No title

JOURNAL OF SEMANTICS

AN INTERNATIONAL jOURNAL FOR THE INTERDISCIPLINARY STUDY OF THE SEMANTICS OF NATURAL LANGUAGE

MANAGING EDITOR: PETER BoscH {IBM Scientific Centre, Heidelberg and University of Osnabriick) ASSOCIATE EDITORS: NICHOLAS AsHER {University of Texas, Austin) R oB VAN DER SANDT {University of Nijmegen) LiiDELING (University of Tiibingen) R EVIEW EDITOR: LiiDELING {University of Tiibingen) ASSISTANT EDITOR:

ANKE ANKE

EDITOR IA L BOARD: HANs KAMP {University of Stuttgart)

MANFRED BIERWISCH {MPG and Humboldt

SEBASTIAN LiiBNER {University of Diisseldorfj

University Berlin)

SIR JoHN LYONS {University of Cambridge) JAMES D. McCAWLEY {University of Chicago)

BRANIMIR BocuRAEv {Apple Computer Inc) MARro BoruLLO {University of Toulouse) KEITH BROWN {University of Essex)

GENNARO CHIERCHIA {University of Milan)

A.N N

CoPESTAKE {Stanford University)

OsTEN DAHL

{University of Stockholm)

PAUL DE

KKI!R {University of Amsterdam)

J

J

CurRE GARDENT {University of Saarbriicken) BART GEURTS {University of Osnabriick) Mr c HAEL HERWEG {IBM Scientific Centre, Heidelberg) LAURENCE R Ho (Yale University) oACHIM ACOBS {University of Wuppertal) PHILIP N. oHNSON-LAIRD ( Princeton University) MEcuMr KAMEYAMA (SRI International Stanford)

RN

J

MARc MoENS (University of Edinburgh) fRANCIS J. PELLETIER {University of Alberta) MANFRED PrNKAL {University of Saarbriicken) ANToNY SANFORD (University of Glasgow) AruNIM VON STECHOW {University ofTiibingen) MARK STEEDMAN {University of Pennsylvania) ANATOLI STRIGIN {Max Planck Gesellschaft, Berlin) HENRIETTE DE SwART {University of Utrecht) BoNNIE WEBBER {University of Pennsylvania) HENK ZEEVAT {University of Amsterdam) THOMAS E. ZIMMERMANN {University of Stuttgart)

EDITORIAL ADDRESS: Journal of Semantics, c/o Dr P. Bosch, IBM Germany Scientific Centre, Vangerowstr. 18, D-69115 Heidelberg, Germany. Phone: {49-6221-) 59-4251/4437. Telefax: (49-6221-) 59-3200. E mail: [email protected]

© Oxford University Press All rights reserved; no part of rhis publication may be reproduced, stored in a retrieval system, or transmitted in any form or by any means, electronic, mechanical, photocopying, recording, or otherwise without eirher rhe prior

UK by the Copyright 9HE, or in the USA by the Copyright Clearance

written permission of rhe Publishers, or a licence permitting restricted copying issued in the Licensing Agency Ltd, Center,

222

90 Tottenham Court

Road, London WI P

Rosewood Drive, Danvers, Massachusetts

01923,

USA.

Journal ofSernantia (ISSN 0167 )133) is published quarterly in February, May, August and November by Oxford University Press, Oxford, UK. Annual subscription is US$140 per year. journal ofSernantia is distributed by MAIL America,

2323

Randolph Avenue, Avenel, New Jersey 07001, USA. Periodical postage paid at Rahway, New Jersey,

USA and at additional entry points. US POSTMASfER: send address corrections to Journal Avenel. New Jersey

0700 r,

USA.

For subscription information pkas� su insid� back covN.

of Smumtia, c/o

MAIL America, 2323 Randolph Avenue,

JOURNAL OF SEMANTICS Volume 14 Number 4

CONTENTS BART.GUERTS

Good News about the Description Theory of Names

319

JAY DAVID ATLAS

Negative Adverbials, Prototypical Negation and the De Morgan Taxonomy

349

RANI NELKEN AND NISSIM FRANCEZ

The Analogy between Nominal and Temporal Anaphora Revisited

369

Book Review

417

Please visit the journal's world wide web site at http://www.oup.eo.uk/semant

JOURNAL OF SEMANTICS Volume

14 (1997)

CONTENTS Articles JAY

DAVID ATLAs Negative Adverbials, Prototypical Negation and the De Morgan Taxonomy

349

ANDREA BoNOMI

The Progressive and the Structure of Events

I7

3

KAI VON FINTEL

Bare Plurals, Bare Conditionals, and only BART GUERTS Good News about the Description Theory of Names

I

3I 9

ALFoNs MAEs

Referent Ontology and Centering in Discourse RANI NELKEN AND NISSIM FRANCEZ

207

.

The Analogy between Nominal and Temporal Anaphora Revisited

369

E. OJEDA A SemantiCs for the Counting Numerals of Latin

I 43

ALMERINDO

MARGA REis AND INGER RosENGREN

A Modular Approach to the Grammar of Additive P articles: the Case of German Auch

237

DmTER WUNDERLICH

Argument Extension by Lexical Adjunction

95

JoosT ZwARTS

Vectors as Relative Positions: A Compositional Semantics of Modified PPs

57

LINGUISTICS

an interdisciplinary journal of the language sciences

Editor-in-Chief:

Wolfgang Klein, Max-Planck-Institut, Nijmegen, The Netherlands LINGUISTICS was founded in 1963 as an international review and reconstituted under a new Editorial Board in 1979. The journal provides an international and interdisciplinary forum for research in the language sciences. LINGUISTICS publishes articles and short notices in the traditional disciplines of linguistics: pragmatics, semantics, syntax, morphology, phonology. In addi­ tion, it will accept articles and notices from neighboring disciplines that are engaged in the study of natural language: experimental phonetics, psycho­ linguistics, and neurolinguistics, and first and second language acquisition. LINGUISTICS also publishes occasional special issues and welcomes pro­ posals for these. LINGUISTICS accepts work from other disciplines, insofar as it is deemed to be of interest to linguists and other students of natural language; for example, cer­ tain studies in logic, artificial intelligence, social interaction, physiology, and neurology would meet this requirement. LINGUISTICS is published in one volume of six issues per year ( approx. 1200 pages). As of 1979 the numbering was changed; Volume 35 corresponds to Issues 347.-352. Subscription rates for Volume 35 Institutions /libraries Individuals* Single issue

(1997):

OM 646.00 I aS 4716.00 I sFr. 575.00 (postage extra) OM 324.00 (includes postage) OM 114.00/oS 832.00/sFr. 101.00 (postage extra)

Prices in US dollars for subscriptions in North America only: $ 430.00 (postage extra) Institutions /libraries $ 208.30 (includes postage) Individuals* $ 72.00 Single issue Subscriptions at the individual rate are for personal use only and must be prepaid and ordered directly from the publisher. Prepayment may be made by check or by credit card: MasterCard [Access]. EuroCard. Visa and American Express. Orders placed for institutions will be invoiced at the institutional rate. The individual rate is not available in Germany. Switzerland or Austria. •

Institutional subscriptions and single or back issues can be ordered from your local bookseller or subscription agent, or directly from MOUTON DE GRUYTER (a division of Walter de Gruyter) at the following addresses:

For North America: Walter de Gruyter, Inc. 200 Saw Mill River Road Hawthorne, NY 10532 USA

For all other countries: Walter de Gruyter & Co. Postfach 30 34 21 D-10728 Berlin Germany

New at SAGE in 1997!

LANGUAGE AND LITERATURE

jOURNAL OF THE POETICS AND UNGUISTICS ASSOCIATION

University of Leeds "LANGUAGE AND LI1'£RATliRE is a pointer to the future."

Edited by Katie Wales

-Times Higher Education Supplement

... covers the latest developments in stylistic analysis, the linguistic analysis of literature and related areas. With its uniquely broad coverage, the journal offers readers easy access to all the important new research relevant to stylistics. Topics covered include: literary and non-literary stylistics, the connection between stylistics, critical theory, linguistics and literary criticism, and their applications to teaching. LANGUAGE AND LmRATURE also includes book reviews, examining new publications in the field. Published three times a year• lSSN: 0963-9470

Order Fonn for New Subscribers- Subscribe at the Introductory Rate � SAGE Publications ltd, 6 Bonhill Street, London EC2A 4PU, UK � Call our Subscription Hotline on +44 (0)171330 1266 USA orders to be sent to. PO Box 5096, ThousAnd Ollc.s, CA 91359

Name

________ __ __________ _

Add�------

1

:J I enclose a cheque {made payable to SAGE PubliaUons Ltd) l o r: L -_J Please invoice my credit card

-------

7Jl6

D Mastercard Q Visa

I

Amount._ ___ _.

I I I I I I II I I I I I I I I I

0 Yes! I want to subscribe to Ltmguagt IUid Littrahlrr starting with Volume 6 (1997)

eard No:

0 lntroductory �te for lndlvldwals O.SIUS$40 (Usual Rate £321USS51)

Expuy Date:

0

0 lnstitution.tl Rate £7SIUSS120

0 Please send me a brochure on the journal

Signature:

______

Date:

I I

CAMBRIDGE Journal of Linguistics Editors: IAN ROBERTS, University ofStuttgart ROBERT D. BORSLEY, Univ"sity ofWaks, Bangor Journal of lingubc:ic.s is concerned with all branch� of thcorcticallinguisrcs i , including syncax, morphology, phonology, phonetics, semantics, pragmadcs :and historical, sociological, computational and psychological aspects of language: and

linguistic theory. The journal also provides an cxcc:llem survey of recent publications in the field whh review articles on major works marking imponam theoretical :advances as wdl

as

about twenty rc.:views

and �honer notices in each issue:. The: full list of publications rc:ccivod and the comprehen-�ivc: index published annually make the journal an c:vc:n more: valuable resource for both general and applied linguists. Forthcoming anides indude: The organization and specification of manner

r----------------------------,

l Take a closer look free! -

I

s,,,..

l Please send me a free sample copy of I Journal of Linguiscia

feamre; Crossover: a unified view;

Name�------

Pragmatics and word meaning; A unified

Address

theory ofscope for quantifiers and 'wh'

------

phrases; Determiner phrase in a l:anguage without determiners; Subject clitiC5 and cliric rccyding: locomotive sentences in some Iberian Romance langugQ; Syntactic constraints on intonational phrasing in

English.

Subscriptions Volume 33 in 1997: Marth and September £61 for Institutions: £45 for Individuals: prices indude delivery by air. ISSN 0022· 2267.

Send CIOUpOft ""Jounuh Mm:tt�ns [)q>anmalr. Cambridp: Univmlry t'ras, The F.
: E-cuil: jout� I

L----------------------------�

� CAMBRIDGE �� UNIVERSITY PRESS

Mobilian Jargon

Referring to Space

linguistic and Sociohistorical Aspects of a Native American Pidgin EMANUELJ.DRECHSEL

Studies in Austronesian and Papuan Languages Edited by GUNTER SENFT

Drechsel demonstrates the significance of language contact in America in this grammatical and sociohistorical study of Mobilian Jargon, an American Indian pidgin which probably originated as a lingua franca among the pre­ Columbian mound-building chiefdoms of the Mississippi valley. It was extensively used from 1700 until the mid-twentieth century both among diverse groups of southeastern Native Americans, and in their interactions with non­ Indians.

Oxford Studies in Language Contact

0-19-824033·3, 412 pp, maps, tables, Clarendon Press, February 1997,£55.00

Parts and Wholes in Semantics FRIEDERIKE MOLTMANN Moltmann provides a unified account of a broad range of English and cross linguistic data involving expressions of the notions of "part" and "whole". She presents a new theory of part structures in which the notion of an integrated whole plays a fundamental role, and in which the part structure of an entity may vary across different situations, perspectives, and dimensions.

0-19-509574-X, 272

pp, OUP USA, October

1997,£32.99

Aspect and Predication The Semantics of Argument Structure GILLIAN CATRIONA RAMCHAND This is an important research monograph which addresses the issues at the centre of current theoretical debates. It investigates the relationship between meaning and grammatical form, focusing on the Scottish Gaelic language.

0-19-823651-4, 260 pp., line 1997,£30.00

Press, March

illustrations, Clarendon

This collection of previously unpublished case­ studies of a number of Austronesian and Papuan languages illustrates the means that these languages offer to their speakers for referring to space. Understanding the differences between the various systems of spatial reference requires not just linguistic, but also cultural, historical, and geographical knowledge.

Oxford Studies in Anthropological Unguistics No. 11 0-19-823647-6, 336 pp, 70 bla ck and white Clarendon Press, October 1997, £50.00

text-ligures,

Understanding Cultures Through Their Key Words English, Russian, Polish, German, and Japanese ANNA WIERZBICKA Wierzbicka demonstrates that every language has its "key concepts" (expressed in key words) and that these concepts reflect the core values of the culture in question. Examining empirical evidence from five languages, and using her own "natural semantic metalanguage" to provide an analytical framework, she shows that cultures can be revealingly studied, compared, and explained to outsiders through their key concepts.

Oxford Studies in Anthropological linguistics No. 8

0-19-508835-2, 464 pp, OUP USA, October 1997,£61.99 0-19-508836-0, paperback, £24.00 Available through good bookshops. For more information, contact . Helen Lawrence, Academic Marketing, Oxford University Press, Great Clarendon Street, Oxford OX2 6DP ir +44 (0)1865 556767 or email: lawrench @oup.co.uk 24-hour credit card hotline B' +44 (0)1536 454534

For information on adverti ng in this journ�l� please contact : ,, ... )

Scope of this Journal The JOURNAL OF SEMANTICS publishes articles, notes, discussions, and book reviews in the area of narural language semantics. It is explicitly interdisciplinary, in that it aims at an integration of philosophical, psychological, and linguistic semantics as well as semantic work done in artificial intelligence and anthropology. Contributions must be of good quality (to be judged by at least two referees) and should relate to questions of comprehension and interpretation of sentences or texts in natural language. The editors welcome not only papers that cross traditional discipline boundaries, but also more specialized contributions, provided they are accessible to and interesting for a wider readership. Empirical relevance and formal correctness are paramount among the criteria of acceptance for publication.

Information for Authors: Papers for publication should be submitted in 3 copies to the

managing editor. They should be typed on A4 (or similar format), one-sided, double-spaced, and with a wide margin and must be accompanied by an approx. �oo word summary. Notes and

bibliographical references must appear at the end of the typescript. All bibliographical references in the text by author's surname and year of publication. Diagrams must be submitted camera-ready. All papers submitted are subject to anonymous refereeing and are considered for publication only on the assumption that the paper has neither as a whole or in part already been published elsewhere nor has elsewhere been submitted or accepted for publication. Unless special arrangements have been made, copyright rests with Oxford University Press. Authors receive 20 offprints of their published articles and 1o offprints of their published reviews, free of charge. Larger numbers can be supplied at cost price by advance arrangement.

Subscriptions: The Journal of Semantics is published UK and Europe

quarterly.

£78.oo; USA and Rest of World US$140.00. (Single issues: UK and Europe £23.00; USA and Rest of World US$40.00.) Personal:* UK and Europe £J8.oo; USA and Rest of World US$71.00. (Single issue: UK and Institutional:

Europe £11.oo; USA and Rest of World US$21.00.)

*Personal rates apply only when copies are sent to a private address and payment is made by personal cheque/credit card.

Prices include postage by surface mail or, for subscribers in the USA and Canada by Airfreight or in Japan, Australia, New Zealand and India by Air Speeded Post. Airmail rates are available on request. Back Issues. The current plus two back volumes are available from the Oxford University Press,

Great Clarendon Street, Oxford OX2 6DP. Previous volumes can be obtained from Dawsons

Back Issues, Cannon House, Park Farm Road, Folkestone, Kent CT19 sEE. tel +44 (o)1303 8so101, fax +44 (o)1303 85 0440. Volumes 1-6 are available from Swets and Zeitlinger, PO Box 830, 2160 SZ Lisse, The Netherlands. Payment is required with all orders and subscriptions are accepted and entered by the volume. Payment may be made by cheque or Eurocheque (made payable to Oxford University Press), National Girobank (account soo 1 os6), Credit cards (Access, Visa, American Express, Diners Club), or UNESCO coupons. Please send orders and requests for sample copies to the Journals Subscriptions Department, Oxford University Press, Great Clarendon Street, Oxford OX2 6DP, UK. Telex 837330 OXPRES, tel +44 (o)1865 267907, fax +44 (o)1865 267485.

Copyright: It

is a condition of publication in the Journal that authors assign copyright to Oxford University Press. This ensures that requests from third parties to reproduce articles are handled efficiently and consistently and will also allow the article to be as widely disseminated

as possible. In assigning copyright, authors may use their own material in other publications provided that the Journal is acknowledged as the original place of publication, and Oxford University Press is notified in writing and in advance.

Advertising:

Advertisements are welcome and rates will be quoted on request. Enquiries

should be addressed to Helen Pearson, Oxford Journals Advertising, PO Box 347, Abingdon SO,

OX14

sXX. UK. Tel/fax:

+44 (o)1235 201904- Email: [email protected].

© Oxford University Press

1997

Good News about the Description Theory of Names BART GEURTS University of Osnabriick and Max Planck Institute for Psycholinguistics, Nijmegen

Abstract

INTRO D U CTI O N At the end of the first lecture of Naming and Necessity, Kripke gives short shrift to Kneale's version of the description theory of names, according to which the meaning of a name N is 'the individual named N'. I have always felt that Kripke's criticism of this view falls wide of the mark, and that Kneale's position is essentially correct. In the following pages I try to justify this assessment, referring to Kneale's theory and its kin as 'quotation theories'; for their central claim is that the content of a name quotes the name itsel£ Although occasionally Russell came quite close to defending the quotation theory, to the best of my knowledge its first proponent was Kneale (1962), who was taken to task for this by Kripke (1980). Subse­ quently the theory received support from Loar (1976), Bach (1981, 1 987), Cresswell (198 5), and Fodor (1987). More recently, a number of presupposi­ tion theorists have swelled the ranks of the quotation theory: it is more or less taken for granted by van der Sandt and Geurts (1991), van der Sandt (1992), Beaver (1993), and Geurts (1995), among others. And Zeevat (1996) argues at some greater length for a presuppositional version of the quotation theory that is more or less the same as mine. My defence of the quotation theory is part of a larger project, which is to provide an alternative for semantic analyses built upon such notions as rigid designation or direct reference. Despite their considerable intuitive appeal, I believe that these notions are red herrings, regardless whether they are deployed in the analysis of names, demonstratives, natural kind terms, or wherever. But in this paper the focus will be entirely on proper names. •

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

This is an attempt at reviving Kneale's version ofthe description theory ofnames, which says that a proper name is synonymous with a definite description of the form 'the individual named so-and-so'. To begin with, I adduce a wide range of observations to show that names and overt defmites are alike in all relevant respects. I then turn to Kripke's main objection against Kneale's proposal, and endeavour to refute it. In the remainder ofthe paper I elaborate . on Kneale's analysis, adopting a theory of presupposition proposed by van der Sandt.

po Good News about the Description Theory of Names 1 NAMES A N D OTHER DEFINITE DESCRIPTI O NS

Names often take the form of definite NPs: 'the United Nations', 'the Goldberg Variations', 'the Netherlands', 'the Annunciation', 'the Holy Spirit', 'the B:mk of England'. These are names, no doubt, but they certainly look like definite NPs. In English, river names always carry a . definite article ('the Mississippi'). In Italian, names of women often have, and sometimes must have, a definite article: 'la Loren', 'la Carolina'. I suspect that all languages which have definiteness markers allow them to occur on proper names, and I wouldn't be surprised if there were languages in which names were always marked for definiteness." • So-called definiteness effects apply to names and definite NPs alike. On the one hand, if a construction selects for indefinite (or weak) NPs, names as well as definite NPs are excluded. This holds for English there-sentences, for example: •

(1) There is {1ohn/*the philosopher/a philosopher} available. On the other hand,

if a construction selects for definite NPs, it will accept names, too, as the case of the partitive construction illustrates:

•

(2) half of {Belgium/the country/*some countries} As Kripke ( 1980) points out, Donnellan's referential/attributive distinc­ tion applies to names just as it applies to definite NPs. In Kripke's

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

I claim that a name is synonymous with a definite NP of the form 'the individual named so-and-so'. For the time being we can leave the proposed analysis in this somewhat underspecified state, because I first want to raise a number of points that don't require a more precise formulation. If the quotation theory is correct, then names must be expected to be used and interpreted like other definite NPs. If. on the other hand, names are rigid designators, then we should expect significant empirical differ­ ences between names and definite NPs. I will now show that it is the former prediction rather than the latter that is borne out by the facts. The main objectives of this section are to take stock of the semantic properties of names and definite NPs, and to show that there do not seem to be any fundamental differences between these two types of expressions. I will freely help myself to whatever technical jargon I find convenient for these purposes. For example, I will sometimes pretend that names and definites have scope but will also talk of their 'referents'. This terminological bric a brae should not be taken too seriously. Taken together, the following observations strongly suggest that names pattern with (other) definite NPs in all relevant respects:

Bart

Geurts

32I

example, two interlocutors observe Smith at a distance and take him to be Jones. Accordingly, they use the name jones' to refer to Smith (although Kripke insists that semantically speaking they are referring to Jones). • Names can be used literally as well as non-literally (this point is clearly related to the previous one, but it is not exactly the same). For example, a man who jocosely refers to his wife as 'the Queen' may use the name 'Cleopatra' to much the same effect (c£ Bach 1987). • Names are like overt definites in that they can be used generically:3

•

Obviously, in these examples the names 'Coca Cola' and 'Pristichampsus' are used generically. It is true that it is difficult to think of a plausible scenario in which a proper name like john' will have a generic interpretation, but this has nothing to do with the fact that john' is a name. It is just that the kind that a generic occurrence of john' would denote is of such limited use. It is precisely in this respect that 'Coca Cola' and 'Pristichampsus' are different from john', and this is why the examples above are felicitous. Definite NPs and names are typically, though not always, used to refer to objects that are part of the common ground between speaker and hearer. We may distinguish two cases here. On the one hand, definites and names may both be used anaphorically to refer to objects that have been explicitly introduced into the discourse. For example,

(s) I have a poodle named 'Horace'. {Horace/My poodle} is three years old.

•

On the other hand, names and definites may also be used non­ anaphorically to refer to individuals that are given in the larger situation in which the discourse takes place. For example, in some contexts either 'the girl from no. 2 1' or julia' can be used to refer to my daughter, and in England 'the Queen' and 'Elizabeth II' will typically refer to one and the same person. Definites and names both have bound-variable uses: (6)

If I can choose between a Mercedes and a BMW, I'll take the BMW. b. If a child is christened 'Bambi', then Disney will sue Bambi's parents. a.

The definite NP in the consequent of (6a) is bound to the indefinite in the antecedent, and the same applies for the second occurrence of'Bambi'

.

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

(3) {The light bulb/Coca Cola} was invented by an American. (4) {The panzercroc/Pristichampsus} hunted during the Eocene Epoch, about 49 million years ago, but it was very rar� . (Bakker 1988: 74)

322

Good News about the Description Theory of Names

(7)

If a child is "christened 'Bambi', and Disney Inc. hear about it, then they will sue Bambi's parents. b. The name of a product is essential to its commercial success. For example, if you want to buy washing powder and are given the choice between Black, White, and Grey, you will choose White, won't you? a.

It seems to me that (7a) is better than (6b), and as far as I can tell there is nothing wrong with (7b). The following examples are cases of binding, too, intuitively speaking:5 (8)

Every time we do our Beatles act, {Ringo/the guy who plays the part of Ringo} gets drunk afterwards. b. Every time John goes to see a performance of Hamlet, he falls in love with {Ophelia/the actress who plays the part of Ophelia}. (9) a. Perhaps Mary has a son named john' and perhaps {her son/John} is the thie£ b. Mary is under the illusion that she has a son named john' and she believes that {her son/John} is the thie£ a.

In (8a, b) the definite NPs as well as the names can be interpreted non­ referentially. On this interpretation, 'Ringo' is more or less equivalent with 'whoever plays the part of Ringo', and the same holds, mutatis mutandis, for 'Ophelia'. In (9a, b) there is a preference for a non­ referential interpretation, and intuitively 'her son' or john' in the second conjunct is 'bound' in the first. As a matter of fact, in Geurts (1995, to appear) I present a treatment of intensional contexts that allows us to view these expressions as bound elements, but this is by the way, because all I want to do at this point is establish the parallels between names and overt definites.

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

in (6b). This observation, which is crucial to my concerns, has not met with unanimous approval from the readers of earlier versions of this paper, so let me dwell on it for a while. One of the referees of this journal claims that (6b) sounds odd, and although I do believe that this sentence is fully acceptable as it stands, I have some sympathy �th this judgement. But if (6b) is vaguely odd, it is because the word 'Bambi' is repeated for no good reason: why use this word when a possessive pronoun would have done just as well?4 If this is correct, then we should be able to come up with better examples if we can somehow motivate the repetition of the name. And we can, for instance by increasing the distance between the name and its antecedent and/or by introducing competing antecedents:

Bart Geurts •

323

Although definite NPs and names normally refer to objects in the common ground, they can be used to introduce new objects: (ro) My best friend is {my sister/John}.

•

Both variants of (ro) can be used in a situation in which the intended referent is new to the hearer. Names generally take wide scope. In this respect, too, they are like many other definite NPs. Compare:

(I I )

( 1 ra) is ambiguous: it can either mean that the person who happens to be Prime Minister could have been rich, or just that we might have had a rich Prime Minister. (ub) and (uc), by contrast, only allow for readings of the first type. The suggestion that, with respect to scope, names are unlike definites in that they always take wide scope is incorrect for two reasons. First, as we have just seen, only some definite NPs alternate relatively freely between wide-scope and narrow-scope interpretations. Secondly, and this is the last point on my list, Names can take narrow scope, too. For reasons to be discussed below, this may require a somewhat outlandish type of context, but it does happen. For one thing, there are the bound-variable uses of names mentioned above. For another, there is the case of Aaron Aardvark: ·

•

( 12) The electoral process is. under attack, and it is proposed, in light-of recent results, that alphabetical order would be a better method of selection. than the present one. Someone supposes that 'Aaron Aardvark' might be the winning name and says, 'If that procedure had been instituted, Ronald Reagan would still be doing TV commercials, and [( 12)] Aaron Aardvark might have been president' (Bach 1987: 146-7). Clearly, in this scenario the speaker need not believe that there is anybody for the name 'Aaron Aardvark' to refer to, yet (12) isn't infelicitous in any way. Further examples of the same type are: ( 1 3)

a. In English, Leslie may be a man or a woman. b. But John is always male.

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

a. The Prime Minister could have been rich. b. The man could have been rich. c. John could have been rich.

324

Good News about the Description Theory of Names 2 P RELIMI N A RY SKIRMISHE S

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

Although it is always hard to prove that something doesn't exist, the foregoing observations do suggest quite strongly that there are no deep semantic or pragmatic differences between names and overt definites. Such distinctions as must be made set some definite NPs apart from others but they don't draw the line between names and overt definites. Furthermore, most of the observations I have mustered are difficult to reconcile With the hypothesis that names are rigid designators. Not to mince words: they falsify it. If names were rigid designators it would be worrying, at the least, that names can have definite articles, that they can be used attributively and non-literally, and can be used to introduce individuals that are new to the hearer, and it would be a mystery that names have bound-variable uses and may take narrow scope. It may be challenged that at least some of these observations are irrelevant because they involve 'metalinguistic' uses of names, the implication being that names are generally not used this way.6 I think this is partly right and partly wrong. What is right about it is that some of the data I have listed are somewhat out of the ordinary. But that is precisely what one should expect if the quotation theory is correct. If it is true that the meaning of a name is 'the individual named N, then the content of a name is special in a way that will make it difficult to construct examples in which names act as bound variables, for example. But the same holds for some overt definites, too. Consider 'the cosmos'. There can be no doubt that this is a definite NP, but due to its truly comprehensive meaning it will be difficult to construct sentences analogous to the ones above in which 'the cosmos' acts as a bound . variable. Difficult, but not impossible. And although such cases will be very 'special', they will surely count as evidence that 'the cosmos' is an ordinary definite NP. Therefore, I see no reason to dismiss the data presented in (6)-(9) and (12)-(13) merely because these are special cases. It is understandable that someone who takes the position that names are rigid designators will want to claim that my counterexamples are irrelevant because they involve deviant uses of proper names.7 However, this is a lame defence unless it can be shown on independent grounds that the semantic values of. say, 'Bambi' in (6 b) or 'Aaron Aardvark' in (12) are non-standard. It will not do to note that these sentences are special or require special contexts, and leave the matter at that. In the absence of such independent evidence, these data stand as evidence against Kripke's theory of names.

Bart Geurts 325 3

N A ME S A N D RE FE RE N CE

Kripke charges that the quotation theory is circular because it actually presupposes what it must explain, namely how names come to refer: Someone uses the name 'Socrates'. How are we supposed to know to whom he refers? By using the description which gives the sense of it. According to Kneale, the description is 'the man called "Socrates" ' . . . We ask, 'To whom does he refer by "Socrates"?' And then the answer is given, 'Well, he refers to the man to whom he refers.' If this were all there was to the meaning of a proper name, then no reference would get off the ground at all (Kripke 1980: 70). Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

But what if the quotation theory didn't explain how names manage to refer? Suppose we analyse definite NPs along Russellian lines.8 Then the quotation theory entails that names are non-referential expressions. This may strike some as an intuitively repugnant conclusion, but Kripke himself has argued persuasively that such intuitions are not necessarily detrimental to a Russellian theory. Critics of Russell's theory of descriptions have often claimed that the theory is falsified by referential uses of definite NPs. Kripke (1977) argues, however, that this is not so, because a Russellian may hold that the theory of descriptions is a semantic theory and need therefore only be concerned with what is said (in Grice's sense); intuitions about referential uses of definite NPs pertain to what is meant by a speaker on a given occasion. Put otherwise, the idea is that although semantically speaking definite NPs are non-referential expres­ sions, speakers may use them for conveying information about specific individuals. But if names are definite NPs, then the same observations should apply to them, too. Hence, Kripke's complaint that the quotation theory is not a theory of reference not only presupposes that any semantical t�eory of names should be a theory of reference, like his own; it also denies the quotation theorist an account of reference that Kripke recommends elsewhere. If NPs of the form 'the individual named N' are to be treated in a Russellian framework, they must probably be viewed as 'incomplete' definites, i.e. as on a par with 'the table', 'the child', and so on.9 As Strawson was perhaps the first to point out, it would seem that such expressions, which rarely if ever describe unique objects in the world yet can be used without apparent difficulties, cause an especially severe problem for Russell's theory of descriptions. Russell's followers have countered Strawson's objection in various ways. According to Bach (1987), for example, a speaker who produces a sentence containing an incomplete definite NP virtually never says what he means (I am again speaking in Gricean terms): strictly speaking such sentences are false, but if all goes well,

326

Good

News about the Description Theory of Names

as it usually does, they will none the less manage to convey information about specific individuals. This may or may not strike one as a plausible strategy for dealing with incomplete NPs, but this is as it may be, since for the moment I merely want to note that the quotation theory is immune to K.ripke's objection that it doesn't explain how names come to refer. This objection clearly presupposes that names are referential expressions, where reference is to be understood as a semantic relation between a term and a real-world individual, and this presupposition cannot be taken for granted. Of

4 BEING NAMED N VS . BEING THE R EFER ENT OF N Kripke's criticism is misguided for a further reason, as well, because it doesn't distinguish between two properties that should be strictly kept apart, viz. being the referent ofN, and being named N.10That N is the name, or one of the names, of an individual a does not entail that N refers to a or that a is among N's referents. Kripke assumes as a matter of course that when we baptize an individual N we eo ipso determine the reference of the name N. This is not so. Beethoven's Missa Solemnis is commonly referred to as 'the Missa Solemnis', but its name is 'Missa Solemnis'. Similarly, an Italian may refer to Maria Callas with 'la Callas', but her name was 'Callas'.11 The difference is slight (a mere two or three letters, after all) but telling: it highlights the fact that although it so happens that in English names can generally be used to refer, it might have been otherwise (and it sometimes is, even in English).The grammar of English might have dictated that in order to refer with the help of a name, it must always be preceded by the definite article, or, for that matter, by 'the individual named'. Bearing a name is like wearing a tie. Like ties, names are seldom unique, but circumstances permitting they may be used for referential purposes. More accurately, just as you can employ the attribute of wearing a tie to identify to your audience the person you have in mind (John, as the case may be), you can use the attribute of being named John' for the same purpose. Taken on its own, however, a name doesn't refer any more than a tie does. It is instructive to compare names with number terms. Indeed, although number terms are rarely classified among the names, they are often used as such. Bearing a name may be likened to having a number: a crate in

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

course, it can hardly be denied that names may be used to convey information about real-world individuals, but the claim that names are referential expressions does much more than restate this banal truth in different terms, and should not be accepted uncritically.

Bart Geurts 327 Rotterdam harbour may have a number stamped on it, for example 6. But of course the number term '6' doesn't refer to the crate in question, although it may be used to refer to it. That is to say, when taken on its own, a number term

by the computer's hardware and the programs running on it: if due to some

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

can sometimes be used to refer, but in most cases it must be part of a larger expression: 'crate number 6', '6 Elm Street', 'Paris VI', 'volume 6', and so on (with names it happens to be the other way round, as we have seen). Another respect in which number terms are like names is that there are countless ways in which an individual may be assigned a number or number term (the distinction is not always clear, but it doesn't seem to matter much, either). Sometimes individuals are inscribed with numbers: houses, banknotes, football players. In many other cases, there is a less immediate relation between a number and an individual that 'has it'. Numbers may be assigned at random or following a rigid procedure, for instance according to some linear order. There are various conventions for assigning house numbers, convicts are numbered as they enter prison, cameras receive a number when they leave the factory, and· in some countries citizens receive a number at birth. Just as there are many ways of assigning a number to an individual, so are there many ways of assigning it a name. There are descriptive names like 'Fatty' or 'Benjamin'. Similarly, the title of a book or film is expected to be somehow related to its content. In my native country, last names are assigned according to strict regulations, but first names are afflicted by whim. Other cultures make use of patronymics, and in still others parents are named after a child. And so on. What all these naming practices have in common is just that some association is established between a name and its bearers, but how this association is initiated and sustained is different from case to case. The name-bearing relation between 'Lolita' and the famous novel was initiated by the author and is sustained, inter alia, by printing the name on the front cover of every copy. The association between my last name and myself is sustained, inter alia, by records at a register office, but this does not apply for my first name. If I decided that I wanted to be called 'Rudolf instead of 'Bart', it would just be a nuisance to the people in my social sphere, but I don't think it would be humanly possible to change my last name into 'Camap' (not in my country anyway). It is not even true that the name-bearing relation is necessarily grounded in a social convention of some sort. For example, the Hies on my computer all have names; some of these I chose myself, others were chosen by various people around the globe, only some of whom I am acquainted with, and yet others, such as 'cachei8J4IS.shtml' for example, were generated by some program. The association between a Hie and its name is entirely sustained

J28

Good News about the Description Theory of Names

s INTR ODUCIN G O T T O 'Ottomobil' is the official name of a domestic robot which i s currently being developed in my department; my colleagues and I usually call him 'Otto' (we prefer to think of Otto as male).'2 A central part of Otto's design is a system for natural language understanding and generation, and my personal pride and joy is the robot's presupposition module, which is based upon a theory that I helped to develop. '3 In fact, Otto's presupposi­ tion module not only deals with presuppositions, but with anaphora as well. Otto treats presuppositions and anaphors alike as elements that want to be bound to an antecedent. If occasionally an appropriate antecedent may not be found, Otto is prepared to accommodate the presupposition in question; but in general he prefers to bind his presuppositions. This preference is

stronger in some cases than in others. For example, while Otto very much likes to bind anaphors and other descriptively attenuate presuppositions, he doesn't mind very much accommodating presuppositions triggered by factive verbs.'4

(14)

a. If someone broke the copier, then the new secretary is the culprit. b. [: [x, g: copier u, x broke u] =} [�, y: culprit z, new secretary v, z = v) J c. [u,

v:

copier u, new secretary v, [x, z: z

=

x, x broke u, culprit z)

[: z = v]] d. [u, v: copier u, new secretary v, [x: x broke u, culprit x) =}

=}

[: x

=

v)]

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

programming error the ftle name 'Bocherini' gets changed into 'Corelli', then the ftle's name is 'Corelli', not 'Bocherini'. The original intention behind name-giving is simply irrelevant. The expression 'bearing a name' covers as many relations as there are naming practices, and it seems to me that Kripke's causal theory of reference is best viewed, not as a theory of reference, but as a partial theory of what it means to bear a name. The theory is only a partial one because there are naming practices that it doesn't account for, such as the one considered in the previous paragraph. Kripke's charge that the quotation theory is circular can now be countered a second time, as follows. Even if a proponent of the quotation theory should want to explain how names can be used to refer, his analysis of names need not be circular, because the notion of reference need not enter into it: 'the individual named N ' is not the same as 'the individual that N refers to', and the former does not presuppose the latter, either.

