mediterranean as Linguistic Areas. Convergencies from a historical and typological perspective. 2007. xxvi, 364 pp. 87 Verhoeven, Elisabeth: Experiential Constructions in Yucatec Maya. A typologically based analysis of a functional domain in a Mayan language. 2007. xiv, 380 pp. 86 Schwarz-Friesel, Monika, Manfred Consten and Mareile Knees (eds.): Anaphors in Text. Cognitive, formal and applied approaches to anaphoric reference. 2007. xvi, 282 pp. 85 Butler, Christopher S., Raquel Hidalgo Downing and Julia Lavid (eds.): Functional Perspectives on Grammar and Discourse. In honour of Angela Downing. 2007. xxx, 481 pp. 84 Wanner, Leo (ed.): Selected Lexical and Grammatical Issues in the Meaning–Text Theory. In honour of Igor Mel'čuk. 2007. xviii, 380 pp. 83 Hannay, Mike and Gerard J. Steen (eds.): Structural-Functional Studies in English Grammar. In honour of Lachlan Mackenzie. 2007. vi, 393 pp. 82 Ziegeler, Debra: Interfaces with English Aspect. Diachronic and empirical studies. 2006. xvi, 325 pp. 81 Peeters, Bert (ed.): Semantic Primes and Universal Grammar. Empirical evidence from the Romance languages. 2006. xvi, 374 pp. 80 Birner, Betty J. and Gregory Ward (eds.): Drawing the Boundaries of Meaning. Neo-Gricean studies in pragmatics and semantics in honor of Laurence R. Horn. 2006. xii, 350 pp. 79 Laffut, An: Three-Participant Constructions in English. A functional-cognitive approach to caused relations. 2006. ix, 268 pp. 78 Yamamoto, Mutsumi: Agency and Impersonality. Their Linguistic and Cultural Manifestations. 2006. x, 152 pp. 77 Kulikov, Leonid, Andrej Malchukov and Peter de Swart (eds.): Case, Valency and Transitivity. 2006. xx, 503 pp.

76 Nevalainen, Terttu, Juhani Klemola and Mikko Laitinen (eds.): Types of Variation. Diachronic, dialectal and typological interfaces. 2006. viii, 378 pp. 75 Hole, Daniel, André Meinunger and Werner Abraham (eds.): Datives and Other Cases. Between argument structure and event structure. 2006. viii, 385 pp. 74 Pietrandrea, Paola: Epistemic Modality. Functional properties and the Italian system. 2005. xii, 232 pp. 73 Xiao, Richard and Tony McEnery: Aspect in Mandarin Chinese. A corpus-based study. 2004. x, 305 pp. 72 Frajzyngier, Zygmunt, Adam Hodges and David S. Rood (eds.): Linguistic Diversity and Language Theories. 2005. xii, 432 pp. 71 Dahl, Östen: The Growth and Maintenance of Linguistic Complexity. 2004. x, 336 pp. 70 Lefebvre, Claire: Issues in the Study of Pidgin and Creole Languages. 2004. xvi, 358 pp. 69 Tanaka, Lidia: Gender, Language and Culture. A study of Japanese television interview discourse. 2004. xvii, 233 pp. 68 Moder, Carol Lynn and Aida Martinovic-Zic (eds.): Discourse Across Languages and Cultures. 2004. vi, 366 pp. 67 Luraghi, Silvia: On the Meaning of Prepositions and Cases. The expression of semantic roles in Ancient Greek. 2003. xii, 366 pp. 66 Nariyama, Shigeko: Ellipsis and Reference Tracking in Japanese. 2003. xvi, 400 pp. 65 Matsumoto, Kazuko: Intonation Units in Japanese Conversation. Syntactic, informational and functional structures. 2003. xviii, 215 pp. 64 Butler, Christopher S.: Structure and Function – A Guide to Three Major Structural-Functional Theories. Part 2: From clause to discourse and beyond. 2003. xiv, 579 pp. 63 Butler, Christopher S.: Structure and Function – A Guide to Three Major Structural-Functional Theories. Part 1: Approaches to the simplex clause. 2003. xx, 573 pp. 62 Field, Fredric: Linguistic Borrowing in Bilingual Contexts. With a foreword by Bernard Comrie. 2002. xviii, 255 pp. 61 Goddard, Cliff and Anna Wierzbicka (eds.): Meaning and Universal Grammar. Theory and empirical findings. Volume 2. 2002. xvi, 337 pp. 60 Goddard, Cliff and Anna Wierzbicka (eds.): Meaning and Universal Grammar. Theory and empirical findings. Volume 1. 2002. xvi, 337 pp. 59 Shi, Yuzhi: The Establishment of Modern Chinese Grammar. The formation of the resultative construction and its effects. 