No title

JOURNAL OF SEMANTICS AN INTERNATIONAL JoURNAL FOR THE INTERDISCIPLINARY STUDY OF THE SEMANTICS OF NATURAL LANGUAGE MAN...

Author: Oxford University Press

9 downloads 352 Views 22MB Size Report

This content was uploaded by our users and we assume good faith they have the permission to share this book. If you own the copyright to this book and it is wrongfully on our website, we offer a simple DMCA procedure to remove your content from our site. Start by pressing the button below!

Report copyright / DMCA form

DOWNLOAD PDF

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) AS SOC lATE EDITORS: MANFRED KRIFKA ( University of Texas, Austin) R o VAN DER SANDT ( University ofNijmegen) REVIEW EDITOR: TIBOR K1ss (IBM Scientific Centre, Heidelberg) ASSISTANT EDITOR: BIANKA BusCHBECK -WOI.F (IBM Scientific Cenrre, Heidelberg)

o

EDITORIAL BOARD : N. AsHER ( University of Texas, Austin) R. BARTSCH ( University of Amsterdam) M. BIERWISCH (MPG and Humboldt University Berlin) B. BoGURAEV (Apple Computer Inc) M. BoRILLO ( University of Toulouse) G. BROWN (University of Cambridge) K. BROWN (University of Essex) G. CHIERCHIA ( University of Milan) 0. DAHL (University of Stockholm) S. C. GARROD ( University of Glasgow) B. GEURTS (University of O snabriick) M. HERWEG (IBM Scientific Centre, Heidelberg) L. R. HoRN (Yale University) J. jACOBS ( University of Wuppertal) P. N. JoHNSON-LAIRD ( Princeton University) M. KAMEYAMA (SRI lnrernational, Menlo Park)

H. KAMP ( University of Stuttgart) S. LEVINSON (MPI Nijmegen) S. UiBNER ( University of Dusseldorf) SIR JoHN LYONS ( University of Cambridge) A. MANASTER-RAMER (Wayne State University) J. M C A LE ( University of Chicago) M. MOENS ( University of Edinburgh) F. J. PELLETIER ( University of Alberta) M. PINKAL ( University ofSaarbriicken) T. SANFORD ( University of Glasgow) R. ScHA (University of Amsterdam) A. VON STECHOW ( University of Tiibingen) M. STEEDMAN ( University of Pennsylvania) W. WAHLSTER (DFKI, Saarbriicken) B. WEBBER ( University of Pennsylvania) H. ZEEVAT ( University of Amsterdam) T. E. ZIMMERMANN ( University of Stuttgart)

c W Y

EDITORIAL ADDRESS: Journal of Semantics, c/ o Dr P. Bosch, IBM Germany Scientific Cenrre, Vangerowstr. 18, D-69115 Heidelberg, Germany. Phone: (49-6.1..1.1-) 59-4.1.51/4483. Telefax: (49-6.1..1.1-) 59-J.I.OO. Email: [email protected], [email protected] ©Oxford University Press 1995 All rights reserved; no part of this 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 either the prior written

permission of the Publishers, or a licence permitting restricted copying issued in the UK by the Copyright Licensing Agency Ltd, 90 Tottenham Court Road, London W1 P

Congress Street, Salem, Mass o 1 970.

9HE,

or in the USA by the Copyright Clearance Center,

27

The Journal of Semantics is published quarterly in February, April, August and November by Oxford University Press. Subscription is S I33·

$120 per

year. Second class postage paid at Newark NJ and at additional mailing offices. ISSN

0167-

POSTMASTER: send address corrections to The Journal of Semantics, c/o Virgin Mailing and Distribution, Cargo Atlantic,

10 Camptown Road,

Irvington NJ

07111-IIO\. USA.

For Subscriptions information please see inside back cover

JOURNAL OF SEMANTICS Volume

12

Number

2

SPECIAL ISSUE: LEXICAL SEMA NTICS PART

2

CONTENTS GEOFFREY NUNBERG

Transfers of Meaning

109

jAMES PusTEJOVSKY AND PIERRETTE BoUILLON Aspectual Coercion and Logical Polysemy

I

33

NICHOLAS AsHER AND PIERRE SABLAYROLLES

A Typology and Discourse Semantics for Motion Verbs and Spatial PPs in French

I63

Book Review

2I I

Linguistics and Philosophy

A Journal of Natural Language Syntax, Semantics, Logic, Pragmatics, and Processing Editor-in-Chief:

Journal

Gregory N. Carlson

Highlight

Associate Editors:

Jeroen Groenendijk, Martin Stokhof, Laurence R. Hom, Pauline Jacobson, William C. Rounds, Erhard Hinrichs, Mark Richard, Scott Soames

Executive Board:

Richmond H. Thomason, Barbara Partee, Robin Cooper

Linguistics and Philosophy is a journal for studies focused on

natural language, and is of interest to practitioners in the disci plines covered in the title. Although the field thus described is so extensive that a complete listing of relevant topics is precluded, at least the following specifically fall within it: - tradional areas in the philosophy of language such as meaning, truth, reference, description, entailment, speech acts - tradi tional areas of linguistics such as syntax, semantics and prag matics (when the studies are of sufficient explicitness and generality to be also of philosophical interest) - aspects of artificial intelligence concerned with language such as computational linguistics and natural language processing - systems of logic with strong connections to natural language: modal logic, tense logic, epistemic logic, intensional logic - philosophical questions raised by linguistics as a science: lin guistics methodology, the status of linguistics theories, the nature of linguistic universals - philosophically interesting problems at the intersection of lin guistics and other disciplines: language acquisition, language and perception, language as a social convention.

Subscription Information 1995, Volume 18 (6 issues) Subscription Rate: NLG 488.00

Private Subscription Rate: NLG

including postage and handling.

ISSN 0165-0157 279.00. 195.00 I USD 85.00. Prices are

I USD

P.O. Box 322, 3300 AH Dordrecht, The Netherlands P.O. Box 358. Accord Station, Hingham,

MA 02018-0358, U.S.A.

Kluwer academic publishers

C

o

m p

u

t

a

t

1 o

_11� a

I

� � r1 g 1.:! 1

s t

i

_

c

s_

Computational linguistics is the foremost journal devoted exclusively to computational analyses of natural language.

It encompasses AI research in language, linguistics, and the psychology of

Julia Hirschberg Editor II

language processing and performance.

1995 lnstirurional Rate: $90.00. Add

With each issue providing applied and

$16.00 posagc and handling ouuide

theoretical papers, book reviews, technical correspondence, and letters to

U.S.A. Canadians add additional 7% GST. Prices subject to change without notice. Prepayment is required. Send

the editor, the journal presents a

chcdt dfllwn

stimulating forum for the exchange of

fund.., MC, VISA, or AMEX number

ideas and trends in computational

MIT Pras Joaraab 55

language.

Computational Linguistics is the

againsr a U.S. bank in U.S. ro:

Hayword Succt

Cambridgr. MA02142-1399

TEL: 617-253-2889

fA)(, 617-258-6779 [email protected]

journal serving a rapidly expanding interdisciplinary field and should be part of every research library colleccion. Recommend CL to your librarian today.

Individual subscriptions arc available only through mcmbcrlhip in

rhc

Associuion for CompurarionaJ Linguistics.

Please con�:aet: Priscilla Rasmussen Associarion for Computarional Linguisrics (ACL)

Published quarterly by The MIT Press for me AJsociarion for Computacional Linguistia ISSN 0891-2017. Volume 21 forlhcoming.

P.O.Box 6090

Some!SCI, NJ 08875 USA

Td: 908-873-3898 [email protected]

Journal of Logic, Language and Information Editor:

Peter Giirdenfors

Dept. ofPhilosophy, Lund University, Sweden Section Editors:

D. Gabbay Daniele Mundici Stanley Peters, R. Stalnaker, Johan van Benthem

Journal Highlight

This is the official publication of the European Association for Logic, Language, and Information. The scope of the journal is the logical and computational foundations of natural, formal, and programming languages, as well as the different forms of human and mechanized inference. It covers the logical, linguistic, and information-theoretic parts of the cognitive sciences. Examples of main subareas are Intensional Logics including Dynamic Logic; Nonmonotonic Logic and Belief Revision; Constructive Logics; Complexity Issues in Logic and Linguistics; Theoretical Problems of Logic Programming and Resolution; Categorial Grammar and Type Theory; Generalized Quantification; Information-Oriented Theories of Semantic Structure like Situation Semantics, Discourse Representa tion Theory, and Dynamic Semantics; Connectionist Models of Logical and Linguistic Structures. The emphasis is on the theoretical aspects of these areas. The purpose of the journal is to act as a forum for researchers interested in the theoretical foundations of the above subjects and their interdisciplinary connections, with an emphasis on general ideas increasing coherence. Journal of Logic, Language and Information is abstracted/Indexed in Mathematical Reviews; Current Mathematical Publications; MathSci; lnspec Database; Zentralblatt fOr Mathematik; Sociological Abstracts; Unguistic and Language Behavior Abstracts; MLA International Bibli ography Subscription

Information

ISSN 0925-8531

1995, Volume 4 (4 issues)

Subscription Rate: NLG 435.00 I USD 249.00, including postage and handling. Special rate for members of the European Foundation for Logic, Language and Information available. For more information, please write to the FoLU Bureau, Muidergracht 24, 1018

TV

FWI,

University of Amsterdam, Plantage

Amsterdam, The Netherlands

P.O. Box 322, 3300 AH Dordrecht, The Netherlands P.O. Box 358, Accord Station, Hingham,

MA 02018-0358,

U.S.A.

Kluwer academic publishers

Highly respected for its sustained excellence in quality of scholarship, Linguistic Inquiry leads the field in research on current topics in linguistic theory. Issue after issue, LI keeps you informed of new theoretical developments by presenting the latest in international research. Sc:Ject Forthcoming Articles 26:1 Feature Geometry and Feature Spreading

Morris Hnllt

26:2 "Somethmg Else" for the Bmding Theory Pctrr Culiclltrtr aud Rnyfacktmlll{f 26:3 Marking Statements, Complexity and Simplirication Procedures A11drta Calnbrcst 26:4 NC. licensing and Underspecification in Optimality Theory Ju11ko /lc1, Annm Mtslcr. aud Jove Pad�ell Unguistic /uquiryWorld-Wide Web (WWW) Home Pase URL: http://www·mitprcss.mit.edu/irnl.s<.Jtalog/linguistic.html

1995 RATES Individual SSO.OO,Jnstitution $108.00, Student (copy of current 10 requirl!d)/Rctll't'd SJ200. Outside U.S.A., add $16.00 postage and

handhng. Canadians add additiona17% CST. Pricts su/litclta chnugt

tvrllrllul IIIliIet PreJXIYmflll rcqrnrtd

Send chcck�rawn against a

U.S. bank tn U.S. funds, payable to linguistic luquiry-AMEX, MC, or VISA numbcrtu: Ctrculatmn �partment

MIT PRESS JOURNALS

55 Hayward Street Cambridge, MA 02142-1399 USA TEL: (617) 253-2889 FAX: (617) 258-6779 )[email protected]

Samuel Jay Keyser

Editor m Chief Volume 26 forthcoming Published quarterly by The MIT Press ISSN 0024-3892

INGUISTIC

IR

[R!]@l�(UJ [j@l� l©liT'iJ� (UJ@l�@ @liT'iJcdl lo IT'i)� (UJ o��o� 1r rtiJ ® (Q) IT]' Editor

Joan IVIaling

Dept. of Psychology, Brandeis University, Waltham, MA, USA Natural Language and Linguistic Theory provides a forum for the discussion of theoretical research that pays close attention to natural language data, so as to provide a channel of communication between

Journal Highlight

researchers of a variety of points of view. The journal actively seeks to bridge the gap between descriptive work and work of a highly theoreti cal, less empirically oriented nature. In attempting to strike this balance, a primary goal of the journal is to encourage work which makes complex language data accessible to those unfamiliar with the language area being studied, and work which makes complex theoretical positions more accessible to those working outside the theoretical framework under review. The journal is structured to include: - theoretically oriented work on the syntax, semantics, phonology and the lexicon of natural language; - interdisciplinary contributions written with theoretical linguists in mind; - surveys of recent theoretical developments which facilitate accessi bility for a graduate student readership; - reactions/replies to recent papers; - brief notes, comments or observations concerning languages or theoretical points; - letters to the editors; - book reviews of important linguistics titles; - special topic issues. Natural Language and Linguistic Theory is surveyed by

MLA Interna

tional Bibliography, Linguistic & Language Behavior Abstracts, Current

Contents/Arts

&

Humanities. Current Contents/Social

Sciences, ASCA. Arts

&

&

Behavioral

Humanities Citation Index, Social Sciences

Citation Index. Linguistic Abstracts, Linguistic Bibliography. IBZ/IBR, Computer

&

Information Systems Abstracts Journal, Sociological

Abstracts. Cambridge Scientific Abstracts, Research Alert, Informa tion Technology and the Law, PASCAL Database, Bibliographie Linguistique. Abstratcs in Anthropology

Subscription Information

1995, Volume 13 (4 issues)

ISSN 0167 -806X

Subscription Rate: NLG 440.00 I USD 252.00.

Private Subscription Rate: NLG 164.00 /USD 62.00. Prices are including postage and handling.

P.O. P.O.

Box 322, 3300 AH Dordrecht, The Netherlands Box 358, Accord Station, Hingham,

MA 02018-0358,

U.S.A.

Kluwer academic publishers

For information on advertising in this publication, please contact:

jane Parker journals Advertising Oxford University Press Walton Street Oxford OX2 6DP, UK Tel: 01865 267917 Fax: 01865 267782

OIDF®D

.IDURtW.S

e

Scope of this Journal The JOURNAL OF SEMANTICS publishes articles, notes, discussions, and book reviews in the area of natural 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 qualiry {to be judged by at least two referees) and should relate to questions of comprehension and interpretation of sentences or rexts 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 accept ance for publication.

Information for Authors Papers for publication should be submitted in 3 copies ro the managing editor. They should be r yped on

A4 (or similar

format), one-sided, double-spaced, and with a wide margin and must be accom

panied by an approx. 200 word summary. Notes and bibliographical references must appear at the end of the rypescript. 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 anony mous 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 Universiry Press. Authors receive 20 offprints of their published articles and 10 offprints of their published reviews, free of charge. Larger numbers can be supplied at cost price by advance arrange ment.