Bart Geurts

329

·

6

ACCOMM ODATI ON AN D C OUNTE RPARTS

I said that, when Otto is presented with sentence (14a) he will accommodate

the presuppositions triggered by 'the copier' and 'the new secretary'. This is

not all he does, however. Even if he decides to accommodate a presupposi­ tion, Otto verifies if it matches his representation of the world, or is at least consistent with such knowledge he has at his disposal. In order to explain

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

(14b) is the DRS that Otto's grammar module computes for the sentence in (14a). This sentence contains three definite NPs, which trigger the presuppositions that in (14b) are marked by underlining. Otto will deal with· these presuppositions as follows. Beginning with the presupposition triggered by 'the copier', Otto first checks if it can be bound to an antecedent in his current DRS. Seeing that there is no suitable antecedent available, he decides that the presupposition must be accommodated, and since he has a preference for global accommodation, he accommodates the presupposition in the main DRS. The second presupposition, triggered by 'the culprit', can be bound to the indefinite in the antecedent of the conditional, which is the solution that Otto favours. Finally, as there is no suitable antecedent for the presupposition triggered by 'the new secretary', this presupposition is accommodated too, again in the main DRS. The resulting reading that Otto assigns to (14a) is (14c), which is equivalent with (14d). In this example, all presupposition-inducing expressions happen to be definite NPs, but I should like to emphasize ·that Otto applies the same handful of principles to all types of presuppositions, regardless whether they are triggered by definite NPs, pronouns, factive verbs, clefts, focusing particles like 'too' or 'even', etc. These principles are, to recapitulate: (i) a presupposition is preferably bound, but (ii) if it cannot be bound it will be accommodated, and (iii) if a presupposition must be accommodated, then it is preferably accommodated in the least embedded DRS, i.e. global accommodation is preferred to local accommodatiotL Definite NPs and pronouns are special only in that they presuppose their entire content: semantically speaking, these expressions are nothing but presupposition inducers. All there is to say about the content of 'the man' or 'he', for example, is that these expressions trigger the presuppositions that there is a man and that there is a male person, respectively. In contrast to these purely presuppositional expressions, as they might be termed, a factive verb like 'regret' not only triggers the presupposition that its complement

3 30

Good News about the Description Theory of Names

this process in more detail, let us consider a simpler example. Suppose I inform Otto that:

(1 s)

The new secretary is Irish.

The

DRS

(16) [!:

that Otto will produce for this sentence looks like this:

new secretary x, Irish x]

( 1 7) [z: new

secretary

z,

Bart believes: [ x: new secretary x, Irish x], . . . ]

{I7) makes explicit the fact that Otto, too, believes that there is a new secretary. It will be obvious that {I7) is only a partial representation of Otto's belief box: apart from the two beliefs explicitly represented in (17), Otto will have a wealth of further beliefs, which are abbreviated here by three dots.

says that Otto believes that there is a new secretary, z, and that he believes that I believe that there is a new secretary, x, but it doesn't establish

(17)

a link between these two individuals. In this particular case, however, such a link may be assumed to exist, . and, more to the point, we may take it that

Otto believes that such a link exists. As he might say, his

z

and my x are the

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

Formally, this represents a purely existential proposition: (16) is true in any given world iff that world contains a new secretary who is Irish. This is correct as far as it goes, but it doesn't go far enough, for we should expect that someone who is processing (I s) will attempt to establish a connection between the presupposition triggered by the subject term, on the one hand, and what he takes to be the case in the world surrounding him, on the other. This is precisely what Otto does. Two cases must be distinguished here depending on whether Otto has already formed the belief that there is a new secretary, or not. In the first case, it is not new to Otto that there is a new secretary, and he has already accumulated information about her. Perhaps they have already met, or maybe Otto has just heard from someone else that there is a new secretary. At any rate, we may assume that in ·this case he has already a mental representation which symbolizes the new secretary (as we might say, if we deemed Otto's belief to be true), and he will link the underlined material in {I6) to this representation. But what exactly do we mean when we say that this presupposition is linked to a representation that was already available beforehand? I have argued elsewhere that a DRS must be seen as a representation of the speaker's commitment slate,'5 and so {I6) is Otto's representation of my commitment slate, which he constructed in response to my utterance of {I s). It follows that {I6) is actually embedded in the context of Otto's beliefs, and if we want to make this context explicit, we get something like the following picture:

Bart Geurts

331

same person. I prefer to say that they are counterparts in the sense of Lewis.'6 If we want to make this relation explicit we need something like the following: ( 1 8) [z: new secretary z, Bart believes [� new secretary x, Irish x], . . . ] Another, more succinct notation for counterpart relations, which I have proposed in Geurts ( 199 5. to appear), is t.o simply use the same reference marker twice, as follows: ( 19) [x: new secretary x, Bart believes: [�: new secretary x, Irish x], . . . ] Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

The intended interpretation of this structure is the same as that of (I 8). This notation of counterpart relations exploits the fact that, logically speaking, the two occurrences of x are independent of one another. · Counterpart relations are to capture our intuitions about identity across possible worlds-or, more accurately: intuitions that we would express in terms of identity relations across world boundaries. Counterpart relations are, in Lewis's words, 'our substitute for identity between things in different worlds' (Lewis I96 8: I 14). When we speculate about what might have happened to John, say, we pretend to speak of the same individual in counterfactual circumstances. We allow that two people are thinking of the same individual, although they may disagree about virtually all of its properties. Even more strikingly, we allow that Hob and Nob have the same witch in mind, although there are no witches.'7 All these cases can be understood in terms of counterpart relations. Unlike identity, counterpart relations are indeterminate in various ways.'8 To say that two individuals are counterparts is to say that they are alike in some respects, and since similarity is a matter of degree, we have to agree on a lower bound of similarity before we can decide whether a counterpart relation obtains in any given case. More importantly, for our present purposes at least, is that similarity depends on how you look at it. Two individuals may be similar in some respects but not in others, and before we can say that they are counterparts or not we have to decide what respects are to count. Now I can explain in more detail what happens when Otto decides that 'the new secretary' in ( Is ) cannot be construed as an instance of binding. The speaker who uttered this sentence, i.e. myself, presupposes that there is . a new secretary, and Otto will accommodate this presupposition in his representation of the speaker's commitment slate, and link it to his own representation of the new secretary by means of a counterpart relation, as shown in ( 19). As Otto himself would put it, he and I have the same person in mind.

3 32

Good News about the Description Theory of Names

Otto is a trustful robot. In most cases he has no doubts that what people tell him is the truth, and he is particularly confident that I am usually well-informed and would never lie to him. If I say that the new secretary is Irish, Otto comes to believe that the new secretary is Irish. So eventually the effect of my uttering (1 s) is that Otto's belief box is updated as follows: (2o) (x: new secretary x, Irish x, Bart believes: (�: new secretary x, Irish x], . . . ]

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

This is what happens. when Otto already believed that there is a new secretary before I made my statement. If Otto did not have this information beforehand, the interpretation process will be different in some respects, but the outcome will be more or less the same. Suppose that Otto is not aware that there is a new secretary. Again I enter the scene, and recite my line. This will have the same net effect on Otto's belief box as in the previous case, because Otto is a trustful robot, and expects that what I tell him is the truth. This is not to say that it makes no difference at all whether or not Otto already believes something that is presupposed by the speaker addressing him. Trustful though he may be, Otto sets greater store by what he believes than what people tell him, and if he already is aware that there is a new secretary his representation of the person in question is presumably richer than when he has just gleaned this information from my statement. The important point, however, is that in either case the relation between Otto's own representation of the new secretary, on the one hand, and his representation of my representation of the new secretary, on the other, will be the same. (2o) is in a sense a purely descriptive representation. Despite their name, reference markers do not refer to individuals in the world (or model), and the reference marker x does not refer to the new secretary. But if (2o) represents Otto's belief box, then surely there must be some connection between this reference marker and the new secretary? And there might be. In fact, there are many types of possible connections between reference markers and real-world individuals. If Otto has already seen the new secretary, he will have a visual memory of her, and · the reference marker will be linked to that. If he has heard her speak, he will have an auditive memory which is �ssociated with the reference marker. He may have read her application, he may have seen her photograph, he may talked about her with people who have seen her or who know people who have seen her, and so on. It is connections such as these that prompt us to say that Otto has beliefs about the new secretary.

Bart Geurts 3 3 3 7 OTT O 'S T REAT MENT OF NAMES

• • • • •

names often carry definiteness markers; names are subject to definiteness restrictions; the referential/attributive distinction applies to names; names can be construed non-literally; and names can be construed generically.

I don't mean to suggest that I can account for definiteness restrictions, the distinction between referential and attributive readings, non-literal construals, and genericity.The point I want to make is merely that, given that definites are subject to definiteness restrictions and have referential, attributive, non­ literal, and generic interpretations, the presuppositional theory of names leads us to expect that the same will hold for names, which it does. The remaining observations from section I require more detailed explanations:

•

Names are typically, though not always, used to refer to objects that are part of the common ground. We distinguished two ways of being part of the common ground: if a name is used anaphorically, its referent was already introduced into the discourse; if it is used non-anaphorically, its referent will generally be given, too, though not in the immediate context. That names prefer their referents to be given is a characteristic they have in common with all other presupposition inducers. It is true that names have a comparatively strong preference for a linking interpretation, which is to say that, on the whole, their presuppositions

are rather difficult to accommodate without being linked to a given

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

In accordance with the quotation theory, Otto treats names as definite descriptions of the form 'the individual named so and so'. This is not to say, of course, that Otto treats names as definite NPs in disguise; he doesn't transform a name into a definite NP before he interprets it. It is just to say that Otto regards names and overt definite NPs alike as purely presupposi­ tional expressions, which presuppose their entire descriptive content. Given that Otto already knew how to handle presuppositions, this was by far the most obvious treatment from a designer's point of view: it is more elegant than any other solution we could think of and enables Otto to interpret names in precisely the right way. To show this, let us consider how the presuppositional version of the quotation theory used by Otto measures up to the observations listed in section 1 . The first five of .these observations need hardly any comment. Given that names are treated on a par with overt definite NPs, it doesn't come as a surprise that:

3 34

•

Good News about the Description Theory of Names

referent, but in this respect names aren't exceptional, either, since they share this characteristic with some other presupposition inducers, such as 'incomplete' definite NPs (I will return to this point below). Names have bound-variable uses. In the DRT framework adopted here, any presupposition that is linked to an antecedent in an embedded DRS will seem to behave like an element bound by a quantifier. So this is just a special case of the general rule that presuppositions prefer to be linked to a given antecedent.

(z r ) a. If a child is christened 'Bambi', then Disney will sue Bambi's

(2 1b) is the DRS that Otto's parser computes for (2ra). The second occurrence of the name 'Bambi' triggers the presupposition that there is someone named 'Bambi' (at its first occurrence the name is mentioned, not used). This presupposition is bound in the antecedent of the conditional, and the resulting interpretation is (2rc). Hence, on the present account there are no relevant differences between (21a) and the following: (22) a. If a farmer owns a donkey, he beats it. b. If a farmer beats his donkey, his wife beats it, too.

•

In (na) the pronoun refers back to the indefinite in the antecedent of the conditional, and similarly, in (22b) the presupposition triggered by the focus particle, that someone other than 'his' wife beats 'it', is bound in the antecedent of the conditional. Hence, the presuppositions triggered by 'it' in the first sentence and 'too' in the second are treated exactly the same as the presupposition induced by 'Bambi' in (2 1 a). Although names typically refer to entities in the common ground, they" may be used to introduce new individuals, as in: (2 3) My best friend is John. (= ( ro) )

•

In general, if a presupposition cannot be bound to a suitable antecedent, it will be accommodated. So unless there is a suitable antecedent for John' in the context in which (23) is uttered (as when the sentence is preceded by, say, 'I have three friends: John, Jack, and Joe'), the presupposition that there is an individual with that name will be accommodated. Names generally take wide scope. It seems a rather safe conjecture that if a presupposition is bound it will usually be bound globally, i.e. in the

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

parents. (= (6b)) b. [: [x: child x, x is christened 'Bambi'] '* [g: u is named 'Bambi', Disney sues u's parents] ] c. [: [x: child x, x is christened 'Bambi', x is named 'Bambi'] '* [: Disney sues x's parents] ]

Bart Geurts

33S

·

(24} John believes that most party members are stupid. . This sentence has at least tWo readings, depending on whether the quantifier 'most party members' takes wide or narrow scope with respect to the attitude verb. But this quantifier also triggers the presupposition that there are party members, and this presupposition will normally be bound or accommodated globally, i.e. outside John's belief context. If we refer to the scope of an expression a, we speak of a unit corresponding with a at some level of analysis (e.g., a's correlate at LF or a's interpretation relative to a given model). Although in some frame­ works this is only a metaphor, the guiding intuition is that a's scope is determined by moving a about. If, on the other hand, we speak of a presupposition triggered by a, the metaphor is a quite different one, the idea being, rather, that a initiates a search for an antecedent meeting certain specifications. The distinction between scope taking and presupposition projection is somewhat obscured by the circumstance that certain expressions are purely presuppositional.'9 This holds in particular for anaphoric pro­ nouns, definite NPs, and names. The semantic contributions made by these expressions coincide with the presuppositions that they induce. Consequently, if these presuppositions are bound or accommodated globally, as they usually are, it may seem as if the expressions that triggered them had taken wide scope, but as a matter of fact the notion of scope doesn't enter into this at all. This is why I said that my analysis explains why names appear to prefer having wide scope. The presuppositions triggered by names seem to have a decidedly stronger tendency to 'take wide scope' than some others. In this respect, too, they are on a par with other descriptively attenuate, 'incomplete', ·

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

principal DRS; for surely cases like (22a, b) are the exception rather than · the rule. Moreover, if a presupposition must be accommodated, global accommodation is preferred to accommodation in an embedded DRS. Thus, on the whole presuppositions appear to have a strong tendency to be projected to the global level of the discourse representation. Since there is no reason to assume that this doesn't hold for names, we thus explain why names appear to prefer having wide scope, I say 'appear' because I don't actually want to maintain that names are scope bearing expressions. There is a subtle but important difference between saying that a presupposition is accommodated globally and saying that an expression takes wide scope. The difference becomes rather obvious when we consider expressions that have scope and induce presuppositions at the same time, like quantifiers, for example:

3 36

Good News about the Description Theo ry of Names

definites like 'the door' or anaphoric pronouns like 'it'. Compare the following sentences, for example: (z s)

(zsa) allows for a reading according to which we might have had a situation s in which the President in s was a woman. In order to obtain this reading, the presupposition triggered by 'the President' must be accommodated locally, i.e. within the scope of the modal operator, and must not be linked to whoever happens to be president in the situation in which the sentence is uttered. This type of interpretation does not appear to be so readily available for (2sb), presumably because 'the car' insists more strongly that its referent be given (either in the previous discourse or in the wider context) than 'the President' does. Van der Sandt ( I 992) suggests that such differences correlate with varying degrees of descriptive richness. This suggestion is perhaps correct, although it may not be the whole story. But no matter how such . differences are to be accounted for, it is evident that in this respect a proper name like john' resembles 'the car' more than 'the President', in that it insists rather adamantly that its presupposition be linked to . an entity that is given. . Names can take narrow scope. By way of exception, our robot Otto is prepared to accommodate the presupposition triggered by a name in an embedded DRS. This is how he will deal with the name 'Aaron Aardvark' in the scenario devised by Bach:

(26) Aaron Aardvark might have been president. (= ( 1 2) ) Supposing that Otto doesn't . know of any person named 'Aaron Aardvark' (which he doesn't), and supposing that for whatever reason he doubts that the speaker knows of any such person, then Otto will decide to locally accommodate the presupposition triggered by the name, thus arriving at the following reading: (27) [:

0 [x: x is named 'Aaron Aardvark', president x]]

In general, local accommodation is only used as a last resort, and as with other semantically attenuate definite NPs, it is exceptional for a name to be interpreted by means of local accommodation. Bach's example really is a rare fmd.

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

•

The President might have been a woman. b. The car might have turned left.

a.

Bart Geurts

3 37

8 W HAT J O H N M I G HT HAVE BEEN Suppose that shortly after John's birth following: (28) a. If John had had red hair b. If John had been a girl c. If John had been a twin d. If John had been a Rolex watch

his

}

mother utters one of the

his father would have been even happier.

(29)

a. [: [2£: x is named 'John', P x] rv> [x's father is happier] ] b. [x: x is named john', [:P x] rv> [x's father is happier] ]

Here 'rv>' symbolizes the counterfactual conditional, and the value of 'P' covaries with the predicates in (28a)-(28d). In (29a) the presupposition is triggered that there is an individual named john', and our presupposition theory predicts that this will project to the main DRS. The resulting

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

We should probably say that (28a) refers to a property John himself might have had. But being a Rolex watch is not such a property, and so (28d) is to be paraphrased rather along the following lines: 'IfJohn's mother had given birth to a Rolex watch instead of to John (whatever that may mean), his father would have been even happier.' In short, while (28a) is about a counterfactual situation involving John, (28d) is about a counterfactual situation in which John's place has been taken by a Rolex watch. (Note that, according to counterpart theory, which I still endorse, this distinction cannot be taken over into our semantic analysis, since no individual can inhabit more than one world. However, it does not follow that counterpart theory cannot account for the intuition, which I am trying to character­ ize, that there is such a distinction.) Examples (28b, c) are in a sense intermediate cases because it is less clear how we should describe them. Speaking only for myself: although I would probably say that John himself might have been a girl, I would rather not say that he himself might have been a twin. Therefore, I would bracket .(28b) with (28a) and (28c) with (28d); but I don't expect that these judgements will prove to be uncontroversial. This is as it may be, however, because the relevant observation is that our intuitions about the relation between John and his counterparts vary at all. Needless to say, this variation causes problems for a Kripkean analysis of names. In the present framework, by contrast, it is only to be expected. According to the theory outlined in the foregoing, the examples in (28) will all be interpreted along the following lines:

338

Good News about the Description Theory of Names

interpretation is (29b), which says that there must a counterpart relation between John and the counterfactual individual satisfying the antecedent of the conditional. (29b) does not specify what kind of counterpart relation this should be, which is precisely what we want, because this relation varies from case to case, as we have seen. In other words, the variation we observed in (28) reflects the context dependence of the counterpart relation.

9 THE I NTUIT I O N O F RIGID ITY

(3o) Mary is happy. If someone understaq.ds this statement correctly, then he grasps a proposi­ tion which is true in a certain range of possible situations: in each of these situations, Mary is happy, where Mary is the referent of 'Mary' in the situation in which (Jo) is uttered. This, Kripke suggests, is the claim that names are rigid designators. This observation appears to establish two points at the same time. First, since it is always the same person that makes the proposition expressed by (3o) true, in any given possible situation, it would seem to prove that the name 'Mary' is rigid. Secondly, it would seem to prove that the property of being named 'Mary' isn't part of the meaning of the name, because the proposition expressed by (3o) might be true even in a possible situation in which Mary had a different name. In brief, ifKripke's observation is correct, then names are rigid and the quotation theory is false. Kripke's observation is not as straightforward as it appears, however. To begin with, it matters a great deal how we frame our initial question. If we ask, as Kripke does, if (3o) might be true in a possible situation in which Mary was called differently, then it will seem as if the name 'Mary' is rigid.

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

In imitation of Bach (1987), I would have entitled this section 'the illusion of rigidity'. But on second thoughts I don't believe that rigidity is an illusion: it is a genuine empirical phenomenon. What is illusory is the notion that it calls for an explanation in semantic terms. I have tried to explain why it is that names appear to have such a strong preference for taking wide scope. But Kripke's most forceful argument in favour of the thesis that names are rigid designators is based on the intuitive truth conditions of simple sentences, i.e. sentences without any relevant scope-bearing expressions (such as modals). The argument, which is only seemingly straightforward, as we will presently see, goes as follows. Consider a simple sentence containing a name, such as:

Bart Geurts

339

If, however, we ask what information is conveyed by this sentence, then it would seem to be part of its content that Mary is named 'Mary'. The point I want to make here is prefigured in Lewis's (I 98 I) commentary on Kaplan's (I989) character/content distinction.2° Kaplan claims that his notion of content captures the pre-theoretical notion of 'what is said' by uttering a sentence in a given context. For example, if two interlocutors simultaneously utter (J i a) and (Jib), respectively, what is said is the same: (3 I )

a. You are sleepy. b. I am sleepy.

Unless we give it some special technical meaning, the locution 'what is said' is very far from equivocal . . . Kaplan's readers learn to focus on the sense of 'what is said' that he has in mind, ignoring the fact that the same words can be used to make different distinctions. For the time being, the words mark a defmite distinction. But why mark that distinction rather than others that we could equally well attend to (Lewis 198 1: 97)?

Lewis does not claim that Kaplan's theory is false. What he objects to is Kaplan's suggestion · that his theoretical notion of content captures the pre-theoretical notion of what is said, and the implication that genuinely alternative theories of meaning cannot capture this notion. This suggestion and its implication are wrong, because there is no single pre-theoretical notion of what is said. Intuitions about what is said vary with one's interests. In a sense, two interlocutors uttering (J ia, b) may have said the same thing in one sense, but in another sense they haven't. A theory oflexical semantics may be expected to explain this shifty behaviour of the verb 'say', but there is no reason to require that central concepts in semantic theory (such as 'character' or 'content') must capture any of our ways of understanding 'what is said'. Lewis's objection applies to Kripke's rigidity thesis, too. It is true that there is a sense in which (3o) correctly describes a possible state of affairs in which Mary happily lives under a different name. But there is also a sense in which (3o) is not correct in such a state of affairs; for if Mary is called 'Gertrude', say, then it is incorrect to call her 'Mary'. Indeed, it has never been denied, as far as I am aware, that someone who utters (3o) conveys the information that Mary is called 'Mary'. What Kripke denies is merely that this information is part of the meaning of 'Mary', and he suggests that this equally holds for the pre­ theoretical and. theoretical notions of meaning. Given the chameleontic quality of the pre-theoretical notion of meaning, it is pointless to disagree with the first half of this claim; but the s�cond half is false, in my view.

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

Or in Kaplan's terms: although these sentences have different characters, there are contexts in which their contents coincide. Lewis's objection is the following:

340

Good News about the Description Theory of Names

1 0 STILL, WHY DOES 'MARY' SEEM TO BE RIGID? Why is it that names appear to be rigid? This is a legitimate question, even if we mustn't presuppose that the intuition of rigidity is a direct reflection of the semantics of proper names. My answer to this question is not a new one. It has been given, independently it seems, by several reliable sources, including Sommers (1982), Bach (1987), and Fodor (1987)."'' Of these, Fodor's version comes closest to the formulation that I prefer:

I would prefer putting it as follows. The meaning of a name N is 'the individual named N', where the semantic contribution of the definite article is to be analysed in presuppositional terms. Given that the property of bearing the name N is an accidental one (I · could name my left ear 'Fortinbras' if I chose), referring to an individual with N will not, in general, be particularly effective unless the hearer already knows the intended referent and that it is named N. Therefore, a name will practically always be used to refer to an individual that was already given to the hearer beforehand, and if a name is thus used it will appear to be rigid. To say that 'the N', where N stands for any nominal head, is a presuppositional expression is to say that its intended referent a is presented as given, and that it is presented as given that a is an N. Someone who utters (3o), for example, presupposes that there is a person a whose name is 'Mary' and asserts that a is happy. So what Kripke calls the 'meaning' of this sentence is just its asserted content, and it is of course correct that if we confine our attention to this part of the information conveyed by (3o), the 'meaning' of 'Mary' will appear to be rigid, and the property of 'being named 'Mary' ' will appear to be irrelevant. As Bach (1987) points out, the strategy which Kripke consistently employs in Naming and Necessity is to first fix the referent of a name before he broaches the issue whether or not the name is rigid. But then it shouldn't come as a surprise that names always tum out to be rigid. Nor is it surprising that, for Kripke, non-referring names eo ipso disqualify as admissible evidence. My suggestion is that names appear to be rigid because they are presuppositional expressions. But surely not all presuppositional expressions appear to be rigid designators. So there must be something else that gives

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

The course of wisdom would be to reiterate the moral-viz., that names are a hard problem for everybody-and then to shut up and leave it alone. Still, how about this: 'Cicero' and Tully' are synonymous but differ in presupposition . . . Then 'Cicero was wet' says, in effect, that he was wet and presupposes that he was called 'Cicero'. Tully was wet' says that he was wet too, but it presupposes that he was called 'Tully'. 'Cicero is Tully' is informative because, although it doesn't say that the guy who was called 'Cicero' was called Tully', it 'carries the information' that he was (Fodor 1987: S s).

Bart GeurtS

341

names their peculiar rigid feel. There is, but it is not peculiar to names. Consider the following situation. Otto and I are standing in the corridor, discussing the weather. We have just witnessed Mary walk past us and enter the kitchen. A moment later we hear cheerful singing emanating from the kitchen. At that point I say: (32) She is happy.

(33) [�: person x, female x, happy x] Since Otto treats anaphora as a special case of presupposition, he assumes that the pronoun triggers the presupposition that there is a given person who is female, and that it is asserted that this person is happy. There is no suitable antecedent for this presupposition to be bound to, so it must be accom­ modated. But even so Otto will try to integrate the presupposition with his other knowledge, which is to say, in the present case, that he will connect it to his representation ofMary. Thus Otto's beliefbox, in which his representation of my commitment slate is embedded, may be pictured as follows: (34) [x: woman x, x is named 'Mary', x is in the kitchen, Bart believes: [x: person x, female x, happy x] ] Otto believes that there is a woman called 'Mary', that she is in the kitchen, and that I believe of this person that she is a female person who is happy. Or rather, the person that Otto's beliefs are about and the person that he takes my statement to be about are counterparts. This is why the pronoun in (32) appears to be rigid. By uttering (32) I have conveyed the information in (3 3), part of which is that Mary is a female person; for this is the descriptive content of the pronoun 'she' (by approximation, at least). But when we ask Otto if what I have said might be true even if it transpired that Mary is a man, his answer will be yes. This is a reasonable answer because one very common way of interpreting 'what is said' is by restricting it to asserted information. If instead of (32) I had uttered (Jo), Otto would construe the name as 'rigid' for the same reason he would construe the pronoun in (32) as 'rigid'. And he would say that the property of being named 'Mary' is not part of what I have said just for the same reason he would say that the property of being a female person is not part of the statement made by (32).

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

To Otto it will seem as if 'she' is a rigid designator. For what I have said is true iffMary is happy, and this is what it means to say, according to Kripke, that an expression is rigid.22 Why does it seem to Otto as if 'she' is rigid? As I have explained, Otto maintains a representation of the speaker's commitment slate, which upon my utterance of (32) might look as follows (I ignore the previous discourse because it was irrelevant to my utterance of (32) ):

342

Good News about the Description Theory of Names

1 1 WHAT I D I D N 'T WANT TO SAY

(3 5) If Mary has a car, it is pink. It doesn't require much inventiveness to think of examples like (3 5). Analogous examples with names are much more difficult to produce. The essential characteristic that names and pronouns share is that they are nearly always used to refer to an object that speaker and hearer take to be given (with names this tendency is even stronger than with pronouns). It is this pragmatic fact which accounts for the intuition of rigidity. Secondly, I didn't claim that names and pronouns appear to be rigid because they generally link up to anteceden� that are. I might have claimed this, because the main point that I want to establish in this paper is that the quotation theory is right, and this is a theory about the meaning of names. But I didn't. Thirdly, I didn't claim that names and pronouns are the only types of expressions that usually seem to be rigid. For example, there are many overt definites that behave exactly the same way, the most obvious class being semantically attenuate descriptions like 'the man', 'the thing', and so on. Also, I believe that the presuppositions associated with quantifiers like 'all' and 'most' are often rigid in the same sense in which names are. (36) Everybody is happy. Typically, this will be uttered in a situation in which the domain of 'everybody' is contextually given. Let c be a context in which (36) is uttered, and let A be the intended domain of 'everybody' in c. The proposition expressed by (36) in c is true in any given situation ( iff all individuals in A are happy in (. Hence; 'everybody' is rigid-or, better perhaps, its interpretation in c has a rigid component. On the other hand, I don't want to claim that all presuppositional expressions engender the unpression of rigidity. To the extent that we ·

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

It is always possible to misconstrue a story, no matter how simple, and I know from experience that my story about names is especially liable to be misunderstood, so let me briefly mention some of the things I didn't claim and don't want to claim, either. First, I don't want to claim that names are pronouns. This is how Sommers (1982) puts it, but not I. I do want to claim that names are very much like pronouns. They both are presuppositional expressions, and have the full complement of possible interpretations that presuppositional expressions generally allow for. But, of course, the descriptive content of a pronoun is of a different category than that of a name, which makes it much easier to construct examples of 'blocking' for pronouns than for names. For example:

Bart Geurts

343

have clear intuitions of rigidity at all they pertain to individuals and sets of individuals. Moreover, the intuition that an expression a is stronger when we feel that the descriptive content of a is inessential. It is for these reasons that we wouldn't want to say that factives are rigid, although qua presuppositional expressions they are quite similar to definite NPs. 12

C O N CL US I O N

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

At the outset I announced that I would try to rehabilitate Kneale's analysis of proper names. But didn't I actually replace it with something completely different? I don't think so. Kneale's crucial insight was that names are synonymous with definite NPs of the form 'the individual named so and so'. Of course, this is not yet a theory of names; but it has the merit of broadening the problem. If at this point Kneale had adopted Russell's theory of descriptions, his theory of names would already have been superior to the rigid-designator account, as I have tried to show. But he actually suggested a presuppositional analysis of definite NPs, and in doing so he broadened the problem even further, although this will not have been evident at the time. What I have done is combine Kneale's insight with a theory of presupposition that is more explicit and, presumably, more comprehensive than what he envisaged. The resulting analysis of names is superior to Kripke's for two main reasons. First, it explains a whole range of empirical facts about names that Kripke's theory cannot account for, if they don't falsify it to begin with. Secondly, it treats names as just another class of presuppositional expressions within the framework of a theory that is amply motivated on independent grounds. In comparison, Kripke's theory is clearly ad hoc. What I have outlined in the foregoing is a theory of interpretation in two senses of the word: it is about the mental processes involved in the interpretation of an utterance, and about the mental representations that they give rise to. As I understand it, this is a semantic theory, but it is obviously not what philosophers like Kripke have in mind when they speak of semantics, meaning, and related notions. It might be objected, therefore, that what I have proposed is a change of subject, rather than just an alternative account of names. In a way this is right. Discussions about names tend to get out of hand in the sense that it is impossible to make substantial claims about the semantics of names without making quite fundamental assumptions about the status and aims of semantic theory. (This is precisely why

344 Good News about the Description Theory of Names

names are such a controversial subject, although it is hard to think of a category of expressions that would appear to be simpler.) But Kripke's theory and mine are not incommensurable, since we start out from the same set of semantic intuitions. Our theories are about the same empirical facts, and if one of them provides a more satisfactory account of these facts, then this is bound to have consequences for our views on the status and aims of semantic theory; but these foundational issues must be left for another occasion. BART GEURTS University

of Osnabriick,

49069 Osnabriick

FB 7

Received: 1 s.o9.97 Final version received: 04.04.98

APPENDIX: EXERCISE To claim that a name N is semantically equivalent with a definite NP of the form 'the individual named N' is not to imply that the two expressions can always be used interchangeably. Of course, John' is not the same :is 'the individual named 'John"'. The latter expression is longer than the former, for example, which is presumably why many attempts at referring to John by means of 'the individual named 'John" ' will be found incongruous. Clearly, however, observations like these are the province of Gricean pragmatics. They do not even begin to prove that the said equivalence doesn't hold. Nevertheless, I believe that in quite a few cases the difference between N and 'the x named N', where x is a hyponym of 'individual', is very slight. In order to convey an impression of how similar the two types of expressions actually are, I have devised the following exercise. Below are eight text passages, collected on an Internet tour. These passages are lightly edited: I have occasionally replaced a name N with a description of the form 'the x named N', or vice versa. The purpose of the exercise is to determine which of the highlighted expressions have been tampered with and which haven't. The correct solution is given as note 23.'3 •

•

Should Mr Banharn become successful in his bid to form this country's next coalition government, there are still a number of obstacles to overcome. Certainly the defeat of Narong Wongwan in Phrae Province yesterday has to a certain extent removed one obstacle blocking his success. But he still has to convince the Bangkok public that he is suitable for the premiership. He would need to somehow bolster public confidence which appears certain to drop once the realisation sinks in that the man named Banharn is expected to be prime minister. In addition, he must answer the question of whether Mr Vattana will be appointed to his Cabinet The woman named Veronica had gone through the various treatments and services of the different physicians without finding any solution. She had spent all of her money and her health deteriorated to find her in a worse situation than she was before. When she heard ofJesus and pushed through the crowd to touch him, Jesus immediately knew that there was a withdrawal of virtue. He told her that her faith had made her whole.