2002. xiv, 262 pp. 58 Maylor, B. Roger: Lexical Template Morphology. Change of state and the verbal prefixes in German. 2002. x, 273 pp. 57 Mel’čuk, Igor A.: Communicative Organization in Natural Language. The semantic-communicative structure of sentences. 2001. xii, 393 pp. 56 Faarlund, Jan Terje (ed.): Grammatical Relations in Change. 2001. viii, 326 pp. 55 Dahl, Östen and Maria Koptjevskaja-Tamm (eds.): Circum-Baltic Languages. Volume 2: Grammar and Typology. 2001. xx, 423 pp. 54 Dahl, Östen and Maria Koptjevskaja-Tamm (eds.): Circum-Baltic Languages. Volume 1: Past and Present. 2001. xx, 382 pp. 53 Fischer, Olga, Anette Rosenbach and Dieter Stein (eds.): Pathways of Change. Grammaticalization in English. 2000. x, 391 pp. 52 Torres Cacoullos, Rena: Grammaticization, Synchronic Variation, and Language Contact. A study of Spanish progressive -ndo constructions. 2000. xvi, 255 pp. 51 Ziegeler, Debra: Hypothetical Modality. Grammaticalisation in an L2 dialect. 2000. xx, 290 pp. 50 Abraham, Werner and Leonid Kulikov (eds.): Tense-Aspect, Transitivity and Causativity. Essays in honour of Vladimir Nedjalkov. 1999. xxxiv, 359 pp. 49 Bhat, D.N.S.: The Prominence of Tense, Aspect and Mood. 1999. xii, 198 pp. 48 Manney, Linda Joyce: Middle Voice in Modern Greek. Meaning and function of an inflectional category. 2000. xiii, 262 pp. 47 Brinton, Laurel J. and Minoji Akimoto (eds.): Collocational and Idiomatic Aspects of Composite Predicates in the History of English. 1999. xiv, 283 pp. 46 Yamamoto, Mutsumi: Animacy and Reference. A cognitive approach to corpus linguistics. 1999. xviii, 278 pp.

45 Collins, Peter C. and David Lee (eds.): The Clause in English. In honour of Rodney Huddleston. 1999. xv, 342 pp. 44 Hannay, Mike and A. Machtelt Bolkestein (eds.): Functional Grammar and Verbal Interaction. 1998. xii, 304 pp. 43 Olbertz, Hella, Kees Hengeveld and Jesús Sánchez García (eds.): The Structure of the Lexicon in Functional Grammar. 1998. xii, 312 pp. 42 Darnell, Michael, Edith A. Moravcsik, Michael Noonan, Frederick J. Newmeyer and Kathleen M. Wheatley (eds.): Functionalism and Formalism in Linguistics. Volume II: Case studies. 1999. vi, 407 pp. 41 Darnell, Michael, Edith A. Moravcsik, Michael Noonan, Frederick J. Newmeyer and Kathleen M. Wheatley (eds.): Functionalism and Formalism in Linguistics. Volume I: General papers. 1999. vi, 486 pp. 40 Birner, Betty J. and Gregory Ward: Information Status and Noncanonical Word Order in English. 1998. xiv, 314 pp. 39 Wanner, Leo (ed.): Recent Trends in Meaning–Text Theory. 1997. xx, 202 pp. 38 Hacking, Jane F.: Coding the Hypothetical. A comparative typology of Russian and Macedonian conditionals. 1998. vi, 156 pp. 37 Harvey, Mark and Nicholas Reid (eds.): Nominal Classification in Aboriginal Australia. 1997. x, 296 pp. 36 Kamio, Akio (ed.): Directions in Functional Linguistics. 1997. xiii, 259 pp. 35 Matsumoto, Yoshiko: Noun-Modifying Constructions in Japanese. A frame semantic approach. 1997. viii, 204 pp. 34 Hatav, Galia: The Semantics of Aspect and Modality. Evidence from English and Biblical Hebrew. 1997. x, 224 pp. 33 Velázquez-Castillo, Maura: The Grammar of Possession. Inalienability, incorporation and possessor ascension in Guaraní. 1996. xvi, 274 pp. 32 Frajzyngier, Zygmunt: Grammaticalization of the Complex Sentence. A case study in Chadic. 1996. xviii, 501 pp. 31 Wanner, Leo (ed.): Lexical Functions in Lexicography and Natural Language Processing. 1996. xx, 355 pp. 30 Huffman, Alan: The Categories of Grammar. French lui and le. 1997. xiv, 379 pp. 29 Engberg-Pedersen, Elisabeth, Michael Fortescue, Peter Harder, Lars Heltoft and Lisbeth Falster Jakobsen (eds.): Content, Expression and Structure. Studies in Danish functional grammar. 1996. xvi, 510 pp. 28 Herman, József (ed.): Linguistic Studies on Latin. Selected papers from the 6th International Colloquium on Latin Linguistics (Budapest, 23–27 March 1991). 