Subscriptions The Journal of Semantics is published quarterly. Institutional: UK and Europe £66.oo; USA and Rest of World US$120.00. {Single issues: UK and

Europe £!9.00; USA and Rest ofWorld US$Js.oo.)

Personal:• UK and Europe £J2.oo; USA and Rest of World US$61.00. {Single issue: UK and Europe

£9.oo; USA and Rest of World US$17.00.)

• 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 Universiry Press, Wal

ton Street, Oxford OX2 6DP. Previous volumes can be obtained from Dawsons Back Issues, Cannon House, Park Farm Road, Folkestone, Kent CTI9 sEE, rei +44 (o)I JOJ 8SOIOI, fax +44 (o)I JOJ 850440. Volumes t-6 are available from Swets and Zeitlinger, PO Box 830, 2160 SZ Lisse, The

Netherlands.

Order Information Payment is required with all orders and subscriptions are accepted and entered by the volume. Pay ment may be made by cheque or Eurocheque (made payable to Oxford Universiry Press), National Girobank (account soo 1056), 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 Universiry Press, Walton Street, Oxford, OX2 6DP, UK. Telex 837330 OXPRES, Tel +44 (o)1 86s 267907.

Advertising Advertisements are welcome and rates will be quoted on request. Enquiries should be addressed to Jane Parker, Advertising Sales Manager, Oxford Universiry Press, Walton Street, Oxford OX2 6DP,

UK. Tel: +44 (o)186s 56767 Fax: +44 (o)186s 267782.

)<>umal <>fSemalltics

©Oxford University Press 1 WS

12: 109-1 Jl

Transfers of Meaning GEOFF R E Y N U NB E R G Xerox Palo Alto Research Center and Stanford University

Abstract

1 INTRO DUC T ION By 'transfers of meaning' I mean the ensemble of productive linguistic processes that enable us to use the same expression to refer to what are intuitively distinct sorts of categories of things. Broadly speaking, transfers involve all the figures that traditional rhetoric describes as metaphors, synesthesias, metonymies, and synecdoches, in all their synchronic manifesta tions. The difference is that transfers are linguistic processes, whereas the rhetorical figures are defined and classified according to the independent con ceptual relations that they exploit. The difference between metaphors and metonymies, for example, is that the first presupposes a resemblance and the second a contiguity. And we can go on to classify figures according to the particular conceptual schemas they rest on, either as general correspondences like 'abstract for concrete', 'part for whole', and 'animate for inanimate', or as more specific schemas like 'life is a journey' or 'polities are bodies'. But taken by themselves, these schemas and principles aren't sufficient to explain the linguistic phenomena of transfer. Granted that there is a salient correspondence between monarchs and crowns, for example, it still has to be explained why the word

crown

can be used to refer to monarchs-or for that matter why this fact

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

'Transfers of meaning' are linguistic mechanisms that make it possible to use the same expression to refer to disjoint sorts of things. Here I discuss predicate transfer, an operation that takes names of properties into new names that denote properties to which they functionally correspond. It is this operation that is responsible for the new meaning of the predicate parked out back in the utterance 'I am parked out back', as well as for the lexical alternations that figure in systematic polysemy. Predicate transfer is subject to two general conditions, which require that basic and derived property stand in a functional correspondence and that the derived property should be a 'noteworthy' feature of its bearer. I argue that by appealing to predicate transfer we can maintain a very strict definition of syntactic identity, which rules out all cases of'sortal crossing', in which a term appears to refer to things of two sorts at the same rime, as in examples like Ringo squeezed himself into a tight space; in such a case, the reflexive is strictly coreferential with its antecedent.

II o

Transfers of Meaning

should have any linguistic consequences at all. For this we have to look to specifically linguistic mechanisms, which is what I will be talking about here.1 These mechanisms exist in the service of the expression of conceptual regu larities, but they are in principle independent of them, and are constrained in ways that don't permit a purely pragmatic explanation. They are the linguistic handmaidens of figuration, but each is specialized in her offices.

2 ME C H ANI S M S O F TR ANS FER The easiest way to appreciate the difference between rhetorical figures and the

transfer. A customer hands his key to an attendant at a parking lot and says

either ( I ) or (2 ):

I. This is parked out back. 2. I am parked out back. Both these utterances involve metonymies. In ( I ), for example, we would be inclined to say that the subject refers not to the key that the speaker is holding, but to the car that the key goes with. And in fact all the linguistic evidence

supports this analysis. For example, the number of the demonstrative is determined by the intended referent, not the demonstratum. So even if the cus tomer is holding up several keys that fit a single car, he would say 'This is parked out back', whereas if he's holding up a single key that fits several cars, he would say, 'These are parked out back'. We can make the same point looking at languages that mark demonstratives and adjectives for grammatical gender. In Italian, for example, the word for key is feminine, Ia chiave, and the word for truck is masculine,

il cam ion . And if a customer gives the attendant the key to a

truck it will be the referent, not the demonstratum, that determines the gender of the demonstrative and the adjective for 'parked', as in (3):

3· Holding up a key (Ia chiave, fern. sg.) to refer to a truck (il camion, masc.) Questa (masc. sg.) e parcheggiato (masc. sg.) in dietro. 'This (masc.) is parked (masc.) in back.'

One final example to the same effect: we can conjoin another predicate that describes the car, but not a predicate that describes the key: 4· This is parked out back and may not start.

S·

??This fits only the left front door and is parked out back.

So there's every reason for saying that the subject of sentences like these refers to the car.

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

linguistic mechanisms is to consider how we can exploit the same sorts of correspondences among things in the world to effect two different kinds of

Geoffrey Nunberg

II I

But what of an utterance like (2 ), 'I'm parked out back'? This too is plainly a

metonymy of some sort, and there may be a temptation to analyze it as we would ( I ), saying that the subject of the sentence refers not to the speaker, but to

the speaker's car.2 But the tests we have used to validate this analysis for the demonstrative in ( I ) give a different answer here. For example, if the speaker has

two cars he wouldn't say:

6. We are parked out back. (though of course this would be an appropriate utterance if there were two people who were waiting for the car). By the same token, an Italian man who was waiting for his car would express this by using a masculine adjective

parcheggiato for 'parked', even though the word for 'car' is feminine, Ia macchina . And in this case, we can conjoin any other predicate that describes the speaker, but not always the one that describes the car:

8. 9·

I am parked out back and have been waiting for I s minutes.

* I am parked out back and may not start.

The conclusion is that the subject of ( I ) refers to the speaker, and the transfer

involves the predicate.J That is, the predicate parked out back contributes a property of persons, the property they possess in virtue of the locations of their

cars. Now the difference between these examples clearly doesn't have anything to do with the kind of relations they exploit. In both cases we assume a correspondence between the things in one domain, the cars parked in various locations, and the things in another domain, keys or drivers as the case may be. But we can take semantic advantage of these correspondences in two different

ways. Sentence ( I ) is a case of deferred ostension or deferred indexical reference, a process that allows a demonstrative or indexical to refer to an object that correspond in a certain way to the contextual element picked out by a

demonstration or by the semantic character of the expression.4 In this connection, note that we can't get this kind of deferred reading when we use a description in place of a demonstrative, as in ( 10):

10. *The key I'm holding is parked out back.s Whereas (2 ) exemplifies another kind of transfer process, which I'll call predicate transfer. The principle here is that the name of a property that applies to something in one domain can sometimes be used as the name of a property that applies to things in another domain, provided the two properties correspond in a certain way. And just to fill out the contrast here, note that unlike deferred ostension, predicate transfer is indifferent to how the bearer

---

-�----

-

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

7. lo sono parcheggiato (*parcheggiata , fern. sg. ) dietro.

II2

Transfers of Meaning

of chis new or derived property is referred to-by an indexical or description or whatever. For example, in chis situation the parking lot manager could say to the attendant: I I . The man with the cigar (Mr. McDowell, etc.) is parked out back.

What these examples show, then, is that unlike rhetorical classifications like metaphor and metonymy, the various mechanisms of transfer can't be distinguished simply by pointing at the types of correspondences they exploit. And for this reason the description of these mechanisms is fundamentally a linguistic problem, rather than a problem of conceptual analysis. That is, there

In the rest of chis paper I will be concentrating on just one of these

mechanisms, predicate transfer. In this section I will schemacize the conditions that license chis operation. In the following sections I will spell out the role of predicate transfer in lexical polysemy, and then discuss its implication for some well-known syntactic puzzles. Finally, I will talk about some of the methodo logical difficulties that predicate transfer raises.

3 CONDI TIONS ON PRE D I C A TE TR ANS FER Predicate transfer is subject to two conditions. The first of these I have already mentioned: the property denoted by the derived predicate has to correspond in a certain way to the property denoted by the original predicate. With an utterance like (2), for example, we begin with a functional correspondence

between the locations of cars in a lot and the properties of the owners, or, more

accurately, the 'havers', of these cars. When two property domains correspond in an interesting or useful way-of which more in a moment-we can schemarize the operation of predicate transfer as follows:

I 2. Condition on predicate transfer Let .9 and .9' be sets of properties that are related by a salient transfer function gt: .9 -+ .9' Then if F is a predicate that denotes a property P E .9, there is also a predicate F', spelt like F, that denotes the property P', where P' = gt(P).6 In the cases we have been talking about, of course, these correspondences between properties are mediated by correspondences between their bearers e.g. the functional relation from cars to their owners-and we might want to represent chis directly. So let h be a salient function from a set of things A to another (disjoint) set of things B. Then for any predicate F that denotes a

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

is nothing we can learn about keys, drivers, or cars that will help us to explain the differences betWeen examples like ( 1 ) and (2 ).

Geoffrey Nunberg

II 3

property P that applies to something in A. we can represent the meaning of a derived predicate F', spelt like F, as in either ( I 3a) or ( 13b):

I3a. A P.A. y(Vx[dom h)· h(x) - y - P(x)) I 3b. A P.A. y(3xdom h)· h(x) - y A P(x)) Note that this entails that utterances of this type are in fact ambiguous between 'universal' and 'existential' readings, depending on whether all or only some of the bearers of the original property are in the inverse image of h for a given value. And in fact both types of reading are generally available. In cases like 'I am parked out back', we would normally assume that the speaker means to say that all the (relevant) cars he owns have the property in question. By contrast, when a painter says, 'I am in the Whitney', she doesn't imply that all her

says of his firm, 'We are in Chicago', he might intend either sort of interpretation, depending on whether he is talking about the firm's head quarters or one of its offices. Still, it is more useful to think of these two types of readings as two ways of representing the general schema given in

( 1 2),

rather than as two distinct

conditions that license predicate transfer. One reason is that not all of the relations between properties that figure in predicate transfers are mediated by relations between their individual bearers. This is the crucial feature that distinguishes metonymies from metaphors, where one property calls up another directly, either in virtue of some similarity between them or through

the perceptual mechanisms that underlie synsthesias like a dark mood . I'll come back to metaphor at the end of this paper; for the moment we just want to bear in mind that metaphors and metonymies aren't different types of transfer, they're different conditions that can support a functional relation between properties.

The schemas in (I 3) do a reasonable job of representing the truth-conditions associated with utterances like 'I am parked out back' and 'I'm in the Whitney', but they miss some important pragmatic conditions on the use of such utterances. For example, suppose my car has the property of having once been driven by Jean Gabin. Then according to the conditions in ( I 3), I should be able to use the name of this property to describe the property that I acquire in virtue . of my relationship to my car. But it would be odd for me to say:

I4.

??I was once driven by Jean Gabin.

even in a context in which it might be relevant to say 'My car was once driven by Jean Gabin'. By the same token, a painter might say with reference to one of her paintings, 'I'm in the Whitney Museum', but not, ordinarily: I 5·

??I'm in the second crate on the right.

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

paintings or even all her relevant paintings are in the Whitney, but only that something she painted is in the Whitney, as in ( 13b). And when an accountant

114

Transfers of Meaning

16. Ringo was hit in the fender by a truck when he was momentarily distracted by a motorcycle. 17. ??Ringo was hit in the fender by a truck two days after he died. Let's assume that these utterances exemplify transfers of the meanings of the relevant relational expressions-that is, that Ringo denotes the singer rather than his car.8 The difference between the two cases is that when a truck hits Ringo's car while he is driving it. the event will have important consequences for him as well: he is likely to have been startled, or annoyed, or put to trouble and expense. Whereas once Ringo is dead, nothing that happens to his car is going to invest him with any properties worth mentioning.9 Another way of putting this would be to say that it is unlikely to be conversationally relevant to

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

Intuitively, the difference is this: when a painting goes into a museum its creator acquires a significant or notable property, whereas when it goes into a crate she doesn't, at least not usually? Let me describe this condition by saying that predicate transfer is only possible when the property contributed by the new predicate is 'noteworthy', which can be reckoned in either of two ways. The first type is exemplified in an utterance like 'I'm parked out back' or 'I'm in the phone book'. The purposes of these utterances is to classify the members of one set of things (e.g. garage customers) relative to the immediate conversational purposes (of discharging the attendant's responsibilities, say), in virtue of their relation to the identifying properties of some other groups of things {the cars). In cases like these we will say that a property is noteworthy if it offers a useful way of classifying its bearer relative to the immediate conversational interests. From the point of view of a garage attendant, for example, a customer is usefully classified in terms of the properties he acquires from the location of his car('I'm parked out back') or its overall appearance('I'm a red Mustang') rather than in terms of the properties he acquires from its provenance or mechanical condition {'I was bought from a friend', 'I shimmy at high speed')-though of course some of these properties might be more useful for other conversational purposes. The second type of noteworthiness is exemplified by utterances like 'I'm in the Whitney', or 'American Express is in Singapore now'. In these cases the derived property has a more abiding interest or consequence for its bearer, beyond the immediate conversational purposes. The difference between the two may depend on fairly subtle features. The utterance 'I'm in the phone book' most likely serves the purposes of immediate classification {'You don't have to write down my number'); the utterance 'Clinton is in the phone book' is more likely offered by way of remarking Clinton's accessibility, or whatever. Noteworthiness can be thought of as a special case of relevance, depending on how we understand the latter notion. To see this relation, we can consider examples(16) and(17), adapted fromjackendoff(1992):