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

Germany e- mail: [email protected]

Bart Geurts 345 •

•

•

•

•

they were looking for someplace to stay. He sat in his car, silently watching them and waiting to see where they went. He was doing a little of his own observation, on his own time. Scully had begun to check the data base for criminal records, Mulder suspected she wouldn't find anything. The women, as he watched them, didn't seem to know where they were going. The one named Vickie walked slowly behind the group, looking about curiously. It was almost as if she were comparing the city to something. Mulder smiled, they definitely weren't lying about not being from this city. The sound of his cell phone beeping was enough to bring Mulder out of his thoughts. 'Mulder.' he muttered, already knowing that it was his partner that was calling. 'I found something interesting.' Scully said and Mulder raised his eyebrows. 'Oh?' 'Yeah, the woman named Anik, she's got quite a lengthy record.' Scully said, the tone of her voice seemed to mock him. Anik, that name was interesting in itsel£ Mulder had never seen a more intriguing woman and that name just topped off the package. Jack is in fact a Nightbane, while he appears human, Jack can shed his human form for a demonic looking body, the following is how Jason describes his morphus . . . Everyone else in the room is nearly blinded by a bright gleam of black 'light.' When their vision clears a few seconds later, the boy namedJack is gone and a monster is in his place. The story so far: A man set out on a mystery date with I-74 with only the vague notion of buying a bowl of chili when he arrived. Instead he met a girl named John Ehrlichman who seduced and then kidnapped the man. She carjacked his '73 Cutlass and headed toward Yorba Linda with the promise that they and Nixon would 'party like it's 1 999.' They did and he did and they returned home to the Watergate Cafe. Unfortunately for them, Nixon died, the parties stopped and the Watergate Cafe fell into disuse. At the Watergate Cafe, where the coffee is always flowing, where the cottage cheese­ and-ketchup platter is always being served and where the tapes are always rolling, life was going on as a cheap game show parody of itsel£ The girl namedJohn Ehrlichman, whose structural perfection was matched only by her hostility, was on the telephone hatching another scheme. Most people know the man namedJohn Elefante. Some have known him since his days as lead vocalist with the rock group Kansas. Others know him as producer of numerous Grammy-winning projects for attists such as Christian rock veterans Petra. When Chambeau Blau died of equine fever in 1971 newspapers proclaimed the end of an era. Blau was the last and best of the horse hypnotists. At right we see Blau with the horse named �illy. Billy was mo.re mule than horse, but he was Blau's favorite subject. There was a grocery store. It could have been any grocery store, but it was bigger than most. It had twelve aisles, a produce section, a deli, a bakery, a liquor section, a magazine bin and a bulk candy section. It had a man named Val working the express lane cash register. We were in the grocery store, larger than most, for half an hour, when the man wearing a beach towel walked in. He went to the produce section, grabbed a bag of apples, a bag of oranges and then went to the bakery and grabbed a loaf of bread. Then he walked out of the larger than most grocery store. The man named Val followed him and asked, 'Why are you stealing from my grocery store?'

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

•

Fox Mulder hummed as he watched the nine ladies walk down the street. He guessed

346 Good News about the Description Theory of Names

NOTES 1 I should like to thank Rob van der Sandt

fact it isn't easy at all. See Geurts ( 1998) for discussion. 7 Presumably, this would be Kripke's position. C£ Kripke (1980: 62, n. 25). 8 Just suppose. I don't endorse Russell's analysis but I want to show that Kripke's criticism can be refuted without pre­ supposing the analysis of definite NPs that I favour. 9 This terminology is tendentious because only in the context of Russell's theory of descriptions is it necessary to con­ sider such NPs to be incomplete. I use the term here merely because it has gained some currency in the Russellian tradition. 10 This distinction is emphasized by Bach ( 1987), who proposes to speak of 'the bearer of N' instead of 'the individual named N', in order to forestall any further confusion. I am not convinced, however, that such terminological measures will prove to be very effective. 1 I Of course, Callas's real name was 'Kalogeropoulos'. I 2 Ottomobil is a continuation of the legendary Immobil project. I 3 See van der Sandt and Geurts (1991), van der Sandt (I992), Geurts (1995, to appear), and Geurts and van der Sandt (to appear). 14 Here and in the following I ignore the possibility that accommodation is not possible because the presupposition clashes with the Otto's beliefs (or because of any other reason). 1 5 See Geurts (1995, to appear). My notion of commitment slate derives from Hamblin (1971). Commitment is cru­ cially different from belief: a speaker who utters I{J accepts that I{J is true, as Stalnaker (1984) would say, in the sense that he commits himself to act, at least for a while, as if he held I{J to be true. Of course, in many cases speakers actually believe what they say, but sincerety is not a prerequisite for

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

and the three anonymous referees of the journal of Semantics for their uncom­ monly elaborate comments on the pre­ vious version of this paper. I have profited greatly from their criticism and advice, and I hope the paper has, too. During the final stage of my work­ ing on this project I was supported by a grant from the Deutsche Forschungs­ gemeinschaft (DFG), which I gratefully acknowledge. 2 Strawson (1950) refers to definites like 'the Glorious Revolution' and 'the Great War' as 'quasi-names' or 'embryonic names', and remarks that an expression may 'for obvious reasons, pass into, and out of, this class (e.g. "the Great War")'. . Although this observation is un­ doubtedly correct, I prefer to express it differently. Instead of postulating a separate class of 'quasi-names', I would say that the distinction between names and other defmites is vague, because the relation 'being named N' does not have a precise meaning (see section 4). 3 Here I am indebted to one ofthe referees, who suggested that this is not possible. 4 This is a familiar situation in pre­ supposition theory. It is widely held, for example, that in a sentence like the following the presupposition that there is a king of France, which is triggered by the definite NP in the consequent of the conditional, is absorbed in the antecedent: If France has a king, then the king of France is bald. Of course, this sentence is slightly awk­ ward, too, and for the same reason: there is no apparent reason why a full definite should be used instead of a pronoun. 5 Here I am indebted to one of the referees for suggesring that examples like (Sa, b) don't exist. 6 It may seem easy to distinguish 'meta­ linguistic' froin 'ordinary' uses, but in

·

Bart Geurts 3 47 successful communication, and it is not part of the notion of commitment.

Having said this, I will go on to speak in tenns of the speaker's beliefs, because this is often more convenient than speaking in terms of commitment, and nothing hinges upon the distinction for my current purposes.

r 6 See Lewis (1968, 1 97 1 , 1983b), Edelberg (1992), Geurts (1995, to appear), Zeevat (1996}. The technical apparatus that I

19 An additional complication is that the notion of scope may be applied at different levels of analysis. To see why this matters, consider the following example: (i) Maybe John has started smoking. Here the presupposition that John didn't smoke (before a given date} is triggered within the scope of the modal expression. In the absence of evidence to the contrary, this presupposition will be accommodated in the main DRS, and the default interpretation for (i) that we predict may be paraphrased as follows: (ii) John didn't smoke, but maybe he has started smoking.

which has obtained wide scope. It is only in the latter sense that the notions of scope taking and presupposition pro­ jection are orthogonal to each other. 20 As pointed out to . me by one of the referees for the Journal. 2 1 More accurately: it seems to me that the intuition underlying the answers given by these authors is the same in each case, although strictly speaking their proposals are incompatible with one another. 22 C£ Sommers (1982}. One of the referees objected that Kripke has never claimed that pronouns are rigid designators, or

that · they can be used as such. That is correct, but I am merely extending Kripke's observations to other types of

expressions. To the extentthat Kripke's is an empirical theory it is based on the observation that the interpretation of a name engenders certain intuitions. But other expressions (pronouns, certain

overt definites, and other presupposi­ tion-inducing expressions) engender the same intuitions, so instead of asking

why names appear to be rigid, it is a sound policy to ask why any expression may appear to rigid. 23 Solution: I lied when I said that I had edited these textS; they are printed here exactly as I found them.

REFERE NCES Bach, K. (198 1), 'What's in a name?'

Dynamic Semantics,

371-86. Bach, K. (1987), Thought and Reference, Oxford University Press, Oxford. Bakker, R. ( 1988), The Dinosaur Heresies,

Amsterdam. Cresswell, M (198 5),

Australasian journal of Philosophy,

59,

Penguin Books, Harmondsworth (ftrst

edition, 1986).

Beaver,

D. I. (1 993), What Comes First in

ILLC Prepublica­ tion Series LP-93 - 1 5, University of

Structured Meanings: The Semantics of Propositional Attitudes,

MIT Press, Cambridge, MA. Edelberg, W. ( 1992), 'Intentional identity and

the attitudes', 1 5 , S6I-96.

Linguistics and Philosophy,

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

can only sketch here is developed in much more detail in Geurts (1995, to appear). 17 See Geach (1967), Edelberg (1992), Geurts (to appear). 1 8 For more extensive discussion of this point, see Lewis (r983b).

Now logically speaking this means that we give wide scope to the presupposition that John didn't smoke, but linguisti­ cally speaking there is no expression

348 Good News about the Description Theory of Names Fodor, J. A. (I987), Psychosemantics: The

Problem of Meaning in the Philosophy of Mind, MIT Press, Cambridge, MA.

Geach, P.

T.

(I 967), 'Intentional identity',

journal of Philosophy, 64. 627-32. Geurts, B. {I99S). 'Presupposing', doctoral dissertation, University of Stuttgart. Geurts, B. (I998), 'The mechanisms of denial', Language, 74, 274-307. Geurts, B. (to appear), 'Presuppositions and anaphors in attitude contexts', m ·

Linguistics and Philosophy.

Focus: Linguistic, Cognitive, and Computa­ tional Perspectives, Cambridge University

Press, Cambridge. Hamblin, C. L. (I97I), 'Mathematical models of dialogue', Theoria, 37, I 30-5 S· Kamp, H. & Reyle, U. (I993), From Discourse to Logic; Kluwer, Dordrecht. Kaplan, D. (I989), 'Demonstratives: an essay on the semantics, logic, metaphysics, and epistemology of demonstratives and other indexicals', · in J. Almog, J. Perry, & H. Wettstein (eds), Themes from Kaplan. Oxford University Press, Oxford, 48 I-563. Kneale, W. (I962), 'Modality de dicto and de re', in E. Nagel, P. Suppes, & A. Tarski (eds), Lagic, Methodology and

Philosophy of Science. Proceedings of the 1960 International Congress, Stanford

University Press, Stanford, 622-3 3 . Kripke, S. A. (I977), 'Speaker's reference and semantic reference', in P. A. French, T. E. Uehling, & H. K. Wettstein (eds), Contemporary Perspectives in the Philosophy of Language, University of Minnesota Press, Minneapolis. Kripke, S. A. (I98o), Naming and Necessity, ·

·

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

Geurts, B. & Sandt, RA. van der (to appear), 'Domain restriction' in P. Bosch and R. A. van der Sandt (eds),

Blackwell, Oxford. First published in D. Davidson & G. Harman (eds), Semantics for Natural Language, Reidel, Dordrecht {I972). Lewis, D. (I968), 'Counterpart theory and quantified modal logic', journal of Philo­ sophy, 65, I 1 3-26. Reprinted in: Lewis (I983a), 26-39. Lewis, D. (I971), 'Counterparts of persons and their bodies', journal of Philosophy, 68, 203-1 1. Reprinted in: Lewis (1983a), 47-54· Lewis, D. (I 98o), 'Index, context, and content', in: S. Kanger & S. Ohman (eds), Philosophy and Grammar, Reidel, Dordrecht, 79- 100. Lewis, D. (198 3a), Philosophical Papers, Vol. 1, Oxford University Press, Oxford. Lewis, D. (198 3b), 'Postscripts to "Counter­ part theory and quantified modal logic" ', in Lewis (1983a), 3 9-46. Loar, B. (1976), 'The semantics of sin­ gular terms', Philosophical Studies, 30, 3 5 3-77Sandt, R.A. van der (I 992), 'Presupposition projection as anaphora resolution', journal of Semantics, 9, 3 3 3-77. Sandt, RA van der & Geurts, B. (1991), 'Presupposition, anaphora, and lexical content', in: 0. Herzog & C.-R. Rollinger (eds), Text Understanding in LILOG, Springer-Verlag, Berlin, 2 s 9-96. Sommers, F. (1982), The Lagic of Natural Language, Clarendon Press, Oxford. Stalnaker, R. C. (I984), Inquiry, MIT Press, Cambridge, MA. Strawson, P. (1 950), 'On referring', Mind, 59, 32D-44· Zeevat, H. (I996), 'A neoclassical analysis of belief sentences', Proceedings of the 1oth Amsterdam Colloquium, ILLC, University of Amsterdam, 723-42.

Journal of&mantics 1 4'

© Oxford University Press 1 997

349-3 67

Negative Adverbials, Prototypical Negation and the De Morgan Taxonomy JAY DAV I D ATLAS Pomona College, Claremont

Abstract

I I NTRODUCTION Zwarts (I986, I996, I998) has suggested a taxonomy for negative Generalized Quantifier Noun Phrases Q. The taxonomy for a language that has sentential connectives conjunction '&' and disjunction 'V' as well as phrasal connectives 'and[p] ' :md 'or[p]' makes use of a generalization of the classical De Morgan relationships for sentential negation in the form: (I)

a.

b.

Q (F or G) I� QF & QG QF & QG I� Q (F or G)

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

Gamut ( � 991) and Atlas (1991, 1993, 1996b) showed that the Generalized Quantifier lonly Proper Name l licenses Negative Polarity Items but fails to be downwards monotonic in Barwise & Cooper's (1981) sense. In Atlas (1996a, in press) I examined Zwarts's {1996, 1998) De Morgan taxonomy for negative No�n Phrases. Two ofthe four De Morgan entailments used by Zwarts to characterize the negation of negative Noun Phrases express downward monotonicity of the Noun Phrase Q, viz. Q{F or G) I� QF & QG, and QF V QG I� Q(F and G). I introduced a connection between a revised Zwartsian De Morgan taxonomy and the familiar concept in Cognitive Linguistics, due to Rosch { 1978), and discussed by Lakoff( 1 987) and Taylor (1995), of 'prototypicality'. I argued on philosophical grounds that anti-additive quantifiers were logically 'prototypical' negation operators; anti-additive quantifiers are downwards monotonic and closed under disjunction, viz. they satisfy a third De Morgan relationship QF & QG I� Q{F or G). I then hypothesized that the logically prototypical operators were realized· in natural language by linguistically and psycho-linguistically prototypical expressions; e.g. I no Count Noun I is 'prototypical', but lonly Proper Namel is not; the latter is not downwards monotonic, though it is closed under disjunction, and it licenses Negative Polarity Items. Thus the questions were raised: (a) Are there other negative expressions in the same logical class as I only Proper Name I? (b) Is any one logical De Morgan condition essential for Noun Phrase and Verb Phrase lexically semantic negativity? In this essay I answer those two questions: (a') Yes. (b') No. The latter answer explains the theoretical importance of a concept like 'prototypical' negation in logical semantics, a concept like 'prototypical' negativity in lexical semantics, and the Hypothesis that claims that logically prototypical quantifiers are realized in natural language by linguistically prototypical negative Noun Phrases {Atlas 1996a, in press).

3 so

Negative Adverbials, Prototypical Negation, and the De Morgan Taxonomy

(2)

Q (F and G) II- QF V QG b. QF V QG II- Q (F and G) a.

(a) No single De Morgan condition is necessary for the negativity of Noun Phrases or Verb Phrases. (b) Corollary. Neither form of Downward Monotonicity, (1a) or (2b), is a necessary condition for negativity. (c) There are other pseudo-mini�al operators besides lonly Proper Name I. The NP hardly anyone is merely downwards monotonic, but the adverbial VP !hardly pl is pseudo-minimal, like the NP lonly Proper

·

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

where the phrasal 'or' has at least the interpretation of set-theoretic union of the extensions of the predicates F and G, and the phrasal 'and' has only the interpretation of set-theoretic intersection of the extensions of the predicates F and G. Zwarts observed that a quantifier that was downwards monotonic in Barwise & Cooper's (1981) sense would satisfy (�a) and (2b), e.g. IFew Nl.' A quantifier that satisfied only (�a, 2b) was called by Zwarts (I996) 'subminimal'. Zwarts observed that I No Nl satisfies (1a, b); such quantifiers are called 'anti-additive', and Zwarts (1996) calls anti-additive quantifiers 'minimal negations'. The Noun Phrase I Not all Nl satisfies (2a, b); such quantifiers are called 'anti-multiplicative', and Atlas (1996b) called Noun Phrases that were anti-multiplicative quantifiers 'outer negations'. Atlas (I991, I99J, 1 996b) observed that the quantifier lonly Proper Namel was not downwards monotonic, since I rejected the analysis I no one other than Proper Namel that is downwards monotonic. My analysis, viz. !Exactly one individual, who is at most Proper Namel satisfies only (1b)-closure under disjunction. Because lonly Proper Namel had seemed to be a minimal (i.e. anti-additive) quantifier to many theorists, I called the class of quantifiers that satisfied only (1b) 'pseudo-minimal' or 'pseudo-anti-additive'. I also observed that lonly Count Nounl was, by contrast, minimal (anti-additive) as expected.2 I (Atlas 1996a, in press) had argued that minimal (anti-additive) negative NPs were 'prototypical' of negative NPs, in the sense of Rosch (1978), Lakoff (1987), and Taylor (1995). Those arguments had raised the question whether merely downwards monotonic NPs had a special claim on being called 'negative'; indeed they raised the question whether any one of the four De Morgan relations had a particular claim to providing an essential criterion for negativity in a language containing '&', 'V', 'and', and 'or'. I saw no way to decide this matter until the work of Klein (1997, 1 998), and her discussion of Abraham & Akkerman (1995), Abraham & Akkerman (I 996), and Akkerman & Abraham ( I 996), suggested to me an examination of negative adverbials. In this essay I shall defend the following claims:

Jay David Atlas 3 S I

Namel, and so it is not even downwards monotonic, though it provides an environment for Negative Polarity Items.

2 HARDL Y VERB PHRASE

(3) Hardly anyone shared Tom's taste for the film 'The Texas Chain-Saw Massacre'. and at least two uses of hardly in adverbial VPs: (4} a. b. c. d.

It is hardly surprising. He's hardly a linguist. I He's hardly a genius. I I'm hardly thrilled with you. It hardly rained. It is hardly conceivable that she married him.

A referee of this essay for this journal suggested that (4b) exemplified an ironic use, by contrast with (4c), which permits some sort of inference to (sc), and by contrast with (4d), which permits an inference to (sd) rather than to (se) {compare the non-metalinguistic (4a)/(sa) ): (s)

a.

b. c. d. e.

It is certainly not surprising. He's definitely not a genius. It only just rained./ It rained just, but at most Uust, a bit, slt;ghtlyJ. 'Conceivable' is inappropriately applied to the proposition that she married him; it overstates the case. It is · not really, completely, or absolutely conceivable that she married him.

The use in (4d) is obviously metalinguistic (see Horn 1985, 1 989). The referee points out that one might subsume the ironic and metalinguistic uses under one category if, as suggested by Robyn Carston (1996), both ironic and metalinguistic uses are fundamentally 'echoic'. I could probably work up more enthusiasm for this view had I not criticized the adequacy of the 'irony as echoic mention' view of Sperber & Wilson (198 1), and the Jorgensen, Miller, & Sperber (1984) psycho-linguistic defense of it, in Atlas (1986).

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

There is a class of adverbials in English that are 'negative' in the sense of licensing Negative Polarity Items but whose semantic properties have not been adequately explored. I have in mind little, hardly, barely, and scarcely. Klein {1997: 1 8, 1998), following van Os (1988: 235, 1989) calls this class, e.g. Dutch {weimg, nauwelijks} and German {went;g, kaumJ, 'quasi-negatives'. In English one notes uses of hardly in quantified NPs:

3 52

Negative Adverbials, Prototypical Negation, and the De Morgan Taxonomy

(6) I hardly VPI is polysemous; it means (a) lyp just, but {at most just VP, nearly not VP}',3 (b) !certainly not VPI. I shall focus on meaning (6a) in the rest of this essay. We have already noted the intuitive fact of some sort of inference from (4c) It hardly rained to It rained , since It hardly rained seems to mean (sc) It only just rained, or, alternatively, It rained only a bit. The symbol 'II > ' is a symbol for a generic inference whose semantic and/or pragmatic character is left open:

(7) It hardly rained I I > It rained. Horn (I996: 19-23), while defending his view that lonly Nl was a downwards monotonic operator, observed the following inferences and NPI-licensing facts about barely in (8) and (9): (8) a. The Celtics almost won the game I I > The Celtics did not win the game. b. The Celtics almost won the game ty> The Celtics won the game. c. The Celtics barely won the game II > · The Celtics won the game. d. The Celtics barely won the game ty> The Celtics didn't win the game. (9) # He almost {budged / slept a wink I touched a drop I spoke to anyone}. b. He barely {budged I slept a wink I touched a drop I spoke to anyone}. a.

Horn claims that the data in (8) show that almost is more negative than barely, since almost permits an inference to a negative sentence while barely only permits an inference to the correlative affirmative sentence, even though his data i� (9) show that barely licenses 'minimizer' NPis (Klooster I994, I995, Atlas I 996b) while almost does not. Further Horn (I 996: 20- I) argues that almost is not downwards monotonic but that barely is, sort of, downwards monotonic. Horn observes that almost is not

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

One difficulty with ironic interpretations of (4b) He's hardly a linguist is their notable multiplicity. They might include He's not at all a linguist but also include He's the compleat linguist. These ironic interpretations indicate that in the former case one's non-ironic base interpretation of (4b) He's hardly a linguist paralleled the interpretation (sc) It only just rained of (4c) It hardly rained, and in the latter case one's non-ironic base interpretation of (4b) He's hardly a linguist paralleled the interpretation (sa) It is certainly not surprising of (4a) It's hardly surprising. This multiplicity of ironic inter­ pretation suggests a multiplicity of non-ironic base senses, viz. that I hardly VPI is polysemous between lonly just ypl and !certainly not VPI. Some lexicographers also seem inclined to such an ambiguity view, e.g. those responsible for Webster's New Collegiate Dictionary (I98I: S I 8), who distinguish between lonly just ypl and !certainly not VPI. I shall adopt such an ambiguity view:

Jay David Atlas

353

downwards entailing, since 'if he's alm�st alive, it doesn't follow that he's almost ambulatory', but that barely is 'quasi' downwards entailing, since 'if he's barely alive, he barely ambulatory, ignoring the case in which he's not ambulatory at all'. Horn (personal communication, 1998) elaborates:

( 1 0) The reason I call this quasi-downward entailment is of course that we

I myself had so much difficulty deciding how to analyze my own linguistic judgments of I hardly VPI and I barely VPI sentences, I sympathize with Horn's uncertainty. But I want to take Horn's original observation seriously; I believe that he has shown that I barely VP I is not strictly downwards entailing, and I added to his observations in (8)-(9) the following argument:

(n) If barely is downwards entailing, john is barely 61 o11 tall I� john is barely [61 o" tall and blond]. That is, in every valuation (possible world) in which the first is true, the second is true. Now, consider a possible world in which all blonds are 61 111 tall or taller, and John is barely 61 o" tall. In that world no one is barely [6'o" tall and blond]. So it is false that every world in which John is barely 6'o" tall is also a world in which John is barely [61o" tall and blond]. Hence john is barely 61 o" tall does not entail john is barely [61 o" tall and blond].

Upon the argument above Horn (personal communication, 1998) commented: (12) Your example is even more blatant [than that someone who is entirely naked is barely dressed], in that the bullet I'd have to bite is the one that would have me defending the (technical) truth of the claim that someone who has dark hair and is [barely] 61 tall is indeed 'barely [61 tall and blond].' Implausible, but no more than many other things each of us [Atlas and Horn] has swallowed before breakfast.

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

don't normally call someone who is entirely paralyzed 'barely ambulatory'. Nor do we say of someone who studied a little phonology and no syntax that she has barely studied syntax. In each case, the claim that 'barely' is downwards entailing amounts to the commitment to saying that the sentences one doesn't normally say are nevertheless true. If I'm right in supposing that one might say 'Th� patients in this ward are (all) barely ambulatory, and three of them are completely paralyzed', or 'The students in the intro class have barely studied syntax, and three of them haven't had any syntax at all', I might in fact · argue that this is in fact downwards entailment in the . strict sense. I'm just not sure whether these are my judgments.

3 54 Negative Adverbials, Prototypical Negation, and the De Morgan Taxonomy

.To my previous argument in ( n ) I would now add observations on instances of De Morgan's entailment claims (2b) and (�a), which, if I barely VP I and I hardly VP I were downwards entailing, would be correct, contrary to fact: (13) Fred [hardly, barely) managed to pass his language examination lfFred {hardly, barely} managed to get a job in Oakland lJL Fred {hardly, barely} [managed to pass his language examination and managed to get a job in Oakland].

(I4) Fred {hardly, barely} [managed to pass his language examination or managed to get a job in Oakland] lJL Fred {hardly, barely} managed to pass his language examination &: Fred {hardly, barely) managed to get a job in Oakland.4 The evidence of (I3) and (I4) suggests that !barely VP I and !hardly VP I are not downwards monotonic. In fact !hardly I barely VP I is pseudo-anti­ additive and fails to be anti-multiplicative, because it satisfies instances of (I h) and fails to satisfy instances of (2a) as shown in ( I S) and (I6): (I S) Fred {hardly, barely} got his Master's degree &: Fred {hardly, barely} paid his rent I � Fred {hardly, barely} [got his Master's degree or paid his rent]. ( I6) Fred {hardly, barely} [patted his head (at t0) and rubbed his stomach (at to)] lJL Fred {hardly, barely} patted his head· (at t0) V Fred {hardly, barely} rubbed his stomach (at t0).5 Thus !barely VPI and !hardly VP I are pseudo-anti-additive operators like lonly Proper Namel. They are not downwards monotonic. If I barely VP I and I hardly VPI are similar to I only Proper Name I, do they also support an entailment claim like (I7b) (Atlas I99I, I993, I996b)? (17)

a. Only Fred can stand Bert I� Fred can stand Bert. b. Fred can hardly stand Bert �� Fred can stand Bert.

Horn (I996: I9) explicitly left it open what kind of inferential relation I I > existed in (8c)/(17b). Hom (I 996: 2 I) reported claims of Ducrot (I 973) and Fillmore, Kay & O'Connor (1988: 529) respectively that a peine P is 'virtually identical to presque non-P' and that there is independent evidence that 'barely may be analyzed semantically as "almost not", but space does not permit reviewing it . . .' The intuitive paraphrases for I hardly VP I I offered earlier, viz. (6a) I only just VPI and I VP but only somewhat I a bitl were reports of my semantic

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

The falsity of the instance of the entailment claim (2b) in (I3) is a particularly clear refutation of the downwards mono tonicity of I hardly ypl and I barely VP I. The failure of the other De Morgan condition (Ia) for downward entailment is less obvious:

Jay David Atlas 3 55

intuitions, backed by a lexicographical authority. I would not have claimed that !hardly VPI was Sy-nonymous with !almost not VPI; rather there are, I suggest, the following true entailment and logical equivalence claims:

(18)

Q hardly VP It- Q [almost, nearly} not VP b. Q hardly VP -1t- Q VP & Q {almost, nearly} not VP.6 a.

(19)

'Tom almost swam the English Channel in record time in his Bjorn Borg swimsuit' I�» Tom did not quite swim the English Channel in record time in his Bjorn Borg swimsuit. b. For the achievement predicate 'swam the English Channel in record time . . . , Tom did not quite swim the English Channel in record time in his Bjorn Borg swimsuit I� Tom did not swim the English Channel . . . c. So, by composition of the implicatural and entailment relations for achievement and accomplishment predicates, 'Tom almost swam the English Channel . . . ' I� o It-» Tom did not swim the English · .7 Channel a.

'

.

.

.

So,. (2o)

a.

asserting Tom {barely, hardly} swam the English Channel in record time in his Bjorn Borg swimsuit,

on Ducrot's (1973) hypothesis becomes: b. asserting Tom almost did not swim the English Channel in record time in his Bjorn Borg swimsuit. But, c. 'Tom almost did not swim the English Channel in record time in his Bjorn Borg swimsuit' It-» Tom didn't quite not swim the English Channel in record time . . .

For an accomplishment or achievement predicate, or lack thereof as in this case, where the 'lack' predicates have the same logical properties as the positive predicates, d.

Tom didn't quite not swim the English Channel . . . It- Tom didn't not swim the English Channel . . .

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

Suppose, for the moment, that Ducrot (1973) were correct, that !hardly p i means !almost not Pl. How would one explain the inference Hom accepts from lQ {hardly, barely} VPI to lQ VPI? That is, how, on Ducrot's hypothesis, would one explain the inference from lQ almost not VPI to lQ VPI? (I just happen to have a little theory handy.) On the analysis in Atlas {1984):

3 56 Negative Adverbials, Prototypical Negation, and the De Morgan Taxonomy

from which classical Double Negation Elimination yields: e. Tom didn't not swim . . . I� Tom swam . . . . So, £ 'Tom (hardly, barely) swam the English Channel in record time in his

·

'Q {barely, hardly}

w· 1� o 1� » Qw.

The conclusion (2of)/(21d) is the result that Horn (1996: 19) wanted in (8c): (8) c. The Celtics barely won the game II> The Celtics won the game, with Horn's (1996) indeterminate inference relation I I > replaced by Atlas's (1984) composition of the implicature and entailment relations I� o I�» for achievement and accomplishment predicates W, like Horn's (1996: 19) won the game. It is important to notice that this result (22): (22) For accomplishment and achievement predicates w,

Q {hardly, barely} W I� o I�» QW. is restricted to a special class of predicates and that Qw is not strictly entailed by Q {hardly, barely} W even for those predicates. The inferential relation is two-staged: an implicatum, and one of its direct entailments (Atlas 1991: 1 37). Since we want the inference: (23) It hardly rained II > It rained, where 'rained' is neither an achievement nor an accomplishment predicate but an activity predicate, and Ducrot's (1973) paraphrase of {hardly, barely} by almost not will not, according to the theory of Atlas (1984), produce the inference for non-accomplishment and non-achievement predicates, I must

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

Bjorn Borg swimsuit' I� o I�» Tom swam the English Channel in record time . . . In general, then, on the analysis in Atlas (1984), on Ducrot's (1973) hypothesis: (21) a. asserting lQ almost VP I I�» (i.e. implicates) lQ not quite VPI; b. for accomplishment and achievement predicates W (Vendler 1967, Ryle 1949: J I), Q not quite W I� Q not W; c. so, by composition of the implicatural and entailment relations for achievement and accomplishment predicates w, 'Q almost w ' I� o I�» Q not W. d. If, as Ducrot suggests, lQ {barely, hardly} VPI means lQ almost not VPl, and l not w l is a failure predicate (where failure is semantically an achievement or accomplishment), and we assume classical Double Negation Elimination, then:

Jay David Atlas

3 57

amend Ducrot's {I973) paraphrase in two respects. First, I shall merely claim a logical equivalence, not synonymy, and second, I shall take the cheap way out, and simply conjoin the desired predicate (having failed to find an alternative logical form that is motivated by a more principled semantical analysis): (24) Q hardly VP --+-- QVP & Q almost not VP. Apart from what Bertrand Russell {I9I 8) once referred to as the advantages of theft over honest toil, can the hypothesized logical equivalence in (24) be justified by the linguistic predictions it makes? For example, is it the case that: The usual intuitive appeal to the supposition of the truth of the left-hand side and the falsity of the right-hand side, as in (26a): (26)

a. It hardly rained, and it didn't rain. b. ?It hardly rained, {and, and in foct, but} it didn;t rain.

does indeed yield what seems to be an inconsistent conjunction. In addition, the sentence It rained is not cancelable in (26b), so the relation in (23) does not meet a necessary condition for being a conversational implicature.8 Ducrot (1973) himself took It rained to be a presupposition of It hardly rained/ It almost did not rain, but Atlas & Levinson (I 98 I) argue that presupposition is best treated as an entailment of the statement and as a generalized conversational implicatum of its main verb negation (see also Boer & Lycan {I976)). One could not take seriously the idea that It rained could be a conventional implicatum, in Grice's {I975) original sense, of It hardly rained. Since these exhaust the obvious alternatives, I shall pursue the entailment analysis without claiming that I have uncovered the most explanatory underlying logical form but merely claiming that I have accurately expressed the truth-conditions. The adequacy conditions on any such analysis are, first, that the entailment in (23) is accounted for, and second, that lQ hardly/ barely fl is seen to be non�downwards monotonic on F,just like lonly Proper Name I. The conjunction in (24) gives just these results. The entailment of lQfl is obvious. The conjunction in (24), when Q is upwards monotonic, consists of a conjunction of an upwards and a downwards monotonic operator on F, and when Q is downwards monotonic, consists of a conjunction of a downwards monotonic and an upwards monotonic operator on F. So the conjunction is, in general, non-monotonic (see Gamut I991: 232-42) ). It is (1b) and (2b) that Atlas (1996a, in press) argues are core or 'basic' De Morgan properties of semantically negative expressions in a language with

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

(25) It hardly rained If- It rained?

358

Negative Adverbials, Prototypical Negation, and the De Morgan Taxonomy

a natural language representation of the meanings of '&', 'V', 'and[pj '. and 'or[p( The adverbial l hardly fl satisfies the closure under disjunction condition ( I b) required for being a 'prototypical' negative operator, but since it is not downwards monotonic it fails to be 'prototypical', and on the standard view is not a semantical negation. The traditional criterion for being a negation in the semantical accounts of the grammar of Negative Polarity Items was mere Downward Monotonicity, the De Morgan form of which is (2b), or alternatively, (1a) (Ladusaw 1980, Linebarger 1981, 1987, 1991, Zwarts 1996, 1998, Hom 1992, 1 996).

HARDL Y A NYONE

By contrast with the adverbial phrase lhardly Fl, the Noun Phrase hardly anyone seems downwards monotonic in the basic Barwise & Cooper (1981) sense. For example, Hardly anyone smokes is true if and only if a small enough subset of the domain of quantification smokes. But that seems to entail that the same or smaller subset smokes in the Netherlands, i.e. entails the truth of Hardly anyone smokes in the Netherlands [(27a)]. Likewise the entailment claim (27b) seems true, and the instance (27c) of the De Morgan condition (2b) is true. (27) a. Hardly anyone smokes �� Hardly anyone smokes in the Netherlands. b. Hardly anyone smokes I� Hardly anyone smokes and drinks. c. Hardly anyone smokes V Hardly anyone drinks I� Hardly anyone both smokes and drinks. But the instances (28) of pseudo-anti-additivity (1b) fail: (28)

Hardly anyone smokes & Hardly anyone drinks lJL Hardly anyone smokes or drinks. b. Almost no one smokes & Almost no one drinks lJL Almost no one smokes or drinks. a.

For example, though (29a) could be true, simultaneously (29b) need not be true: (29)

Hardly anyone P1 & Hardly anyone Pz & . . . & Hardly anyone P0• b. Hardly anyone P1 or Pz or . . . or P0• a.

Since Pi can partition the Domain of Quantification, unless the Domain is itselfjust a few individuals at most, (29b) will be false even though (29a) is true. As expected, hardly anyone satisfies the downward monotonicity conditions (�a) and (2b):

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

3

Jay David Atlas 3 59

(3o) a. Hardly anyone smokes or drinks I� Hardly anyone smokes & Hardly anyone drinks.9 b. Hardly anyone smokes V Hardly anyone drinks �� Hardly anyone smokes and drinks.

(3 I)

Hardly anyone came to the party-{in fact, because} no one came. b. Few individuals came to the party-{in fact, because} no one came. a.

So, (32) Hardly anyone Fs lJL Someone Fs. By analogue .with the analysis of lalmost which:

fl

in Atlas (I984), according to

(33) a. 'almost F' I�» not quite F, b. for achievement and accomplishments predicates G, not quite G I� not G, c. so, for achievement and accomplishment predicates G, 'almost G' I� o I�» not G, the analysis for hardly anyone would be (34): (34)

'almost no one Fs' I�» not quite no one Fs; b. for an 'absolute' or 'endpoint' NP, e.g. no one, not quite no one Fs �� not no one Fs;'0 c. not no one Fs I� someone Fs; d. so, 'almost no one Fs' I� o I�» someone Fs. a.