1994. ix, 421 pp. 27 Abraham, Werner, T. Givón and Sandra A. Thompson (eds.): Discourse, Grammar and Typology. Papers in honor of John W.M. Verhaar. 1995. xx, 352 pp. 26 Lima, Susan D., Roberta L. Corrigan and Gregory K. Iverson: The Reality of Linguistic Rules. 1994. xxiii, 480 pp. 25 Goddard, Cliff and Anna Wierzbicka (eds.): Semantic and Lexical Universals. Theory and empirical findings. 1994. viii, 510 pp. 24 Bhat, D.N.S.: The Adjectival Category. Criteria for differentiation and identification. 1994. xii, 295 pp. 23 Comrie, Bernard and Maria Polinsky (eds.): Causatives and Transitivity. 1993. x, 399 pp. 22 McGregor, William B.: A Functional Grammar of Gooniyandi. 1990. xx, 618 pp. 21 Coleman, Robert (ed.): New Studies in Latin Linguistics. Proceedings of the 4th International Colloquium on Latin Linguistics, Cambridge, April 1987. 1990. x, 480 pp. 20 Verhaar, John W.M. S.J. (ed.): Melanesian Pidgin and Tok Pisin. Proceedings of the First International Conference on Pidgins and Creoles in Melanesia. 1990. xiv, 409 pp. 19 Blust, Robert A.: Austronesian Root Theory. An essay on the limits of morphology. 1988. xi, 190 pp. 18 Wierzbicka, Anna: The Semantics of Grammar. 1988. vii, 581 pp. 17 Calboli, Gualtiero (ed.): Subordination and Other Topics in Latin. Proceedings of the Third Colloquium on Latin Linguistics, Bologna, 1–5 April 1985. 1989. xxix, 691 pp. 16 Conte, Maria-Elisabeth, János Sánder Petöfi and Emel Sözer (eds.): Text and Discourse Connectedness. Proceedings of the Conference on Connexity and Coherence, Urbino, July 16–21, 1984. 1989. xxiv, 584 pp.

15 Justice, David: The Semantics of Form in Arabic. In the mirror of European languages. 1987. iv, 417 pp. 14 Benson, Morton, Evelyn Benson and Robert F. Ilson: Lexicographic Description of English. 1986. xiii, 275 pp. 13 Reesink, Ger P.: Structures and their Functions in Usan. 1987. xviii, 369 pp. 12 Pinkster, Harm (ed.): Latin Linguistics and Linguistic Theory. Proceedings of the 1st International Colloquium on Latin Linguistics, Amsterdam, April 1981. 1983. xviii, 307 pp. 11 Panhuis, Dirk G.J.: The Communicative Perspective in the Sentence. A study of Latin word order. 1982. viii, 172 pp. 10 Dressler, Wolfgang U., Willi Mayerthaler, Oswald Panagl and Wolfgang Ullrich Wurzel: Leitmotifs in Natural Morphology. 1988. ix, 168 pp. 9 Lang, Ewald and John Pheby: The Semantics of Coordination. (English transl. by John Pheby from the German orig. ed. 'Semantik der koordinativen Verknüpfung', Berlin, 1977). 1984. 300 pp. 8 Barth, E.M. and J.L. Martens (eds.): Argumentation: Approaches to Theory Formation. Containing the Contributions to the Groningen Conference on the Theory of Argumentation, October 1978. 1982. xviii, 333 pp. 7 Parret, Herman, Marina Sbisà and Jef Verschueren (eds.): Possibilities and Limitations of Pragmatics. Proceedings of the Conference on Pragmatics, Urbino, July 8–14, 1979. 1981. x, 854 pp. 6 Vago, Robert M. (ed.): Issues in Vowel Harmony. Proceedings of the CUNY Linguistics Conference on Vowel Harmony, May 14, 1977. 1980. xx, 340 pp. 5 Haiman, John: Hua: A Papuan Language of the Eastern Highlands of New Guinea. 1980. iv, 550 pp. 4 Lloyd, Albert L.: Anatomy of the Verb. The Gothic Verb as a Model for a Unified Theory of Aspect, Actional Types, and Verbal Velocity. (Part I: Theory; Part II: Application). 1979. x, 351 pp. 3 Malkiel, Yakov: From Particular to General Linguistics. Selected Essays 1965–1978. With an introduction by the author, an index rerum and an index nominum. 1983. xxii, 659 pp. 2 Anwar, Mohamed Sami: BE and Equational Sentences in Egyptian Colloquial Arabic. 1979. vi, 128 pp. 1 Abraham, Werner (ed.): Valence, Semantic Case, and Grammatical Relations. Workshop studies prepared for the 12th International Congress of Linguists, Vienna, August 29th to September 3rd, 1977. 1978. xiv, 729 pp.