Geoffrey Nunberg

11s

know that the deceased Ringo has the property that he acquires when his (former) car is hit on the fender, even though it may be relevant to know of the car that it has been hit. But our ability to make this distinction rests on our being able to distinguish between the relevance of a proposition (e.g. that Ringo's car was hit) and the relevance of its trivial entailments (e.g. that Ringo has the property of having had his car hit). The question is of some interest for the theory of conversation, but I won't pursue it here. For the present we can just take noteworthiness in an intuitive way: they are properties that matter. 4

P REDI CATE T R A N S FE R O N C O M M O N N O U N S

I

8. Who is the ham sandwich?

The process of transfer is pretty straightforward here; from the point of view of the waiter, at least, customers acquire their most usefully distinctive properties in virtue of their relations to the dishes they order. But in this case, unlike the 'parked out back' examples, the relevant property is expressed by a common noun, which can equally well be used as the content of an NP in referential position in a sentence like ( I 9): 1 9·

The ham sandwich is at table 7·

In earlier discussions several people, myself among them, have talked about examples like {I9) as involving a kind of 'reference transfer', or 'deferred reference', with the implication that an actual ham sandwich must figure in the interpretation of the utterance-that is, that the transfer operates on the NP interpretation. 10 But there are a number of reasons for concluding that the transfer here takes place on the common noun meaning-that is, that this is a case of meaning transfer, rather than reference transfer. One way to make this point is to look at examples where there is a clash between the number of the dish and the number of the patron, as in (2o): 20. That {*those) french fries is (*are) getting impatient. (though the felicity of examples like this depends as well on specific principles of English morphosynrax)." A related example is They played lots of Mozart,

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

The examples o f predicate transfer that I've been talking about up to now have all involved the meanings of predicates or verb phrases, but I don't want to give the impression that the process is essentially a matter of function coercion, rather than argument coercion. In fact predicate transfer is available for the property names that are expressed by common nouns, as well, whether they appear in predicate position or referring position. Take ( I 8), as uttered by a restaurant waiter:

1 16

Transfers of Meaning

where a transferred predicate co-occurs with a quantifier or article that is inconsistent with its literal meaning-how could Mozart used as a mass term refer in a well-formed way to the composer?12 Or we can consider the interpretation of the determiner in example (19), which doesn't presuppose the existence of a unique ham sandwich (think of a waiter in a fast-food restaurant who is standing in front of a table piled with ham sandwiches), but does presuppose the existence of a unique ham-sandwich orderer. Finally, we can consider some examples involving anaphors (a matter I'll take up in greater detail later on). Fauconnier (1985) gives examples (2 1 ) and (22). 21. The mushroom omelet was eating with chopsticks. 22. *The mushroom omelet was eating itself/himself with chopsticks.

23. *Je suis en train de manger la tarte flambee qui est assise ala table 20. '*I am eating the flaming pie who - that is sitting at table 20.' For (23 ) to have an interpretation, we would require that the phrase Ia tarte flambee qui est assise a Ia table 20 refer to a dish. But it doesn't. 5 PREDI C A TE TRANSFE R I N S Y S TE M A T I C P O LYSEMY The availability of transfer for common nouns, adjectives, and the rest is what underlies the patterns of lexical alternation that go by the names of 'regular polysemy', 'systematic polysemy', 'logical metonymy', and the like. Of course, not all transfers that apply lexical categories are instances of lexical processes. Many are too much dependent on the availability of a specialized context. The correspondence between the properties of dishes and customers provides a useful means of identification only in the specialized domain of a restaurant, and then only relative to the interests of waiters-this isn't a good way of classifying people in general, or even of classifying customers when the discourse isn't directly concerned with the relations between customers and orders (I doubt whether a waiter would say, 'I saw the ham sandwich at table 7 the other day driving a Mercedes'). Cases like this we can think of as 'occurrent metonymies', where the relation between two domains can only be exploited in a restricted range of situations.'4 So we would be reluctant to say that the item ham sandwich is polysemous, with lexical sense along the lines of'ham sandwich

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

The use of a reflexive in (22) would presume that the object of eat was introduced by the subject NP-that is, that the subject NP actually referred to a mushroom omelet on the route to its ultimate interpretation. B ut inasmuch as the transfer actually takes place at the level of the common noun, which contributes only a property of persons, the example is ill formed. IJ Kleiber ( 1 984) raises a similar puzzle with regard to (23 ):

Geoffrey Nunberg I I 7

orderer'; w e might better class i t with Clark's 'contextual expressions' (see e.g. Clark I 99 3). But often the property correspondences that license predicate transfer hold across a wider range of situations, providing a more context-independent way of classifying the bearers of derived properties, which answers to more general interests. In these cases we may very well want to say that the transferred predicate represents a lexical sense of the item in question. To a certain extent,

black white senses like 'In chess, the person playing a white'. I expect the idea here

this is a relative matter. For example dictionaries often assign the words and

is that even though this is a context-dependent correspondence, the derived property is much more generally useful for classifying players than the property

come to the general patterns of lexical alternation commonly described by means of the formula 'x for y'. For example:

24.

Examples of systematic polysemy transmissions for cars: 4 speed ,

automatic , etc.

texts for inscriptions: a Webster's Third, a Guide Bleu, etc. painters for works: a Picasso , a Derain , etc.

containers for volumes of stuff:

She drank two classes , etc.

writer for oeuvre: fifty pages ojWordsworth

Indianapolis votedfor the referendum The table is made ofoak .

place for inhabitants: tree for wood:

Now from one point of view, we might want to say that there is no qualitative difference between these alternations and the cases of occurrent metonymy we were looking at a moment ago. Certainly we should bear in mind that just as with occurrent metonymies, the availability of a transferred predicate in the alternations in

(24)

is subject to the general conditions of

noteworthiness that I discussed earlier. Take the well-known rule of 'grinding',

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

of having ordered a ham sandwich as a means of classifying restaurant customers, since so many other things follow from which color a player takes.15 And as usages become progressively more useful and less context-specific, we

which turns count nouns into mass nouns that denotes the stuff they are made of, as in 'We ate chicken', 'They wear beaver', or 'The table was made of oak'. As it happens, we don't ordinarily apply grinding to the names of breeds of animals to derive the names of their meats or their fur or skin, as in 'a plate of Rhode Island Red' or 'a bag made of Hereford', for the simple reason that breeds are not usually regarded as gastronomically or sartorially natural kinds; this is not a useful way of identifying a type of meat or fur. But there are exceptions, and in just these cases breed names can be used as mass terms-for example, when we are talking about the breeds of sheep from which wool is derived, as in 'His sweater was made of angora', or when the government issues a warning telling consumers to 'stay away from Hereford until the insecticide danger diminishes'.

-------

118

Transfers of Meaning

Even if we call these alternations 'lexical', then, they are still subject to the same pragmatic constraints that determine the acceptability of occurrent metony mtes. At the same time, there are also reasons for assigning alternations like these a specifically lexical status. They tend to have a high text frequency; they often become habitual or 'idiomatic'; and they may develop specialized senses that aren't entirely predictable on the basis of general principles of transfer (for example, the specialized meanings of mass terms like

balsam

or

lavender ). Then,

too, we might want to argue that the features that figure in these particular correspondences are present in the lexical meanings of the items themselves: that it is the lexicon that tells us that a tree yields wood, a painter makes paintings, a glass is a container, and so on.'6 Finally, and I think most telling, the

some reason or another, for example, the English rule of grinding seems not to apply to yield the names of liquids:

25. 26.

??We fried the chicken in safflower (olive, com, etc.). ??I enjoy a glass of orange (pear, apple, etc.) with my breakfast.'7

Of course, it may very well be that in some kitchens cooks routinely use

sa.fllower for 'safflower oil', but this is not the general practice. And analogously,

there seems to be no good reason why French-speakers should use the names of fruits to refer to the brandies made from them ( une prune , un poire ) whereas English-speakers do not.'8 Another example drawn from Apresjian

( 1 973): in Russian you can use the name of an organ to denote a disease of that organ, as in 'She has a kidneys', whereas in English you can't, though presumably the usage would be comprehensible enough. And according to

Jerrold Sadock (personal communication), Greenlandic Eskimo does not per mit grinding of animal names to produce mass terms for either meat or fur, though you can apply grinding to tree names to get terms for wood. So you can't say 'He eats walrus' or 'He wears walrus', but you can say 'His boat is made of oak'. All of this suggests that we need a second kind of device to explain how particular languages lexicalize particular cases of predicate transfer-what are

described in Nunberg and Zaenen ( 1992) as 'lexical licenses'. These are not the same things as lexical rules, which are defined over particular lexical classes and subject to synchronically arbitrary syntactic or morphological constraints, though of course licenses may become lexical rules diachronically.'9 What licenses do, rather, is to index specific types of correspondences (e.g. between animals and their meat, or singers and the songs they sing) as available for exploitation to produce new lexical items. In this regard they can be thought of on the model of what Morgan

( 1 978)

calls 'conventions of use'-rules that

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

lexical alternations that involve predicate transfer are often subject to language specific constraints that seem to have no general pragmatic explanation. For

Geoffrey Nunberg

1 19

6

P REDI C ATE T R A N S F E R AND SEMA N T I C COMPOSITION

A related problem with this emphasis on the lexicon (and I think that with most writers this is only a question of emphasis, rather than of theoretical dispositions) is that it can obscure the fact that transfer is essentially a phrasal process, which operates on predicates of any complexity, or more generally on categories up to the type X' in the Jackendoff framework. And as such it is subject to familiar constraints on semantic composition. Take the sentence: 27. The article contains a description of a widely-studied Peruvian virus . Let fd be a transfer function that takes the names of agents into the names of the diseases they cause. Then we could represent the four possible interpretations of the NP a widely-studied Peruvian virus as in (27): 28. (widely-studied (Peruvian (virus))) = 'a microorganism that is native to Peru and widely studied.' 29. (widely-studied (Peruvian fd(virus))) - 'a virally-caused disease that is endemic to Peru and widely-studied.' 3 0. (widely-studied fd(Peruvian (virus)))= 'a widely-studied disease caused by a virus native to Peru.' 3 1 . fd(widely-studied (Peruvian (virus)))='a disease that is caused by a widely studied virus native to Peru.' But the phrase can't have any interpretation as in (4 I ) : 32. fd(widely-studied) (Peruvian fd(virus))) ='a disease endemic to Peru that is caused by a widely studied virus.'

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

determine which practices are 'idiomatic', in the common acceptation of the term, as opposed to being merely possible. There is much more to be said about rules like these, but for our purposes the important thing to keep in mind is that both lexical and nonlexical transfers are subject to the same range of pragmatic constraints and exploit the same semantic mechanisms. This is one reason why it is a mistake to think of transfer itself as an intrinsically lexical process, a picture that is implicit in the tendency to talk about this phenomenon in the context of discussions of lexical polysemy, as well as in the names of many of the more-or-less formal devices that have been proposed to deal with it, such as 'lexical networks' (Norvig & Lakoff I 987), 'lexical subregularities' (Wilensky I 99I), and 'lexical implication rules' (Ostler & Atkins 1 99I). Even when a rule like grinding has been explicitly lexicalized, what is conventionalized is not the mechanism involved, but only the particular semantic domains the mechanism is allowed to operate on. 20

1 20 Transfers of Meaning

Peruvian would apply to the widely studied and virus . The point may be clearer if we note the oddness of the phrase ??a tiny incurable virus, which could only mean 'an incurable disease caused by a tiny virus', with That is, the sentence can't have a reading where

output of the transfer function on the composition of

the adj ectives again applying in a way that would violate the ordinary process of composition. (Note that it is not the order of adjectives itself that is responsible here: we can say a tiny incurable wart.) There are several ways to obtain the effect of compositionality here. One approach would be to try to accomplish all of this lexically, though the kind of coercion mechanism proposed by Pustejovsky and developed, among others, by Briscoe and Copestake in this volume. They discuss the phrase

corn-fed chicken ,

of underived items like

veal

and

beef).

In that case

corn-fed

has its normal,

unextended sense, and no phrasal transfer is involved. And we could tell an analogous story about Peruvian virus under the reading 'virally-caused disease endemic to Peru' in which the adjective applies in its literal sense to a feature in some element (relic?) in the qualia structure of

virus .

But this approach has a number of problems. For one thing, it still leaves us with the job of explaining the restrictions on compositionality that we noticed in connection with the phrase ??a tiny incurable virus ; if both adjectives have their literal sense here, why shouldn't they compose as they do in tiny incurable wart ? We could introduce constraints on the operation of coercion, of course, but these would wind up simply recapitulating the process of phrasal composition. And in any case, a lexical approach can't explain many cases of phrasal transfer, as in the example The south side ofCambridge voted Conservative, given by Briscoe and Copestake in this volume, where the content of a subject

NP

refers to a

group of people, but where there is no obvious way to derive this interpretation lexically. There's a third difficulty with this approach that has a greater methodo logical interest for us: note that not every modifier that could apply to a chicken from which meat has been derived can apply happily to the mass term chicken :

3 3· They served meat from corn-fed (Arkansas, happy, beheaded) chickens. 34· They serve corn-fed (Arkansas, ??happy, ??beheaded) chicken. The restriction in these examples is roughly that the property of the animals contributed by the adjective has to be one with consequences for the cons titution or quality of the meat derived from them-the meat from chickens may vary according to what they were fed, but not how they were slaughtered. This makes sense if we take the adjectives here to be transferred predicates sensitive to the requirements of noteworthiness that I mentioned earlier, but it

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

where chicken is treated as a mass noun denoting chicken meat; they suggest that we can take corn-fed here as applying to an ORIGIN feature char appears in the derived qualia structure of the mass term chicken (or in the qualia structure

Geoffrey Nunberg

121

i s inexplicable if they have their literal meanings and apply to a hidden feature in the noun: why should some adjectives be better than others in combination with the mass noun chicken in ( 3 3 ), when all apply equally well to the count noun

chickens

in (34)?21 The examples in ( 3 S) and ( 36) make the same point:

3 S·

The people in the huge (overflowing, domed, old) stadium rose to give the

36.

team a standing ovation. The huge (overflowing, ??domed, ??old) stadium rose to give the team a standing ovation.