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

Thus even though on the standard view hardly anyone is a semantically negative NP, because it is downwards monotonic, it is not in my sense a 'prototypical' negation, because it fails to be closed under disjunction [(28)]. By contrast, lhardly VPI is closed under disjunction, but it is not downwards monotonic. It is neither semantically negative on the standard view nor a 'prototypical' negation. The only way to avoid missing the semantical generalization about hardly itself is by appealing to the concept of 'prototypical' negation (Atlas I996a, in press): neither hardly anyone nor lhardly VPI is a 'prototypical' negation. It has been claimed, e.g. in Webster's New Collegiate Dictionary (I 98 I: 5 I 8), that hardly ever just means almost never, or very seldom. Since seldom means, roughly, at Jew times, and !Jew Nl is well known to be a merely downwards monotonic generalized quantifier, the observations above on the properties of hardly anyone, i.e. of very Jew individuals, are to be expected. In fact, the relation between hardly anyone and someone seems not to be an entailment but an implicature, as the instances of denying someone in (J ia, b) are consistent, though some speakers might find the sentences slightly odd:

360 Negative Adverbials, Prototypical Negation, and the De Morgan Taxonomy

Thus on this extension (34) of the analysis of Atlas (I984), it is predictable that (32) Hardly anyone Fs UL Someone Fs but that there is a 'strong' implicature from asserting IHardly anyone Fs l to !Someone Fs l. Even though (3Ia, b) are consistent, and so !Someone Fs l is cancelable, the slight oddity some speakers find in the cancellation in (3 I a, b) is attributable to the presence of a generalized conversational, scalar implicature. That the implicature is a generalized one, which is a default implicature, and is a scalar one, which relies on a highly structured lexical field, tends to make the inference psychologically salient and the cancellation correspondingly more difficult, though the sentences are not unacceptable or inconsistent.

Among the De Morgan entailments (1a, b) and (2a, b) below: (I) a. b. (2) a. b.

Q (F or G) I� QF & QG QF & QG I� Q (F or G) Q (F and G) I� QF V QG QF V QG I� Q (F and G)

I shall refer to condition (Ib) as Pseudo-Anti-Additivity (PAA), condition (2b) as 'prototypical' Downward Monotonicity (PDM), and condition (�a) as Mere Downward Monotonicity (MDM) (Atlas I996a, in press). The Zwartsian anti-additive NPs are the ones that satisfy conditions (�a, b); the Zwartsian anti-multiplicative NPs are the ones that satisfy conditions (2a, b), and the 'prototypical' negations are the ones that satisfy conditions (Ib) and (2b)." We can then taxonomize generalized quantifier NPs and Adverbial VPs in the following Table (3 s): (3 5) GENERALIZED

l__ Prototypical _j L_ Anti-Additive _j

QUANTIFIER NOUN

PDM

PAA

MDM

PHRASES no Count Noun none of the Count Noun only Count Noun at most Count Noun at most Proper Name few Count Noun

(2b) I I

(I b) I I I I I

(�a) I I I I I I

I I

0

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

4 A DE M O RGAN TAXO NOMY O F NPS AND ADVERBIAL VPS

Jay David Atlas

·

hardly anyone not all Count Noun only Proper Name some but not all Count Noun

ADVERBIAL VERB PHRASES

0

0 0

0 I 0

I I

I

I

I

0

o'2

0

I 0 I I

I

I 0 0

I

I I 0 0 0

5 I A T MOST VERB PHRASE I AND THE D ILEMMA Notice that by the scale for decreasing strength of negation in the Table (3 5): PROTOTYPICAL > DOWNWARDS MONOTONIC > PSEUDO­ MINIMAL, the NP hardly anyone is of the same strength of negation as I not all CN I but less . strong than lat most CNI, lat most Proper Name I and stronger than lonly Proper Name l.'3 The adverbial lhardly yp l has the same De Morgan properties as lat most VP I and l only VPI. It is also notable that lat most CN I and lat most VP I are divergent in their De Morgan profiles, like hardly anyone and I hardly VP I, with I at most CN I more negated than lat most ypl and hardly anyone more negated than lhardly VP I. On my account l not all CNI and hardly anyone are not 'prototypical' negations. In order to consider these NPs as negations, we should have to follow the now standard suggestion that the downwards monotonic De Morgan condition (2b) and the downwards monotonic condition (1a) be the semantical conditions individually necessary for negation. But lat most yp l fails to satisfy ( 1a) Mere Downward Monotonicity. If ( 1a) were to be a semantical condition necessary for negation, lat most VPI would not be a negation. That, I believe, would be a highly counter-intuitive result.'4 I At most VPI satisfies only pseudo-anti-additivity ( I b), like I hardly VP I, lonly VPI, and lonly Proper Name l. If closure under disjunction (Ib) were a semantical condition necessary for negation, l not all CNI, lfew CNI, and hardly anyone would not be negations. Surely that would also be a counter-intuitive result. If any other one of the four De Morgan conditions were necessary for negation, the upper-bounding lat most yp l would not be a negation. That again would be strongly counter-intuitive. So this is the De Morgan Theorist's Dilemma: any one of the four De Morgan conditions leads to

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

not Verb Phrase never Verb Phrase seldom Verb Phrase hardly Verb Phrase at most Verb Phrase only Verb Phrase

361

362 Negative Adverbials, Prototypical Negation, and the De Morgan Taxonomy

Acknowledgement For discussion of prototypical negative Noun Phrases, see Atlas (1996a, in press}. I am grateful to Peter Bosch, Penny Brown, Jack Hoeksema, Henny Klein, Sjaak de Mey, Thomas J. Rankin, Frans Zwarts, Pim Levelt, Stephen Levinson, and the staff of the Max Planck Institut fiir Psycholinguistik, Nijmegen, The Netherlands. I am particularly grateful to the anonymous referee for this journal for his or her most helpful comments. JAY DAVID ATLAS Department of Philosophy Program in Linguistics Pomona College, Claremont 55 1 North College Avenue Claremont, CA 91711-6355, USA e-mail: [email protected]

Received: o6.10.97 Final version received: 20.07.98

N OTES I

The symbols 'I' and ' I• are left and right quasi-quotation marks, also known as corner quotes, used by Quine (198 1: 3 3-7). I Q VPI is true just in case the extension II VPII of VP belongs to the

extension II Q II of Q, i.e. II VPII E IIQII. The monotonically increasing quanti­ fiers are ones that contain every superset of every set that is an element of the quantifier, i.e. if II VP, II � IIVP� II and

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

linguistically counter-intuitive consequences if chosen as a necessary condition for negation. CONCLUSION: No single De Morgan condition is necessary for Noun Phrases or adverbial Verb Phrases to be negations. COROLLARY: Neither form of Downward Monotonicity, (�a) or (2b), is a necessary condition for negation. Hence, if a phrase licenses the occurrence of a Negative Polarity Item,' it does not follow that the licensing phrase is downwards monotonic.' 5 My hypothesis {Atlas 1996a, in press) that the logically 'prototypical' De Morgan negations-downwards monotonic (2b) and closed under disjunction (Ib)­ are expressed in a natural language by linguistically 'prototypically' negative Noun and adverbial Verb Phrases tells us less than we want to know about the negativity of Noun Phrases and Verb Phrases (e.g. why the logical concept of downward monotonicity is not more linguistically explanatory), but it tells us (a) that the anti-additive negation operators are expressed by the linguistically and psycho-linguistically 'prototypical' negative NPs and adverbial VPs and (b) that the logical concept of 'prototypical' negation saves a linguistic generalization for hardly, barely, scarcely, et al., viz. that both the NPs and the adverbial VPs containing hardly, barely, scarcely, et al. are not 'prototypical' negations.

Jay David Atlas 363 s

I owe this · nice example, an improve­ ment on my original one, to a discussion with Thomas J. Rankin. 6 For van Os (I988, I989) and Klein {I997, I 998) almost is an 'approximative', an adverb that indicates closeness to an endpoint, e.g. almost ready, nearly ready, virtually ready. Van Os {I988: 23 5} and Klein ( I997: I 8, 20) classify hardly and barely in their category of 'quasi­ negatives' and make no claim of the sort made by Ducrot {I973), Fillmore, Kay, & O'Connor {I988), Horn (I996), and . myself that there is a semantical rela­ tionship between their two categories. This is remarkable because Klein (I997: 9 3) vaguely gestures at a relationship when she writes that the 'quasi-negatives mirror the characteristics of the quasi­ congruent approximatives', without making it clear what she meant by 'mirror'. She {I997: 94) does claim that adverbs of both categories are Positive Polarity Items. (Quasi-congruency was a notion introduced by Vandeweghe {I988} for adverbs modifying an absolute predicate in assertions that implicated that the difference between the denotation of the modified pre­ dicate and the denotation of the unmodified predicate was so small as to be unimportant, e.g. almost I nearly acceptable (Klein I 997: 63).) Klein (I997: 89, 93) makes these judgments for unstressed not : ?It is not hardly acceptable, ?It is not almost acceptable, but she observes that nearly is an exception: It is not nearly acceptable enough. On the Ducrot paraphrase of I hardly F I by !almost not f l, the former becomes ?It is not almost not acceptable. On my para­ phrase of I hardly f l by I only just fl, i.e. I f just, but at most Fl and nearly not fl, ?It is not hardly acceptable becomes ?It is not [only just] acceptable, i.e. lit is not [accep. table just, but at most acceptable I and nearly not acceptable] !, in which I find deviant the clause ?not . . . nearly not acceptable that is part of the compound Verb Phrase.

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

l Q VP, I is true, then l Q VP,l is true. Some men is upwards monotonic, since Some men walked rapidly II- Some men walked but not conversely. A monoto­ nically decreasing quantifier is one that preserves truth under the subset relation of the extensions of the predicates. For example, Few men walked II- Few men walked rapidly. Since Few men smoke or drink II- Few men smoke & Few men drink, an instance of ( u), and Few men smoke II- Few men drink II- Few men smoke and drink, an instance of (2b), it is easy to see that in general a downwards monotonic quantifier must satisfy ( u) or (2b). 2 So, as Jerry Sadock pointed out to me, it is not surprising that languages other than English would distinguish the two 'only' quantifiers lexically, e.g. German lallein Proper Name I and l nur VP I. 3 For my purposes here nearly not is illus­ trative, but I do not assume that at most just and nearly not are linguistically or logically the same; see Atlas (I 996b: 304). At most is the fundamental operator. 4 The claim in (I4} is less intuitively obvious than the claim in (I 3 ). Fred {barely, hardly} [smoked or drank] seems intuitively to entail Fred {barely, hardly} smoked & Fred {barely, hardly} drank. But this is a linguistic feature of the coor­ dinating VP conjunction 'or', not a logical fact. Fred rarely smoked or drank is interpreted to mean Fred rarely smoked & Fred rarely drank. Also, notice that in (I 3} or ·(I4} evaluating a denial of the alleged entailment would require evaluating an unacceptable, ungramma­ tical sentence *Fred hardly smokes or drinks but doesn't hardly drink that cannot by virtue of its ungrammaticality be evaluated for necessary falsehood. Atlas {I996a, in press) argues that · instances like (13) of (2b) are the more 'basic' conditions for downward mono­ tonicity, which is a happy circumstance, in that {I 3} seems more easily inter­ pretable than {I4} since (I 3} does not involve 'or' in the Verb Phrase.

364 Negative Adverbials, Prototypical Negation, and the De Morgan Taxonomy

·

=

=

It HARDLY rained {0 ?, ,eh?}-{in fact, indeed) it didn't rain, whose first clause is uttered with a question intonation con­ tour, is an acceptable statement second clause whose retracts or corrects the first 'echoic' clause. Also, the following statement whose first clause is uttered with an irony contour in English: It

hardly rained alright-{in fact, indeed) it didn't rain, is an acceptable statement.

Finally, on another . matter, suspension of what is inferred by embedding it in an if-clause will not distinguish an entailed proposition from an irnplica­ tum or a presupposition, so we need not consider suspension data. ·

9 The problem of interpreting hardly as a VP modifier does not, of course, arise here; see note 4· IO For example, often would not be an 'endpoint' quantifier, so not quite often ljL not often. So It would be consistent to claim He came to visit her often enough but not quite often enough, which it is. I I Of course, if an NP satisfies (2b), one expects that it will satisfy (�a). I2 The intuitive judgment on this entail­ ment is sometimes unclear, as some find processing at most a bit awkward: (2b)

John at most smokes V John at most drinks ljL John at most [smokes and drinks]. The

satisfaction of (I b) and the failure of ( 1a) seem clear: (�a) John at most [smokes or

drinks] ljL John at most smokes & John at most drinks; (Ib) John at most smokes & John at most drinks I� John at most [smokes or drinks]. I 3 Interestingly one may distinguish I not all CN I from hardly anyone by the De

Morgan condition (2a) Pseudo-Anti­ Multiplicativity (PAM). ! not all CNI is PAM; hardly anyone and l hardly Verb Phrase I are not (see also Atlas I 996b: n. 48, 3 23-4). One might have thought, as did Zwarts (I 996, I998), that satisfy­ ing an extra De Morgan condition would make I not all CN I 'more negated' than hardly anyone. But that is not necessarily the case. Of all the negative NPs in the table, I not all CNI is the only one that satisfies (2a). An upwards monotonic NP like lsome N l also satis­ fies (2a), and it is clear from the form of (2a) that it is essentially ah upwards ·monotonic principle. This fact should raise doubts that satisfying (2a) should make l not all CNI 'more negated' than hardly anyone or 'few CNI merely because it satisfies one more De Morgan condition than they do. Since hardly anyone is paraphrasable by" almost no one, the question is the relation between the latter and I not all CN I or lsome CN notl. Both are internal nega­ tives. Despite !not all CNI being PAM and hardly anyone not PAM, !some CN

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

7 I use double quotation marks to indi­ cate that the expression is an assertion. Though the composition I� o I�» is well defined, the composition I�» o I� is not defined, since the domain of the implicatural relation must be assertions, not propositions. 8 Larry Horn has caused me to mention the following considerations. The question of the truth�value of (26a) is one thing; the question of the felicity of speech-acts of asserting it and related sentences is something else. In general I do not think that logical inconsistency entails infelicity, as I have discussed at length elsewhere (Atlas I990, I995, I997). If one thinks that asserting an 'obvious' inconsistency is to perform an infelicitous speech-act, then one should, if one thinks as I do that It hardly rained entails It rained, put a question mark '?' before (26b), as I have done; •but the question mark is not there merely because I think the conjunction (26a) is necessarily false. I don't think, for I example, that every asserting of o is to perform an infelicitous speech -act. (A: Clinton is telling the truth. B: So o 1.) I do think that asserting (26b) is a deviant assertion; (26a) is also, on my view, a necessary falsehood, a contra­ diction. There are similar assertions that are not deviant. For example, the fol­ lowing contrastively stressed statement:

Jay David Atlas 365

does not fl seems less (!) negated than lalmost no one fsl, but that datum is inconsistent with Zwarts's (1996, 1998) hypothesis that the more De Morgan conditions a phrase satisfies, the more negated the phrase is. For this empirical reason I again propose abandoning Zwarts's hypothesis as, · for additional philosophical reasons discussed in Atlas (I996a, in press), I have proposed before. (My thanks to Sjaak de Mey for discussion of this point.) I4 The phrase lat most VPI which natu­ . rally fails (2b) Prototypical Downward ,

Monotonicity as well, is analytically equivalent to l no more than VP I, an obviously negative phrase. I 5 In Atlas (I99I, 1993, I996b) I showed that though the NP I only Proper Name I licensed Negative Polarity Items, it was not downwards monotonic; it was only closed under disjunction (Ib). The Conclusion stated in the body of the text formulates the general theorem merely suggested by the fact that lonly Proper Namel is a counter-example to the claim that every NPI licensor is a downwards monotonic NP.

Abraham, W. & Akkerman, L. (I995), 'Die Monotonieeigenschaften der Intensivier­ wi:irter im Deutschen. Ein Beitrag zur Textsemantik des Deutschen (I):' Deutsch als Fremdsprache, Zeitschrifi zur Theorie und Praxis des Deutschunterrichts Jiir Aus­ lander, 32, 3, I 5 3-60. Abraham, W. & Akkerman, L. (I 996), 'Zu den Monotonieeigenschaften der hohen und extrem hohen Intensivier­ wi:irter im Deutschen', I]GLSA, I, 2,I75-2I7. Akkerman, L. & Abraham, W. (I996), 'Die Monotonieeigenshaft�n der Intensivier.­ wi:irter praktisch und so gut wie im Deutschen', in E. Weigand & F. Hundsnurscher (eds), Lexical Structures and LAnguage Use, Proceedings of the International Conference on Lexicology and Lexical Semantics, Munster, I 3-I 5 Sep­ tember I 994, Max Niemeyer Verlag, Tiibingen, 3-I4. Atlas, J. D. (I984), 'Comparative adjectives and adverbials of degree: an introduc­ tion to radically radical pragmatics', Linguistics and Philosophy, 7, 347-77. Atlas, J. D. (I986), 'Irony: the Sperber­ Wilson-Miller Theory', Colloquium

Lecture, Department of Philosophy, University of Hong Kong, Hong Kong, November I 986. Atlas, J. D. (I991), 'Topic/comment, pre­ supposition, logical form, and focus stress implicatures: the case of focal particles only and also', journal of Semantics, 8: I27-47· Atlas, J. D. (I993), 'The Importance of Being "only": testing the Neo-Gricean versus nee-entailment paradigms', journal of Semantics, 10, 30I-I 8. Atlas, J. D. (I996a, in press), 'Negative quantifier noun phrases: a typology and an acquisition hypothesis', Invited Lecture, Pionier Conference on Per­ spectives on Negation, sponsored by the NWO Pionier Project 'Reflections of Logical Patterns in Language Struc­ ture and Language Use', University of Groningen, The Netherlands. 24-26 August . I996. To appear in the Pro­ ceedings of the Conference, edited by V. Sanchez & J. Hoeksema, Benjamins, Amsterdam. Atlas, J. D. (I996b), ' "Only" noun phrases, pseudo-negative generalized quanti­ fiers, negative polarity items, and ·

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

RE FERE NCES

366

Negative Adverbials, Prototypical Negation, and the De Morgan Taxonomy

monotonicity',

265-329·

journal of Semantics,

1 3,

Atlas, J. D. & Levinson, S. C. {r98 1), 'It­ clefts, informativeness, and logical form: radical pragmatics (revised stan­

dard version)', in P. Cole (ed.), Radical Academic Press, New York, I-6 I. Barwise, J. & Cooper, R (r98 r), 'General­ ized quantifiers and natural language',

Pragmatics,

Linguistics and Philosophy, 4. 1 5 9-219.

309-30.

Ducrot, 0. {1973), La Preuve et le dire, Maison Marne, Paris. Fillmore, C., Kay, P., & O'Connor, M. (1988), 'Regularity and Idiomaticity in Grammatical Constructions: the case of let alone', Language 64, 501-38. Gamut, L. T. F. (1991), Lagic, Language, and

Meaning, Vol. 2, Intensional Logic and Logical Grammar, University of Chicago

Press, Chicago. Grice, H. P. (1975), 'Logic and conversation', in P. Cole & J. L. Morgan {eds), Syntax and Semantics 3, Speech Acts, Academic Press, New York, 41-58. Hom, L. R (r985), 'Metalinguistic negation and pragmatic ambiguity', Language, 61,

121-74·

Hom, L. R

Negation,

{1989), A Natural History of

University of Chicago Press,

Chicago. Hom, L. R {1992), The said and the unsaid,' in C. Barker & D. Dowty (eds),

SALT II: Proceedings of the Second Con­ ference on Semantics and Linguistic Theory, May 1-3, 1992, Ohio State University, Columbus, Ohio.

Horn, L. R (1996), 'Exclusive company: only and dynamics of vertical inference'

journal of Semantics,

1 3,

1-40.

Jorgensen, J., Miller, G. A., & Sperber, D.

(1984),

Test of the mention theory of

General,

of Experimental Psychology: I I2-20. (1997), 'Adverbs of degree in I I J,

Klein, H. Dutch', Ph.D. dissertation, University of Groningen, The Netherlands.

Klein, H. ( 1998), 'Adverbs ofdegree in Dutch and related languages', Linguistic Aktuell/ Linguistics Today, 21, John Benjamins, Amsterdam. Klima, E. (1964), 'Negation in English', in J. A. Fodor & J. J. Katz (eds), The Structure

of Language: Readings in the Philosophy of Language, Prentice-Hall, Englewood Cliffs, NJ, 246-323. Klooster, W. (1994), 'Syntactic differentia­

tion in the licensing of polarity items', Pionier Colloquium on Negation and Polarity, University of Groningen, The Netherlands, June 1994. Klooster, W. (1995), 'Monotonicity and scope of negation', MS, University of Amsterdam, The Netherlands, first draft, June I 994· Ladusaw, W. (r98o), Polarity Sensitivity as

Inhere�?t Scope Relations,

York. Lakoff, G.

Garland, New

(1987), Women, Fire, and Dangerous Things, University of Chicago

Press, Chicago. Lewis, D. K. quantification',

(1975), in

E.

'Adverbs of Keenan (ed.),

Formal Semantics of Natural Language, Cambridge University Press, Cambridge,

3-1 5.

Linebarger, M.

(r98r), The Grammar of Negative Polarity, Indiana University

Linguistics Club, Bloomington, Indiana. Linebarger, M (1987), 'Negative polarity and . grammatical representation', Linguistics and Philosophy, ro, 325-87. Linebarger, M. (1991), 'Negative polarity as linguistic evidence', in L. M Dobrin et al. (eds), CLS 27, Papers from the 27th

Regional Meeting of the Chicago Linguistics Society 1991: Part Two: The Parasession on Negation, Chicago Linguistics Society, Chicago, 165-88. McCawley, J. (1988), The Syntactic Pheno­ mena of English, Vol. 2, University of Chicago Press, Chicago.

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

Boer, S. A & Lycan, W. G. (1976), The myth of semantic presupposition', · Ohio State Working Papers, No. 21. Depart­ ment of Linguistics, Ohio State Uni­ versity, Columbus, Ohio, 1-90. Carsten, R (1996), 'Metalinguistic negation and echoic use', journal of Pragmatics, 1 3 ,

irony', journal

Jay David Atlas 367 (ed.), Radical Pragmatics, Academic Press, New York, 295-3 I 8. Taylor, ]. R (I995), Linguistic Categorization: Prototypes in Linguistic Theory, second edn., Clarendon Press, Oxford. Vandeweghe, W. (I988), 'Omgevingen van vrijwel', TABU, 18, 93-I05. Vendler, Z. (I967), Linguistics and Philo­ sophy, Cornell University Press, Ithaca,

NY. Webster's New Collegiate Dictionary (I 98 I), G. & C. Merriam, Springfield, MA. Zwarts, F. (I 99 I), 'The syntax and semantics of negative polarity', MS, University of Groningen, The Netherlands. Zwarts, F. (I996), 'Negation: a notion in focus', in H. Wansing (ed.), Perspektiven der Analytischen Philosophie (Perspectives in Analytical Philosophy), Vol. 7, Walter de Gruyter, Berlin, I 69-94· Zwarts, F. (I998), 'Three types of polarity', in E. Hinrichs & F. Hamm (eds), Plural Quantification, Kluwer, Dordrecht, I77-238.

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

Moxey, L M. & Sanford, A. J. (I993), Communicating Quantities, Erlbaum/ Hove, London. Os, C. V. van (I988), 'Intensivierung im · Deutschen', Ph.D. dissertation, Univer­ sity of Groningen, The Netherlands. Os, C. V. van.(I989), Aspekte der Intensivier­ . ung im Deutschen, Tiibingen. Quine, W. V. 0. (I98 I), Mathematical Logic, rev. edn., Harvard University Press, Cambridge, MA. Rosch, E. (I978), 'Principles of categori­ zation', in E. Rosch & B. Lloyd (eds), Cognition and Categorization, Erlbaum, Hillsdale, NJ, 27-48. Russell, B. A. W. (I985/I9I8), The Philo­ sophy of Logical Atomism, Open Court, LaSalle, IL. Ryle, G. (I949), Concept of Mind, Barnes & Noble, New York. Seuren, P. A. M. (I985), Discourse Semantics, Blackwell, Oxford. . Sperber, D. & Wilson, D. (I98 I), 'Irony and the use-mention distinction', in P. Cole

Journal ofSmuzntics

© Oxford University Press

14= 369-4 1 6

1997

Splitting the Reference Time: The Analogy between Nominal and 1 Temporal Anaphora Revisited RANI NELKEN and NISSIM FRANCEZ The Technion,

Haifa

Abstract

1 I NTRODUCTION In Partee (1973), Partee introduced the notion of temporal anaphora to account for ways in which temporal expressions depend on surrounding elements in the. discourse for their semantic contributidn to the discourse. This analogy between NP anaphora and temporal anaphora has been an important factor in the analysis of temporal expr�ssions in natural language semantics and a driving force in the early stages of the development of Discourse Representation Theory (DRTY (Kamp 1981, Heim 1982, Kamp & Reyle I993) in particular. A factor vital to the analysis of temporal anaphora which is absent from the analysis of NP anaphora is the well-known notion of reference time (Reichenbach 1 947). Thus the analysis of temporal anaphora has an added level of complexity relative to analyses of nominal anaphora. This may be seen in the unified treatment of temporal expressions in discourse given by Hinrichs (198 1), Partee (1984), and Hinrichs (1986). In these works, a heavily burdened reference time plays an important role. In this paper, we examine temporal anaphora under two kinds of special circumstances: the interaction of temporal anaphora with quantifi­ cation over eventualities and anaphoric links to multiple eventualities. Such circumstances, while interesting in their own right, are also good test-beds for theories of the semantic interpretation of temporal expressions.

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

The analysis in Partee (1984) of quantified sentences, introduced by a temporal connective, gives the wrong truth conditions when the connective is before or after. In this paper, we show how splitting the different roles of Reichenbach's reference time may be used in order to solve this problem. We further enhance the analogy between pronominal and temporal anaphora, by proposing an analog of plural NP-anaphora in the form of temporal anaphora involving multiple event antecedents and an analog of an E-type analysis of pronouns in the analysis of quantified narrative discourse.

370

The Analogy between Nominal and Temporal Anaphora Revisited

In order to investigate temporal anaphora under quantification over eventualities we examine the following examples:

{I} Before John makes a phone call, he always lights up a cigarette. (Partee I 984) (2) When he came home, he always switched on the tv. He took a beer and · sat down in his armchair to forget the day. {de Swart I99I) (3) When John is at the beach, he always squints when the sun is shining. (de Swart I99I)

(4) When John spilled his coffee and Bill tripped on the rug, Mary sighed. (5) Before John spilled his coffee and Bill tripped on the rug, Mary sighed. (6} John gave a lecture on Sunday and on Monday. He became very tired. What is common to these examples is a temporal anaphoric link between an eventuality and multiple eventuality antecedents. In (4) and (5), Mary's sighing follows or precedes both events of John's spilling the coffee and Bill's tripping on the rug. In (6), John's becoming tired is linked to either one of the events of his giving a lecture or to both. We show that dealing with such cases requires as operation which combines individual eventua­ lities into compound ones. We introduce such an analysis based on the analysis of plural pronominal anaphora. This enhances the relation between nominal and temporal anaphora.

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

The analysis of sentences such as {I) in Partee {I984) yields wrong truth­ conditions for the temporal connectives before or after. In Partee's DRT­ based analysis, such sentences trigger box-splitting with the eventuality of the subordinate clause and an updated reference time located in the antecedent box, and the eventuality of the main clause located in the consequent box. Figure I shows a DRS obtained by Partee's analysis for sentence (I). As can be seen in this DRS, the placement of the updated reference time, r1 0 in the antecedent box, causes undesirable universal quantification over it. We show how this problem can be solved by adopting a refinement of the notion of reference time adopted from Kamp & Reyle {I99 3 )· This refinement involves splitting the roles of the reference time between several different temporal indices. We show how this split provides a straight­ forward solution t9 this problem and also gives a satisfactory account of examples such as (2) and (3). To examine temporal anaphora involving anaphoric reference to multiple eventualities we discuss the following examples.3

Rani

Nelken and Nissim Francez

37I

n x r0

John (x)

e1 r1

e 1 � r0

r1 < e 1

e 1 : j make_a_phone_call (x) I Figure

I

�

e2 e2 � rl

e2 : j iight_up_a_cigarette(x) I

Partee's analysis of sentence (r).

2 BACKGROUND analysis of the mechanism of temporal anaphoric reference hinges upon an understanding of the ontological and logical foundations of temporal reference. Different concepts have been used in the literature as primitives. These include temporal instants as in Tense logic (Prior I967), intervals of time (Bennett & Partee I972) as in the analysis of temporal connectives in (Heinamaki I 978), event structures (Kamp 1979) as in Hinrichs' (r981, 1986) analysis of temporal anaphora, and event lattices as in Krifka (I989). An important factor in the interpretation of temporal expressions is the classification of situations into different aspectual classes (or Aktionsarten), which is based on distributional and semantic properties. In this paper, we only consider events and states, termed together eventualities in Bach (I 98 I). In narrative sequences, event clauses seem to advance the narrative time, while states block its progression. The mechanism used to account for this phenomena in Hinrichs (I98 I) and Partee (I984) is based on the notion of reference time, originally proposed by Reichenbach (I 947). Reichenbach's well-known account of the interpretation of the different tense forms uses the temporal relations between three temporal indices: the utterance time, event time and reference time. These temporal indices are points of time, the relation between which determines the interpretation of the tense form. The reference time according to Reichenbach is determined either by context, or by temporal adverbials.

2. I

A unified analysis of temporal anaphora

Hinrichs' and Partee's use of the reference time provides for a unified treatment of temporal anaphoric relations in discourse, including narrative progression (especially in sequences of simple past tense sentences), temporal adverbs and temporal adverbial clauses, introduced by a temporal

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

An

372

The Analogy between Nominal and Temporal Anaphora Revisited

connective. This reference time is no longer an instant of time, but rather an interval. This approach can be summarized as follows: in processing a , Sn). S1 requires a contextually-determined initial discourse: (S1 , S2 , reference time r0• Subsequent clauses S;, for i � o are. interpreted relative to r; 1 and update it to r;, as follows: •

•

•

_

•

An

event clause: -Introduces a new event e, s.t. -Updates r; to be just after' r; - 1

•

A

�

r; .

r; 1 r; _ 1 {Partee 1 984). This ts denoted by

e�

_

)

r; = r;- 1 .

As

an example of . such an analysis, consider the following narrative discourse {Partee 1 984):

(7) John got up, went to the window, and raised the blind. It was light out. He pulled the blind down and went back to bed. He wasn't ready to face the day. He was too depressed. Figure 2 shows a DRS for the first two sentences of this discourse, according to Hinrichs' and Partee's analysis. It contains the following components: • •

• •

The 'n' in the top DRS is a mnemonic for 'now'-the utterance time. e1 stands for the first event in the discourse-John's getting up. This event is interpreted relative to a contextually-provided reference time, r0, whereby e 1 � r0• r1 -a new reference time marker is then introduced with r0 � '� " r1 serves as the current reference time for the following event e2• r0 e 1 r1 e2 r2 e3 s1 n x

John (x) r0 < n e 1 !;;;;; r0 r0 ::5 r1 r1 < n e2 !;;;;; r1 r1

::5

r2

r2 < n r2 !;;;; s 1

e1

:I

e2 : . .

get_up (x)

.

sl : . . .

Figure

2

A DRS for a narrative discourse.

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

state clause: -Introduces a new state s, s.t. r;- 1 � s. -The reference time is not updated (i.e.

Rani Nelken and Nissim Francez •

373

Updating the reference time continues in this fashion, until the second sentence is processed. This sentence introduces a state, sl > with r2 � sl > and so on . . .

Adverbial · modifications, whether phrasal (e.g. 'On Sunday') or clausal (e.g. 'When Bill left'), introduce a new reference time, which overrides the current reference time. An important assumption referred to here as Hinrichs} processing-order assumption (POA) is the following: I (POA) Preposed temporal subordinate clauses are processed before the main clause.

Assumption

•

•

The subordinate clause eventuality, esub: 1 . is not constrained to occur within the current reference time rcum but rather follows. it, and 2. introduces a new reference time rnew• which in cases of simple linear progression follows r'""' regardless of the temporal connective. 3· The temporal connective determines the relation between esub and rnew· For example, in 'when' clauses, esub j fnew· The main clause eventuality emain is interpreted with respect to rnew• as discussed earlier, regarding narrative progression: 1. A state includes fnew· 2. An event is included in it and updates it further.

For example, consider the following discourse (Partee 1984). Following Partee, instead of giving the full DRS for it, we illustrate the analysis using the diagram in Figure J, with circles denoting inclusion. (8) Mary turned the comer. When John saw her, she crossed the street. She hurried into a store. 2.2

Quantification over events

Partee (1984) extends Hinrichs' treatment of temporal anaphora to the analysis of sentences that contain a temporal adverbial and quantification

Figure

3

Partee's interpretation of (8).

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

Temporal subordinate clauses are interpreted according to the POA as follows:

374 The Analogy between Nominal and Temporal Anaphora Revisited over eventualities. A major assumption included in her analysis is the Box­ Splitting assu mption (BSA): 2 (BSA) Temporal adverbials and quantification over eventualities trigger box-splitting as do if or every clauses in DR T

Assumption

Consider the following example from (Partee Figure 4:

(9)

1984)

with its DRS in

Whenever Mary telephoned, Sam was asleep.

n x y r0 Mary (y) Sam (x) �

e l rl e 1 � r0 e1 < n e1 :5 r1 r1 < n

e1

:I telephone (y) I

sl rl � s l

=>

s1

: I sleep (x) I

Figure 4 DRS for (9).

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

The subordinate clause 'Mary telephoned' cannot be interpreted relative to a reference time denoting a single ti'me period, since Mary's telephoning is not specified to occur at some specific time. Still, the sentence has to be interpreted relative to some reference time. This reference time can be taken to be a large interval, and should contain each of the relevant occurrences of Mary's telephoning, during which Sam was asleep. This reference time is represented in Figure 4 whenever as T0 in the top sub-DRS. By the BSA, the quantifier 'whenever' triggers box-splitting. The event marker e, is added to the antecedent box, with the condition that it is temporally included in the current reference time, r0 and is prior to n. 'Whenever' also causes the introduction of r, , a new reference time marker, which lies just after' e' " The stative clause 'Sam was asleep' causes the introductio� of a state s 1 0 which includes the reference time '' " The embedding conditions for the whole construction are just like those for a regular 'if' or 'every' clause, i.e. the sentence is true, if every proper embedding of the antecedent box can be extended to a proper embedding of the combination of the antecedent and the consequent boxes. This means, as desired, that for each choice of an event e, of Mary's telephoning,

Rani Nelken and Nissim Francez

37S

and reference time r1 just after' it, there is a state of Sam's being asleep, that surrounds r1 • A sentence such as (9a), which has the same meaning as (9), except the 'whenever' is replaced by 'when', and 'always' is added in the main clause, would get the same DRS.4 ·

(9a)

.

When Mary telephoned, Sam wa5 always asleep.

2. 3

Extending the analysis

up a cigarette. Paraphrasing this, we could say that John lights up cigarettes at all times preceding each phone call, not just once preceding each phone call. This problem is not present in the DRS in Figure 4, since although the reference time r1 is universally quantified over in that DRS as well, it is also restricted to immediately follow e 1 • It is similarly restricted if 'before' is replaced with 'just before' or 'ten minutes before'. But (unrestricted) 'before' is analyzed as 'some time before', and thus the problem arises. We henceforth informally refer to this problem as 'Partee's Quantification Problem' (PQP) . Partee (I 984) suggests that in these cises we somehow have to ensure that

the reference time marker, r1 , appears in the universe of the consequent causing it to be existentially quantified over, giving the desired interpretation. As noted by (de Swart I99I), the trivial solution of moving the reference time marker to the right . box does not agree with Hinrichs' POA. According to the POA, the subordinate clause is processed before the main clause. Simply moving the reference time marker to the right box would seem to defer part of the processing of the subordinate clause, namely, the updating of the reference time, to the processing stage in which the right box is updated, i.e. the processing of the main clause. In our proposed solution, the 'reference time' is indeed moved to the right box, but

DRS,

it is a different notion of reference time, and is exempt (as will be shown) from this criticism.

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

As noted by Partee (I 984), this analysis does not extend in a straightforward manner to cases in which the operator when is replaced by (an unrestricted) before or after, in such quantified contexts. Constructing a similar DRS for such sentences yields wrong truth conditions. For instance, Figure I shows a DRS for ( I ), adhering to the principles above. The reference time marker r. , used for the interpretation of the main clause, is placed in the universe of the antecedent box. Because the temporal connective is 'before', r1 is restricted to lie before e 1 • The embedding conditions determine that this reference time be universally quantified over, causing an erroneous reading in which for each event e 1 , ofJohn's calling, for each earlier time r1 , he lights

376 The Analogy between Nominal and Temporal Anaphora Revisited

3 THE PROPORT ION PROBLEM De Swart (I99I) sees .PQP as a temporal manifestation of the notorious proportion problem, which arises in donkey sentences such as: ( n ) Most women who own a cat are happy. (Kadmon I99o)

(12) { (e1 , e2)i e.

<

ez}

The quantificational structure of such sentences can be analyzed either by an iteration of monadic quantifiers, or as a single dyadic quantifier of type (I, I , 2). In the first approach, adverbs of quantification (Q-adverbs) are assigned the structure: (I3) Q(Ss, {e. I 3 (Sm, rc�. )}) In (I 3 ), Ss and Sm denote, respecti�ely, the sets of events described by the subordinate and the main clause, ret, denotes the image set of e. under the temporal connective TC, i.e. the set of events e2, which are related to e1 via the relation TC. The Q-adverb establishes a relation between the set S1 and the set { e1 1 3 (Sm, TC�, ) }. According to the second approach, the structure is the one given in (I4). The dyadic quantifier [Q, 3] establishes a relation between two sets of events: Sn Sm, and a binary relation between events: TC. {I4) [Q, 3] (Ss, Sm, TC) De Swart's solution does overcome PQP, although not within DRT. As such, the existential quantification in (I 3) and in {I4) has to be stipulated,

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

The sentence is false in the case where out of ten women, one owns fifty cats and is happy, while the other nine women own only one cat each, and are miserable. This is not predicted by the unselective binding of quantifiers in DRT, which quantify over all the free variables in their scope, in this case women-cat pairs. According to de Swart {I99I), PQP is similar. The universal quantifier in sentences such as (I) binds pairs of events and updated reference times, whereas the desired quantificational scheme is universal quantification over the event and existential over the reference time. De Swart (199I) offers a solution from a Generalized Quantifier approach, based on the analysis of quantified NPs in transitive sentences. In this analysis, the reference time is an implicit variable, which is needed in the interpretation of the temporal relation, but is not part of the quantificational structure. The denotation of a temporal connective is a relation, TC � E X E, between two sets of events. For example, the denotation of 'before' is given in (1 2).

Rani Nelken and Nissim Francez

377

whereas the analysis presented here acquires this existential quantification · 'for free'.

4 SPLITTI NG T HE ROLES O F THE REFE RE N CE TIME

We will follow Reichenbach in what we consider to be the spirit of his proposal, though not in its details. For some of the specific proposals he made do not seem to stand up to scrutiny. In fact, we believe that Reichenbach werit astray when he wanted his notion of reference point to do too many things at once. In particular, he writes as if the reference time which is required for the interpretation of past perfects like the one in . . . is of a kind with the reference times we discussed in Section . . . where they were needed in the interpretation of sequences of sentences in the past progressive and simple past tense. In this section we briefly summarize the analysis of Kamp & Reyle (1993). The interpretation of temporal expressions makes use of the following mechanisms: .

1.

Event time Location time 3· Reference Point (Rpt) 4· Perfect 5· Temporal Perspective Point (TPpt) 2.

Each of these indices (besides the event time) takes on one role of the original monolithic reference time. We briefly . summarize the use of each of these temporal indices. 1. The event time The event time of an eventuality is defined as the the smallest interval that includes it. For an eventuality e, the event time t is recorded in the DRS as t = ET(e).5

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

The DRT-based analysis of PQP in this paper uses a different notion of reference time than that used by the accounts in the exposition above. This notion of reference time is based on the analysis of temporal expressions in Kamp & Reyle (1993). This analysis distributes the different roles, which were attributed to the reference time in previous accounts, between several different temporal indices. This split allows a more fine-grained and precise analysis of temporal phenomena than the use of a single monolithic reference time. In one place {p. 594) Kamp & Reyle write:

378 2.

The Analogy between Nominal

and Temporal Anaphora Revisited

The Location time

For example, in (I s), the event triggers the introduction of an event marker e, and location time marker t into the DRS, with the DRS­ condition e � t. The past tense of the verb adds the condition t < n. In (I6), the location time of the event in the main clause is restricted to fall Gust) after the event time of the event of the subordinate clause.

(I S) Mary wrote the letter. (I6) Mary wrote the letter when Bill left.

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

The location time is used according to these principles: • The location time is an interval t, used to temporally locate eventualities, in accordance with their aspectual classification: events: are included in t (recorded in the DRS as a condition e � t on the respective markers), states: temporally overlap t (recorded as s 0 t ) . • The verb tense determines the relation between the location time t and the utterance time n. For example: simple past: t lies anteriorly to n, simple present: t coincides with n.6 • Temporal adverbials restrict the location time t: -Temporal adverbs introduce a DRS-condition on t. -Temporal subordinate clauses introduce a relation between two different indices: (a) The location time tmain of the eventuality of the main clause. (b) The event time tsub of the eventuality esub of the subordinate · clause. -The exact temporal relation denoted by a temporal connective depends on the aspectual cla5ses of the eventualities related by it. In this paper, we shall focus only on the temporal connectives: 'before', 'after' and 'when'. Following Kamp & Reyle (I993), we assign the following relations for 'when' c£ Hinrichs (I 986): (a) If both the when-clause and the main clause denote states, then their respective time indices overlap. (b) If both are events, then the times are temporally close, with the exact relation undetermined. For concreteness, we will assume here that the time index of the event in the main clause is just after' the time index of the event in the subordinate clause, bearing in mind that this is not always the correct interpretation. (c) When one is a state and the other an event, then the time index of the state includes that of the event.

Rani Nelken

and

Nissim

Francez 379

(17) John was happy on Sunday. In much the same way, a temporal adverbial clause also restricts the location time of the main clause eventuality. This is achieved by relating this location time with the time span of the subordinate clause eventuality, i.e. its event time. One side-effect of this is an asymmetry between the temporal indices used in main and subordinate clauses, using the location time for the main clause but the event time for the subordinate clause. This can be seen in the analysis cited above for (17). We will encounter this asymmetry several times during the paper. · 3·

The Rpt

Narrative progression in discourse is dealt with in this analysis by using the Rpt as follows.' • The Rpt can be either an event or a time discourse marker, already present in the DRS (recorded as an assignment Rpt : = e). • Eventualities are interpreted with respect to the Rpt: events: follow the current Rpt, states: include the Rpt. • The Rpt is updated during the processing of the discourse. • In a 'terminal' DRS (ready for an embedding test), all the auxiliary Rpts 'disappear' (do not participate in the embedding). For example, Discourse ( 1 8) would introduce an event marker, ekavt• for the event of the first sentence, and restrict it to follow the Rpt, which is chosen to be some event already present in the DRS (if there is one).

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

At this point, it is fruitful to compare the event time with the location time. The event time is a much more obvious temporal index than the location time, since it denotes the exact temporal span of the eventuality. The location time, on the other hand, imposes a much less stringent constraint on the temporal whereabouts of the eventuality. The location time takes on one of the roles of Reichenbach's original reference time, in the interpretation of temporal adverbials. A temporal adverbial restricts the location time of the eventuality, not its event time. A temporal adverb adds a condition on the location time of the eventuality. For example, in constructing a DRS for (17), we would introduce a state marker s for the state of John's being happy, and a time marker t for the location time of s. As a result of the DRS construction rules, a condition would be added to the effect that s and t overlap. The added temporal adverbial on Sunday would restrict the location time t. By following this strategy, we are not restricting the whole of the state of John's being happy to fall on Sunday. Rather, we are restricting its location time, i.e. some temporal interval overlapping this state to do so.

380 The Analogy between Nolll.inal and Temporal Anaphora Revisited Then, when processing the second sentence, ekaw may be chosen as the : = ekaw• and the event of the second sentence, ewrit� ·

Rpt, recorded as Rpt would follow e1�a v�· (r8)

Bill left. Mary wrote the letter.

4·

The Perfect The perfect is dealt with in Kamp & Reyle (199 3 ) as an aspectual operator and analyzed by using the notion of a nucleus (Moens & Steedman 1988) to account for the inner structure of an eventuality. A nucleus is defined as a structure containing the following elements: • preparatory process • culmination • consequent state The categorization of verb phrases into different aspectual classes can be given in terms of the part of the nucleus they refer to. The eventualities described by the perfect of a verb refer to the consequent state of its nucleus. For example, (19) denotes the state s holding at the present, that Mary has met the president. This state is a result of the event e, in which Mary met the president. Temporally, the state s starts just when e ends, or as it is put in (Kamp & Reyle 199 3 ): e and s abut, (represented as e X s ) . (r 9) Mary

S·

has met the

president.

The 'IPpt The temporal perspective point (TPpt) will not be discussed in this paper for the sake of simplicity.

s A N ALTERNATIVE S O LUTI O N O F P Q P

(1984) analysis of sentences that contain quantification over eventualities led to what we have informally termed Partee's

As we saw, Partee's

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

We should point out at this point, that the mechanism for handling Discourse (18) is quite different than that used to handle ( r6). Thus, while in processing this discourse, eleave acts as the Rpt for the interpretation of the second sentence, it does not do so in ( 1 6). In the processing of that sentence, we use the event time of ekaw to restrict the location time of ewrit�· Thus, although the end result is similar in terms of truth conditions, the · construction rules are quite different. These two different mechanisms are · used instead of a single reference time, which was supposed to account for both kinds of phenomena in previous analyses.

Rani Nelken and Nissim Francez 3 8 1 quantification problem (PQP). This analysis was based on the interpretation rules of non-quantified sentences. By adopting the split between the different roles of the reference time, as developed in K.amp & Reyle (1993), an alternative DRT solution to PQP is proposed. The processing of sentences that contain a {preposed) temporal subordinate clause and quantification over events proceeds as follows: •

•

•

This analysis is illustrated by constructing a DRS for (1) (repeated here). To allow easy comparison with Partee's analysis; we give this DRS in Figure sb together with the DRS according to Partee's analysis for the same sentence shown in Figure sa.

(1) Bifore John makes a phone call, he always lights up a cigarette. The adverb 'always' in this sentence triggers box-splitting. The subordinate clause 1ohn makes a phone call' is interpreted in the left box. It causes the introduction of an event marker, e, with its corresponding event time marker t = ET(e). The main clause 'he lights up a cigarette' triggers the introduction8 of an event marker e', and its location time marker t ', with the DRS-condition e1 � t '. Since the temporal connective in this sentence is 'before', the relation between these two markers is one of precedence, represented via the condition t ' < t. In this DRS, n denotes the utterance time. The whole implication is considered a state s. This state is no longer an atomic eventuality. Rather, it is a complex state denoting John's habit. This encompassing state holds during the present, and so its location time is n, according to the regular construction rules regarding the location times of states.

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

•

By the BSA, such sentences trigger box-splitting, with the temporal subordinate clause interpreted in the antecedent box, and the main clause in the consequent box. The temporal subordinate clause introduces an eventuality esub with an event time lsub· . The main clause introduces an eventuality e""'in with a location time tmain • and, in addition, a condition asserting a temporal relation between lsub and /""'in· The exact temporal relation depends on the temporal connective. The asymmetry in using the event time for esub and the location time for e""';" arises from the interpretation · rules of temporal connectives based on Kamp & Reyle (1993) {as summarized in section 3). We adopt a suggestion by Chierchia, cited in Partee (1984), that the whole implication be rendered a complex state. This state introduces a location time of its own and a condition linking the state to the location time. The rules for this condition are just the same as for simple states.

382 The Analogy between Nominal and Temporal Anaphora Revisited .n x r0 John (x) e1 '1 e 1 � 'o 't < e 1

a:

el :

I

I

phone(x)

e2 =>

l

e2 � 't

e2 : light_up (x)

I

n xs John (x) n � s e ' t'

et s :

e

:I

Figure

t = Ef(e)

phone(x)

s

I

=>

j

t' < t e ' � t'

e ' : light_up (x)

I

a: Partee's analysis, b: our analysis.

In this analysis, the event time of the eventuality in the subordinate clause serves as the antecedent for the location time of the eventuality in the main clause. Since the subordinate clause and the main clause are each interpreted in their own sub-DRS, each of the relevant temporal markers resides in its appropriate box, yielding the correct quantificational structure: for every t (a time ofJohn's phoning) there exists a t' < t (a time ofJohn's smoking). This quantificational structure does not need to be stipulated as part of the Q-adverb's meaning (as in de Swart 1991), but arises directly from an extension of the principles used in the analysis of non-quantified sentences. According to this solution, the temporal relation in the sentence is captured as a relation between the event time of the subordinate clause and the location time of the main clause, instead of using the reference time for both. It is well known that the proportion problem may be solved in DRT by stipulating a uniqueness presupposition on the dependent argument. In using the event time instead of the reference time as the argument of the temporal relation, we achieve this uniqueness naturally, since there is, by ·

definition, a unique event time per event.9 This solution to PQP is not prone to de Swart's criticism against the naive solution of moving the reference time to the right DRS. The (preposed) temporal clause may be processed before the main clause, adhering to Hinrichs' POA This is so because t 1, the location time of e 1, which 'replaces' r. , the reference time of Partee's analysis, as the temporal

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

b:

Rani Nelken and Nissim Francez 3 8 3

6 O T H E R APPLI CA T I O N S

I n this section we apply the analysis described above to related con­ structions, the past perfect and iterated quantifiers. We show how these phenomena, previously analyzed by using the ·original notion of reference time, can be given a more satisfactory analysis by adopting the split in the roles of the reference time.

6. I

Past perftct

First, let us consider the past perfect tense, as in (2o). (2o)

Often, when Anne

had slept late, she had a headache. (de Swart

1991)

De Swart gives this example to illustrate the necessity of using reference times in the interpretation of temporal connectives. This is done to refute any approach that would try to overcome Partee's quantification problem by getting rid of reference times altogether. The past perfect tense in the subordinate clause has the Reichenbachian schema E-R-S (meaning that the reference time (R) is in the past of the speech time (S), and the event

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

index of the eventuality in the main clause, arises from processing the main clause, and not from updating the reference time of the subordinate clause. As for the model-theoretic interpretation of complex states, such as s in Figure sb, de Swart & Molendijk (1994)'0 point out that Kamp & Reyle (1993) propose a construal of simple states as triples (t, x, P), where t is a period of time, x an individual and P a property, such that t is an interval of time, during which x has property P. Generalizing this, they propose to construe states as pairs ( t, 1 ), where t is a period of time and 1 is a possibly complex condition holding during time t. Thus we reach a solution of PQP. This solution is based solely on applying Kamp & Reyle's split of the reference time to sentences containing quantification over eventualities. Note that the crucial property of the analysis that solves PQP is the fact that an event introduces its own location time. This contrasts with the way in which Hinrichs' and Partee's reference time is updated during discourse, where the introduction of an event updates the reference time for the next event. Note that we did not have to stipulate this property of the location time particularly for the solution of PQP. Rather, it was an independent property of the general cons�ction rules of Kamp & Reyle (1993) for temporal adverbials."

384 The Analogy between Nominal and Temporal Anaphora Revisited

6.2

Iterated quantifiers

De Swart (1991) analyzes (3) (repeated here) as containing an iteration of quantifiers: an implicit generic quantifier and the explicit adverbial 'always'.

x s 1 t1 n s 1 0 t1 t1 < n

Anne(x)

e 1 s2 t2 t2 = ET(s2) e 1 ::::c s2

s, :

e, :

I sleep_late(x) I

19

s3 t3 53 0 t3 tz 0 t3

2

s3 :

j have_headache(x) l

Figure 6 Often, when Anne had slept late, she had a headache.

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

time (E) precedes R). The state in the main clause has to be temporally linked not to E, the event time of the subordinate clause, but to R, its reference time. Thus, goes the argument, the reference time is essential to the analysis of such sentences. This point would seem to be troublesome for our approach, which uses event times and location times and not the 'monolithic' reference time. But this is only seemingly difficult, since our analysis of perfect tenses is not Reichenbachian. Following Kamp & Reyle (1993), the perfect is analyzed by using the perfect operator. According to this analysis, the eventuality referred to by the subordinate clause is the resultant state of a previous event. In this case, it is the result state of the event of Anne's sleeping late. The temporal relation in the sentence is between this result state and the state in the main clause. This analysis is illustrated by the DRS in Figure 6 for (2o).'2 Since the sentence contains the quantifier 'often', we construct a DRS with a duplex condition. The subordinate clause introduces the event, e n of Anne's sleeping late and its resultant state, s2• The main clause introduces the state s3, of Anne's having a headache. The temporal relation is between the event time t2 of 52, and the location time t3 of s3 • The whole duplex condition is considered a complex state, Sn the location time of which is t,.

Rani Nelken and Nissim Francez 3 8 5

Adopting Chierchia's previously mentioned suggestion of introducing complex states, (1991) proposes a representation as in (21): (3) When John is at the beach, he always squints when the sun is shining. (21) Gen(S1 , { e , / 3 (Sm , TC�, ) } ) In (2 1), the following symbols are used: • •

•

where: -Sshine is the set of events denoted by the embedded subordinate clause 'the sun is shining' -S.squin1-the set of events denoted by the embedded main clause 'he squints' . -TC, �3 -the image set of the nested temporal connective 'when'. Note that in this analysis there are two instances of existential quantifica­ tion, which have to be stipulated. The DRT analysis, which is a straightforward extension of our solution to Partee's quantification problem achieves the existential quantification (in both cases) for free, because of the embedding conditions of the DRS. Using a conditional structure of box-splitting embedded within another conditional structure, we get the DRS'3 in Figure 7· In this DRS, the situation described by John's always squinting when the sun is shining is analyzed as a complex state s3• This state holds whenever John is at the beach, recorded by the condition that the location time, t2, of s3 overlaps the event time t, of John's being at the beach, s2•

7

PLURAL TE MPORAL A NAPHORA

In the account presented in section 3 of temporal anaphora, based on Kamp & Reyle (1993), eventualities and their temporal indices served as the basic components of the analysis of temporal expressions in discourse. In this section, an account of temporal anaphora involving simultaneous reference to multiple eventualities is proposed. This account reveals more

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

•

Gen is the Q-adverb corresponding to the implicit generic quantifier Ss is the set of events denoted by the subordinate clause John is at the beach' TCe, is the image set of the temporal connective 'when' Sm is the set of complex states in which the following holds:

386 The Analogy between Nominal and Temporal Anaphora Revisited n x y s1

John (x) the sun (y) n s;;; s1 s3 t2 3 S 0 t2 t2 0 1 1

s2 1 1 sl : s2

:I

ET(s2)

11

=

at_beach (x)

I

,, I I

�

zooM

ET(s4) = 13 s4 :

I

PROG(shine (y))

I

�

el

:I

el t4 el s; 14 t4 !;; t3

squint(x)

I

Figure 7 When John is at the beach, he always squints when the sun is shining.

extensive connections between NP-anaphora and temporal anaphora than the connections arising in previous work. 7.1

The problem posed by multiple eventualities

Consider examples (4}-(6} (repeated here). The common denominator to these examples is the existence of a temporal anaphoric link between an eventuality in the main clause and multiple eventualities in the subordinate clause. (4) When John spilled his coffee and Bill tripped on the rug, Mary sighed. (s) Before John spilled his coffee and Bill tripped on the rug, Mary sighed. (6} John gave a lecture on Sunday and on Monday. He became very tired. In (4), the main clause eventuality follows both events in the subordinate clause. In (s), it precedes them. In the short discourse (6), there is an ambiguity between a 'collective' and a 'distributive' reading. This ambiguity stems from the question of whether John became very tired after giving the two lectures or after each lecture. Such examples are not discussed in the previous DRT work on temporal anaphora. Furthermore, attempting to accommodate such examples within these frameworks immediately encounters difficulties. This is because temp�ral anaphoric links in these frameworks always link pairs of

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

t3 s4

Rani

Nelken and Nissim Francez 387

7.2

Combined eventualities

Enhancing the construction rules with a method for combining individual events would offer an elegant solution for the problems raised. For example, in sentences (4) and (s), the temporal anaphoric link would be between the main clause event and the combined event ofJohn's spilling his coffee and Bill's tripping on the rug. The temporal relation in such sentences should

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

eventualities. In these frameworks, the temporal relation contained in (4) and (s) would have to be captured as a temporal anaphoric link between the main clause eventuality and one of the subordinate clause eventualities. Thus, accounting for these examples would require rules to determine which of the two eventualities in the subordinate clause is the temporal anaphoric antecedent in each of the two sentences. Such an approach suffers from two problems. First, we do not know the temporal ordering of the subordinate clause eventualities. This ordering is not necessarily reflected by the syntactic ordering of the conjuncts in the subordinate clause, as can be seen in (4). In this discourse, we cannot really tell which event happened first '4 Second, even if we would allow a semantic operation of choosing the first or last event of a set of events, this approach remains highly artificial. This is because it would seem to determine the event antecedent according to the temporal connective. This requires a stipulation of a different rule for choosing the event antecedent for each temporal connective. Thus, for example, for a 'when' clause, the event antecedent should be the latest event This would give the required reading, according to which, the main clause eventuality follows the latest event and ·therefore all the events in the subordinate clause. For 'before' clauses, a converse rule would be required. It would choose the earliest event as the temporal antecedent. This would give the required reading, according to which, the main clause event precedes the earliest event and therefore all the subordinate clause events. A related problem occurs in cases such as (6). Trying to analyze this example according to the 'narrative sequence' constructions discussed above would place the event of John's becoming tired after either one of the events of his giving a lecture. None of these options is compatible with the distributive reading, according to which John became tired after each lecture. Such a reading is available for the discourse. Thus trying to analyze these examples using a temporal anaphoric relation between a pair of eventualities raises many problems. Intuitively, the temporal relation in these examples is between the main clause eventuality and a combined event antecedent in the subordinate clause.

388 The Analogy between Nominal and Temporal Anaphora Revisited

7·3

The similarity to plural pronominal anaphora

As it turns out, there is great similarity between the problem of temporal anaphoric reference involving multiple eventuality antecedents and plural pronominal anaphora, as arises in the following sentences (adapted from Kamp & Reyle 1993):

(23) John took Mary to Acapulco. They had a lousy time. (24) John is a lawyer and Mary is a lawyer. They hired a new secretary. The second sentence in both these example discourses contains the plural pronoun 'They', which anaphorically refers to John and Mary, together. The short discourse (24) exhibits an ambiguity between a collective and a distributive reading. According to the collective reading, John and Mary hired a new secretary together, while the distributive reading asserts that each of them hired a new secretary separately. An account of plural pronominal anaphora is provided in Link (198 3). A version of this account is incorporated in DRT in Kamp & Reyle (1 993). The basis of the analysis is the construction of 'individual sums' (i-sums) from the individuals. Such i-sums can be anaphorically referred to by a plural pronoun either collectively or distributively. Model-theoretically, i-sums are based . on a lattice-theoretical construction. The details of this analysis will be reviewed in the sequel. We can now see the similarity we have been alluding to between temporal anaphoric links to multiple eventuality antecedents and plural pronominal anaphora. In both cases, we have anaphoric links to entities comprised of combined individual entities (persons or events):

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

be between the location time of the event in the main clause and the event time of the combined event in the subordinate clause.' 5 Similarly, constructing a combined event for the pair of events in the first sentence of Example (6), would give the collective/distributive distinction by referring to the combined event either collectively or distributively. This idea of combining individual events into compound events is not new. It has been proposed in (Krifka 1989, Lasersohn 1992) within the framework of lattice-theoretical event-based semantics. Indeed, we shall make use of some of the mechanisms proposed by Krifka . (1989), incorporating them within DRT. What is new in this paper, to our knowledge, is our outlook on these phenomena as temporal analogs of plural pronominal anaphora. In the following section, we point out the analogy between the problems we encountered in analyzing temporal anaphora involving multiple eventualities and plural pronominal anaphora.

Rani Nelken and Nissim Francez 389 • •

In Examples (23) and (4), the antecedent is the combined individual/event. In (24) and (6) there is an anaphoric reference to the combined entity exhibiting a collective/distributive ambiguity.

7·4

Plural individuals and combined events

Kamp & Reyle (1993) present a DRT analysis of plurals, which models the domain of individuals using a complete free atomic upper semi-lattice with bottom element. This account is based on Link's (198 3) i-sums. The lattice-theoretic accounts ofKrifka (1989) and Lasersohn (1992) are a framework in which combined events may be constructed from individual events. Krifka (1989) defines the event and time domain as follows: • •

•

·

An event structure is a complete join semi-lattice without bottom element. The time domain is an atomic complete join semi-lattice without bottom element. It has the following features: -The atoms are time points. -The set of time points is linearly ordered by a transitive relation of temporal precedence <. -< is extended to a relation between times in general, by defining t < t' iff for every part s of t and s' of t', s < s'. A temporal trace function, 1R., mapping an event to its temporal trace (which is what we have been calling the event time). 1R. is a homomorphism relative to the joins of the two semi-lattices.

Thus model-theoretically, there is indeed a great similarity between the way in which plural NPs are analyzed in DRT and the analysis of combined events of Krifka (1989).

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

Of course, a major difference between the pronominal and temporal anaphora examples is the occurrence of a plural pronoun ('they') in the sentence. The occurrence of a plural pronoun provides strong evidence for a plural construction. Nevertheless, there is evidence for the need for a plural-like construction in the temporal case as well. In section 7·5· we will see how the lack of pronoun (or pronoun analog) comes into effect in the construction rules. Besides the similarity between the two linguistic phenomena of plural pronominal anaphora and plural temporal anaphora, there is also a great similarity in the model-theoretic approaches proposed for plural pro­ nominal anaphora Link (1983) and compound events (Krifka 1989, Lasersohn 1992). It therefore makes sense to model an analysis of multiple event temporal anaphora based on the analysis of plural pronominal anaphora. This is precisely what is done in the sequel.

390 The Analogy between Nominal and Temporal Anaphora Revisited 7·5

Triggering plural temporal anaphora

1. Processing the antecedent: this causes the introduction of a discourse 2.

marker for the antecedent. Processing the anaphoric element: this includes introducing a discourse marker for the anaphoric element and choosing an appropriate antecedent discourse marker, with which the newly introduced marker is equated.

These two steps are clearly distinguished in both the classical DRT treatment of nominal anaphora, and in the handling of temporal anaphora in Kamp & Reyle (1993). ' 7 How are these two stages realized in plural pronominal anaphora? Plural pronominal anaphora involves two 'new' operations: summation over individuals and optional distributive expansion. When the antecedent is comprised of several individuals, anaphora handling operates according to the two following stages (details are given in the next section): 1.

2.

Processing the antecedent: this causes the introduction of separate discourse markers for the antecedent individuals, and an additional plural discourse marker for their sum. Processing the anaphoric element: this includes introducing a discourse marker for the plural anaphoric element and choosing an appro­ priate antecedent discourse marker. Now, either the newly introduced marker is equated with the antecedent's marker (giving the collective

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

In this paper, we wish to consistently see both temporal relations induced by temporal connectives and narrative discourse as forms of temporal anaphora. This view leads us to model the analysis of temporal anaphora on the more familiar analysis of pronominal anaphora. We have already seen an important difference between the two kinds of anaphora, i.e. the reference time. The reference time (or more accurately, the different manifestations of it) is an added element of temporal anaphora, which has no analog in pronominal anaphora. Another important difference between the two kinds of anaphora, is the use of pronouns in nominal anaphora. Partee (1 984) sees the tense morpheme of a clause as the temporal analog of the pronoun. However, while pronouns come in both singular and plural varieties, the tense morpheme doesn't have such a distinction. In particular, there is no temporal analog for a plural pronoun. To see how this affects the plural-like construction rules for temporal anaphora involving multiple eventuality antecedents, we look more closely at the stages of anaphora handling. The handling of both nominal and temporal anaphora within DRT can be divided into the following two stages:'6

Rani Nelken and Nissim Francez 391

r�ading) or an optional rule of distributive expansion is applied (giving the distributive reading).

7.6

Incorporating combined events into DRY

In order to incorporate the analysis of combined events into DRT, we review the analysis of plural pronominal anaphora in DRT. In Kamp & Reyle (1 993), a summation operator symbol, ffi, is introduced into the DRT language. The corresponding function maps a pair of individual entities in the model onto their sum, a non-atomic entity. This operator is utilized in the DRS construction for sentences such as (23) as shown in Figure 8. This DRS contains the following 'new' elements: •

•

Discourse markers for combined individuals: -Such discourse markers are given capital letters, e.g. Z, U. -They may serve as arguments to (collective) · predicates, e.g. had_lousy_time (U). The summation operator ffi, which maps a pair of individual markers to their join. For instance, in this DRS the individual markers for John and Mary, u and v are summed using ffi, . to give Z.

By analogy with these constructions for the individual domain, we introduce the following extensions to the temporal DRT language. We postpone the model,.theoretic interpretation of these elements to the

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

The importance of this distinction between the stages in which each operation is performed is the following. While the occurrence of a plural pronoun provides strong evidence for a plural construction, on deeper examination, we see that it is not the trigger for the plural construction. Thus the lack of a temporal analog for a plural pronoun is not such a limiting factor. The main operation of the plural construction, i.e. joining together the individual elements does not depend on the occurrence of the plural pronoun. During the second stage of the construction, when the plural pronoun or the tense morpheme are encountered, an appropriate compound antecedent is already present. At this stage, all that is left to do is to refer to this compound antecedent. We will introduce a temporal analog of the optional distribution rule. This rule can optionally be applied, giving the distributive reading, or not applied, giving the collective reading. The only difference is that instead of having two different kinds of pronouns: singular ones, which can refer to single entities and plural ones, which can refer to compound entities, we have a single kind of tense morpheme. This morpheme behaves as a temporal anaphoric element which can refer either to a singular eventuality antecedent or a compound one.

392

The Analogy between Nominal and Temporal Anaphora Revisited u vy Z U John (u) Mary (v) Acapulco (y) take(u, v, y) Z= u $ v U=Z had_l ousy_time ( U) Figure 8

John took Mary to Acapulco. They had a lousy time.

·

•

•

•

·

Discourse markers for compound eventualities, usmg capital letters, e.g. E. EBE-the symbol for the summation operator on eventuality discourse markers, mapping the markers to their join. Compound eventualities have temporal traces, just as regular events do. These are regular temporal intervals (minimal covers of the traces of the component events). We usually denote them by capital letters, e.g. T.

Now that we. have introduced a new operation of summation on eventualities, we need to see when this operation is applied. Again, we turn to the construction rules for plural individuals for hints. Summation over individual markers is done in Kamp & Reyle (1993) in the following cases: •

•

The occurrence of k accessible individual discourse markers causes the introduction of a new plural discourse marker Z and a DRS-condition asserting that Z is the suni of the k individuals. NP conjunction causes the introduction of a plural marker, which is the sum of the two individuals of each of the conjuncts.

By analogy, we introduce the following construction rules. In these rules, we employ the following notation taken from Kamp & Reyle (1993): •

• •

A DRS K is an ordered. pair ( UK, ConK ), where UK is the universe of the DRS and ConK are its DRS-conditions. We write K ' � K if K ' is a sub-DRS of K (K ' can also be K itself). We write S(e) (VP(e)) in case e is the eventuality described by the sentence S (verb phrase VP).

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

appendix. In essence, this interpretation follows Krifka (1989), as summarized in section 7 4 ' 8

Rani Nelken and Nissim francez

393

Sentence conjunction Triggering configurations:

Operations:

( 1 ) Introduce a new plural eventuality marker, E, into UK (2) Replace y by S(E) (3) Introduce into ConK the condition: E = e 1 $E e2

Triggering configurations:

Operations :

( 1 ) Introduce a new plural eventuality marker, E, into UK (2) Replace y by V P (E) (3) Introduce into ConK the condition : E = e 1 $E e2

Eventuality summation Triggering configurations:

K' :5; K and e 1 , e2 , . . . , ek (k "2:. 2) are eventuality discourse markers occurring in K and accessible from K'

Operations:

( 1 ) Introduce a new pluni.l eventuality E into UK. (2) Introduce into ConK' the condition: E=

e l $E e2 $E . . . $E ek

Employing these construction rules, (4) is represented by the DRS in Figure 9· This DRS contains the following elements: •

events: -e1 -John's spilling his coffee -e2-Bill's tripping on the rug -E-the join of the events e1 and e2, formed as a result of the eventuality summation construction rule -e3 -Mary's sighing

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

Verb Phrase conjunction

394

The Anal ogy between Nominal and Temporal Anaphora Revisited x y z e 1 e2 e3 E t 1 t2 t3 t4 t5 T n John (x)

Bill (y)

Mary(z)

t1 < n t2 < n t3 < n e 1 �: t 1 e2 �: t2

e1

:· I spill_coffee (x) I

I

e2 : trip_on_rug (y)

t4 = ET(e 1 ) t5

=

I

ET(e2)

E = e1 $£ e2 T = ET(E) e3 �: t3 T < t3

I

e3 : sigh (z)

•

9

When John spilled his coffee and Bill tripped on the rug, Mary sighed.

times: -t4-the event time of e1 -t5 -the event time of e2 -T-the event time of the combined event E -t3-the location time of e 3 •

The temporal relation between the combined event E and the event e3 of Mary's sighing is recorded as a relation between the combined event time, T, and t3 ' the location time of e3• This relation is recorded in the DRS as a result of employing the general construction rule for temporal subordinate clauses. As in the case of single eventualities, the temporal relation is (asymmetrically) recorded as a relation between the event time of the subordinate clause (T) and the location time of the main clause eventuality (tJ Because of the way the temporal precedence relation, <, was defined, any time that follows the compound event time automatically also follows the event times of the individual events that compose the compound event. Since t3 is later than t, it is also later than both t4 and t5• Note that this DRS does not contain any information about the relative ordering of the component events, e1 and e2• The fact that Mary's sighing follows both . these events is independent of their temporal order. A similar DRS can be constructed for (s), where the temporal relation is again between the compound event time of the subordinate clause and the location time of the main clause. The only difference being the fact that in a DRS for this sentence, the temporal relation is reversed, i.e. the location time of the main clause event precedes the combined event time of the subordinate clause, and thus precedes both the event times of the component events.

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

Figure

I

Rani Nelken and Nissim Francez 395

By thus extending the semantic construction rules, we arrive at DRSs that give the right truth conditions for the examples, without having to stipulate different construction rules for each temporal connective. Rather, the only differences between the construction rules for different temporal connectives are a direct application of the regular (i.e. singular event) construction rules for the temporal connectives combined with the summation construction rules.

·

7.6. 1 Distributive vs. collective readings

(6) John gave a lecture on Sunday and on Monday. He became very tired. (24) John is · a lawyer and Mary is a lawyer. They hired a new secretary. Once again, we wish to model the construction rules for dealing with the collective/distributive ambiguity in the temporal anaphoric case according to the rules for the plural pronominal anaphoric case. Kamp & Reyle (1993) give an optional distribution construction rule. This rule operates on NPs in the subject or object position. If the discourse marker assigned to the NP is X, the S subtree is replaced by a duplex condition. This con,dition quantifies over a fresh marker x, restricts x to be part of X, and adds a new DRS-condition, applying the predicate of the VP to x. This gives a quantificational structure, according to which the predicate applies to every part x of X. In this construction rule the following 'new' symbols are used: •

•

The superscript pl on the discourse referent x denotes the fact that the referent was introduced by a plural NP. Only such referents may serve as antecedents for a plural pronoun. The symbol II is used for the 'part of relation.'9

Using this construction rule, the collective/distributive ambiguity in (24) (repeated here) is captured as follows. The pair of DRSs is given in (10). (24) John is a lawyer and Mary is a lawyer. They hired a new secretary. •

•

The first sentence is analyzed using the individual summation construction rule, introducing a new plural discourse referent Y. The pronoun 'they' introduces a new discourse referent X, which is identified with Y.