That is, we assume that the group of people who rose to give an ovation acquires a noteworthy property from the dimensions of the structure in which they were housed, but not from its physical constitution or age. But again, this is

huge and ovetjlowing have transferred senses in ( 36), as opposed

to applying literally to a hidden feature (say, 'structure') in the headword

stadium ,

for in the latter case we are at a loss to explain why

domed

is not

acceptable here as well. We have to assume, then, that transfer is a phrasal process that works to a large degree in concert with the process of semantic composition. One way of dealing with this would be to permit transfer to take place independently on any simple or complex predicate or relational term, and then filter the output via constraints charged with maintaining consistency and some measure of compositionality. This approach has the advantage of providing for some violations of strict compositionaliry. There is in fact some evidence that this is required. For example, consider sentences where a metaphorically used subject and verb combine with a nonmetaphorical object, either idiomatically, as in

(37), or nonidiomatically, as in ( 3 8):22 37· 38.

The bottom dropped out of the bonds market. The Austrians managed to elude Napoleon for a week, but the eagle finally swooped down on them at a small town near the border.

If transfer were constrained to adhere to strict compositionaliry, we would have to say that the phrases

the bottom dropped out of and the eagle swooped down

are

constituents in these examples. But a proponent of the existence of VPs could rightfully argue that this is slim evidence for abandoning the VP hypothesis; better to assume that the metaphorical transfers in these examples are not constrained to apply in a strictly compositional way. In any case, at this time it is largely an open question as to how figuration and semantic composition interact.

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

explicable only if

122 Transfers of Meaning

7

S Y N T A C T I C C O N S E Q UE N CE S : ' S O R T A L C R O S S I N G S'

With this general understanding of the operation of predicate transfer, we can tum to some of its syntactic implications. Consider sentences like (39), suggested by Jackendoff, and (4o):

39·

Ringo squeezed himself into a narrow space.

40. Yeats did not like to hear himself read in an English accent. (Contrast: I am often read in an English accent.) On the standard way of thinking about these phenomena, the reflexives in their antecedents. In

(39), that is, the subject R ingo

refers to a person, whereas

the reflexive appears to refer to his car. Let me describe sentences like these as involving apparent 'sortal crossings'. This difficulty has been known for some time, and people have suggested a number of solutions to it. In the heyday of imperial syntax, a sentence like (39) would have been handled by a syntactic operation like 'car deletion' on an underlying structure like 'Ringoi squeezed Ringoi's car into a narrow space'. And other people have suggested various semantic and pragmatic approaches.23 But all of these approaches have the same failing, which is that they presume that it is the meaning of the reflexive that has been transferred, rather than the

meaning of the verb. That is, they all try in one way or another to get himself to refer to a car. Whereas what I am suggesting here is that himself in a sentence

like

(39)

actually refers to Ringo, so that the reflexive and its antecedent are

coreferential in a strict sense. What is transferred is the sense of the expression 'squeeze x into a tight space'. That is, in virtue of having performed a certain operation on his car, Ringo has also performed a corresponding (and note worthy) operation on himself, and this is what permits the predicate (or more accurately, relation) transfer.24 And by the same token, we assume that people who read Yeats' poetry aloud are doing something to the poet as well. But here again, the transfer is only available when the derived property or relation provides some useful or noteworthy information about its bearer, in accord with modem assumptions about authorship. So we wouldn't ordinarily say: 4 1 . ??Yeats wrote a lot of himself in sprung rhythm. When Yeats writes a poem in a particular metrical pattern, that is, he is not performing an operation on himself in the way he is when he reads one of his poems aloud to an audience. All of this leads to a strong hypothesis: natural language permits no sortal I_

crossings in any of the rules or constructions that ordinarily impose conditions

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

these examples present a problem, since they seem not to be coreferential with

Geoffrey Nunberg 1 23

of identity-not just with reflexives, but with pronouns, relative constructions, and so forth. A related hypothesis would stipulate that there need be no relaxing of the coordinate structure constraint to deal with sentences like:

42.

Roth is Jewish and widely read.

Jewish

Instead, we will interpret both

and

widely-read

as predicates that

contribute properties of persons.'s The object in all this is to honor the intuitions that originally motivated semanticists to appeal to zeugma as a test for determining whether a word has one or more senses. Predicate transfer allows us to treat syntactic identity as sortally chaste, by imputing all the promiscuity to the predicates involved. This is what the syntax has been trying to tell us all along, I think, but we weren't

8

T WO METHO DO LO G I C A L I S SUE S

Predicate transfer raises two methodological questions for semantic analysis. First, how do we determine when an expression is being used in a transferred sense? As we saw in connection with the example 'I am parked out back', this is by no means always easy to decide. In the general case, we have a sentence of the form 'a is G', where a is a word that ordinarily denotes something of sort A (e.g. people), and where

G

is a predicate that ordinarily applies to things of some

disjoint sort of things B (e.g. cars). We assume that either a or

G

must have a

transferred meaning, but how do we tell which? Let's take a very simple example: Billy's shoes were tied . We can analyze this in either of two ways: either

shoes

has a transferred reading where it refers to

shoelaces, or tied has a transferred reading where it denotes the property that shoes acquire when their laces have been tied. As we saw, there is no theoretical

reason for choosing between these analyses: predicate transfer can operate on

both functors and arguments. Nor is intuition a sure guide (or at least people seem to be of different opinions about the matter when you put the question to them baldly). But there is support for the second analysis, where it is

tied

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

paying close enough attention.

that

has the transferred reading, in the observation that we can conjoin tied with another expression that ordinarily applies to shoes, but not with an expression that ordinarily applies to laces:

46. 47· 48.

Billy's shoes were neatly tied but dirty. ??Billy's shoes were neatly tied but frayed. Billy tied his shoes, which were dirty (??frayed).

This seems conclusive, but why? The general assumption is that where a predicate is used in a transferred sense, there will be only limited possibilities of

-I

I

!

1 24 Transfers of Meaning

substirution-i.e. we can say the shoes were tied but not the shoes werefrayed -since other predicates will be subject to conditions of noteworthiness. {In this case, a shoe acquires an important property from its laces being tied-it is less likely to slip off, for example-but not from their being frayed.) Recall that this is the same form of argument we used above in arguing that corn-Jed must have a transferred reading in the phrase com-fed chicken , since we could not replace it with another word like beheaded or happy, even though all of these would have to be construed as applying to the animals from which the meat was derived. Let's see now how this procedure works in the slightly more complicated case like that in {49): 49· Yeats is still widely read.

so. . . . though he has been dead for more than so years.26 The restriction on sortal crossing rules out the possibility that the pronoun could refer to the poet while its antecedent referred to his works. On the second reading, yeats is a mass term that refers to the poet's works, and the predicate has a literal reading where it contributes a property of the works. O n this reading we would use the pronoun it in subsequent reference: 5 1 . . . . even though most of it is out of print.

and here the restriction on crossing forces the analysis of the subject as a mass term. But we don't have to rely only on the hypothesis about sortal crossing to make this point; we can also appeal independently to the requirements of noteworthiness that we made use of a moment ago. If the predicate has a transferred reading in the version of (49) that is continued with the pronoun he (i.e. where the subject denotes the poet himself), then we will expect that the possibilities of substituting other predicates in this version will be constrained by requirements of noteworthiness. Whereas the choice of predicates should be unconstrained in the version that is continued with the pronoun it (i.e. where the subject denotes the works): 52. Yeats is the second book from the right on the top shelf, if we still have it {??him).

The assumption is that the property of having one's books on such-and-such a shelf isn't particularly noteworthy in a poet, the way that, for instance, having one's books widely read or on the bestseller list might be (recall the painter's

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

Like Billy's shoes are tied, this example admits of two analyses, but in this case the sentence is in fact ambiguous. O n one reading, yeats refers to the poet himself, and the predicate has a transferred reading in which it contributes a property of poets. On this reading we would make subsequent references to the subject using the animate pronoun he, as in (so):

Geoffrey Nunberg 1 2 5

utterance of 'I'm i n the second crate o n the right'). But this restriction is explained only if we assume that the predicate has a transferred reading in the version of this sentence that is continued with the animate pronoun. This takes me to the second methodological difficulty that predicate transfer raises. Up to here we have been assuming that predicate transfer is essentially a unidirectional process. Shoes acquire noteworthy properties from the proper ties of their laces, meats from the properties of the animals they are derived from, and drivers from the properties of their cars, bur not, in general, vice versa. So there are a number of modifiers that apply literally to the count noun chickens on the hoof that can also apply to the mass term chicken ( corn-fed , range,

Arkansas), but none that applies literally to meat that can also be applied to the count noun (??sauteed chickens). This is one reason why it is usually easy to

the properties of the bird, bur not vice versa. In other cases, though, distinct categories may be interdefined in a way that makes extensive bi-directional property transfer possible. One example is the class of words that includes

newspaper , magazine , directory, travel guide , and so

on-basically any individual type of publication that is prepared or published by a single dedicated organization.27 These words can refer to a token of the publication, to an edition of the publication, or to the publishing organization itself And predicates that contribute properties of one of these sorts can frequently be applied with the transferred reading to an entity of some other sort, as in:

5 3·

The newspaper Mary works for was featured in a Madonna video.

Example

( 5 3)

involves no sortal crossing. Rather, the subject denotes an

organization, and the predicate in the main clause has a transferred reading where it contributes a property that the organization acquires from the fact that one of its copies was featured in a video. Of course, not every property of a newspaper token imparts a noteworthy property to the organization that publishes it; alongside of Pustejovsky:

54·

(5 3),

it would be odd to say

(54),

??The newspaper Mary works for fell off the table. (where

suggested by

newspaper

=

'newspaper publisher') But note that

(54) does have a possible reading if we take the subject to denote,

not the company that Mary works for, bur a copy of the newspaper it produces. Here, that is, we have predicate transfer operating on the content of the relative clause, rather than of the main clause, and going in the opposition direction, from the properties of publishers to the properties of their products. (Note that

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

determine which use of a given expression is basic and which are derived-e.g. that it is the mass use of chicken that is created by predicate transfer. All of this reflects a basic definitional dependency: chicken meat is defined by reference to

126

Transfers of Meaning

S S · Dense Metonymy: Given several disjoint sorts of things A, B, . . . and several classes of predicates F, G . . . such that members ofF literally apply only to things of sort A, members of G literally apply only to things of sort B, and so on, a word W is densely metonymous if£ a. W has distinct uses to refer to things of sorts A, B . . . and b. When W is applied to something of sort A, it often happens that predicates belonging to G can be applied to W under transferred readings, and when W is applied to something of sort B it often happens that predicates belonging to F can be applied to W under transferred readings, and so on. One problem with truly dense metonymies is that we may not be able to assign one or the other use a prior place in the lexicon, or to say in which direction the transfer operates. In this sense a word like newspaper is unlike an item like yeats or chicken , where the direction of transfer is clear. This creates some difficulties for lexical description, though I suspect that these are in large measure the artefacts of our theoretical approaches-;-the need to say which uses of a word are 'basic' and which are 'derived'; the desire to distinguish sharply

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

here the noteworthiness of the transferred predicate is reckoned by its utility as a means of identifying the newspaper copy, rather than by its intrinsic interest. The point may be clearer ifwe change Pustejovsky's example to something like, 'She handed me an envelope containing all the newspapers she had worked for', where the object NP clearly denotes a collection of copies identified in terms of the properties of the companies that produce them.) A word like newspaper we can think of as densely metonymous . Its various denotations are interdefined: equivalence classes of copies and editions are individuated in large part by reference to the distinguishing properties of the organizations that produce them, and vice versa. ('Which newspaper did Mary burn?' we ask, and the answer might well be, 'the one John works for'; Which newspaper does John work for?' we ask, and the answer might well be, The one Mary burned'). And the same correspondences that license the multiple uses of the word newspaper license widespread predicate transfer from the properties of one of its denotations to another: publishers acquire a number of noteworthy properties from copies of their publications or from editions of their publications, and each of these acquires many noteworthy properties from the publishers, and so on. As a result, the phrase a newspaper that supported the regime is trebly ambiguous; depending on the context it can refer to a company, an edition, or a copy. And likewise for the phrases a newspaper that the demonstrator burned, the newspaper that hit the streets yesterday morning, and so on. We can think of dense metonymy schematically as follows:

Geoffrey Nunberg

127

between lexical representation and knowledge representation in general. Informally, we might think of newspaper as listed in the lexicon in association with a constellation or ring of interrelated concepts. The central problem, it seems to me, is notjust how to set this up formally, but also how to structure the knowledge representations for the relevant entities-newspaper copies, editions, publishers, etc.-so that we can predict that we will get this pattern of transfer with newspaper and dictionary but not, say, with novel or user's manual. But I won't have any substantive proposals to offer here.2x 9

METAP H O R

56. ??The second line of the proof begins with the horseshoe that Deb pinned over her door. But we shouldn't make too much of these differences. Both types of figura tion are licensed by a repertory of conceptual schemas of greater or lesser generality. And as I've suggested here, both types are available for exploitation by linguistic processes like predicate transfer, and are subject to the same kinds of functional constraints.