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

We now turn to the question of distributive and collective readings of sentences involving plural eventualities. We have already seen such ambiguity in discourses such as (6) and related it to the distributive/ collective ambiguity seen in discourses such as (24). Both discourses are repeated below:

396 The Analogy between Nominal and Temporal Anaphora Revisited NP Optional Distribution (Kamp & Reyle 1993) Triggering configurations:

(i) Operations:

VP

s

�

X

�

V P'

(ii) V

X

y may be replaced by the duplex condition

lxPilAQ �yu •

Now, there are two options: -The collective reading is obtained by applying the predicate to X. This reading means. that John and Mary hired a new secretary together. -The · distributive reading is obtained by applying the 'optional distribution' construction rule. This causes the introduction of a duplex condition, quantifying over a new discourse referent x, restrict­ ing it to be a part of X, and applying the predicate to x in the consequent box. This reading means that John and Mary each separately hired a new secretary.

Based on the plural NP optional distribution construction rule, we introduce the following optional distribution construction rule for eventualities: Optional Distribution for Eventualities Triggering configurations: y e () ConK Operations:

S(e') Introduce the duplex condition

�AIRpt:=e j �YI

S(e')

where e is a new eventuality discourse referent and E is a discourse marker for a compound eventuality accessible from y

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

where x is a new individual discourse referent ' and y is obtained from y by replacing X by x

...

Rani

yzX

yzXY John(y) Mary (z) Y = y E9 z

a :

Nelken and Nissim Francez 397

John(y) Mary (z) Y = y E9 z X= Y

b:

X= Y hire_new_secretary (X) Figure IO

hire_new_secretary (x)

a:

Collective reading, b: Distributive reading.

(6) John gave a lecture on Sunday and on Monday. He became very tired. •

• •

The first sentence introduces two event discourse markers for the two events of John's giving a lecture. Using the summation construction rule, their join, E, is constructed. . Now, there are two options: -E may be chosen as the Rpt, and the second sentence analyzed according to this Rpt. Since the second sentence is eventive, the event follows the Rpt, in this case the compound event, and therefore both its component events.20 x t1 t2 T e 1 e2 E n

x t1 t2 T e 1 e2 E s t n

John (x)

John (x)

e 1 !: t 1 t 1 < n e2 !: t2 t2 < n

I

I e2: I give_lecture (x) I

e 1 !: t1 t1 < n e2 !: t2 t2 < n

I

I

I

I

e 1 : give_lecture (x) a:

e2 : give_lecture(x)

e 1 : give_lecture(x)

b:

E = e 1 EBE e2

�

E = e 1 E9E e2 Rpt : = E

I

E !: s s O t t < n

s : become_very_tired(x)

a:

Rpt : = e e !: s s O t t > n

very

I

e

. Figure I I

st

s

: I become_very_tired (x) I

Collective reading. b: Distributive reading.

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

Based on this construction rule, the collective/distributive ambiguity in (6) {repeated here) is captured as follows. The pair of DRSs is given in Figure 1 1 .

398 The Analogy between Nominal and Temporal Anaphora Revisited

-The optional distribution construction rule may be applied. This will give a duplex condition, quantifying over a new event referent e, restricted to be a part of E. e may now be chosen as the Rpt for the second sentence. This gives a reading according to which each such event e is followed by an event of the type described by the second sentence.

8 QUANTIFIED NARRATIVE D ISC OU RSE

8.1

The quantification pattern of quantified narrative discourse

We have already discussed at some length the interaction of temporal anaphora and quantification over eventualities. We now wish to examine how narrative progression behaves in such quantified contexts. As an example, consider Example (2) repeated here.

(2) When he came home, he always switched on the TV. He took a beer and sat down in his armchair to forget the day. (de Swart 1 991)

This discourse contains a narrative progression of events. As in regular narrative discourse (of the typed studied extensively by Hinrichs 1 98 1, Partee 1 984, Hinrichs 1 986, and others), events seem to move time forward. What makes this discourse different than the regular narrative progression examples, is the presence of the quantifier 'always' in the first sentence. De Swart (1991) points out that in such cases there is some influence of the quantifier of the first sentence on the events described by the second sentence. In fact, the first sentence describes a habitual sequence of events. The second sentence seems to continue the sequence. De Swart (1991) does not suggest how to account for this pattern. . This sort of discourse is closely related to the 'generic passages' discussed in Carlson & Spejewski (1997). The main focus of that paper is on discourses that consist of a 'summary sentence', which describes a generically recurring event, and subsequent sentences which describe a sequence of events detailing how the recurring event occurred. An example of this kind is the following discourse (25). While the main focus of Carlson

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

In the previous sections, we have looked at several temporal anaphora phenomena. We have given an analysis of temporal anaphora, based on the analogy to pronominal anaphora, and seen how this analogy can give insight into the temporal anaphora phenomena. In this section, we look at another such phenomenon.

Rani Nelken and Nissim Francez 399 & Spejewski (I997) is this sort of discourse, they note that their analysis also carries over to discourses closer to (2), in which there is no summary sentence describing the overall recurring event.

(25)

a.

b. c. d. e.

My grandmother used to bake the most wonderful pies on Saturdays. She would go to the orchard on Sltady Lane early in the morning. She used to pick a basket each of apples and peaches. Then she would go into the kitchen and shoo everyone else away. About 4 o'clock an irresistible aroma wafted through the entire howe. (Carlson & Spejewski I997)

Hypothesis I Thefirst sentence of the discourse introduces a universal/generic quantifier. The narrative sequence of subsequent sentences falls within the scope of this quantifier.

Applying this hypothesis to the analysis of (2), would let the quantifier 'always' in the first sentence of the discourse take scope over the narrative sequence of the second sentence. This hypothesis would analyze (2) as asserting the following:

(26) For each event of coming home, there is a sequence of events of switching on the TV, taking a beer and sitting in the armchair. A similar analysis can be given for (25). While Hypothesis I gives intuitively correct truth conditions for these discourses, it also suffers from several shortcomings. Some problems with this hypothesis are shown in Carlson & Spejewski (I997)· These factors lead them to present an alternative hypothesis for such discourses. Basically, their analysis is a refinement of the following hypothesis. We will first consider the hypothesis per se and defer the treatment of the refinement to section 8.8. Hypothesis 2 The first sentence of the discourse introduces a universal/generic quantifier. Further events in the narrative sequence each introduce their own generic quantifier. In addition, previous events in the narrative sequence are added to the restrictor of the quantification through a process of accommodation.

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

Carlson & Spejewski (I 997) show that in such discourses the sequence of sentences, ex�luding the first one, forms a narrative sequence. One of the main differences between this kind of sequence and regular narrative discourses is that the events described in the narrative are understood to be recurring. How should the quantificational pattern that appears in discourses such as (2) and (25) be accounted for? The simplest hypothesis is the following:

400

The Analogy between Nominal and Temporal Anaphora Revisited

An argument against quantifiers scoping over narrative discourse

8.2

Hypothesis I cannot be maintained, if we consider the truth conditions of (27). This discourse is created from (2) by replacing the quantifier 'always' by 'usually' . (27) When he came home, he usually switched on the TV. He took a beer and sat down in his armchair to forget the day. Hypothesis I gives this discourse a reading according to which, most events of coming home are continued by the sequence of events described in .the remainder of the discourse: (28) Most situations in which he came home, are such that he did the following things: • he switched on the TV. • he took a beer. • he sat down in his armchair to forget the day. While this reading is possible, this discourse has another, more salient reading, which may be paraphrased as in (29). According to this reading, the scope of the quantifier 'usually' is restricted to the first sentence. (29) Most situations in which he came home are such that he switched on the TV. Furthermore, all the situations in which he switched on the TV. after coming home, are such that he took a beer and sat down in his armchair to forget the day. This reading is not predicted by Hypothesis 1 . Thus it cannot be the case that the quantificational force in the discourse arises solely from the

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

In this paper, we independently present a different argument against Hypothesis 1 . The consideration of different factors, lead us to develop a temporal analog ofE-type anaphora (Evans I98o, Chierchia I99S) in the NP domain. This E-type approach will lead to a solution of the problem, closely related to Hypothesis 2. While this analysis turns out to be very similar to that of Carlson & Spejewski {I 997), it may still be of independent interest for several reasons. Mainly, the introduction of temporal E-type anaphora serves to strengthen the ties between nominal and temporal anaphora even further. As we have tried to show throughout this paper, these ties provide an important link and an insightful and fruitful line of research. In addition, our E-type analysis naturally leads to the discussion of additional contexts, not previously considered by Carlson & Spejewski {I997), such as down­ wards monotonic quantifiers over events.

Rani Nelken and Nissim Francez

40 I

quantifier in the first sentence. The reading in (2 9) would be predicted by Hypothesis 2. In the next subsection we give some motivation for this hypothesis. 8.3 A temporal analog of E-type anaphora: motivation As

8 . 3 . 1 E-type anaphora An

E-type analysis of pronouns was suggested in Evans (198o) to account for sentences such as the following: (3o) a. Few congressmen admire Kennedy and are very junior. b. Few congressmen admire Kennedy and they are very junior.

The two sentences above are not equivalent. Sentence (3oa) says that the number of congressmen who both admire Kennedy and are very junior is small. Sentence (3ob), on the other hand, is different. It says that the number of congressmen who admire Kennedy is small, and that every congressmen who admires Kennedy is very junior. Evans (1980) analyzed this difference as being based on the interpretation of the pronoun 'they'. Classical accounts, which hold that (non-deictic) pronouns should be interpreted as bound variables fail to capture this difference. For instance, such an account for (Job) would introduce a variable for the pronoun 'they', within the scope of the quantifier 'few'. This would give a reading equivalent to that of (3oa), which is not the intended reading. As an alternative to the bound variable account, Evans (1 980) suggested that there is another type of pronouns, which he termed E-type pronouns. Instead of introducing a variable, these pronouns should be interpreted as definite descriptions, which select the individuals that satisfy the antecedent clause. 8.4 A DR T implementation of E-type anaphora Within the DRT framework, Kamp & Reyle (1993) deal with E-type anaphora using an abstraction operator. This operator maps a set of individuals satisfying some DRS onto their i-sum:

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

we have seen, the Hypothesis I fails to predict the required reading of discourses such as (27). How can we account for these readings? Once again, we turn to NP anaphora in search of an analogous phenomena. Such an analog is found in the form of E-type anaphora2' (Evans 1 980, Chierchia 1 995). We introduce a temporal analog of E-type anaphora, after giving a brief review of E-type anaphora in the NP domain.

4oi The Analogy between Nominal and Temporal Anaphora Revisited NP Abstraction (Kamp & Reyle 1 993)

T•;ggoring configu.-ations: y r;;;!y e ConK :

� 1 CJV

K2

Form the union K0 = K1 u K2 of the two component DRSs of the condition. Choose a discourse referent w from UK · o Introduce into UK a new discourse referent Y and add to ConK the condition Y = I:w : K0

Operations:

(3 1 ) Susan has found most books which Bill needs. They are on his desk. In the DRS for this discourse, abstraction has been applied to the duplex condition. This is represented by using the 2: operator. The discourse marker Y is the i-sum of individuals y that satisfy the sub-DRS to the right xzYUwt Susan (x) Bill (z) .

�

bo (y) need (z, y)

� Y

� have_found(� y) l y

Y = I:y

book (y) need (z, y) have_found (x, y)

U= Y on(U, w) t's desk (w) t=z

Figure

12

·

Susan has

found most books which Bill needs. They are on his desk.

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

Using this operation, it is possible to represent discourses such as (3 r ). In this discourse, the plural pronoun 'they' refers to the books that Susan found. What this discourse asserts is not that Susan found most books that are both needed by Bill and are on his desk. Rather, she found most of the books that Bill needs, and all of these are on his desk.

Rani Nelken and Nissim Francez 403

of y. The previously discussed rule of distributive expansion may also be applied to i-sums which have been introduced through an operation of abstraction. For example, if we continue Discourse (3 1) with (32), we add another duplex condition to the DRS as shown in Figure 1 3, asserting that each u s.t uiiU was. read by Bill. (32) He

has

read them.

This DRT analysis captures the· spirit of E-type prono�ns. In the discourses discussed, the i-sum of the individual elements that satisfy the antecedent clause is constructed. This i-sum can be referred to either collectively, as in (12) or distributively as in (1 3).

E-type temporal anaphora

After seeing how E-type anaphora in the NP domain can be handled in DRT, we present an analogous construction for the temporal domain. It is our view that the phenomenon encountered in the short discourse (27), is similar to that encountered with E-type pronouns as in (3o). In both cases, an analysis that sees the anaphoric element (whether pronominal or temporal) as a bound variable inside the scope of a quantifier fails to give the required readings. In analogy to the nominal domain, we view the temporal anaphoric element not as a bound variable, but rather as a definite description. The basis of this analysis is the introduction of an abstraction operator on events, EE. The model-theoretic interpretation of this operator is given in the appendix. In sequences such as (2), we first introduce a duplex condition for the first sentence, according to the BSA We then apply abstraction on this duplex condition. Abstraction introduces a new discourse marker for the sum of the events that satisfy the duplex condition. Once this new discourse marker is introduced, it can be referred to either collectively or distri­ butively, using the construction rules presented in section 7.6.1. The construction rule for abstraction is the following. It is based on the similar rule for NP abstraction above.

JulA, �� Figure

I3

have_read (�

He has read them.

•> I

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

8.5

404

The Analogy between Nominal and Temporal Anaphora Revisited Event Abstraction

Triggering configurations: y r; ye C onK : Operations:

Fonn the union K0 = Ki u K2 of the two component DRSs of the condition. Choose an event discourse referent e ' from UK . Introduce into UK a new discourse ;eferent E and add to ConK the condition E = I:.Ee ' : Ko

(33) When he came home he always switched on the TV. He took a beer.

e l tz

1

ez t3

t2 = ET(e1 )

e 1 : come_home (x)

ez r; t3 tz < t3

�

I

e2: I sw._on_tv.(x)) I e 1 12 ez '3

E I = r.E ez

l

t2 = ET(e 1 )

e 1 : come_home (x)

l

ez r; t3 tz < t3

ez: sw._on_tv. (x)

I j

e3 t4

�� .

Figure

14

e3 r; t4 Rpt : = e2 ez < e3

ez

e3:

1

take_beer(x)

I

When he came home he always switched on the TV. He took a beer.

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

For the sake of simplicity, we illustrate this analysis by constructing a DRS not for (2) but for the shorter discourse (3 3), in Figure 1 4.

Rani Nelken and Nissim Francez 405

This DRS

in

is constructed as follows:

1. By the BSA, a duplex condition is constructed for the first sentence. 2. Abstraction is applied over the marker ez, using the operator :EE. This operator maps the set of events ez that satisfy the DRS to its right onto their sum, E1• 3· The sum E1 is now chosen as the antecedent of the temporal anaphora. 4· Applying the optional rule of distributive expansion as given in 7.6.1 , · another duplex condition is added. This condition asserts that each event ez that is a part of E1 is followed by an event e3 of taking a beer. s. The temporal relation between the events ez and e3 is restricted using the Rpt, as is done in cases of regular narrative progression.

(34) Gohn had a dental appointment yesterday.) Whenever the dentist operated the drill, John started yelling. He became hoarse. In this discourse, the event of John's becoming hoarse follows the compound event composed of all the component events in which John yelled. Note also that by the Event Abstraction construction rule we are allowed to abstract over any event discourse referent in the union of the two parts of the duplex condition.22

8.6

Applying the analysis

to

(usually'

The E-type analysis presented above gives the correct truth conditions for discourses, in which the quantifier 'always' is replaced with the quantifier 'usually'. We illustrate this by constructing a DRS for the following simplification of Example (27).

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

In this example, we have applied abstraction, giving the compound event E1 only to later distribute over each of E1 's component events. Thus, abstraction plays only an intermediate role in the construction. However, this role is still important since we assume that the temporal relation between the two sentences involves temporal anaphora. In order for anaphora to be applied, there must be a suitable accessible antecedent, represented by a discourse marker in the DRS. The event markers in the duplex condition are not accessible from outside the duplex condition. The intermediate step of applying abstraction makes them accessible. There are cases similar to the collective constructions of the previous section, where the compound event plays a bigger role. One such example is the following discourse in which we need to collectively refer to the compound event:

406 The Analogy between Nominal and Temporal Anaphora Revisited

. e l tz ez t3

I

e 1 : come_home (x) e2 :

I

ez !;; t3 tz < t3

sw._on_tv. (x)

I I

e3 t4 e3 !;; t4 Rpt : = e2 ez < e3

take_beer(x)

Figure

I5

When he came home, he usually switched on the TV. He took a beer.

(3 5) When he came home, he usually switched on the TV. He took a beer. The DRS for this discourse would be similar to that of Figure I 4 except that the first conditional DRS-condition of the type K1 =? K2 would be replaced by a duplex condition as shown in Figure I 5· In this DRS, the event in the second sentence of taking a beer falls outside the scope of the quantifier 'usually'. Thus the wrong prediction of Hypothesis I is avoided.

8.7

Extending the analysis for additional quantifiers

we have seen above, the proposed E-type analysis gives the required truth conditions for both 'always' and 'usually'. A natural question ts whether the same pattern is shared by other quantifiers as well. As

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

£1 = r.Eez

t2 = ET(e 1 )

Rani

Nelken and Nissim Francez 407

8 .7. 1 An apparent problem with

downwards monotonic quantifiers

it turns out, downwards monotonic quantifiers over events sometimes yield a different quantificational pattern. If we place a downwards monotonic quantifier in the first sentence of such discourses, the narrative sequence is not understood as following the main clause event in the first sentence, but rather as occuring instead of it. This is illustrated by (36).

As

(36) When he came home he rarely switched on the TV. He took a beer and sat down in his armchair to forget the day.

(37) Few situations i� which he came home are such that he switched on the TV. Furthermore, all situations in which he switched on the TV. after coming home, are such that he took a beer and sat down in his armchair to forget the day. (38) Few of the situations in which he came home are such that he switched on the TV. Instead, most situations in which he came home, are such that he took a beer and sat down in his armchair to forget the day. Also, note that the events of the second clause cannot be understood as falling within the scope of the quantifier in the first sentence. The same quantificational pattern appears if we replace 'rarely' with 'never'. Proceeding according to the construction rules introduced for upwards monotonic quantifiers, would thus seem to give wrong truth conditions for downwards monotonic quantifiers. We would introduce a duplex condition with the quantifier 'few', and then abstract over those events that satisfy both the left and the right boxes of the duplex condition. This would give reading (37). . There is a genuine difference between the quantificational pattern exhibited by (3 6), and that exhibited by (3oa). In both cases, a downwards monotonic quantifier is present. However, the quantificational pattern is different In (3oa), the pronoun 'they' referred to these few congressmen who admire Kennedy. In (36), we cannot accept those few situations in which he came home and switched on the TV. as the antecedent of the temporal anaphora. On first sight, this difference appears to be a difference between the nominal and temporal domain with respect ·to downwards monotonic quantifiers. However, we wish to show that both quantificational patterns may be found both in the nominal and the temporal domains. Also, while

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

This discourse cannot have the paraphrase in (37). It should rather be understood as in (38). We call these readings the E-type and non E-type readings respectively.

408 The Analogy between Nominal and Temporal Anaphora Revisited

the problem becomes most apparent with downwards monotonic quanti­ fiers, both kinds of quantificational patterns occur with all kinds of quantifiers. We show such examples for both domains. 8.7.2 Ambiguity in discourses containing quantification over NPs

(39) a. Few Computer Science students are on the volleyball team. They just don't have time to practice. b. Few piano players practice Karate. Their hands are too delicate. c. Most people who live in the suburbs have a barbecue on Sunday. They commute to work on Monday. ·

In (39a) the pronoun 'they' is most naturally understood as referring not to the set of few CS students who are on the volleyball team, but to the set of all such students. Similarly, in (39b), 'their hands' more naturally refers not to the hands of those few piano players that do practice Karate, but rather to those of the entire set of piano players. In (39c), 'they' most naturally refers to the set of people who live in the suburbs, not necessarily those who have barbecues. The E-type reading is also acceptable in these discourses. Replacing the quantifier with 'no' blocks the E-type reading altogether. 8.7.3 Ambiguity in discourses containing

quantification over events

Discourse (27), which contained an · upwards monotonic quantifier was ambiguous between an E-type and a.non E-type reading. We have already seen examples of discourses with downwards monotonic quantification that exhibit the E-type reading. The following discourse has the complementary quantificational pattern: (4o) When John was an alcoholic, he rarely accepted a drink. He drank it in one shot and felt guilty all evening.

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

In the classical examples ofE-type anaphora the plural pronoun 'they' refers to the set of elements that satisfy the DRS composed of both parts of the duplex DRS-condition. However, there are also discourses with similar structure, which have the non E-type reading. In these discourses, it seems the pronoun 'they' refers to those elements that satisfy just the antecedent box. Here are two such discourses:

. Rani

Nelken and Nissim Francez

409

8 .7.4 Our solution As shown above, it is possible to find examples exhibiting both quantifica­

8.8

Long sequences of quantified narrative discourse

As we have already mentioned, Carlson & Spejewski (1997) do not adopt Hypothesis 2 per se. In fact, they introduce a refinement based on the conclusion that this hypothesis is too strong: We will now examine this refinement and its implications for our analysis. Carlson & Spejewski (1997) believe that Hypothesis 2, which requires each different sentence to introduce its own generic quantifier and to include in its restrictor all the preceding events, is too strong. The reasoning

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

tional patterns, both in the NP domain and in the temporal domain, regardless of the direction of monotonicity of the quantifier. We have already seen how the E-type reading can be accounted for using abstraction. How should we account for the non E-type readings? One way to solve this would be to stipulate a rule that would allow the abstraction operation defined above to operate not only on the union of both boxes of the duplex condition, but optionally just on the left one. This rule would allow readings in which the antecedent, constructed by the abstraction operation, would be just the set that satisfies the restrictor of the quantification. Optional application of this rule would give the non E-type readings. Allowing such a rule would also explain the case of 'no' (for NPs) or 'never' (for events). In discourses containing these quantifiers, this optional rule must be applied since the alternative E-type interpretation leads to vacuity. The regular abstraction rule calls for constructing the union of both sides of the duplex condition. However, when the quantifier in the duplex condition is 'no' there can be no individual (or event), which satisfies . both boxes. Therefore, it would be vacuous to claim something about such individuals (events). We assume that such vacuous readings are ruled out.23 While we saw some examples exhibiting both kinds of quantificational patterns, it is not the case that the different readings are equally salient in all cases. For example, it seems that there is a clear preference for non E-type readings in the temporal domain when the quantifier is downwards monotonic. The question of when this optional rule should be applied is a matter for further research. However, we believe that there is a strong connection between the · understanding of the principles behind such a rule in the NP domain to those in the temporal domain. A better understanding of one is bound to shed light on the other.

410 The Analogy between Nominal and Temporal Anaphora Revisited is as follows: consider situations in which an event described by one of the sentences does not occur. In such cases, we still have relatively strong intuitions regarding the occurrence of the other events. For example, they consider the following variant of the 'Grandma passage' (25) above:

(41)

a.

b. c. d. e.

Now, consider a situation in which (e) does not hold, i.e. the grandmother bakes a cake going through the steps (b-d), but does allow the dog to stay. In such a situation, sentence (f) is still required to hold, i.e. even if the dog stays, it is still the case that an irresistible aroma wafted through the house. Note that whether or not (f) is required to hold depends on the relation between the two events described by (e) and (f). If (f) were a subordinate event to (e), then (f) would be required to hold just when (e) does occur. This can be seen by considering the following continuation (f') instead of (f): (42) f'. She would wield the broom in a menacing fashion and yell 'Get out!'. This consideration leads Carlson & Spejewski (1997) to refine their analysis. The revised analysis does not require a quantificational dependency between each pair of subsequent events in the narrative. Instead, they claim that such a dependency is required only between subordinate and superordinate events and is left optional between pairs of adjacent events. The framework of Spejewski ( 1994), and Carlson & Spejewski (1997) allows the representation of such subordination relations between events. Based on this information, it is possible to deduce quantificational dependencies. The determination of subordination relations between events is based on human reasoning and is not formulated as part of the computational DRS-construction algorithm. In the account presented in this paper, we do not have access to this information. How can we account then, for the quantificational dependencies in the previous discourse? In the consideration of short quantified narrative discourse above, we chose the temporal anaphoric antecedent of the second sentence as the main event of the first sentence.2.4

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

£

My grandmother used to bake the most wonderful pies on Saturdays. She would go to the orchard on Shady Lane early in the morning. She used to pick a basket each of apples and peaches. Then she would go into the kitchen and shoo everyone else away. Then she would sometimes shoo the dog out too. About 4 o'clock an irresistible aroma wafted through the entire house. (Carlson & Spejewski 1997)

Rani Nelken and Nissim Francez 41 1

(43)

a.

b. c. d. e.

£

My grandmother used to bake the most wonderful pies on Saturdays. She would go to the orchard on Shady Lane early in the morning. She used to pick a basket each of apples and peaches. Then she would go into the kitchen and shoo everyone else away. Then she would shoo the dog out too. About 4 o'clock an irresistible aroma wafted through the entire house. ·

Consider a situation in which (e) doesn't hold, i.e. Grandma did not shoo the dog out. In such a situation, we still expect (f) to happen. This is explained by Carlson & Spejewski (1997) once again by their use of . temporal subordination information. Since (f) is not subordinate to · (e), there is no quantificational dependency between them. In our framework, we do not have access to this subordination information. Again, we can solve this problem using the non-deterministic choice of temporal anaphoric antecedent. Here, we could choose the event described by (d) as the ante�edent for (f). We do not provide a procedure for choosing between antecedent candidates. Such a procedure could well take into account such subordination information.

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

As in the general procedures for anaphora resolution, the choice of proper antecedent in case there are many possible candidates is non-deterministic. Here too, in case of a long narrative discourse, we allow a non-deterministic choice of temporal antecedent over which abstraction is applied. In such cases, when we interpret some sentence in the narrative, instead of picking out its immediate predecessor as the antecedent of the temporal anaphora, we may choose a previously described event in the discourse. The determination of how the choice is made depends on general discourse mechanisms. For example, the mechanisms of Spejewski (1994) and Carlson & Spejewski (1997) for the representation of event subordina­ tion relations can be taken into account here in order to correctly choose the suitable events over which to apply abstraction. Thus, in essence we are adopting a refined version of Hypothesis 2, similarly to the choice made by Carlson & Spejewski (1997). Another point raised by Carlson & Spejewski (1997) has to do with the truth conditions of discourses such a5 the 'Grandma passage' above. They claim that for such discourses, even in situations in which one of the sentences does not hold, we still have pretty clear intuitions about the occurrence of the rest. For example, consider the following variant of the passage, which is the same as (41), omitting the quantifier 'sometimes' from (e):

412 The Analogy between Nominal and Temporal Anaphora Revisited

In conclusion of this section, we have presented a temporal analog of E­ type anaphora. This analysis further strengthens the ties between nominal and temporal anaphora. 9 C ON CLUSI O N

Acknowledgments We wish to thank two anonymous referees for their comments on this paper. The work of the second author was partially supported by a grant from the Israeli Ministry of Science 'Programming Languages Induced Computational Linguistics', and by the fund for the Promotion of Research in the Technion. RANI

NELKEN and NISSIM FRANCEZ Computer Science Department The Technion Haifa JZOoo, Israel e-mail: {nelken,Jrancez}@cs.technion.ac.il

Received: 19.01.1996 final version received: 26.06.1998

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

In this paper we have presented an analysis of sentences and short discourses that contain temporal anaphora and quantification over events. Accepting both the BSA and Hinrichs' POA as basic plausible assumptions, we have shown how they can be integrated into a single account which is a direct extrapolation of the analysis of temporal anaphora in non-quantified sentences. We adopt Kamp & Reyle's split in the roles of the reference time. The apparent 'overloading' of too many roles on the single reference time mechanism seems to have complicated things rather than making them simpler. Based on this split, we propose a DRT solution for PQP. This solution achieves the existential quantification, which has to be explicitly stipulated in de Swart's solution of PQP, 'for free'. This split further allows an enhancement of the analogy ofNP-anaphora and temporal anaphora. Plural temporal anaphora is presented as an analog of plural NP-anaphora. We show how the similarity between Link's analysis . of NP plurals using i-sums, and Krifka's lattice-theoretical analysis of events, allows a new analysis of plural temporal anaphora. This view of plural temporal anaphora better explains how one eventuality can be anaphorically linked to more than one eventuality or time. Such an analysis · is difficult within previous accounts. A final issue discussed in this paper is narrative discourse in quantified contexts. In discourses where the first sentence is quantified, following sentences that continue the narrative cannot be construed to be within the scope of the quantifier. By a direct analogy with Evans' E-type analysis of NP anaphora, we are able to analyze such cases.

Rani Nelken and Nissim Francez 4 I 3

APPENDIX: M ODEL THEORY The model-theoretic interpretation. of the constructions considered here is based on the model theory in Kamp & Reyle (I993: part 2, 676-8), with a few modifications, which reflect the semi-lattice based structure of the event and temporal domains, taken &om Krifka (I989). A Model M should include the following: The event domain EV (EV, ffiE) is a complete join semi-lattice without bottom element,'1 where EV is a set of eventualities and ffiE is the join operation of the semi­ lattice. For all eventualities e. , e, E EV, the following relations are defined: -part of: e,IIEe2 iff e, ffiE e, e, -overlap: e, 0 e2 iff 3e3 (e3IIEe, 1\ e3IIEe,) The time domain T is an atomic complete join semi-lattice ( T, ffiT, <, at(-)), where · T is a set of periods, at(t) holds for an atomic period t, called an instant. Instants are linearly ordered by the relation < (temporal precedence), and ffiT is the join relation of the semi-lattice. Let t, , 12 be elements of T, then: -ITT is the corresponding 'part of relation. -The temporal precedence relation is extended from instants to arbitrary periods by: t, < t, iff 'v't�t� E T [at(t� ) 1\ at(t�) 1\ t�IITt, 1\ t�IIT12 -+ t� < t�J -11 � 12 iff 'v't' E T[at(t') 1\ t'IITt, t'IITt,J ET is a function from EV to T, s.t. -ET is a homomorphism relative to the joins of EV and T. -e 0 e' 3t E T[at(t) 1\ tiiTET(e) 1\ tiiTET(e')J Using ET, we define the following relations: -temporal precedence-for all e, , e, E EV: e, < e, iff ET(e, ) < ET(e,) -temporal overlap-for any eventuality e and time t: e 0 t iff 3t' E T[at(t') 1\ t'IITt 1\ t'IITET(e)J -temporal inclusion-for any eventuality e and time t: e � t iff ET(e) � t. It is a simple matter to check that the event domain with the definitions of temporal precedence and temporal overlap satisfies a set of seven axioms which appear in Kamp & Reyle (I993: 667) and go back to Russell (I956). The embedding conditions of DRSs is the same as in Kamp & Reyle (I993) with the following additions, for model M and embedding functionf, DRS K, and events e, e' E EV: M F! E e ffiE e' iffJ(E) J(e) ffiEJ(e') in M M F! eiiEe1 iffJ(e)IIEJ(e') in M M FJ E I:EeK iffJ(E) = EeE {"e' : e' E EV 1\ M FJ U (•,•') K } •

=

=

•

-+

•

•

=

=

• •

=

N OTES I

Preliminary version presented at EACL '95, the Seventh Meeting of the Euro­ pean Chapter of the Association for Computational Linguistics, Dublin, Ireland, 99 5. I

Some familiarity with DRT is assumed. We thank an anonymous referee for suggesting example (4). 4 Note, however, that such sentences are sometimes ambiguous as is the following: 2 3

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

-+

•

4I4 The Analogy between Nominal and Temporal Anaphora Revisited (Io) Before John left town, he always wore a tie.

8

For simplicity, we disregard here the usual aspect of NP anaphora represen­ tation, which in full detail introduces another marker, y and the condition

x = y.

9 We thank an anonymous referee for emphasizing this point. Io de Swart {I99I) also adopts Chierchia's

suggestion. I I Since this property of the location time is the crucial factor in the solution, it may be possible to solve PQP using a unified reference time that has this

property. While this may be possible,

it requires the formulation of construc­ tion rules not only for the problem at

here. I 3 Because of typographical considerations, we have split the DRS into two. The sub-DRS labeled sl should replace

'ZOOM'.

.

I 4 There are discourses in which a sequence of eventualities, mentioned in the subordinate clause, do form a narrative progression. Such cases are well handled by our analysis. Since the narrative progression is included in the subordinate clause, it is handled by a ·

sequence of progressing Rpt's within the antecedent box. Because of the universal quantificational force conferred by the antecedent box, this would seem to lead once more to PQP. . However, there is no such problem, because the universal quantification is only on the sub­ ordinate clause events (unlike Partee's analysis which yields universal quanti­ fication over the reference time of the main clause event). Thus the quantifi­ cational pattern may be paraphrased as follows: 'for each set of events/times of the subordinate clause, there is an event of the main clause', thus avoiding the problem.

I 5 The justification of the asymmetry of using the location time for the main clause event vs. the event time of the subordinate clause event has been discussed in section 3· I6 The

o�der

of these

operations

is

reversed in kataphoric constructions. I7 Note however, that the analysis of Hinrichs and Partee differs slightly from this template. Hinrichs' (see

POA

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

This sentence can refer to one event of John's leaving town, asserting that before it John always wore a tie. Alter­ natively, it can quantify over all events in which John left town, asserting that each one is preceded by an event of his wearing a tie. The ambiguity sterns from the different possible relative scopes of the temporal connective (before) and the quantifying adverbial (always). In this paper we will concen­ trate only on the latter reading. We believe the former reading is well handled by an analysis along the lines of Kamp & Reyle {I993)· 5 Kamp & Reyle {I993) use the relation name 'loc' where we use 'ET'. We prefer to use this relation name to avoid confusion between the event time and the location time. 6 Since the utterance time, n, is a point in Kamp & Reyle (I993), the overlap relation between a state that holds in the present and n degenerates to inclusion. 7 One of the consequences of splitting the different roles of the reference time is that the Rpt is used in narrative pro­ gression, but not in other constructions, e.g. not in the relation between main and subordinate clause eventualities.

hand but also for the array of pheno­

mena discussed by Kamp & Reyle {I 993)· We believe such rules will tum out to be more complex than the ones presented here. I2 Note however, that this analysis does not sufficiently take care of discourses comprised of sequences of sentences in the past perfect. Such discourses are handled by using Karnp & Reyle's TPpt. For the sake of simplicity, we omit the details of these constructions

Rani Nelken and section 2. I separates the processing of the antecedent clause (containing the antecedent) from the processing of the main clause (containing the anaphoric

I8 I9

20

In this presentation, we focus on events. We believe that states may be handled along similar lines. We prefer to use the symbol II instead of the symbol E, used by Kamp & Reyle (I993) for the 'part of' relation. Strictly speaking, not only is E available as a candidate Rpt, but ·so are e, and e,. Recall that, according to Kamp & Reyle (I993), the Rpt is chosen non­ deterministically as some event or time discourse marker already present in the DRS. Thus strictly speaking, both e, and e, remain viable options for the Rpt, and indeed there may be contexts in which such readings are available. The reason for preferring E as an Rpt

2I 22

event is chosen as the Rpt, here too we may choose only E. In discourses in

which the narrative progression is vio­ lated, we may indeed use other events as the Rpt. The resemblance to E�type anaphora was brought to our attention by Shalom Lappin. Recall that the rule allowed the choice of any e' from UK.. This non­

determinism is copied from the parallel NP abstraction rule, where the NP

abstracted over may be in the object position, e.g. 'Few swimmers wear a wristwatch. They are ruined by water'. . In this discourse, 'they' refers to the wristwatches, not the swimmers. In the temporal case, since the discourse is a narrative progression, we usually choose the latter event as the antecedent. This

23

·

is that we are focusing on narrative sequences. Just as in ordinary narrative sequences the immediate predecessor

4I s

freedom becomes useful in cases where there is anaphoric reference to the subordinate clause event. Kibble (I996) independently discusses the similarity between pronominal E-type anaphora, temporal adverbials and modal subordination. He also notices the quantificational pattern arising most commonly with down­ wards monotonic quantifiers, but attri­ butes it to the controversial operation of anaphoric reference to the complement set. In our solution, we do not allow anaphoric reference to the complement set, but allow reference to the set over

24 25

which quantification is applied. We focus here on the E-type reading.

The presence of a bottom element would

undesirably

cause

eventualities to overlap.

any

two

REFERENCES Bach, E. (I98 I), 'On time, tense and aspect: an essay in English metaphysics', in Peter

MS, reproduced by the Indiana Univer­ sity Linguistics Club, I978.

Cole (ed.), Radical Pragmatics. Academic Press, New York, 63-81.

Carlson, G. N. & Spejewski, B. (I997), 'Generic passages', Natural Language

the logic of tense and aspect in English',

Chierchia, G.

Bennett, M & Partee, B .

(I972),

'Towards

Semantics,

5,

2, IOI-6s. (I995), Dynamics of Meaning:

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

element). However, according to their analysis, an event clause advances the reference time. This is done during the processing of the antecedent (stage I). In Kamp & Reyle (I993), the Rpt is chosen when processing the anaphoric element (stage 2), as a time or event marker already present in the DRS. Thus, the analysis of of Kamp & Reyle (I993) is more consistent with the general scheme of anaphora handling.

Nissim Francez

416 The Analogy between Nominal and Temporal Anaphora Revisited

Kamp, H. & Reyle, U. ( 1993), From Discourse to Logic, Vol. 42, Studies in Linguistics and Philosophy, Kluwer Academic Publishers, Dordrecht. Kibble, R ( 1996), 'Modal subordination, focus and complement anaphora', in Proceedings of the Tbilisi Symposium on Language, Logic and Information. Kri£ka, M. ( 1989), 'Nominal reference, temporal constitution and quantification in event semantics', in R Bartsch et al. (eds), Semantics and Contextual Expression, Foris, Dordfecht, 75-1 16. Lasersohn, P. ( 1992), 'Generalized con­ junction and temporal modification', Linguistics and Philosophy, 1 5 , 4. 38 1-410. Link, G. ( 1983), 'The logical analysis of plurals and mass terms: a lattice­ theoretical approach', in R Bauerle et al. (eds), Meaning, Use and Interpretation of · Language, de Gruyter, Berlin, 302-23. Moens, M. & Steedman, M ( 1 988), 'Tem­ poral ontology and temporal reference', Computational Linguistics, 14, 15-28. Partee, B. ( 1973 ), 'Some structural analogies between tenses and pronouns in Eng­ lish', journili of Philosophy, LXX, 601-9. Partee, B. ( 1984), 'Nominal· and temporal anaphora', Linguistics and Philosophy, 7, 243-86. Prior, A ( 1967), Past, Present and Future, Oxford University Press, Oxford. Reichenbach, H. ( 1947), Elements ofSymbolic Logic, reprinted by the Free Press, New York, 1966. Russell, B. ( 1956), 'On order in time', in R C. Marsh (ed.), Logic and Knowledge: Essays 1901-1950, G. Allen & Unwin, London. Spejewski, B. ( 1994). 'Temporal sub­ ordination in discourse', Ph.D. thesis, University of Rochester, NY.

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

Anaphora, Presupposition and the Theory of Grammar, University of Chicago Press, Chicago. de Swan, H. ( 1991 ), Adverbs of Quantifica­ tion: a Generalized Quantifier Approach, dissertation, University of Groningen. Published ( 1993 ) by Garland, New York de Swan, H. & Molendijk, A ( 1994), 'Negation in narrative discourse', in H. Bunt, R Muskens, & G. Rentier (eds), Proceedings of the International Workshop on Computational Semantics, Institute for Language Technology and Artificial Intelligence, Tilburg, 191-200. Evans, G. ( 1980), 'Pronouns', Linguistic Inquiry, I I , 3 37-62. Heim, I. ( 1982), 'The semantics of defi­ nite and indefinite noun phrases', Ph.D. thesis, University of Massachusetts, Amherst. Heinamiiki., 0. ( 1978), 'Semantics of English temporal connectives', Indiana University Linguistics Club, Bloomington. Hinrichs, E. W. ( 1981 ), 'Temporale anaphora im Englischen', unpublished staatsexamen thesis, University of Tue­ bingen. Hinrichs, E. W. ( 1986), 'Temporal anaphora in discourses of English', Linguistics and Philosophy, 9, 63-82. Kadmon, N. ( 1990), 'Uniqueness', Linguistics and Philosophy, 1 3, 273.:.324. Kamp, H. ( 1979), 'Events, instants, and temporal reference', in R Bauerle, U. Egli, & A von Stechow (eds), Semantics from Different Points of View, Springer, Berlin, 376-417. Kamp, H. ( 198 1 ), 'A theory of truth and semantic representation', in T. Janssen, J. Groenendijk, & M. Stokhof (eds), Formal Methods in the Study ofLanguage, Mathe­ matical Center, Amsterdam, 277-322.

© Oxford University Press

1997

Good News about the Description Theory of Names BART GEURTS University of Osnabriick and Max Planck Institute for Psycholinguistics, Nijmegen

Abstract

INTRODUCTION At the end of the first lecture of Naming and Necessity, Kripke gives short shrift to Kneale's version of the description theory of names, according to which the meaning of a name N is 'the individual named N'. I have always felt that Kripke's criticism of this view falls wide of the mark, and that Kneale's position is essentially correct. In the following pages I try to justify this assessment, referring to Kneale's theory and its kin as 'quotation theories'; for their central claim is that the content of a name quotes the name itsel£ Although occasionally Russell came quite close to defending the quotation theory, to the best of my knowledge its first proponent was Kneale (1962), who was taken to task for this by Kripke (1980). Subse­ quently the theory received support from Loar (1976), Bach (198 1, 1 987), Cresswell (198 5), and Fodor (1 987). More recently, a number of presupposi­ tion theorists have swelled the ranks of the quotation theory: it is more or less taken for granted by van der Sandt and Geurts (1991), van der Sandt (1992), Beaver (1993), and Geurts (1995), among others. And Zeevat (1996) argues at some greater length for a presuppositional version of the quotation theory that is more or less the same as mine. My defence of the quotation theory is part of a larger project, which is to provide an alternative for semantic analyses built upon such notions as rigid designation or direct reference. Despite their considerable intuitive appeal, I believe that these notions are red herrings, regardless whether they are deployed in the analysis of names, demonstratives, natural kind terms, or wherever. But in this paper the focus will be entirely on proper names. •

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

This is an attempt at reviving Kneale's version ofthe description theory ofnames, which says that a proper name is synonymous with a definite description of the form 'the individual named so-and-so'. To begin with, I adduce a wide range of observations to show that names and overt defmites are alike in all relevant respects. I then turn to Kripke's main objection against Kneale's proposal, and endeavour to refute it. In the remainder ofthe paper I elaborate . on Kneale's analysis, adopting a theory of presupposition proposed by van der Sandt.

po Good News about the Description Theory of Names 1 NAMES AND OTHER DEFINITE DESCRIPTIO NS

•

•

Names often take the form of definite NPs: 'the United Nations', 'the Goldberg Variations', 'the Netherlands', 'the Annunciation', 'the Holy Spirit', 'the B:mk of England'. These are names, no doubt, but they certainly look like definite NPs. In English, river names always carry a . definite article ('the Mississippi'). In Italian, names of women often have, and sometimes must have, a definite article: 'la Loren', 'la Carolina'. I suspect that all languages which have definiteness markers allow them to occur on proper names, and I wouldn't be surprised if there were languages in which names were always marked for definiteness." So-called definiteness effects apply to names and definite NPs alike. On the one hand, if a construction selects for indefinite (or weak) NPs, names as well as definite NPs are excluded. This holds for English there-sentences, for example: (1) There is {1ohn/*the philosopher/a philosopher} available. On the other hand, if a construction selects for definite NPs, it will accept names, too, as the case of the partitive construction illustrates:

•

(2) half of {Belgium/the country/*some countries} As Kripke ( 1980) points out, Donnellan's referential/attributive distinc­ tion applies to names just as it applies to definite NPs. In Kripke's

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

I claim that a name is synonymous with a definite NP of the form 'the individual named so-and-so'. For the time being we can leave the proposed analysis in this somewhat underspecified state, because I first want to raise a number of points that don't require a more precise formulation. If the quotation theory is correct, then names must be expected to be used and interpreted like other definite NPs. If. on the other hand, names are rigid designators, then we should expect significant empirical differ­ ences between names and definite NPs. I will now show that it is the former prediction rather than the latter that is borne out by the facts. The main objectives of this section are to take stock of the semantic properties of names and definite NPs, and to show that there do not seem to be any fundamental differences between these two types ofexpressions. I will freely help myself to whatever technical jargon I find convenient for these purposes. For example, I will sometimes pretend that names and definites have scope but will also talk of their 'referents'. This terminological bric a brae should not be taken too seriously. Taken together, the following observations strongly suggest that names pattern with (other) definite NPs in all relevant respects:

Bart Geurts 3 2 I

•

•

example, two interlocutors observe Smith at a distance and take him to be Jones. Accordingly, they use the name jones' to refer to Smith (although Kripke insists that semantically speaking they are referring to Jones). Names can be used literally as well as non-literally (this point is clearly related to the previous one, but it is not exactly the same). For example, a man who jocosely refers to his wife as 'the Queen' may use the name 'Cleopatra' to much the same effect (c£ Bach 1987). Names are like overt definites in that they can be used generically:3

•

Obviously, in these examples the names 'Coca Cola' and 'Pristichampsus' are used generically. It is true that it is difficult to think of a plausible scenario in which a proper name like john' will have a generic interpretation, but this has nothing to do with the fact that john' is a name. It is just that the kind that a generic occurrence of john' would denote is of such limited use. It is precisely in this respect that 'Coca Cola' and 'Pristichampsus' are different from john', and this is why the examples above are felicitous. Definite NPs and names are typically, though not always, used to refer to objects that are part of the common ground between speaker and hearer. We may distinguish two cases here. On the one hand, definites and names may both be used anaphorically to refer to objects that have been explicitly introduced into the discourse. For example, (s) I have a poodle named 'Horace'. {Horace/My poodle} is three years old.

•

On the other hand, names and definites may also be used non­ anaphorically to refer to individuals that are given in the larger situation in which the discourse takes place. For example, in some contexts either 'the girl from no. 21' or julia' can be used to refer to my daughter, and in England 'the Queen' and 'Elizabeth II' will typically refer to one and the same person. Definites and names both have bound-variable uses: (6)

If I can choose between a Mercedes and a BMW, I'll take the BMW. b. If a child is christened 'Bambi', then Disney will sue Bambi's parents. a.

The definite NP in the consequent of (6a) is bound to the indefinite in the antecedent, and the same applies for the second occurrence of'Bambi'

.

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

(3) {The light bulb/Coca Cola} was invented by an American. (4) {The panzercroc/Pristichampsus} hunted during the Eocene Epoch, about 49 million years ago, but it was very rar�. (Bakker 1988: 74)

322

Good News about the Description Theory of Names

(7)

a.

If a child is "christened 'Bambi', and Disney Inc. hear about it, then they will sue Bambi's parents. b. The name of a product is essential to its commercial success. For example, if you want to buy washing powder and are given the choice between Black, White, and Grey, you will choose White, won't you?

It seems to me that (7a) is better than (6b), and as far as I can tell there is nothing wrong with (7b). The following examples are cases of binding, too, intuitively speaking:5 (8)

a.

Every time we do our Beatles act, {Ringo/the guy who plays the part of Ringo} gets drunk afterwards. b. Every time John goes to see a performance of Hamlet, he falls in love with {Ophelia/the actress who plays the part of Ophelia}. (9) a. Perhaps Mary has a son named john' and perhaps {her son/John} is the thie£ b. Mary is under the illusion that she has a son named john' and she believes that {her son/John} is the thie£ In (8a, b) the definite NPs as well as the names can be interpreted non­ referentially. On this interpretation, 'Ringo' is more or less equivalent with 'whoever plays the part of Ringo', and the same holds, mutatis mutandis, for 'Ophelia'. In (9a, b) there is a preference for a non­ referential interpretation, and intuitively 'her son' or john' in the second conjunct is 'bound' in the first. As a matter of fact, in Geurts (1995, to appear) I present a treatment of intensional contexts that allows us to view these expressions as bound elements, but this is by the way, because all I want to do at this point is establish the parallels between names and overt definites.

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

in (6b). This observation, which is crucial to my concerns, has not met with unanimous approval from the readers of earlier versions of this paper, so let me dwell on it for a while. One of the referees of this journal claims that (6b) sounds odd, and although I do believe that this sentence is fully acceptable as it stands, I have some sympathy �th this judgement. But if (6b) is vaguely odd, it is because the word 'Bambi' is repeated for no good reason: why use this word when a possessive pronoun would have done just as well?4 If this is correct, then we should be able to come up with better examples if we can somehow motivate the repetition of the name. And we can, for instance by increasing the distance between the name and its antecedent and/or by introducing competing antecedents:

Bart Geurts 323 •

Although definite NPs and names normally refer to objects in the common ground, they can be used to introduce new objects: (ro) My best friend is {my sister/John}.

•

Both variants of (ro) can be used in a situation in which the intended referent is new to the hearer. Names generally take wide scope. In this respect, too, they are like many other definite NPs. Compare:

(I I )

( 1 ra) is ambiguous: it can either mean that the person who happens to be Prime Minister could have been rich, or just that we might have had a rich Prime Minister. (ub) and (uc), by contrast, only allow for readings of the first type. The suggestion that, with respect to scope, names are unlike definites in that they always take wide scope is incorrect for two reasons. First, as we have just seen, only some definite NPs alternate relatively freely between wide-scope and narrow-scope interpretations. Secondly, and this is the last point on my list, Names can take narrow scope, too. For reasons to be discussed below, this may require a somewhat outlandish type of context, but it does happen. For one thing, there are the bound-variable uses of names mentioned above. For another, there is the case of Aaron Aardvark: ·

•

( 12) The electoral process is . under attack, and it is proposed, in light -of

recent results, that alphabetical order would be a better method of selection. than the present one. Someone supposes that 'Aaron Aardvark' might be the winning name and says, 'If that procedure had been instituted, Ronald Reagan would still be doing TV commercials, and [( 12)] Aaron Aardvark might have been president' (Bach 1 987: 146-7).

Clearly, in this scenario the speaker need not believe that there is anybody for the name 'Aaron Aardvark' to refer to, yet ( 12) isn't infelicitous in any way. Further examples of the same type are:

( 1 3)

a. In English, Leslie may be a man or a woman. b. But John is always male.

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

a. The Prime Minister could have been rich. b. The man could have been rich. c. John could have been rich.

324

Good News about the Description Theory of Names 2 P R E L I M I NARY S K I R M I S HES

this is a lame defence unless it can be shown on independent grounds that the semantic values of. say, 'Bambi' in (6b) or 'Aaron Aardvark' in ( 1 2) are non-standard. It will not do to note that these sentences are special or require special contexts, and leave the matter at that. In the absence of such independent evidence, these data stand as evidence against Kripke's theory of names.

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

Although it is always hard to prove that something doesn't exist, the foregoing observations do suggest quite strongly that there are no deep semantic or pragmatic differences between names and overt definites. Such distinctions as must be made set some definite NPs apart from others but they don't draw the line between names and overt definites. Furthermore, most of the observations I have mustered are difficult to reconcile With the hypothesis that names are rigid designators. Not to mince words: they falsify it. If names were rigid designators it would be worrying, at the least, that names can have definite articles, that they can be used attributively and non-literally, and can be used to introduce individuals that are new to the hearer, and it would be a mystery that names have bound-variable uses and may take narrow scope. It may be challenged that at least some of these observations are irrelevant because they involve 'metalinguistic' uses of names, the implication being that names are generally not used this way.6 I think this is partly right and partly wrong. What is right about it is that some of the data I have listed are somewhat out of the ordinary. But that is precisely what one should expect if the quotation theory is correct. If it is true that the meaning of a name is 'the individual named N, then the content of a name is special in a way that will make it difficult to construct examples in which names act as bound variables, for example. But the same holds for some overt definites, too. Consider 'the cosmos'. There can be no doubt that this is a definite NP, but due to its truly comprehensive meaning it will be difficult to construct sentences analogous to the ones above in which 'the cosmos' acts as a bound . variable. Difficult, but not impossible. And although such cases will be very 'special', they will surely count as evidence that 'the cosmos' is an ordinary definite NP. Therefore, I see no reason to dismiss the data presented in (6)-(9) and (1 2)-(1 3) merely because these are special cases. It is understandable that someone who takes the position that names are rigid designators will want to claim that my counterexamples are irrelevant because they involve deviant uses of proper names.7 However,

Bart Geurts 3

325

NAMES A N D REFERENCE

Kripke charges that the quotation theory is circular because it actually presupposes what it must explain, namely how names come to refer: Someone uses the name 'Socrates'. How are we supposed to know to whom he refers? By using the description which gives the sense of it. According to Kneale, the description is 'the man called "Socrates" ' . . . We ask, 'To whom does he refer by "Socrates"?' And then the answer is given, 'Well, he refers to the man to whom he refers.' If this were all there was to the meaning of a proper name, then no reference would get off the ground at all (Kripke 1980: 70). Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

But what if the quotation theory didn't explain how names manage to refer? Suppose we analyse definite NPs along Russellian lines.8 Then the quotation theory entails that names are non-referential expressions. This may strike some as an intuitively repugnant conclusion, but Kripke himself has argued persuasively that such intuitions are not necessarily detrimental to a Russellian theory. Critics of Russell's theory of descriptions have often claimed that the theory is falsified by referential uses of definite NPs. Kripke (1977) argues, however, that this is not so, because a Russellian may hold that the theory of descriptions is a semantic theory and need therefore only be concerned with what is said (in Grice's sense); intuitions about referential uses of definite NPs pertain to what is meant by a speaker on a given occasion. Put otherwise, the idea is that although semantically speaking definite NPs are non-referential expres­ sions, speakers may use them for conveying information about specific individuals. But if names are definite NPs, then the same observations should apply to them, too. Hence, Kripke's complaint that the quotation theory is not a theory of reference not only presupposes that any semantical t�eory of names should be a theory of reference, like his own; it also denies the quotation theorist an account of reference that Kripke recommends elsewhere. If NPs of the form 'the individual named N' are to be treated in a Russellian framework, they must probably be viewed as 'incomplete' definites, i.e. as on a par with 'the table', 'the child', and so on.9 As Strawson was perhaps the first to point out, it would seem that such expressions, which rarely if ever describe unique objects in the world yet can be used without apparent difficulties, cause an especially severe problem for Russell's theory of descriptions. Russell's followers have countered Strawson's objection in various ways. According to Bach (1987), for example, a speaker who produces a sentence containing an incomplete definite NP virtually never says what he means (I am again speaking in Gricean terms): strictly speaking such sentences are false, but if all goes well,

326

Good

News about the Description Theory of Names

as it usually does, they will none the less manage to convey information about specific individuals. This may or may not strike one as a plausible strategy for dealing with incomplete NPs, but this is as it may be, since for the moment I merely want to note that the quotation theory is immune to K.ripke's objection that it doesn't explain how names come to refer. This objection clearly presupposes that names are referential expressions, where reference is to be understood as a semantic relation between a term and a real-world individual, and this presupposition cannot be taken for granted. Of

4 B E I N G NAME D N V S . B E I N G T H E REFERENT O F N Kripke's criticism is misguided for a further reason, as well, because it doesn't distinguish between two properties that should be strictly kept apart, viz. being the referent ofN, and being named N. 10 That N is the name, or one of the names, of an individual a does not entail that N refers to a or that a is among N's referents. Kripke assumes as a matter of course that when we baptize an individual N we eo ipso determine the reference of the name N. This is not so. Beethoven's Missa Solemnis is commonly referred to as 'the Missa Solemnis', but its name is 'Missa Solemnis'. Similarly, an Italian may refer to Maria Callas with 'la Callas', but her name was 'Callas'. 1 1 The difference is slight (a mere two or three letters, after all) but telling: it highlights the fact that although it so happens that in English names can generally be used to refer, it might have been otherwise (and it sometimes is, even in English). The grammar ofEnglish might have dictated that in order to refer with the help of a name, it must always be preceded by the definite article, or, for that matter, by 'the individual named'. Bearing a name is like wearing a tie. Like ties, names are seldom unique, but circumstances permitting they may be used for referential purposes. More accurately, just as you can employ the attribute of wearing a tie to identify to your audience the person you have in mind (John, as the case may be), you can use the attribute of being named John' for the same purpose. Taken on its own, however, a name doesn't refer any more than a tie does. It is instructive to compare names with number terms. Indeed, although number terms are rarely classified among the names, they are often used as such. Bearing a name may be likened to having a number: a crate in

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

course, it can hardly be denied that names may be used to convey information about real-world individuals, but the claim that names are referential expressions does much more than restate this banal truth in different terms, and should not be accepted uncritically.

Bart Geurts 327 Rotterdam harbour may have a number stamped on it, for example 6. But of course the number term '6' doesn't refer to the crate in question, although it may be used to refer to it. That is to say, when taken on its own, a number term

by the computer's hardware and the programs running on it: if due to some

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

can sometimes be used to refer, but in most cases it must be part of a larger expression: 'crate number 6', '6 Elm Street', 'Paris VI', 'volume 6', and so on (with names it happens to be the other way round, as we have seen). Another respect in which number terms are like names is that there are countless ways in which an individual may be assigned a number or number term (the distinction is not always clear, but it doesn't seem to matter much, either). Sometimes individuals are inscribed with numbers: houses, banknotes, football players. In many other cases, there is a less immediate relation between a number and an individual that 'has it'. Numbers may be assigned at random or following a rigid procedure, for instance according to some linear order. There are various conventions for assigning house numbers, convicts are numbered as they enter prison, cameras receive a number when they leave the factory, and · in some countries citizens receive a number at birth. Just as there are many ways of assigning a number to an individual, so are there many ways of assigning it a name. There are descriptive names like 'Fatty' or 'Benjamin'. Similarly, the title of a book or film is expected to be somehow related to its content. In my native country, last names are assigned according to strict regulations, but first names are afflicted by whim. Other cultures make use of patronymics, and in still others parents are named after a child. And so on. What all these naming practices have in common is just that some association is established between a name and its bearers, but how this association is initiated and sustained is different from case to case. The name-bearing relation between 'Lolita' and the famous novel was initiated by the author and is sustained, inter alia, by printing the name on the front cover of every copy. The association between my last name and myself is sustained, inter alia, by records at a register office, but this does not apply for my first name. If I decided that I wanted to be called 'Rudolf instead of 'Bart', it would just be a nuisance to the people in my social sphere, but I don't think it would be humanly possible to change my last name into 'Camap' (not in my country anyway). It is not even true that the name-bearing relation is necessarily grounded in a social convention of some sort. For example, the Hies on my computer all have names; some of these I chose myself, others were chosen by various people around the globe, only some of whom I am acquainted with, and yet others, such as 'cache i 8 J 4 I S.shtml' for example, were generated by some program. The association between a Hie and its name is entirely sustained

J28

Good News about the Description Theory of Names

s INTROD U C I N G OTTO 'Ottomobil' is the official name of a domestic robot which is currently being developed in my department; my colleagues and I usually call him 'Otto' (we prefer to think of Otto as male).'2 A central part of Otto's design is a system for natural language understanding and generation, and my personal pride and joy is the robot's presupposition module, which is based upon a theory that I helped to develop. ' 3 In fact, Otto's presupposi­ tion module not only deals with presuppositions, but with anaphora as well. Otto treats presuppositions and anaphors alike as elements that want to be bound to an antecedent. If occasionally an appropriate antecedent may not be found, Otto is prepared to accommodate the presupposition in question; but in general he prefers to bind his presuppositions. This preference is stronger in some cases than in others. For example, while Otto very much likes to bind anaphors and other descriptively attenuate presuppositions, he doesn't mind very much accommodating presuppositions triggered by factive verbs.'4

( 1 4)

a. If someone broke the copier, then the new secretary is the culprit. b. [: [x, g: copier u, x broke u] =} [ �, y: culprit z, new secretary v, z = v) J c. [u,

v:

copier u, new secretary v, [x, z: z

=

x, x broke u, culprit z)

[: z = v] ] d. [u, v: copier u, new secretary v, [x: x broke u, culprit x) =}

=}

[: x

=

v) ]

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

programming error the ftle name 'Bocherini' gets changed into 'Corelli', then the ftle's name is 'Corelli', not 'Bocherini'. The original intention behind name-giving is simply irrelevant. The expression 'bearing a name' covers as many relations as there are naming practices, and it seems to me that Kripke's causal theory of reference is best viewed, not as a theory of reference, but as a partial theory of what it means to bear a name. The theory is only a partial one because there are naming practices that it doesn't account for, such as the one considered in the previous paragraph. Kripke's charge that the quotation theory is circular can now be countered a second time, as follows. Even if a proponent of the quotation theory should want to explain how names can be used to refer, his analysis of names need not be circular, because the notion of reference need not enter into it: 'the individual named N' is not the same as 'the individual that N refers to', and the former does not presuppose the latter, either.

Bart Geurts 329

·

6 ACCOMMODATI O N AND COUNTERPARTS I said that, when Otto is presented with sentence (14a) he will accommodate the presuppositions triggered by 'the copier' and 'the new secretary'. This is not all he does, however. Even if he decides to accommodate a presupposi­ tion, Otto verifies if it matches his representation of the world, or is at least consistent with such knowledge he has at his disposal. In order to explain

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

(14b) is the DRS that Otto's grammar module computes for the sentence in (14a). This sentence contains three definite NPs, which trigger the presuppositions that in (14b) are marked by underlining. Otto will deal with · these presuppositions as follows. Beginning with the presupposition triggered by 'the copier', Otto first checks if it can be bound to an antecedent in his current DRS. Seeing that there is no suitable antecedent available, he decides that the presupposition must be accommodated, and since he has a preference for global accommodation, he accommodates the presupposition in the main DRS. The second presupposition, triggered by 'the culprit', can be bound to the indefinite in the antecedent of the conditional, which is the solution that Otto favours. Finally, as there is no suitable antecedent for the presupposition triggered by 'the new secretary', this presupposition is accommodated too, again in the main DRS. The resulting reading that Otto assigns to (14a) is (14c), which is equivalent with (14d). In this example, all presupposition-inducing expressions happen to be definite NPs, but I should like to emphasize ·that Otto applies the same handful of principles to all types of presuppositions, regardless whether they are triggered by definite NPs, pronouns, factive verbs, clefts, focusing particles like 'too' or 'even', etc. These principles are, to recapitulate: (i) a presupposition is preferably bound, but (ii) if it cannot be bound it will be accommodated, and (iii) if a presupposition must be accommodated, then it is preferably accommodated in the least embedded DRS, i.e. global accommodation is preferred to local accommodatiotL Definite NPs and pronouns are special only in that they presuppose their entire content: semantically speaking, these expressions are nothing but presupposition inducers. All there is to say about the content of 'the man' or 'he', for example, is that these expressions trigger the presuppositions that there is a man and that there is a male person, respectively. In contrast to these purely presuppositional expressions, as they might be termed, a factive verb like 'regret' not only triggers the presupposition that its complement

3 30

Good News about the Description Theory of Names

this process in more detail, let us consider a simpler example. Suppose I inform Otto that: (1 s) The new secretary is Irish. The DRS that Otto will produce for this sentence looks like this: (16) [!: new secretary x, Irish x]

( 1 7) [z: new secretary z, Bart believes: [ x: new secretary x, Irish x], . . . ] {I7) makes explicit the fact that Otto, too, believes that there is a new secretary. It will be obvious that {I7) is only a partial representation of Otto's belief box: apart from the two beliefs explicitly represented in (17), Otto will have a wealth of further beliefs, which are abbreviated here by three dots. (17) says that Otto believes that there is a new secretary, z, and that he believes that I believe that there is a new secretary, x, but it doesn't establish a link between these two individuals. In this particular case, however, such a link may be assumed to exist, . and, more to the point, we may take it that Otto believes that such a link exists. As he might say, his z and my x are the

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

Formally, this represents a purely existential proposition: (16) is true in any given world iff that world contains a new secretary who is Irish. This is correct as far as it goes, but it doesn't go far enough, for we should expect that someone who is processing (I s) will attempt to establish a connection between the presupposition triggered by the subject term, on the one hand, and what he takes to be the case in the world surrounding him, on the other. This is precisely what Otto does. Two cases must be distinguished here depending on whether Otto has already formed the belief that there is a new secretary, or not. In the first case, it is not new to Otto that there is a new secretary, and he has already accumulated information about her. Perhaps they have already met, or maybe Otto has just heard from someone else that there is a new secretary. At any rate, we may assume that in ·this case he has already a mental representation which symbolizes the new secretary (as we might say, if we deemed Otto's belief to be true), and he will link the underlined material in {I6) to this representation. But what exactly do we mean when we say that this presupposition is linked to a representation that was already available beforehand? I have argued elsewhere that a DRS must be seen as a representation of the speaker's commitment slate,' 5 and so {I6) is Otto's representation of my commitment slate, which he constructed in response to my utterance of {I s). It follows that {I6) is actually embedded in the context of Otto's beliefs, and if we want to make this context explicit, we get something like the following picture:

Bart Geurts

331

same person. I prefer to say that they are counterparts in the sense of Lewis.'6 If we want to make this relation explicit we need something like the following: (1 8) [z: new secretary z, Bart believes [� new secretary x, Irish x], . . . ] Another, more succinct notation for counterpart relations, which I have proposed in Geurts (1995. to appear), is t.o simply use the same reference marker twice, as follows: (19) [x: new secretary x, Bart believes: [�: new secretary x, Irish x], . . . ]

are, in Lewis's words, 'our substitute for identity between things in different worlds' (Lewis I 968: I 14). When we speculate about what might have happened to John, say, we pretend to speak of the same individual in counterfactual circumstances. We allow that two people are thinking of the same individual, although they may disagree about virtually all of its properties. Even more strikingly, we allow that Hob and Nob have the same witch in mind, although there are no witches.' 7 All these cases can be understood in terms of counterpart relations. Unlike identity, counterpart relations are indeterminate in various ways. ' 8 To say that two individuals are counterparts is to say that they are alike in some respects, and since similarity is a matter of degree, we have to agree on a lower bound of similarity before we can decide whether a counterpart relation obtains in any given case. More importantly, for our present purposes at least, is that similarity depends on how you look at it. Two individuals may be similar in some respects but not in others, and before we can say that they are counterparts or not we have to decide what respects are to count. Now I can explain in more detail what happens when Otto decides that

'the new secretary' in ( Is ) cannot be construed as an instance of binding. The speaker who uttered this sentence, i.e. myself, presupposes that there is . a new secretary, and Otto will accommodate this presupposition in his representation of the speaker's commitment slate, and link it to his own representation of the new secretary by means of a counterpart relation, as shown in (19). As Otto himself would put it, he and I have the same person in mind.

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

The intended interpretation of this structure is the same as that of (I 8). This notation of counterpart relations exploits the fact that, logically speaking, the two occurrences of x are independent of one another. · Counterpart relations are to capture our intuitions about identity across possible worlds-or, more accurately: intuitions that we would express in terms of identity relations across world boundaries. Counterpart relations

332

Good News about the Description Theory of Names

Otto is a trustful robot. In most cases he has no doubts that what people tell him is the truth, and he is particularly confident that I am usually well-informed and would never lie to him. If I say that the new secretary is Irish, Otto comes to believe that the new secretary is Irish. So eventually the effect of my uttering ( 1 s) is that Otto's belief box is updated as follows: (2o) (x: new secretary x, Irish x, Bart believes: (�: new secretary x, Irish x], . . . ]

about the new secretary.

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

This is what happens. when Otto already believed that there is a new secretary before I made my statement. If Otto did not have this information beforehand, the interpretation process will be different in some respects, but the outcome will be more or less the same. Suppose that Otto is not aware that there is a new secretary. Again I enter the scene, and recite my line. This will have the same net effect on Otto's belief box as in the previous case, because Otto is a trustful robot, and expects that what I tell him is the truth. This is not to say that it makes no difference at all whether or not Otto already believes something that is presupposed by the speaker addressing him. Trustful though he may be, Otto sets greater store by what he believes than what people tell him, and if he already is aware that there is a new secretary his representation of the person in question is presumably richer than when he has just gleaned this information from my statement. The important point, however, is that in either case the relation between Otto's own representation of the new secretary, on the one hand, and his representation of my representation of the new secretary, on the other, will be the same. (2o) is in a sense a purely descriptive representation. Despite their name, reference markers do not refer to individuals in the world (or model), and the reference marker x does not refer to the new secretary. But if (2o) represents Otto's belief box, then surely there must be some connection between this reference marker and the new secretary? And there might be. In fact, there are many types of possible connections between reference markers and real-world individuals. If Otto has already seen the new secretary, he will have a visual memory of her, and · the reference marker will be linked to that. If he has heard her speak, he will have an auditive memory which is �ssociated with the reference marker. He may have read her application, he may have seen her photograph, he may talked about her with people who have seen her or who know people who have seen her, and so on. It is connections such as these that prompt us to say that Otto has beliefs

Bart

7

Geurts

333

OTTO'S TREATMENT O F NAMES

• • • • •

names often carry definiteness markers; names are subject to definiteness restrictions; the referential/attributive distinction applies to names; names can be construed non-literally; and names can be construed generically.

I don't mean to suggest that I can account for definiteness restrictions, the distinction between referential and attributive readings, non-literal construals, and genericity. The point I want to make is merely that, given that definites are subject to definiteness restrictions and have referential, attributive, non­ literal, and generic interpretations, the presuppositional theory of names leads us to expect that the same will hold for names, which it does. The remaining observations from section I require more detailed explanations: •

Names are typically, though not always, used to refer to objects that are part of the common ground. We distinguished two ways of being part of the common ground: if a name is used anaphorically, its referent was already introduced into the discourse; if it is used non-anaphorically, its referent will generally be given, too, though not in the immediate context. That names prefer their referents to be given is a characteristic they have in common with all other presupposition inducers. It is true that names have a comparatively strong preference for a linking interpretation, which is to say that, on the whole, their presuppositions are rather difficult to accommodate without being linked to a given

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

In accordance with the quotation theory, Otto treats names as definite descriptions of the form ' the individual named so and so'. This is not to say, of course, that Otto treats names as definite NPs in disguise; he doesn't transform a name into a definite NP before he interprets it. It is just to say that Otto regards names and overt definite NPs alike as purely presupposi­ tional expressions, which presuppose their entire descriptive content. Given that Otto already knew how to handle presuppositions, this was by far the most obvious treatment from a designer's point of view: it is more elegant than any other solution we could think of and enables Otto to interpret names in precisely the right way. To show this, let us consider how the presuppositional version of the quotation theory used by Otto measures up to the observations listed in section 1 . The first five of .these observations need hardly any comment. Given that names are treated on a par with overt definite NPs, it doesn't come as a surprise that:

3 34

•

Good News about the Description Theory of Names

referent, but in this respect names aren't exceptional, either, since they share this characteristic with some other presupposition inducers, such as 'incomplete' definite NPs (I will return to this point below). Names have bound-variable uses. In the DRT framework adopted here, any presupposition that is linked to an antecedent in an embedded DRS will seem to behave like an element bound by a quantifier. So this is just a special case of the general rule that presuppositions prefer to be linked to a given antecedent.