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

I want to conclude with a word about metaphors. Earlier on I said that the difference between metaphors and metonymies is that the former were predicate transfers that were licensed by a direct correspondence between properties.29 Take the logicians' practice of referring to the symbol written as ·�· as horseshoe . We can speak here of a relation of resemblance between the set of horseshoes and horseshoe-shaped-symbols, or between prototypical horse shoes and prototypical horsehoe-shaped-symbols, but obviously there is no existential relation between any particular horseshoe and any horseshoe symbol, the way there is, say, between a particular newspaper copy and a newspaper publisher. As a result the particular things in the range of the transfer function don't acquire properties from anything that happens to the things in its domain-nothing that happens to any real horseshoe has any consequences (noteworthy or otherwise) for any particular horseshoe-shaped symbol. So from a metaphorical use of a predicate there is no inference to the truth of any utterance involving the literal use of the predicate. This is the difference between an utterance like 'I'm parked out back' and an utterance like 'I'm cruising down the fast lane of life'; with the latter you wouldn't ask the speaker, 'What are you driving?' (at least not unless drive itself is used metaphorically). Hence with metaphor we don't see the same kinds of apparent sortal crossings that we saw with metonymies: a logical symbol doesn't acquire properties from anything that happens to an instance of the equine footwear that it resembles:

1 28 Transfers of Meaning

Acknowledgements Earlier versions of this paper were presented at the Conference on Lexical Universals in Dagstuhl in April 1 993, at the Annual Meeting of the Association for Computational Linguis tics in Columbus, Ohio in June 1 993, at the University of S trasbourg and at the CREA in Paris in December of 1 99 3 · For comments on this version I am grateful to Sue Atkins, Ted Briscoe, Anne Copesrake, John Lamping, Fran�ois Recanati, Ivan Sag, Annie Zaenen, and the rwo anonymous reviewers for this journal. Received: 1 8.0 1 .94

Addressfor correspondence

Revised version received: 2 5 .08.94

GEOFFREY NUNBERG

Xerox Palo Alto Research Center 3333 Coyote Hill Road

NOTES 1 When I say that the conceptual relations that underlie metaphor and the rest are

of agreement phenomena, for which they

'nonlinguistic', I don't mean that it isn't

Bur as we have seen, the predicate transfer

provide a mass of independent evidence.

the business of linguists to study them,

analysis, and hence their analysis of the

b'Ut as independent principles that we

agreement here, is also independently

invoke to explain linguistic regularities

justified. 4 Of course the subject of '! am parked our

the form of explanation goes, language is as it is because we perceive the world in

back' is in fact an indexical, and we might

such-and-such a way. Whereas what I'm

think that here a deferred use to refer to

interested in here is the specifically

the car should be possible. Bur as I show

linguistic mechanisms that mediate these

elsewhere ( 1 993), the deferred use of I are

usages.

subject ro a number of constraints, which

2 This

was

the

analysis

proposed

in

among other things require that like

I

cannot re fer

Nunberg ( 1 979) and in a number of

other personal pronouns,

articles that drew on this (see e.g. Sag

in a d e ferred way to inanimates. In any

1 9!1 1 ; Schaefer & Clark 1 992). The obser

event, for the present purposes an NP like

vation-and its misanalysis-appears to

the man with the cigar could have been used

have been independently discovered in

in place of I to make the same point.

Smith ( 1 9ll9). Langacker ( 1 984) discusses the related example I'm in the phone book .

Similarly for an utterance like 'This key is

As I understand his analysis, it is more in the spirit of the one I'm suggesting here

cases can apply only to the referent of the

for

I'm parked out back .

Note also that the verb in a sentence like I

am parked out back

is first-person, nor

parked out back'. The descriptor in such expression. For discussion, see Nunberg ( 1 993)· 6 The 'salience' of a function depends on a complex set of factors. Among other

third-person as we would expect if the

things, the properties in the domain of

subject referred to a car. Sag & Pollard

the function (here, car-locations) have to

{ 1 994) argue that an analysis of this

be discriminable, and the relation itsel f

sentence in terms of predicate rransfer is necessary to preserve a semantic account

has to be manifestly familiar to partici pants. These conditions are schemarized

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

Palo Alto, CA G4 304 USA

Geoffrey Nunberg ar some lengrh in Nunberg ( 1 979), rhough

chis view see Sag (198 1 ), Kleiber (1 984). See also Stallard ( 1 99 3 ), who is one of the

in char paper these transfer functions were defined over domains of individuals

few people to have noticed chat rhese

rather than properties, and no distinction

meconymic usages can occur in boch

was made berween deferred indexical

argument and predicate positions. He

reference and predicate transfer. Wich appropriate adjustments, however, rhe

distinguishes them, however, as different

presentation of rhe conditions on these

'referential',

transfers is still valid.

interpretative

7 Of course

rhe noreworchiness

rypes, which he calls 'predicative' and

of a

the conversational interests, and you could fiddle with rhe unerance or rhe

rhese don't work in either rhe plural or

painters in her gallery, for example, mighr

u.

more noreworchy chan ochers, a dif ference char will be imporcant when we come co calk abouc rhe lexicalizarion of predicate transfer. Jackendoff analyzes rhese as involving transfer of rhe sense of Ringo , rarher than of the relational expressions, buc we will see below char rhere are a number of reasons for rejecting chis approach. He assumes, moreover, char we only see this kind of transfer in cases where rhe owner is acrually driving rhe car. Buc thar assumption has no independent motiva tion, and in any case it wouldn't explain an example like 'I am parked ouc back' or 'every rime I park ar rhe long-rerm lot, I wind up gerring broken into'.

9

( 1 992).

Several people have observed to me thar they can in facr imagine someone saying a related sentence like (i): i. I am rhe ham sandwich and I'd like ir right now. I don't find this an odd rhing co say, hue I don't rhink this means that the referent of rhe NP

the ham sandwich

in (i) is itself the

antecedent of the subsequent pronoun. Rather, we should think of this token of it as a pronoun of laziness, analogous to the anaphors in (ii)-(iv): u.

111.

They enjoy eating rabbit, when there are any to be found. I don't speak Italian, bur I'd love to go there.

iv. You'd better put on some mosquico repellent, just in case there are any around.

Though death need nor pur a stop co

In each of these sentences, a pronoun is

As one of the

used to refer co an entity that is seman

transfer in rhe general case.

10

*La rripe esr assise a Ia table 20. 'The tripe is seared at table 20.'

12 For discussion of this case, see Kleiber

13

properties are 'usually' or 'ordinarily'

*Les tripes som assises a Ia cable 20. 'The tripes are seared ar rable 20. '

reviewers of this article noted, one could

tically or materially connected co the

say, 'Picasso had to wait until after he died

referent of an expression in a previous

co get inco rhe Louvre'. Buc of course the

clause-as rabbits co rabbit meat, Italy to

transferred predicate here applies co the

Italian, mosquitoes to mosquito repellent.

personage, who survives the person.

This phenomenon is well known, as are

See e.g. Nunberg

( 1 979).

For critiques of

the constraints on this sore of usage. And

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

t.

say:

The mosr we can say, chen, is chat certain

laner

rhe singular:

chese examples acceptable. A painter who

basement.

distinct

kind of reference transfer.

feels she is being slighted in favor of ocher

while I'm relegated co a crate in the

with

mechanisms; rhe

11 In French, by contrast, sentences like

context in such a way as to make most of

i. Those daubers gee one-person shows

associated

involves whar I am describing here as a

property depends among ocher rhings on

8

1 29

1 30 Transfers of Meaning (i) permits the same kind of analysis: when we use the ham sandwich to identify a person who has ordered a ham sand wich, we introduce a ham sandwich into the discourse context that is available for pronominal reference. In support of this analysis, note that unlike genuinely anaphoric pronouns, these pronouns of laziness tend to be less acceptable when they precede their 'antecedents':

I 4 Another, more theory-laden way of purring this is to say that the information you need to establish the relevant relation is drawn from the knowledge representa tions associated with this particular kind of situation or activity, rather than the representations associated with the denotations of the expressions them selves. That is, you look up 'ham sand wich' and you find nothing about customers; that information is available under 'restaurant'. I s We can contrast these uses of black and white with the practice of identifying a poker player according to the highest or lowest card that he or she has showing on the table, as in 'King bets'. Why would we be surprised to find a dictionary listing this use of king ? Partly because the properties that figure in identification are very evanescent-the player identified as 'king' on one deal may be identified as ace on the next one-and partly because in any case the property of having a king up in poker has fewer general consequences than the property of playing black in chess. I6 This view would be consistent with the approach to lexical qualia structures developed by Pustejovsky and his co workers (see e.g. Pustejovsky I 99 I , 1 993), '1 :::

I

I

I

i

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

v. If ever you get there, you'll find Italy a lovely country. vi. ??If ever you get there, you'll find Italian surprisingly easy to learn. vii. By the rime it got to the table, the roast turkey was stone cold (??boiling mad).

as well as with various approaches to frame semantics and prototype semantics. 1 7 The only possible exception to this involves the use of the names ofgrapes for the wines derived from them: a glass of chardonnay, pinot noir, etc. But it could be argued that oenologically unsophisticated English speakers do not actually analyze these names as derived from the names of grapes, or what is more plausible, at least with regard to Northern Californian dialects, that the exception is tolerated because it is appropriately unnativized. I 8 By contrast, French does not allow this usage for wines; e.g. •un ( une) pomme for 'apple wine'. By the same token, one can use the expressions une menthe to refer to a syrup made from mint, but one cannot use une (or une) pomme to refer to a juice made from apples. This last example shows that these are not simple cases of ellipsis of the head noun, which in the case of syrups (un syrop ) would be masculine. 19 For examples of some of these restric tions, see e.g. Ostler & Atkins ( 1 99 1 ). 20 It is important to distinguish predicate transfer from type-shifting. In the standard cases of type-raising and type lowering that were introduced to facili tate the formal treatment of conjunction and predication, for example, the content of the predicate is unchanged. The pre dicate genius has the same meaning in the sentences I met a genius and We consider her a genius -what is changed is rather the role this predicate plays in the inter pretation. And at the same rime, the shift in meaning of parked out back from a predicate that applies to cars to one that applies to people doesn't involve any change of its type, at least in the Montegovian sense; parked-out-back isjust a different predicate. But the two processes can sometimes coincide, for example in cases where a verb that is ordinarily sub categorized as taking one kind of comple ment appears with a complement of another type. as in begin the book, enjuy the

Geoffrey Nunberg I 3 I

2I

23

24

25

26 c£ ??1 /ikeyeat.s, but most ofhim is out ofprint, which also shows that this could not be a 'mass' animate pronoun that refers to the works. 27 The crucial point here is that we recog nize a unique individual publication or work in the product of the organization, whether it is issued serially, sporadically, or in an isolated instance. The rip-off is usually that the 'publication' has a single ride ( The New York Times , Webster's Third). It is a contingent matter of fact which publication types support this usage: we say 'He works for a newspaper (magazine, travel guide, dictionary, etc.)' bur not '??He works for a book (novel, atlas).' I don't know enough about TheJoy ofCooking to say whether one can say 'He works for a cookbook'. 28 It is a little sobering to realize that the properties relevant to predicting these particular alternations have never figured in the knowledge representations people have offered for items in this domain, considering how thoroughly notions like 'books', 'novel', and the rest have been worked over in the literature. 29 You sometimes hear it said that meta phors and metonymies are distinguished by their rhetorical functions or cultural roles. Metonyies merely identify things; metaphors explain them, so to speak. (Or as one nineteenth-century rhetorician put it, metaphors blaze new trails in the imagination, whereas metonymies merely tidy up old ones.) As a general claim this is wrong. There are plenry of metaphors that are chiefly idenrifica rional, as with the logicians' 'horseshoe'; there are plenty of metonymies that are charged with a rhetorical significance, as when a musician describes record company executives as 'the suits'. Of course, the impression of a rhetorical difference might be based on some as yet-uncalculated statistical generaliza tion, bur I suspect that it has more to do with a tacit privileging of romance over allegory.

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

22

paper, sneeze the napkin offthe table. In case like these, it can be argued that the rype shifring ofa verb or NP is accompanied by a predicate transfer on the content of the V' or N ' it contains. This latter operation is what Pustejovsky (1991, 1993) describes as coercion, but while it is attendant on rype-shifting in these particular cases, it isn't the same thing. Analogously, while the phrase a rare virus can have the reading 'disease caused by a rare microorganism', it is harder to get an equivalent reading for the phrase a tiny virus to mean 'disease caused by a tiny microorganism'. In the latter case the adjective would have to contribute the properry that a disease acquires in virtue of the size of the microorganism that causes it, bur this is in fact unlikely to be noteworthy, given folk-etiological assumptions. Wasow, Sag, & Nunberg ( I 9X2) and Nunberg, Sag, & Wasow ( I 994), among others, have provided a good deal of evid ence that dioms like that in ( 37) must be regarded as compositional. And in any event, no other analysis is possible for the novel metaphors in (38). I suggested one in (I 979), and Fauconnier has put together a more systematic approach in his work on mental spaces (1985). More recently Jackendoff (I 992) has proposed an extremely ingenious semantico-syntactic approach to these difficulties, which unfortunately comes a cropper on the same empirical difficulties as its predecessors. Technically this isn't 'predicate transfer', of course, since what is at stake here is the rwo-place relation 'squeeze into a tight space'. But it would take us far afield right now to try to extend the account of predicate transfer to relational terms. For the present I'll just assume that such an extension is possible, and I will keep using 'predicate transfer' in a loose way. c£ 'I had been reading the man [Chomsky) for years, but had never seen him in the flesh' (The New Republic).

I 32 Transfers of Meaning

RE F E R E N CES Lin guistics' 142, s-JZ. Clark, H. H. ( 1 993), 'Making sense of nonce sense', in Arenas of Language Use , Center Apresjan, J. ( 1 97 3), 'Regular polysemy',

for the Study of Language and Informa tion and University of Chicago Press, Chicago, 305 -�0.

Mental Spaces ,

Fauconnier, G. ( 1 98 5).

MIT

Cahiers de Lexicologie, 14, X s- 1 o 3. Kleiber, G. { 1 992), 'A propos de du Mozart ', i n Gertrud Greciano & G . Kleiber (eds), Systemes interactijs: melanges en /'honneur de Jean David, Centre d'Analyse Syntaxique, Langacker, R. ( I 984), 'Active zones', in C.

& M. Macaulay (eds), Papersfrom the Tenth Annual Meetinji, Berkeley Linxuis tics Society , Berkeley, I S7.

Brugman

Morgan, J. ( 1 978), 'Two types of convention in i ndirect speech acts', in P. Cole (ed.), g:

Pragmatics ,

Aca

demic Press, New York.

Lakoff, G. ( 1 98 7), Taking: A

Pro ceedings of the Thirteenth Annual Meeting of the Berkeley Linguistics Society, 1 S 5-2o6.

Nunberg, G., Sag, 1., & Wasow, T. ( 1 99�).

Language , 70,

3·

Nunberg, G. ( 1 979), The Non-uniqueness of semantic solutions: polysemy',

and Philosophy , J,

Computational Linxuistics,

1 7, �·

lexical selection', in J. Pustejovsky (ed.),

Semantics and the Lexicon ,

Kluwer, Dor

Sag, I. ( 1 9S 1 ), 'Formal semantics and extra

Radical Pragmatics, Academic Press, New York. Sag, I. & K. Pollard ( 1 994), Head-Driven Phrase Structure Grammar, University of Chicago linguistic context', in p. Cole (ed.),

Press, Chicago and CSLI Publications, Schaefer, E. F.

Linguistics

1.

Nunberg, G. { I 993), 'Indexicality and deixis',

Linguistics and Philosophy, 16, I -4 3 · & A . Zaenen ( I 992), 'Systematic

Nunberg, G.

&

Clark,

H. C.

( 1 992),

'Concealing one's meaning from over

hearers', in Arenas ofLanxuage Use, Univer

sity of chicago Press and Center for the Study of Language and Information, Chi cago. Smith,

Q. ( I 9S9), The multiple uses of

i ndexicals',

Study in Lexical Nerwork Theory'.

' Idioms',

con',

Stanford.

Universite de Metz, Metz.

&

for Computational Linguistics.

drecht.

10, 1 -3 2.

Kleiber, G. ( 1 9X�). 'Polysemie et reference',

Norvig, P.

Lexical Semantics and Knowledge Representations, Association

jovsky & S. Bergler (eds),

Synthese, 78,

1 67-9 1 .

Stallard, D. (I 993), Two kinds of metonymy',

Proceedings of the J ISt Annual Meetin)i of the Association for Computational Linxuistics ,

in

Association for Computational Linguis tics, Columbus, Ohio, pp. S7-9�. Wasow, T., Sag, 1.,

&

G. Nunberg ( 1 982),

Proceedinxs of Plenary Sessions, XIII International Congress
UCB/CSD 9 I /6 1 8), Computer Science

polysemy in lexicology and lexicography',

Division, University of California, Berke

in K. V. Tolonen

ley.

H amm Tommola, T. Salmi & J. Schopp (eds), Proceedings of Eurale."l." II, Tampere, Finland.

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

Natural Language and

Syntax and Semantics,

Atkins, B. ( 1 99 1 ), ' Predictable

Pustejovsky, J. ( 1 993), Type coercion and

Jackendoff, R. ( 1 992), 'Mme. Tussaud meets

Linguistic Theory ,

&

meaning shift: some linguistic properties of lexical implication rules', in J. Puste

Pustejovsky, J. ( 1 99 1 ), The generative lexi

Press, Cambridge, MA.

the binding theory',

Ostler, N.

Journal ofSemantics

12: I J J-I62

©Oxford Universiry Press I 995

Aspectual Coercion and Logical Polysemy J A M E S P U S T EJ O V S K Y Computer Science Department, Brandeis University P I E RRE TTE B O U I L L O N ISSCO, University ofGeneva

Abstract

1

P O LY S E M Y A N D S E M A N T I C E X P RE S S I VE NE S S

Recently, work in computational semantics and lexical semantics has made an interesting shift. Motivated by a concern for lexical organization and global coherence in the structure of the language lexicon, some researchers have moved towards more expressive semantic descriptions, as well as more powerful methods of composition. 1 There has been some concern expressed, however, as to the general applicability of type-changing operations such as coercion, as well as concerning the power of generative mechanisms operating in the lexicon and semantics. In this paper, we address these concerns directly, and show that, upon closer examination, these critiques are not substantiated by the linguistic data. Without a proper notion of constraints on coercion, however, there can indeed be overgeneration of interpretations in the semantics, and in fact, the notion of conditions on coercion has always been integral to the basic spirit of generative lexicons. In particular, we examine the

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

Recent work in computational semantics and lexical semantics has made an interesting shift. Motivated by a concern for lexical organization and global coherence in the structure of the lexicon, some researchers have moved towards more expressive semantic descriptions, as well as more powerful methods of composition. There has been some concern expressed, however, as to the general applicability of type-changing operations such as coercion, as well as concerning the power of generative mechanisms operating in the lexicon and semantics. In this paper, we address these concerns directly, and show that, upon closer examination, these critiques are not substantiated by the linguistic data. Without a proper notion of constraints on coercion, however, there can indeed be overgeneration of interpretations in the semantics, and in fact the notion of conditions on coercion has always been integral to the basic spirit of generative lexicons. In particular, we examine the consrraints on type coercion in complement constructions of aspectual predicates in English and French. What we discover is a natural explanation for the behavior of coercion that makes reference to different types of event selection while also addressing the polysemous nature of aspectual verbs.

134

Aspecrual Coercion and Logical Polysemy

Argument structure: Specification of number and type of logical arguments. 2. Event structure: Definition of the event type of a lexical item or phrase. Sorts include STATE, PROCESS, and TRANSITION. J. Qualia structure: Composed of FORMAL, CONSTITUTIVE, TELIC and AGENTIVE roles.2 4· Lexical inheritance structure: Identification of how a lexical structure is related to other structures in the type lattice. r.

A set of generative devices connects these four levels, providing for the compositional interpretation of words in context. The exact nature of these devices will determine the polymorphic expressiveness of the semantics in fairly definite ways. The best-studied illustration of this is the phenomenon of type coercion , but it is by no means the only one. Copestake & Briscoe ( 1 992) model the mechanisms of generative lexicon

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

constraints on type coercion in complement constructions of aspectual predicates in English and French. What we discover is a natural explanation for the behavior of coercion that makes reference to different types of event selection while also addressing the polysemous na.ture of aspectual verbs. We will assume some general familiarity with the framework of a generative lexicon (GL), as outlined in Pustejovsky ( 1 99 1 a, 1993), and Copestake ( 1 993). We feel it is important, however, to clarify the motivating principles and general methodology behind this work, since it is crucial to the analysis taken in this paper. The empirical study of the range and limits of type change and co composition operations in natural language is an essential part of research in formal semantics. The advantages accompanying generative mechanisms and the characterization oflanguages as polymorphic in well-defined ways allow us to overcome the explanatory inadequacies inherent in traditional approaches to lexical design and semantic projection, what Pustejovsky & Boguraev ( 1 993) call word sense enumeration approaches. In order to help characterize the generative power of natural languages in terms of semantic expressiveness, it is natural to think in terms of semantic systems with increasing functional power. Furthermore, a natural way of capturing this might be in terms of the type system which the grammar refers to for its interpretation. There are reasons for describing semantic systems as falling on a hierarchy of increasing expressive power and it seems clear at this point that the current enumerative techniques for lexical description are too impoverished adequately to describe the richness of semantic data, much less to explain either how word senses relate to one another or the creative use of words in novel contexts. A generative lexicon of the type we assume can be characterized as a system involving at least the following four levels of representations:

James Pustejovsky and Pierrette Bouillon

r35

theory as a type system for a lexical knowledge base. Pustejovsky & Boguraev ( I 993) extend this view into the compositional semantics by having the operations make direct reference to the types within the system. The qualia structure along with the other representational devices (event structure and argument structure) can be seen as providing the building blocks for possible object types. The typing information mentioned above comes together in the lexical representation for verbs as well. 2 L I N G U I S T I C E V I D E N CE F O R C OERC I O N

(I) a. John began to read the book. (vP[+ INF] ) b. John began reading the book. (vP[+ PRG]) c. John began the book. (NP) (2) a. Jean a commence a lire le livre. (vP[+INF]) b. Jean a commence le livre. (NP) This deep type is able to project to one of three possible surface forms in English, and two forms in French, depending on which coercion rule applies. There is, however, only one semantic type being selected for, and the clustering of the particular syntactic forms appearing as surface complement types in (I) and (2) are systematically projected by virtue of this semantic type. That is, a verb such as begin or commencer, selecting for an event, will paradigmatically allow for the expression of the grammatical forms shown above, assuming surface syntactic constraints are satisfied. For this reason, the structuring of this kind of linguistic knowledge, where this event type has syntactic expression as any one of the surface types in (I) or (2) is called a lexical conceptualparadigm ( lcp ) (c£ Pustejovsky & Anick, I 98 8). In this view, the NPs, the book in ( I c) and le livre in (2b), are coerced to the appropriate type required by its governing verb, in this case an event. What makes coercion possible in these cases is the availability of the selected type, given as part of the NP's qualia structure , indicating, for example, that the TELIC role for book is the event of reading, while the AGENTIVE role is an event of writing. The result of applying this coercion operator to an NP is effectively to

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

The phenomenon of multiple subcategorization has motivated much of the type changing literature. The approach taken in generative lexicon theory builds on the ideas developed by Partee & Rooth ( I 98 3) and Klein & Sag ( I 98 5 ), while attempting to derive the syntactic expression of a verb's complement from a deep semantic type assignment, together with syntactic constraints. For example, in the well-studied case of aspectual verb complementation in (I) and (2) below, the verbs begin and commencer carry a 'deep type' selecting for an event in complement position.

1 36

Aspecrual Coercion and Logical Polysemy

create an extension of the NP meaning, called a metonymic reconstruction . In the case of NP, the book , for example, the coercion operators provide two event interpretations, namely, reading the book and writing the book . This interpretation is produced by virtue of the type of the selected complement and the availability of such types in the qualia structure of the complement itsel£ It is important to point out that this is a semantic 'reconstruction' rather than a syntactic one.

3

PROBLEMS W I T H U N C O N ST R A I N E D C OE R C I O N

complement coercion. The first apparent counterexamples to the general application of type changing operations argue that commencer does not universally allow NP complements with a coerced interpretation (c£ Goddard & Jayez, 1 993). For example, the NPs in (3) below do not appear to allow the analogous coerced readings that the sentences in (I) and (2) allow. Although the qualia for highway and dictionary presumably make reference to the events of driving and referencing, respectively, these interpretations are not available for the sentences in (3) and (4). Thus, the system would appear to overgenerate interpretations when no constraints on the application of type coercton operations are imposed. NP

(3 ) a. b. (4) a. b.

*Mary began the highway. (driving on ) *John began the dictionary. (referencing) *Marie a commence l'autoroute. *Jean a commence Ia dictionnaire.

What these data show is that the acceptability of coercion with aspectual predicates such as commencer and begin is conditioned largely by the telicity of the event which is metonymically reconstructed in the complement position. Essentially, these verbs select for an event of the sort TRANSITION, ruling out the coerced complement interpretations of drive on the highway for (3b) and (4b), and consult the dictionary for (3c) and (4c), which are both PROCESS events.J

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

There are several phenomena discussed in the literature which would apparently suggest that type coercion is not a general interpretative strategy for compositional semantics. In this section, we review these apparent counter examples and discuss each briefly to show why they are nonproblematic. We will concentrate, however, on the selectional properties of aspectual verbs such as commencer and begin , in order to show very clearly that sense enumerative approaches, such as that taken by Godard & Jayez ( 1 993) are unable to capture linguistic and computational generalizations, with respect to how the lexicon contributes to the compositional semantics of natural language.

James Pustejovsky and Pierrette Bouillon

1 37

Observe that the sentences in (3) and (4) do have legitimate coerced readings with transitional 'create' interpretations, such as building a highway and compiling a dictionary. Constraints due to 'boundedness' of the predicate (sentences (sa) vs. (sb) and (6a) vs. (6b)) are also consistent with conditions on coercion. Observe that mass noun and bare plural NPs in complement position are not acceptable as possible coercions. (s) a. *John began the cheese (eating) I the book (reading). b. John began cheese (eating) I books (reading). (6) a. Jean a commence le fromage I le livre. b. *Jean a commence du fromage I des livres.

(7) a. *The acid is beginning the marble. (corroding)

b. *John is beginning the noise. (being annoyed by) (8) a. *L'acide commence le marbre. b. *Jean commence le bruit.

In the discussion below, we show that the sentences in ( 7) and (8) are actually raising constructions rather than control structures, and that such constructions do not allow coercion at all. Experiencer predicates. A second argument against coercion might appear to come from the selectional properties of experiencer predicates. Consider the sentences in (9) below. (9) a. Books bore me. b. The movie frightened Mary. c. John's face scared me. d. Listening to Mary bores Alice. While it seems straightforward to admit that verbs like begin select an activity of some sort in all their subcategorization forms, what common selectional

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

In these sentences, the homomorphic relation between the NP type in (sb) and (sa) (mass vs. count) and the event type gives rise to process and transition interpretations of event structures, respectively (corresponding roughly to the amorphous and bounded readings within Krifka's 1 992 analysis). That is, the mass NPs in (sa) and (6a) (cheese , books , du fromage , and des livres ), when metonyrnically extended with their qualia, emerge as PROCESS or amorphous event types (i.e. 'eating cheese' and 'reading/writing books'). The count NPs in (sb) and (6b), on the other hand, emerge as TRANSITIONS or bounded events (i.e. 'eating the cheese' and 'reading/writing the book'). Finally, there would appear to be constraints on coercion which suggest that agentivity may play an important role in licensing the operation. For example, as pointed out by Godard & Jayez (1993), sentences such as (7) and (8) are ill formed:

- --------

1 38

- -- �-

-------

Aspecrual Coercion and Logical Polysemy

( 10) D'Vx'Vy'V e [anger(e, y, x )] - 3 e13 e23Exp [Exp (e1, x, y ) 1\ -. angry (e1, x) 1\ angry (e2, x) 1\ ....., e2 < et] This states that a verb such as anger involves someone directly experiencing something, and as a result becomes angry. What is interesting about examples such as (9a) and (9b) is that the semantics of the NP in surface subject position contributes information to the interpretation of what kind of experiencing event is involved. That is, the qualia structure projected by the NP books contributes to the particular manner in which I became bored in (9a), namely the NP's TELIC role of reading. Similarly, our knowledge of movies as something that we watch and experience in a particular manner is encoded in the TELIC role of movie in sentence (9b). The event projected from the noun movie , viz. watch , in tum satisfies the selectional requirements of the verb anger on its subject. Adjectival selection. A third argument against coercive operations involves examples such as a long novel and a bright bulb . If adjectives such as long in this case are analyzed as event predicates, as suggested in Pustejovsky & Anick ( I 988), where long modifies the activity of reading a novel, then there would appear to be a problem with selection in sentences such as (9), where the verb acheter (buy) selects for an individual while long selects for an event. The question is, how can the head of the NP possibly denote both an event and an individual, since such sentences are perfectly natural? ( I I ) Jean a achete un long roman. John bought a long novel.'