(z r )

a. If a child is christened 'Bambi', then Disney will sue Bambi's parents. (= (6b) )

(2 1b) is the DRS that Otto's parser computes for (2 ra). The second occurrence of the name 'Bambi' triggers the presupposition that there is someone named 'Bambi' (at its first occurrence the name is mentioned, not used). This presupposition is bound in the antecedent of the conditional, and the resulting interpretation is (2 r c). Hence, on the present account there are no relevant differences between (2 1 a) and the following: (2 2) a. If a farmer owns a donkey, he beats it. b. If a farmer beats his donkey, his wife beats it, too. In (na) the pronoun refers back to the indefinite in the antecedent of the conditional, and similarly, in (22b) the presupposition triggered by the focus particle, that someone other than 'his' wife beats 'it', is bound in the antecedent of the conditional. Hence, the presuppositions triggered by 'it'

•

in the first sentence and 'too' in the second are treated exactly the same as the presupposition induced by 'Bambi' in (2 1 a). Although names typically refer to entities in the common ground, they" may be used to introduce new individuals, as in:

(2 3)

My best friend is John. (= (ro) )

In general, if a presupposition cannot be bound to a suitable antecedent, it will be accommodated. So unless there is a suitable antecedent for John' in the context in which (23) is uttered (as when the sentence is

•

preceded by, say, 'I have three friends: John, Jack, and Joe'), the presupposition that there is an individual with that name will be accommodated. Names generally take wide scope. It seems a rather safe conjecture that if a presupposition is bound it will usually be bound globally, i.e. in the

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

b. [: [x: child x, x is christened 'Bambi'] '* [g: u is named 'Bambi', Disney sues u's parents] ] c. [: [x: child x, x is christened 'Bambi', x is named 'Bambi'] '* [: Disney sues x's parents] ]

Bart Geurts

33S

·

(24} John believes that most party members are stupid. . This sentence has at least tWo readings, depending on whether the quantifier 'most party members' takes wide or narrow scope with respect to the attitude verb. But this quantifier also triggers the presupposition that there are party members, and this presupposition will normally be bound or accommodated globally, i.e. outside John's belief context. If we refer to the scope of an expression a, we speak of a unit corresponding with a at some level of analysis (e.g., a's correlate at LF or a's interpretation relative to a given model). Although in some frame­ works this is only a metaphor, the guiding intuition is that a's scope is determined by moving a about. If, on the other hand, we speak of a presupposition triggered by a, the metaphor is a quite different one, the idea being, rather, that a initiates a search for an antecedent meeting certain specifications. The distinction between scope taking and presupposition projection is somewhat obscured by the circumstance that certain expressions are purely presuppositional.'9 This holds in particular for anaphoric pro­ nouns, definite NPs, and names. The semantic contributions made by these expressions coincide with the presuppositions that they induce. Consequently, if these presuppositions are bound or accommodated globally, as they usually are, it may seem as if the expressions that triggered them had taken wide scope, but as a matter of fact the notion of scope doesn't enter into this at all. This is why I said that my analysis explains why names appear to prefer having wide scope. The presuppositions triggered by names seem to have a decidedly stronger tendency to 'take wide scope' than some others. In this respect, too, they are on a par with other descriptively attenuate, 'incomplete', ·

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

principal DRS; for surely cases like (22a, b) are the exception rather than · the rule. Moreover, if a presupposition must be accommodated, global accommodation is preferred to accommodation in an embedded DRS. Thus, on the whole presuppositions appear to have a strong tendency to be projected to the global level of the discourse representation. Since there is no reason to assume that this doesn't hold for names, we thus explain why names appear to prefer having wide scope, I say 'appear' because I don't actually want to maintain that names are scope bearing expressions. There is a subtle but important difference between saying that a presupposition is accommodated globally and saying that an expression takes wide scope. The difference becomes rather obvious when we consider expressions that have scope and induce presuppositions at the same time, like quantifiers, for example:

3 36

Good News about the Description Theory of Names

definites like 'the door' or anaphoric pronouns like 'it'. Compare the following sentences, for example: (zs)

(zsa) allows for a reading according to which we might have had a situation s in which the President in s was a woman. In order to obtain this reading, the presupposition triggered by 'the President' must be accommodated locally, i.e. within the scope of the modal operator, and must not be linked to whoever happens to be president in the situation in which the sentence is uttered. This type of interpretation does not appear to be so readily available for (2sb), presumably because 'the car' insists more strongly that its referent be given (either in the previous discourse or in the wider context) than 'the President' does. Van der Sandt ( I 992) suggests that such differences correlate with varying degrees of descriptive richness. This suggestion is perhaps correct, although it may not be the whole story. But no matter how such . differences are to be accounted for, it is evident that in this respect a proper name like john' resembles 'the car' more than 'the President', in that it insists rather adamantly that its presupposition be linked to . an entity that is given. . Names can take narrow scope. By way of exception, our robot Otto is prepared to accommodate the presupposition triggered by a name in an embedded DRS. This is how he will deal with the name 'Aaron Aardvark' in the scenario devised by Bach: (26) Aaron Aardvark might have been president. (= (1 2) ) Supposing that Otto doesn't . know of any person named 'Aaron Aardvark' (which he doesn't), and supposing that for whatever reason he doubts that the speaker knows of any such person, then Otto will decide to locally accommodate the presupposition triggered by the name, thus arriving at the following reading: (27) [: 0 [x: x is named 'Aaron Aardvark', president x] ] In general, local accommodation is only used as a last resort, and as with other semantically attenuate definite NPs, it is exceptional for a name to be interpreted by means of local accommodation. Bach's example really is a rare fmd.

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

•

The President might have been a woman. b. The car might have turned left.

a.

Bart Geurts

3 37

8 WHAT J O H N M I G HT HAVE BEEN Suppose that shortly after John's birth following: (28) a. If John had had red hair b. If John had been a girl c. If John had been a twin d. If John had been a Rolex watch

h is

}

mother utters one of the

his father would have been even happier.

(29)

a. [: [2£: x is named 'John', P x] rv > [x's father is happier] ] b. [x: x is named john', [: P x] rv > [x's father is happier] ]

Here 'rv > ' symbolizes the counterfactual conditional, and the value of 'P' covaries with the predicates in (28a)-(28d). In (29a) the presupposition is triggered that there is an individual named john', and our presupposition theory predicts that this will project to the main DRS. The resulting

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

We should probably say that (28a) refers to a property John himself might have had. But being a Rolex watch is not such a property, and so (28d) is to be paraphrased rather along the following lines: 'IfJohn's mother had given birth to a Rolex watch instead of to John (whatever that may mean), his father would have been even happier.' In short, while (28a) is about a counterfactual situation involving John, (28d) is about a counterfactual situation in which John's place has been taken by a Rolex watch. (Note that, according to counterpart theory, which I still endorse, this distinction cannot be taken over into our semantic analysis, since no individual can inhabit more than one world. However, it does not follow that counterpart theory cannot account for the intuition, which I am trying to character­ ize, that there is such a distinction.) Examples (28b, c) are in a sense intermediate cases because it is less clear how we should describe them. Speaking only for myself: although I would probably say that John himself might have been a girl, I would rather not say that he himself might have been a twin. Therefore, I would bracket .(28b) with (28a) and (28c) with (28d); but I don't expect that these judgements will prove to be uncontroversial. This is as it may be, however, because the relevant observation is that our intuitions about the relation between John and his counterparts vary at all. Needless to say, this variation causes problems for a Kripkean analysis of names. In the present framework, by contrast, it is only to be expected. According to the theory outlined in the foregoing, the examples in (28) will all be interpreted along the following lines:

338

Good News about the Description Theory of Names

interpretation is (29b), which says that there must a counterpart relation between John and the counterfactual individual satisfying the antecedent of the conditional. (29b) does not specify what kind of counterpart relation this should be, which is precisely what we want, because this relation varies from case to case, as we have seen. In other words, the variation we observed in (28) reflects the context dependence of the counterpart relation.

9 THE I NTUITI O N O F RIGIDITY

(3o) Mary is happy. If someone understaq.ds this statement correctly, then he grasps a proposi­ tion which is true in a certain range of possible situations: in each of these situations, Mary is happy, where Mary is the referent of 'Mary' in the situation in which (Jo) is uttered. This, Kripke suggests, is the claim that names are rigid designators. This observation appears to establish two points at the same time. First, since it is always the same person that makes the proposition expressed by (3o) true, in any given possible situation, it would seem to prove that the name 'Mary' is rigid. Secondly, it would seem to prove that the property of being named 'Mary' isn't part of the meaning of the name, because the proposition expressed by (3o) might be true even in a possible situation in which Mary had a different name. In brief, ifKripke's observation is correct, then names are rigid and the quotation theory is false. Kripke's observation is not as straightforward as it appears, however. To begin with, it matters a great deal how we frame our initial question. If we ask, as Kripke does, if (3o) might be true in a possible situation in which Mary was called differently, then it will seem as if the name 'Mary' is rigid.

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

In imitation of Bach (1987), I would have entitled this section 'the illusion of rigidity'. But on second thoughts I don't believe that rigidity is an illusion: it is a genuine empirical phenomenon. What is illusory is the notion that it calls for an explanation in semantic terms. I have tried to explain why it is that names appear to have such a strong preference for taking wide scope. But Kripke's most forceful argument in favour of the thesis that names are rigid designators is based on the intuitive truth conditions of simple sentences, i.e. sentences without any relevant scope-bearing expressions (such as modals). The argument, which is only seemingly straightforward, as we will presently see, goes as follows. Consider a simple sentence containing a name, such as:

Bart Geurts

339

If, however, we ask what information is conveyed by this sentence, then it would seem to be part of its content that Mary is named 'Mary'. The point I want to make here is prefigured in Lewis's (I 98 I) commentary on Kaplan's (I989) character/content distinction.2° Kaplan claims that his notion of content captures the pre-theoretical notion of 'what is said' by uttering a sentence in a given context. For example, if two interlocutors simultaneously utter (J i a) and (Jib), respectively, what is said is the same: (3 I )

a. You are sleepy. b. I am sleepy.

Unless we give it some special technical meaning, the locution 'what is said' is very far from equivocal . . . Kaplan's readers learn to focus on the sense of 'what is said' that he has in mind, ignoring the fact that the same words can be used to make different distinctions. For the time being, the words mark a defmite distinction. But why mark that distinction

rather than others that we could equally well attend to (Lewis

1981: 97)?

Lewis does not claim that Kaplan's theory is false. What he objects to is Kaplan's suggestion · that his theoretical notion of content captures the pre-theoretical notion of what is said, and the implication that genuinely alternative theories of meaning cannot capture this notion. This suggestion and its implication are wrong, because there is no single pre-theoretical notion of what is said. Intuitions about what is said vary with one's interests. In a sense, two interlocutors uttering (J ia, b) may have said the same thing in one sense, but in another sense they haven't. A theory oflexical semantics may be expected to explain this shifty behaviour of the verb 'say', but there is no reason to require that central concepts in semantic theory (such as 'character' or 'content') must capture any of our ways of understanding 'what is said'. Lewis's objection applies to Kripke's rigidity thesis, too. It is true that there is a sense in which (3o) correctly describes a possible state of affairs in which Mary happily lives under a different name. But there is also a sense in which (3o) is not correct in such a state of affairs; for if Mary is called 'Gertrude', say, then it is incorrect to call her 'Mary'. Indeed, it has never been denied, as far as I am aware, that someone who utters (3o) conveys the information that Mary is called 'Mary'. What Kripke denies is merely that this information is part of the meaning of 'Mary', and he suggests that this equally holds for the pre­ theoretical and. theoretical notions of meaning. Given the chameleontic quality of the pre-theoretical notion of meaning, it is pointless to disagree with the first half of this claim; but the s�cond half is false, in my view.

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

Or in Kaplan's terms: although these sentences have different characters, there are contexts in which their contents coincide. Lewis's objection is the following:

340

Good News about the Description Theory of Names

1 0 STILL, WHY DOES 'MARY' SEEM TO BE RIGID? Why is it that names appear to be rigid? This is a legitimate question, even if we mustn't presuppose that the intuition of rigidity is a direct reflection of the semantics of proper names. My answer to this question is not a new one. It has been given, independently it seems, by several reliable sources, including Sommers (1982), Bach (1987), and Fodor (1987)."'' Of these, Fodor's version comes closest to the formulation that I prefer:

I would prefer putting it as follows. The meaning of a name N is 'the individual named N', where the semantic contribution of the definite article is to be analysed in presuppositional terms. Given that the property of bearing the name N is an accidental one (I · could name my left ear 'Fortinbras' if I chose), referring to an individual with N will not, in general, be particularly effective unless the hearer already knows the intended referent and that it is named N. Therefore, a name will practically always be used to refer to an individual that was already given to the hearer beforehand, and if a name is thus used it will appear to be rigid. To say that 'the N', where N stands for any nominal head, is a presuppositional expression is to say that its intended referent a is presented as given, and that it is presented as given that a is an N. Someone who utters (3o), for example, presupposes that there is a person a whose name is 'Mary' and asserts that a is happy. So what Kripke calls the 'meaning' of this sentence is just its asserted content, and it is of course correct that if we confine our attention to this part of the information conveyed by (3o), the 'meaning' of 'Mary' will appear to be rigid, and the property of 'being named 'Mary' ' will appear to be irrelevant. As Bach (1987) points out, the strategy which Kripke consistently employs in Naming and Necessity is to first fix the referent of a name before he broaches the issue whether or not the name is rigid. But then it shouldn't come as a surprise that names always tum out to be rigid. Nor is it surprising that, for Kripke, non-referring names eo ipso disqualify as admissible evidence. My suggestion is that names appear to be rigid because they are presuppositional expressions. But surely not all presuppositional expressions appear to be rigid designators. So there must be something else that gives

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

The course ofwisdom would be to reiterate the moral-viz., that names are a hard problem for everybody-and then to shut up and leave it alone. Still, how about this: 'Cicero' and Tully' are synonymous but differ in presupposition . . . Then 'Cicero was wet' says, in effect, that he was wet and presupposes that he was called 'Cicero'. Tully was wet' says that he was wet too, but it presupposes that he was called 'Tully'. 'Cicero is Tully' is informative because, although it doesn't say that the guy who was called 'Cicero' was called Tully', it 'carries the information' that he was (Fodor 1987: S s).

Bart GeurtS

341

names their peculiar rigid feel. There is, but it is not peculiar to names. Consider the following situation. Otto and I are standing in the corridor, discussing the weather. We have just witnessed Mary walk past us and enter the kitchen. A moment later we hear cheerful singing emanating from the kitchen. At that point I say: (32) She is happy.

(33) [�: person x, female x, happy x] Since Otto treats anaphora as a special case of presupposition, he assumes that the pronoun triggers the presupposition that there is a given person who is female, and that it is asserted that this person is happy. There is no suitable antecedent for this presupposition to be bound to, so it must be accom­ modated. But even so Otto will try to integrate the presupposition with his other knowledge, which is to say, in the present case, that he will connect it to his representation ofMary. Thus Otto's beliefbox, in which his representation of my commitment slate is embedded, may be pictured as follows: (34) [x: woman x, x is named 'Mary', x is in the kitchen, Bart believes: [x: person x, female x, happy x] ] Otto believes that there is a woman called 'Mary', that she is in the kitchen, and that I believe of this person that she is a female person who is happy. Or rather, the person that Otto's beliefs are about and the person that he takes my statement to be about are counterparts. This is why the pronoun in (32) appears to be rigid. By uttering (32) I have conveyed the information in (3 3), part of which is that Mary is a female person; for this is the descriptive content of the pronoun 'she' (by approximation, at least). But when we ask Otto if what I have said might be true even if it transpired that Mary is a man, his answer will be yes. This is a reasonable answer because one very common way of interpreting 'what is said' is by restricting it to asserted information. If instead of (32) I had uttered (Jo), Otto would construe the name as 'rigid' for the same reason he would construe the pronoun in (32) as 'rigid'. And he would say that the property of being named 'Mary' is not part of what I have said just for the same reason he would say that the property of being a female person is not part of the statement made by (32).

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

To Otto it will seem as if 'she' is a rigid designator. For what I have said is true iffMary is happy, and this is what it means to say, according to Kripke, that an expression is rigid.22 Why does it seem to Otto as if 'she' is rigid? As I have explained, Otto maintains a representation of the speaker's commitment slate, which upon my utterance of (32) might look as follows (I ignore the previous discourse because it was irrelevant to my utterance of (32) ):

342 Good News about the Description Theory of Names 1 1 WHAT I D I D N 'T WANT T O SAY

(3 5)

If Mary has a car, it is pink.

It doesn't require much inventiveness to think of examples like (3 5). Analogous examples with names are much more difficult to produce. The essential characteristic that names and pronouns share is that they are nearly always used to refer to an object that speaker and hearer take to be given (with names this tendency is even stronger than with pronouns). It is this pragmatic fact which accounts for the intuition of rigidity. Secondly, I didn't claim that names and pronouns appear to be rigid because they generally link up to anteceden� that are. I might have claimed this, because the main point that I want to establish in this paper is that the quotation theory is right, and this is a theory about the meaning of names. But I didn't. Thirdly, I didn't claim that names and pronouns are the only types of expressions that usually seem to be rigid. For example, there are many overt definites that behave exactly the same way, the most obvious class being semantically attenuate descriptions like 'the man', 'the thing', and so on. Also, I believe that the presuppositions associated with quantifiers like 'all' and 'most' are often rigid in the same sense in which names are.

(36)

Everybody is happy.

Typically, this will be uttered in a situation in which the domain of 'everybody' is contextually given. Let c be a context in which (36) is uttered, and let A be the intended domain of 'everybody' in c. The proposition expressed by (36) in c is true in any given situation ( iff all individuals in A are happy in (. Hence; 'everybody' is rigid-or, better perhaps, its interpretation in c has a rigid component. On the other hand, I don't want to claim that all presuppositional expressions engender the unpression of rigidity. To the extent that we ·

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

It is always possible to misconstrue a story, no matter how simple, and I know from experience that my story about names is especially liable to be misunderstood, so let me briefly mention some of the things I didn't claim and don't want to claim, either. First, I don't want to claim that names are pronouns. This is how Sommers (1982) puts it, but not I. I do want to claim that names are very much like pronouns. They both are presuppositional expressions, and have the full complement of possible interpretations that presuppositional expressions generally allow for. But, of course, the descriptive content of a pronoun is of a different category than that of a name, which makes it much easier to construct examples of 'blocking' for pronouns than for names. For example:

Bart Geurts

343

have clear intuitions of rigidity at all they pertain to individuals and sets of individuals. Moreover, the intuition that an expression a is stronger when we feel that the descriptive content of a is inessential. It is for these reasons that we wouldn't want to say that factives are rigid, although qua presuppositional expressions they are quite similar to definite NPs. 12

C O NCLUSI O N

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

At the outset I announced that I would try to rehabilitate Kneale's analysis of proper names. But didn't I actually replace it with something completely different? I don't think so. Kneale's crucial insight was that names are synonymous with definite NPs of the form 'the individual named so and so'. Of course, this is not yet a theory of names; but it has the merit of broadening the problem. If at this point Kneale had adopted Russell's theory of descriptions, his theory of names would already have been superior to the rigid-designator account, as I have tried to show. But he actually suggested a presuppositional analysis of definite NPs, and in doing so he broadened the problem even further, although this will not have been evident at the time. What I have done is combine Kneale's insight with a theory of presupposition that is more explicit and, presumably, more comprehensive than what he envisaged. The resulting analysis of names is superior to Kripke's for two main reasons. First, it explains a whole range of empirical facts about names that Kripke's theory cannot account for, if they don't falsify it to begin with. Secondly, it treats names as just another class of presuppositional expressions within the framework of a theory that is amply motivated on independent grounds. In comparison, Kripke's theory is clearly ad hoc. What I have outlined in the foregoing is a theory of interpretation in two senses of the word: it is about the mental processes involved in the interpretation of an utterance, and about the mental representations that they give rise to. As I understand it, this is a semantic theory, but it is obviously not what philosophers like Kripke have in mind when they speak of semantics, meaning, and related notions. It might be objected, therefore, that what I have proposed is a change of subject, rather than just an alternative account of names. In a way this is right. Discussions about names tend to get out of hand in the sense that it is impossible to make substantial claims about the semantics of names without making quite fundamental assumptions about the status and aims of semantic theory. (This is precisely why

344

Good News about the Description Theory of Names

names are such a controversial subject, although it is hard to think of a category of expressions that would appear to be simpler.) But Kripke's theory and mine are not incommensurable, since we start out from the same set of semantic intuitions. Our theories are about the same empirical facts, and if one of them provides a more satisfactory account of these facts, then this is bound to have consequences for our views on the status and aims of semantic theory; but these foundational issues must be left for another occasion. BART GEURTS

Received: 1 s.o9.97 Final version received: 04.04.98

APPENDIX: EXERCISE To claim that a name N is semantically equivalent with a definite NP of the form 'the individual named N' is not to imply that the two expressions can always be used interchangeably. Of course, John' is not the same :is 'the individual named 'John" '. The latter expression is longer than the former, for example, which is presumably why many attempts at referring to John by means of 'the individual named 'John" ' will be found incongruous. Clearly, however, observations like these are the province of Gricean pragmatics. They do not even begin to prove that the said equivalence doesn't hold. Nevertheless, I believe that in quite a few cases the difference between N and 'the x named N', where x is a hyponym of 'individual', is very slight. In order to convey an impression of how similar the two types of expressions actually are, I have devised the following exercise. Below are eight text passages, collected on an Internet tour. These passages are lightly edited: I have occasionally replaced a name N with a description of the form 'the x named N', or vice versa. The purpose of the exercise is to determine which of the highlighted expressions have been tampered with and which haven't. The correct solution is given as note 23.'3 •

•

Should Mr Banharn become successful in his bid to form this country's next coalition government, there are still a number of obstacles to overcome. Certainly the defeat of Narong Wongwan in Phrae Province yesterday has to a certain extent removed one obstacle blocking his success. But he still has to convince the Bangkok public that he is suitable for the premiership. He would need to somehow bolster public confidence which appears certain to drop once the realisation sinks in that the man named Banharn is expected to be prime minister. In addition, he must answer the question of whether Mr Vattana will be appointed to his Cabinet The woman named Veronica had gone through the various treatments and services of the different physicians without finding any solution. She had spent all of her money and her health deteriorated to find her in a worse situation than she was before. When she heard ofJesus and pushed through the crowd to touch him, Jesus immediately knew that there was a withdrawal of virtue. He told her that her faith had made her whole.

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

University of Osnabriick, FB 7 49069 Osnabriick Germany e-mail: [email protected]

Bart Geurts 345 •

•

•

•

•

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

•

Fox Mulder hummed as he watched the nine ladies walk down the street. He guessed they were looking for someplace to stay. He sat in his car, silently watching them and waiting to see where they went. He was doing a little of his own observation, on his own time. Scully had begun to check the data base for criminal records, Mulder suspected she wouldn't find anything. The women, as he watched them, didn't seem to know where they were going. The one named Vickie walked slowly behind the group, looking about curiously. It was almost as if she were comparing the city to something. Mulder smiled, they definitely weren't lying about not being from this city. The sound of his cell phone beeping was enough to bring Mulder out of his thoughts. 'Mulder.' he muttered, already knowing that it was his partner that was calling. 'I found something interesting.' Scully said and Mulder raised his eyebrows. 'Oh?' 'Yeah, the woman named Anik, she's got quite a lengthy record.' Scully said, the tone of her voice seemed to mock him. Anik, that name was interesting in itsel£ Mulder had never seen a more intriguing woman and that name just topped off the package. Jack is in fact a Nightbane, while he appears human, Jack can shed his human form for a demonic looking body, the following is how Jason describes his morphus . . . Everyone else in the room is nearly blinded by a bright gleam of black 'light.' When their vision clears a few seconds later, the boy namedJack is gone and a monster is in his place. The story so far: A man set out on a mystery date with I-74 with only the vague notion of buying a bowl of chili when he arrived. Instead he met a girl named John Ehrlichman who seduced and then kidnapped the man. She carjacked his '73 Cutlass and headed toward Yorba Linda with the promise that they and Nixon would 'party like it's 1 999.' They did and he did and they returned home to the Watergate Cafe. Unfortunately for them, Nixon died, the parties stopped and the Watergate Cafe fell into disuse. At the Watergate Cafe, where the coffee is always flowing, where the cottage cheese­ and-ketchup platter is always being served and where the tapes are always rolling, life was going on as a cheap game show parody of itsel£ The girl namedJohn Ehrlichman, whose structural perfection was matched only by her hostility, was on the telephone hatching another scheme. Most people know the man named John Elefante. Some have known him since his days as lead vocalist with the rock group Kansas. Others know him as producer of numerous Grammy-winning projects for attists such as Christian rock veterans Petra. When Chambeau Blau died of equine fever in 1971 newspapers proclaimed the end of an era. Blau was the last and best of the horse hypnotists. At right we see Blau with the horse named �illy. Billy was mo.re mule than horse, but he was Blau's favorite subject. There was a grocery store. It could have been any grocery store, but it was bigger than most. It had twelve aisles, a produce section, a deli, a bakery, a liquor section, a magazine bin and a bulk candy section. It had a man named Val working the express lane cash register. We were in the grocery store, larger than most, for half an hour, when the man wearing a beach towel walked in. He went to the produce section, grabbed a bag of apples, a bag of oranges and then went to the bakery and grabbed a loaf of bread. Then he walked out of the larger than most grocery store. The man named Val followed him and asked, 'Why are you stealing from my grocery store?'

346 Good News about the Description Theory of Names N OTES

fact it isn't easy at all. See Geurts ( 1998) for discussion. 7 Presumably, this would be Kripke's position. C£ Kripke (1980: 62, n. 25). 8 Just suppose. I don't endorse Russell's analysis but I want to show that Kripke's criticism can be refuted without pre­ supposing the analysis of definite NPs that I favour. 9 This terminology is tendentious because only in the context of Russell's theory of descriptions is it necessary to con­ sider such NPs to be incomplete. I use the term here merely because it has gained some currency in the Russellian tradition. 10 This distinction is emphasized by Bach (1987), who proposes to speak of 'the bearer of N' instead of 'the individual named N', in order to forestall any further confusion. I am not convinced, however, that such terminological measures will prove to be very effective. 1 I Of course, Callas's real name was 'Kalogeropoulos'. I 2 Ottomobil is a continuation of the legendary Immobil project. I 3 See van der Sandt and Geurts (1991), van der Sandt (I992), Geurts (1995, to appear), and Geurts and van der Sandt (to appear). 14 Here and in the following I ignore the possibility that accommodation is not possible because the presupposition clashes with the Otto's beliefs (or because of any other reason). 1 5 See Geurts (1995, to appear). My notion of commitment slate derives from Hamblin (1971). Commitment is cru­ cially different from belief: a speaker who utters I{J accepts that I{J is true, as Stalnaker (1984) would say, in the sense that he commits himself to act, at least for a while, as if he held I{J to be true. Of course, in many cases speakers actually believe what they say, but sincerety is not a prerequisite for

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

1 I should like to thank Rob van der Sandt and the three anonymous referees of the journal of Semantics for their uncom­ monly elaborate comments on the pre­ vious version of this paper. I have profited greatly from their criticism and advice, and I hope the paper has, too. During the final stage of my work­ ing on this project I was supported by a grant from the Deutsche Forschungs­ gemeinschaft (DFG), which I gratefully acknowledge. 2 Strawson (1950) refers to definites like 'the Glorious Revolution' and 'the Great War' as 'quasi-names' or 'embryonic names', and remarks that an expression may 'for obvious reasons, pass into, and out of, this class (e.g. "the Great War")'.. Although this observation is un­ doubtedly correct, I prefer to express it differently. Instead of postulating a separate class of 'quasi-names', I would say that the distinction between names and other defmites is vague, because the relation 'being named N' does not have a precise meaning (see section 4). 3 Here I am indebted to one of the referees, who suggested that this is not possible. 4 This is a familiar situation in pre­ supposition theory. It is widely held, for example, that in a sentence like the following the presupposition that there is a king of France, which is triggered by the definite NP in the consequent of the conditional, is absorbed in the antecedent: If France has a king, then the king of France is bald. Of course, this sentence is slightly awk­ ward, too, and for the same reason: there is no apparent reason why a full definite should be used instead of a pronoun. 5 Here I am indebted to one of the referees for suggesring that examples like (Sa, b) don't exist. 6 It may seem easy to distinguish 'meta­ linguistic' froin 'ordinary' uses, but in

·

Bart Geurts 3 47 successful communication, and it is not part of the notion of commitment.

Having said this, I will go on to speak in tenns of the speaker's beliefs, because this is often more convenient than speaking in terms of commitment, and nothing hinges upon the distinction for my current purposes.

point, see Lewis (r98 3b). 19 An additional complication is that the notion of scope may be applied at different levels of analysis. To see why this matters, consider the following example: (i) Maybe John has started smoking. Here the presupposition that John didn't smoke (before a given date} is triggered within the scope of the modal expression. In the absence of evidence to the contrary, this presupposition will be accommodated in the main DRS, and the default interpretation for (i) that we predict may be paraphrased as follows: (ii) John didn't smoke, but maybe he has started smoking.

which has obtained wide scope. It is only in the latter sense that the notions of scope taking and presupposition pro­ jection are orthogonal to each other. 20 As pointed out to . me by one of the referees for the Journal. 2 1 More accurately: it seems to me that the intuition underlying the answers given by these authors is the same in each case, although strictly speaking their proposals are incompatible with one another. 22 C£ Sommers (1982}. One of the referees objected that Kripke has never claimed that pronouns are rigid designators, or that · they can be used as such. That is correct, but I am merely extending Kripke's observations to other types of expressions. To the extentthat Kripke's is an empirical theory it is based on the observation that the interpretation of a name engenders certain intuitions. But other expressions (pronouns, certain overt definites, and other presupposi­ tion-inducing expressions) engender the same intuitions, so instead of asking why names appear to be rigid, it is a sound policy to ask why any expression may appear to rigid. 23 Solution: I lied when I said that I had edited these textS; they are printed here exactly as I found them.

REFERE NCES Bach, K. (198 1), 'What's in a name?'

Dynamic Semantics,

371-86. Bach, K. (1987), Thought and Reference, Oxford University Press, Oxford. Bakker, R. ( 1988), The Dinosaur Heresies,

Amsterdam. Cresswell, M (198 5),

Australasian journal of Philosophy,

59,

Penguin Books, Harmondsworth (ftrst edition, 1986). Beaver,

D.

I. (1 993),

What Comes First in

ILLC Prepublica­ tion Series LP-93 - 1 5, University of

Structured Meanings: The Semantics of Propositional Attitudes,

MIT Press, Cambridge, MA. Edelberg, W. ( 1992), 'Intentional identity and the attitudes', Linguistics and Philosophy, 1 5 , S6I-96.

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

r 6 See Lewis (1968, 1 97 1 , 1983b), Edelberg (1992), Geurts (1995, to appear), Zeevat (1996}. The technical apparatus that I can only sketch here is developed in much more detail in Geurts (1995, to appear). 17 See Geach (1967), Edelberg (1992), Geurts (to appear). 1 8 For more extensive discussion of this

Now logically speaking this means that we give wide scope to the presupposition that John didn't smoke, but linguisti­ cally speaking there is no expression

348 Good News about the Description Theory of Names

·

·

Blackwell, Oxford. First published in D. Davidson & G. Harman (eds), Semantics for Natural Language, Reidel, Dordrecht {I972). Lewis, D. (I968), 'Counterpart theory and quantified modal logic', journal of Philo­ sophy, 65, I 1 3-26. Reprinted in: Lewis (I983a), 26-39. Lewis, D. (I971), 'Counterparts of persons and their bodies', journal of Philosophy, 68, 203-1 1. Reprinted in: Lewis (1983a), 47-54· Lewis, D. (I98o), 'Index, context, and content', in: S. Kanger & S. Ohman (eds), Philosophy and Grammar, Reidel, Dordrecht, 79-100. Lewis, D. (1983a), Philosophical Papers, Vol. 1, Oxford University Press, Oxford. Lewis, D. (1983b), 'Postscripts to "Counter­ part theory and quantified modal logic" ', in Lewis (1983a), 39-46. Loar, B. (1976), 'The semantics of sin­ gular terms', Philosophical Studies, 30, 35 3-77Sandt, R.A. van der (I 992), 'Presupposition projection as anaphora resolution', journal of Semantics, 9, 3 33-77. Sandt, RA van der & Geurts, B. (1991), 'Presupposition, anaphora, and lexical content', in: 0. Herzog & C.-R. Rollinger (eds), Text Understanding in LILOG, Springer-Verlag, Berlin, 2 s 9-96. Sommers, F. (1982), The Lagic of Natural Language, Clarendon Press, Oxford. Stalnaker, R. C. (I984), Inquiry, MIT Press, Cambridge, MA. Strawson, P. (1950), 'On referring', Mind, 59, 32D-44· Zeevat, H. (I996), 'A neoclassical analysis of belief sentences', Proceedings of the 1oth Amsterdam Colloquium, ILLC, University of Amsterdam, 723-42. ·

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

Fodor, J. A. (I987), Psychosemantics: The Problem of Meaning in the Philosophy of Mind, MIT Press, Cambridge, MA. Geach, P. T. (I 967), 'Intentional identity', journal of Philosophy, 64. 627-32. Geurts, B. {I99S). 'Presupposing', doctoral dissertation, University of Stuttgart. Geurts, B. (I998), 'The mechanisms of denial', Language, 74, 274-307. Geurts, B. (to appear), 'Presuppositions and anaphors in attitude contexts', m Linguistics and Philosophy. Geurts, B. & Sandt, RA. van der (to appear), 'Domain restriction' in P. Bosch and R. A. van der Sandt (eds), Focus: Linguistic, Cognitive, and Computa­ tional Perspectives, Cambridge University Press, Cambridge. Hamblin, C. L. (I97I), 'Mathematical models of dialogue', Theoria, 37, I 30-5 S· Kamp, H. & Reyle, U. (I993), From Discourse to Logic; Kluwer, Dordrecht. Kaplan, D. (I989), 'Demonstratives: an essay on the semantics, logic, metaphysics, and epistemology of demonstratives and other indexicals', · in J. Almog, J. Perry, & H. Wettstein (eds), Themes from Kaplan. Oxford University Press, Oxford, 48 I-563. Kneale, W. (I962), 'Modality de dicto and de re', in E. Nagel, P. Suppes, & A. Tarski (eds), Lagic, Methodology and Philosophy of Science. Proceedings of the 1960 International Congress, Stanford University Press, Stanford, 622-33. Kripke, S. A. (I977), 'Speaker's reference and semantic reference', in P. A. French, T. E. Uehling, & H. K. Wettstein (eds), Contemporary Perspectives in the Philosophy of Language, University of Minnesota Press, Minneapolis. Kripke, S. A. (I98o), Naming and Necessity,