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

property would relate the different subject types in (9)? The answer is in fact very similar to that for verbs such as begin . We can view these sentences as involving a metonymic reconstruction of the subject to an event and, in particular, to an experiencing event between the surface object and the surface subject. That is, in (9), it is (my reading) books which bores me, (Mary's watching) the movie which frightened her, (my seeing) John's foce which scared me, and (Alice's ) listening to Mary which bores her.4 Thus, it seems that the linguisric evidence supports an underlying semantic type of an event as the subject, which would directly explain what the connection between the subject and object of the experiencing relation is. In Pustejovsky ( I 99 I b), it is suggested that the underlying semantics of psychological predicates such as bore, anger, and frighten is a causative structure where the surface subject is the logical object of an experiencing event. On this view, the lexical representation for the verb anger has something like the following form, where Exp (x, y) is a sortally restricted relation of experiencing (e.g. hearing, seeing, watching, etc.), and < is a strict partial order of temporal precedence.

James Pustejovsky and Pierrette Bouillon

1 39

What these adjectives demonstrate is not a violation or puzzle for coercion and selection; rather, they serve to illustrate the selectional properties of different classes of adjectives, as modifying different facets or qualia of the head. Modification by an adjective such as long, rapide (fast), or brillant {bright), can be seen as event predication, submodifying the appropriate quale of the head.s The adjectives in these cases modify a distinguished event predicate (i.e. the TELIC quale) associated with the head,

long book

read

for

book , and illuminate for bulb . Thus, a bright bulb is a

is interpreted as one taking a long time to read, while a

bulb which shines brightly when illuminated. These adjectives, on the other hand, should be contrasted with modifiers such as expensive in an expensive book and

opaque

in

an opaque bulb ,

both of which refer to the physical object rather

discussion, it should not be surprising that an entire NP can appear in an environment typed for an individual (e.g. as the object of buy), while its head is modified by an event predicate within the NP (e.g. as modified by

that the lexical semantics for the noun

( 1 2)

novel

is that given in ( 10) below:

.Ax (novel(x) 1\ CONST = narrative '(x) 1\ FORM = book '(x) 1\ TELIC = Ay, eT(read '(x) (y)(eT)] 1\ AGENTIVE = Ay , eT(write '(x)(y)(eT)]] a.

The analysis of adjectives such as

long

and

bright

overall type of the NP, as illustrated in ( I I ) below:

( I 2)

long). Assume

b.

does not change or shift the

.Ax (novel(x) . . . I\ [Telic = Aw , eT(read (x)(w)(eT) 1\ long(eT)]] . . .]

The reading in ( I I ) specifies that the event of reading is modified by the event predicate

long .

The resulting compositional structure is still the type of the

whole NP, and has no effect whatsoever on selection by an outside governor such as acheter (buy) as in ( I I ). That is, verbal selection and event modification are operating within different predicative domains. While the matrix verb

acheter

NP, an long can submodify the TELIC event associated with the object (namely, reading) while not affecting the overall type of the NP. selects for an individual such as a physical object, within the

event predicate such as

In what follows, we demonstrate how the apparent violations of the coercive behavior of aspectual predicates actually reveal a much deeper semantic distinction between two logically related senses of the verb in all the complement forms they select, and not just NP complement cases, which can be applied mutatis mutandis to commencer. We will show that this distinction is due to the event structure of the complements selected by the aspectual

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

than an activity or state associated with the object. These adjectives can be seen as modifying the FORMAL role of the qualia for these nouns. Given this

1 40

Aspectual Coercion and Logical Polysemy

predicates. We also demonstrate why commencer andfinir(finish) allow coercion while cesser (cease) and a"eter (stop) do not.

4 THE RELEVANCE O F EVENT S T R U C T U RE

An event tree structure can be defined in terms of (at least) one of three relations: ( I ) that of'exhaustive ordered part of', <"'' if there is strict precedence on the ordering of the events involved (such as the verb break , for example); (2) that of'exhaustive overlap part of', o"'' where the lexical item encodes an event containing rwo completely simultaneous subevents; and (3) chat of 'exhaustive ordered overlap',
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

In order to understand better the behavior of aspectual coercion, we review our assumptions regarding the structure of events as encoded by lexical items. We begin with the uncontroversial assumption that events can be subclassified into at least three sorts: processes , states , and transitions. Furthermore, we assume, following Pustejovsky ( I 99 I b), a subeventual structure to these event sorts as well. This has the advantage of allowing principles of predicate-argument binding to refer to subevents in the semantic representation, a move which has significant theoretical consequences (c£ Moens & Steedman I 98 8; Grimshaw I 990; Pustejovsky I 99 I b; and Moens 1 987). Evidence from unaccusativicy, aspeccual predicates, and the varied nature of causative constructions show, however, that even this notion of event structure does not fully explain the underlying behavior of semantic projection to argument positions in the syntax. We will adopt an event structure analysis employing the notion of headedness in order to account for the richer range of syntactic behavior. In Pustejovsky & Busa ( 1 994), the event constructions first introduced in Kamp (I 98o) and van Benthem (1983) are extended to a model called an 'extended event structure' in order to represent the relation between an event and its proper subevents. An extended event structure is a tuple, <E, �. <, o, !: , *>, where E is the set of events, � is a partial order ofpart-of, < is a strict partial order, o is overlap, !: is inclusion, and • designates the 'head' of an event, to be defined below. This model provides a formal interpretation for the event representations with structured subevents such as those in Pustejovsky ( 1 99 I b), shown in ( I 3) below.6

James Pustejovsky and Pierrette Bouillon

141

*( e;. ei) = thf [ ei, . . . e; . . .] Headedness is a property of all event sorts, but acts to distinguish the set of transitions, specifying what part of the matrix event is being focused by the lexical item in question. Given that transitions have a binary event structure, there are four possible head configurations, where we mark the head with an asterisk: (14) a. b. c. d.

build: e• T arrive: eT* give : e· T· break : ET

Intuitively, structure ( 1 4a) represents accomplishment verbs, where the focus is on the action bringing something about; (qb) represents achievement verbs, for which the persistence of the final state is the focus of interpretation; ( r 4c) illustrates events involving relational statements on each subevent; and (r 4d) is the representation which is crucial to the analysis of verbs as logically polysemous.7 In terms of event structure, polysemy occurs where the expression is lexically unspecified with respect to headedness, i.e. headless. A headless event is an underspecified representation which admits of two possible inter pretarions. More generally, a predicate should be as many ways ambiguous as there are potential heads in the associated event structure. Support for this view comes from data on unaccusatives and causatives in Italian and other languages. Unaccusatives such as the verb arrive are specified as lexically right-headed

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

and it is just such a relation that is involved with aspectual predicates such as begin and commencer , as we shall argue below. The other notion which is critical to understanding the selectional properties of aspectul predicates is that of event headedness . Event headedness, first introduced in Pustejovsky ( 1 988), provides a way of augmenting the event structure with some way of indicating a type of foregrounding and back grounding of the subevents. We can view an event structure as providing a configuration where events are not only ordered by temporal precedence, but also by relative prominence. One notion of prominence for an event we will entertain is that of a head , *. The conventional role of head in a syntactic representation is to indicate prominence and distinction. Rules of agreement, government, etc. militate in favor of marking structures in terms of heads of phrases. Within the interpretative domain of events when viewed in a structural or configurational manner, the possibility of referring to heads also becomes available. Informally, the head can be defined as the most prominent subevent in the event structure of a predicate, which contributes to the 'focus' of the interpretation. We can view * as a relation between events, *( e;, ej), where < eJ.: e1. -

1 42 Aspectual Coercion and Logical Polysemy

events and have no lexical causative counterpart. Verbs such as sink and break, however, appear in both unaccusative and causative constructions, due to the headless nature of the underlying event structure associated with the verbal semantics.

s

C OE R C I O N W I T H A S PE C T U A L VERBS

( I ) a. John began to read the book. (vP(+INF]) b. John began reading the book. (vP(+PaG]) c. John began the book. (NP) (2) a. Jean a Commence a lire le livre. (vP[+INF]) b. Jean a Commence le livre. (NP) The aspectual verbs finir and finish also allow multiple complements types, patterning in a similar fashion (c£ ( I s ) and ( I 6)): ( I S) a. He has finished reading the book. (vP(+PaG]) b. He has finished the book. (NP) ( I 6) a. II a fini de lire son livre. (vP[+INF]) b. II a fini le livre. (NP) Like commencer and begin , they also permit certain bounded NPs as com plements but disallow coercion with the same class of NPs prohibited by commencer and begin , as illustrated in ( I 7) and (I 8) below. ( I 7) a. *I finished the symphony. (listening) b. I finished the chocolate. (eating) c . *I finished chocolate. (eating)

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

In this section, we examine in more detail the behavior of the three French and English aspectual verbs commencer (begin) , finir (finish) , and arreter (stop) with respect to complement selection and coercion. In particular, we observe that commencer and finir and begin and finish permit NP coercion while arreter and stop do not. We see that both commencer and begin and arreter and stop are polysemous between both raising and control senses, and we show why this polysemy exists.8 Interestingly, however, while the English verb finish exhibits only a control reading, we observe that the French verb finir appears to be polysemous between control and raising senses, but in fact is strictly a control verb. We saw in Section 3 how commencer and begin allow both VP and NP complements, the latter which we analyzed as resulting from type coercion (c£ ( I ) and (2) repeated below):

James Pustejovsky and Pierrette Bouillon 1 4 3

( I 8) a. *]'ai fini la symphonie. b. J'ai fini le chocolat. c. *J'ai fini du chocolat. Now consider the verbs arreter and stop , which do not allow this type of coercion at all (c£ the ungrammaticality of (19b) and (2ob)).

Upon examination of the above data, the immediate question that comes to mind is this: why does coercion, if it indeed applies in these contexts, seem so idiosyncratic in its application with the verbs commencer, begin ,finir, and finish , and why does it not apply at all with the verbs arreter and stop ? In order to answer this question, we will tease apart the syntactic contexts in which coercion is applying by differentiating two senses of aspectual predicates. We will see that this distinction plays an important role in determining whether a predicate licenses coercion or not. We will argue that there are two different but logically related types of aspectual verbs, control and raising verbs, and that only the former allows coercion. The idea of analyzing aspectual verbs as essentially ambiguous is not new, but was already proposed by Perlmutter (I 970) for English and Lamiroy (I 987) for French. The traditional method for distinguishing between control and raising verbs involves a battery of diagnostics testing for selection, agentivity, and controllability (c£ Dowty I 979 and Zaenen I 993). Perhaps the best indicator of a raising predicate is that it imposes no selectional restrictions on its subject, as illustrated with the verb seem in (2 I). (2 I ) a. The lake seems to have frozen. b. A riot seems to have happened yesterday. c. This fact seems to have escaped Mary's attention. The subject in each sentence in (2 I ) is restricted by the embedded predicate in the VP selected for by the verb seem . A control predicate, on the other hand, imposes clear and obvious restrictions on the subject NP (c£ (22)): (22) a. Mary tried to leave the party. b. *A riot tried to happen yesterday.

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

(I 9) a. Madame a arrete de verser le the. Madame stopped steeping the tea. b. *Madame a arrete le the. *Madame stopped the tea. (2o) a. Jean a arrete de lire le livre. Jean stopped reading the book. b. *Jean a arrete le livre. *Jean stopped the book.

144

Aspecrual Coercion and Logical Polysemy

There are also syntactic constraints imposed by control predicates that are absent in raising constructions:9 (23) a. There seems to be a riot going on now. b. *There attempted to be a riot. Perlmutter (1970) mentions force -complement constructions as another clear indication of a control verb. Compare the sentences in (24a) to that in (24b). (24) a. Mary forced John to begin writing his thesis. b. *Mary forced it to begin raining yesterday.

sentence in (24b) illustrates that a 'raised' NP cannot satisfy the selectional con straints imposed by force . The same facts hold with the verb obliger (force) in French, as well, where (25a) shows a legitimate control relation with the embedded commencer-predicate, while the raising construction in (2 5 b) results in an ungrammatical structure. (25) a. Je t'oblige a commencer a lire le livre de Proust. I am forcing you to begin reading the book by Proust. b. *Je ]'oblige a commencer a pleuvoir. *I am forcing it to begin to rain. The well-formedness of object complement coercion with aspectual predicates such as commencer is conditioned by the event son of the qualia associated with the NP itsel£ Thus, only NPs having associated transition events will allow coercion and control. This is not to say, however, that begin selects only for transition events. There are, of course, perfecdy grammatical examples of process or state complements, as shown in (26) and (27) below: (26) a. L'acide commence a corroder le marbre. The acid is beginning to corrode the marble. b. Il commence a pleuvoir. It is beginning to rain. c. La neige a commence a tomber a minuit. The snow began to fall at midnight. d. La guerre commence a atteindre la Bosnie. The war is beginning to reach into Bosnia. e. Jean commence a perdre du sang. John is beginning to bleed. £ Jean commence a etre ennuye par le bruit. John is beginning to be annoyed by the noise.

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

The sentence in (24a) illustrates that the matrix object stands in a control relation to the embedded VP. Observe, however, that the ungrammatical

James Pustejovsky and Pierrette Bouillon

145

g. Jean commence a etre malade. John is beginning to be ill. The above examples illustrate the use of begin as a raising verb. The two senses of the verb begin conform to the distinction that Perlmutter originally made, namely, as either a raising or a control verb.•o As a raising verb, the event sort specified as the complement to commencer and begin may be any sort. As a control verb, it appears that the complement must be a TRANSITION. There do appear to be some counterexamples to this basic distinction. Sentences such as (27a) and (27b) appear to have a control component to their interpretation, even though the complement event-type in each case is atelic.

Notice, however, that although verbs such as lookfor in (27) and work on in (28) are indeed atelic, (28) a. Mary is working on a book. b. Mary is beginning to work on a book. their semantics incorporates an obvious 'telicity' in the qualia sense. That is, the intensional context involves mention of the goal state, or TELIC role of the activity. For example, the relation look-for{x, y) modally incorporates the relation have{x, y) ; similarly, in (28), work-on(x , y) modally incorporates the goal state of exist(y) . Hence, in some sense, the aspectual classification of these predicates as simple processes does not reflect this goal-oriented property. In order to explain the behavior of these verbs with respect to controllability in the sentences in (27a) and (28a), we refer to this class of predicates as intensionally telic , and suggest that this is why control readings are acceptable.' ' That is, it is due to this implicit goal state that these verbs pass the tests for controllability. Let us turn finally to sentences such as (27b). Is this a true counterexample to the claim that control with begin involves a transition event-type? We argue that it is not, and that this is actually an example of raising with a potentially agentive subject. For notice that raising does not preclude all agentive force on the predicate. Even canonical cases of raising, such as (29a) below, can be construed as involving a certain amount of potential agency on the part of the subject. Yet, we still would not want to claim that they are control contexts, as (29b) makes clear. (29) a. John began to feel ill from eating too much food. b. *Mary forced John to begin to feel ill from eating too much food.

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

(27) a. Jean commence a chercher du travail. John is beginning to look for a job. b. James a commence a travailler a Brandeis en 1 986. James began to work at Brandeis in 1 986.

146

Aspectual Coercion and Logical Polysemy

Interestingly, (3oa) also has only a marginal force-complement construction compared to the same construction with start : (3o) a. ?Mary forced James to begin working at Brandeis. b. Mary forced James to start working at Brandeis.

(3 I ) a. "'John began to bleed and Mary began, too. b. John began to read the book, and Mary began, too. What this discussion illustrates is that there are indeed two constructions at play here-control and raising-which are teased apart by certain diagnostics, namely VP-ellipsis and theforce -complement construction. In order to understand the significance of this distinction more clearly, let us review some of our theoretical assumptions. We assume a system of types similar to Montague's intensional type system, augmented with event types with sortal specifications. While the type (e, r) is conventionally interpreted as an unsaturated proposition (i.e. a propositional function ), for purposes of this discussion, we will treat the sentences below as denoting events of some sort, E"; unsaturated events will be analyzed in an analogous fashion as eventualfunctions , (e, E").U We will argue that the following typing assignments characterize the distinction between these two senses:13 (32) a. begin as a Control verb: ((e, eT), (e, eT)) b. begin as a Raising verb: (E", e T) The type in (32a) specifies that the verb selects for an unsaturated event of sort TRANsmoN in object position, and an individual in subject position. The resulting type is an event of sort TRANsmoN. The type in (32b), on the other hand, specifies that the verb selects for a saturated event. Something more needs to be said, however, to explain the sentences in (3 3). Only left-headed TRANSmONS, which make reference to a predicate opposition and the activity of an individual bringing about this change seem possible with the control sense of these verbs. Notice that right-headed TRANSmONS are only possible with the raising interpretation and not the control reading, and preferably ifthey shift to PROCESS readings:

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

What this suggests is that, given sufficiently strong reasons for accepting the analysis of control begin as selecting for transitions only, then such cases as (27b) can be legitimately classified as raising cases, as argued above. Finally, as pointed out in Jacobson ( I990), VP-ellipsis can be used as a diagnostic for determining whether a complement is part of a raising or control construction in English; namely, only control complements enter into the construction. Notice that in (3 I ), the only fully grammatical sentence involves an overt control interpretation of begin , (3 I b).

James Pustejovsky and Pierrette Bouillon

1 47

( 3 3 ) a. ??Je commence a arriver.

?I am beginning to arrive.

b.

Les invites commencent a arriver.

The guests are beginning to arrive. c. *Je commence a trouver Ia de de Ia maison. *I am begining to find the house key. d. Je commence a trouver des poux sur mon chien. I am beginning to find fleas on my dog.

(34) Defoult Causative Paradigm (DCP): VR , P, x, y [[R (e1 , x , y) 1\ -.P(e1 , y) 1\ P(e2, y) 1\ -. e2 <e1] > cause(e1 , e2)] We take this to be one paradigm for the semantics of causal relations as encoded in lexical items. In a lexical entry embodying this conceptual paradigm, all the arguments as well as the subevent types of each relation in the DCP will be coherently bound in the qualia to the AGENTIVE role, which denotes the cause, or to the FORMAL, which denotes the effect:15

( 3 5) A.yhle1 A.e2 3 P 3 R (a : FORMAL = (P(e2, y)] 1\ AGENTIVE = (R (e1 X , y)]] This says that a predicate a is a relation between two subevents and two ,

individuals such that some relation exists between x and y in the 'bringing

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

This shift is possible with degree-achievements (c£ Dowty I 979) or in the presence of a plural indefinite or mass noun in the subject or the object (c£ Krifka I 992; Moens I 987; Pustejovsky I 99 I ). These seem much worse in French than in English as raising constructions, but the significant observation is that the control reading is impossible with these verbs. These data would suggest that the control sense of begin selects for a complement that is a left-headed TRANsmoN, e·T, while the raising sense selects for a complement of any event sort. It is natural at this point to ask why aspectual verbs such as commencer and begin have both these two senses, and whether it is an accidental lexical ambiguity or a logical polysemy. We argue that these two senses are not arbitrary types but are logically related to one another in the same way that the different senses of unaccusative/causative verbs, such as break and sink are related. In Pustejovsky & Busa ( I 994) it is shown that verbs such as break and the Italian ajfondare (sink) are logically polysemous in predictable ways, and do not need to be assigned multiple lexical entries. For example, verbs exhibiting the causative/unaccusative alternation (c£ Levin 1993) are analyzed as cases of logical polysemy. That is, both intransitive and transitive forms ofverbs such as break are taken as underlyingly causative. The statement in ( 34) captures the underlying semantics of causation involved in those predicates that enter into the causative/inchoative paradigm:14

1 48 Aspecrual Coercion and Logical Polysemy

(36) A.xA.P.k2A.e1 3 R 3 P [commencer: ES = e1
If the initial event is headed, a control structure results. If, however, the final event is headed, a raising construction results. In the next section, we demonstrate how this underspecified semantic representation is responsible for both control and raising constructions for aspecrual pred{cates such as commencer, and the constraints on coercion in these constructions. 6 THE S E M A N T I C S O F ASPECTUAL C O E R C I O N Let us now examine in more detail the semantics of aspectual raising and control constructions. We assume that there are two rypings for commencer, as shown above in (28), and that they are logically related senses by virtue of the semantic representation given above in (36). The phenomenon of coercion described above is similar, in some respects, to subtyping polymorphism as encountered in examples such as (37) below.19 (37) a. Mary drives a Honda to work. b. Tom read the Tractatus on holiday. Assuming that the internal type selected by the verb drive in sentence (37a) is vehicle, as illustrated in the typing for the verb:

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

about' (the AGENTIVE role) of the resulting state of y (the FORMAL role), where this state, P(e2, y) did not hold before.16 This representation is underspecified, however, in that there is no event head in the structure. In fact, it is this semantic underspecification which gives rise to the polysemy exhibited by these predicates. By heading (or focusing) the initial event, associated with AGENTIVE, a causative template arises. By heading the final event, however, associated with FORMAL, an unaccusative structure arises (c£ Pustejovsky & Busa 1 994 for details). A similar analysis holds for verbs such as commencer, which exhibit a logical polysemy between control and raising senses. We will view commencer as the lexical version of an unaccusative marker, but for events rather than for entities. 1 7 Underlying the lexical representation for verbs that exhibit control and raising behavior is a deep or underlying causative. The alternation displayed above is licensed by the headless nature of the event structure representation of the predicate commencer. Whether it surfaces as a raising verb or a transitive control verb will be determined by which subevent is headed. In this view, then, the core lexical representation for commencer is given below, where the variables P and R are unspecified predicates, and there is no headed event.18

James Pustejovsky and Pierrerre Bouillon

1 49

TYPE{drive) - (vehicle, (person, t:l))

This says that, given an expression a of type a1 , which is a subtype of�. there is a coercion possible between a1 and az, which changes the type of a in this composition, from a1 to az (c£ Beierle et a/ . 1 992) We will illustrate the further application of this coercion operation below, in conjunction with metonymic reconstruction coercion. It should be pointed out that, although subtyping polymorphism and metonymic reconstruction coercion are similar, in that they enable the variable functionality of a lexical item to be expressed in a single form, they are formally quite different. Subtype coercion follows the inferences available in a singly typed lattice, while metonymic reconstruction requires reference to a multiply typed lattice or feature structure, making use of information available through the qualia. Given that evenr-headedness acts to foreground or 'focus' a single quale of the verbal semantic representation, let us first consider the effects of heading the final event from the lexical structure in (3 5), namely, that in the FORMAL role. This corresponds to the raising interpretation, where what is asserted is simply the initiation of an event, without reference to causal or control preconditions to the event. Consider once again the sentence in (26a), repeated below: .

(26) a. L'acide commence a corroder le rnarbre. The acid is beginning to corrode the marble. We will assume that raising is accomplished by function composition {FC), in the manner ofJacobson ( 1990). In particular, the raising verb commencer, of type (�. cT), imposes the type � on its complement. Assuming the VP to corrode the

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

then the selectional requirements can be satisfied just in case there exists a subtyping relation, Honda :s:; vehicle, formally relating the type of the actual object to the lexically specified type. Let us call G the typing judgements with respect to a grammar. Then, by convention, G 3 a : r represents a type assignment of r to the expression a . The typing relation between the subtype Honda and the type selected by the governing verb drive is respected by the coercion relation E>, giving e [Honda :s:; vehicle] : Honda .... vehicle .20 Similarly, in (36b), Tractatus :s:; book :s:; text defines a relation between the type selected by the verb read and the actual individual. This is an instance of the more general subtyping coercion operation, illustrated below:

1 so

Aspectual Coercion and Logical Polysemy

marble , for example, in (26a), is (e, eP), then, FC(begin , VP) = A.9' [begin (corrode(9' , the-marble)]. Following Pustejovsky (1994b) we can view the basic composition of the sentence in (26a) as type inference, where E9 represents type application as inference according to the typing judgements, G , in the grammar. The type inference tree for this construction is shown in (39). FC (commencer : c 0 -

,J: wrrodrr lr marbrr: e - £'1)

L'acide : e E9 commencrra corrodrr le marbrt : r- ,T (39) L 'acide commencer a corroder le marbre : eT

(4o)

a : a $ Q[a, r] : a - r Q[a, r](a ) : r

This says that, given an expression a of type a, there is a coercion possible between a and r, which changes the type of a in this composition, from a to r. We will illustrate the further application of this coercion operation below, as used in the commencer examples. In (4 1 a) we see how the aspectual verb commencer selects the complement VP, and how in (4 rb) an NP is coerced into an event interpretation. Both sentences involve left-headed event structures, resulting in a control interpretation for the verb commencer. (4 1) a. Marie a commence a lire le livre. (VP) Mary began to read the book. b. Marie a commence le livre. (NP) Mary began the book. The type inference tree for (4 r a) is given below: commencer : (e - cT) - (e - cT) G)

;, lire /e livre :

e- ,T

commrnma lirt le livre : e - cT (42) Marie : e $ Marie commencer a lire le livre : fT

For the derivation of(41 b), reference is made to the qualia structure of the noun book , as shown below, where ARGSTR refers to the argument structure of the nominal book, treated as a type of implicit relational noun (c£ Nunberg 1979 and Pustejovsky I 994b):

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

In Pustejovsky & Boguraev (1993) a general mechanism is defined which makes the appropriate type available for a coercion operation. The qualia can be seen as partial functions, returning the value of a particular quale for an NP. The combined set of qualia provide a set of type aliases for the expression containing them. One particular mechanism, type pumping, has been explored as a means to generating the alias set, and we will make use of this device below. Thus for example, the type available to an expression a with quale Q of type r, can be seen as the following type inference:

James Pustejovsky and Pierrene Bouillon

book ARGSTR =

(43) QUALIA =

[

J

1s1

information . y : phys obJ infonnation-phys obj-container-lcp FORM = hofd(y , x) TELIC = read (eT, W , X ) AGENT = write(eT, V, y) ARGI = x :

ARG2 =

_

[

_

]

c
eT) - (e - cT) $

le livre : e $ (.?yje. (e -

e1)j: e - (c - ;r) .

: _

(.?yje . (e - e T)l(e ) (e - ef) �_ _:__ .:..._ ...:.:.._ -e : .:...: · r_:. cee lee _-.:.... l r rv ,T C
(44) Marie : e €9 ------

_

Marie commencer le livre : t:T This states that the TELIC role of le livre , A.x A.eT[ read (er, x , le-livre)), is available as an alias for shifting the type of the NP. This metonymically reconstructed type, (e, e'r), is identical to that selected by the verb commencer in complement position. Mter the coercion operation, the derivation proceeds as before in (42). A more conventional semantic derivation associated with the type inference tree in (44) is given in (45 ) below.22

(45 ) a. Marie commence le livre. b. c. d. e. £

A.eT[commencer'(Qr(le livre))(Marie)(eT)) => A.eT[commencer'(A.x , e [lire(le livre)(x)(e)])(Marie)(eT)) => Marie{A.x, eT[commencer'(A.x, e [lire(le livre)(x)(e)](x •))(x•)(eT))} => Marie{A.x, eT[commencer'(A.e [lire(le livre)(x •)(e )])(x•)(eT))} => 3eT[commencer'(3e [lire (le livre)(Marie)(e )])(Marie)(el)]

The syntactic structure associated with this sentence can be illustrated in (46):

(46) Control (left-headed transition):

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

This representation illustrates how the qualia make reference to two events associated with the noun, reading and writing. As illustrated below in (44), coercion applies to the complement NP, where reconstruction with either the TELIC or AGENTIVE qualia will result in the appropriate type selected by the verb. We illustrate the derivation with the TELIC role selected.21

152

Aspecrual Coercion and Logical Polysemy

[NP ------[J �] [VP s

VP

lex Marie TYPE ind

� �

�

AUX

lex a TYPE aux

.

v

ex commence A.x k 3P 1

�

I

A.x A£, 3 e2 F

[

F

=

A*

=

]

L�NP J . J] L J

P(x)(e2) R (e2)(x)(e ) 1

=

l

P ' (le-livre)(x)(e2)

��;�

Let us now return to the discussion of raising constructions and coercion in such structures. We observed in sentence (6a) that type coercion is unacceptable with this predicate: (6) a. *L'acide commence le mabre. (corroding) *The acid is beginning the marble. There would appear to be a possible derivation involving type coercion in this sentence where we choose the raising sense of commencer, imposing the type I!' on the complement. But notice that coercion will be successful only if the appropriate type exists in the alias set of the complement. Metonymic reconstruction on the complement in (6a) returns an eventual function of type (e, e'f) rather than the type selected by the verb, 1!'. Since function composition is an operation at the level of the VP, there is no point in the derivation at which the appropriate type is available for the rule to apply, and the sentence is not semantically well-formed. As we saw above, this is not the case with control verbs. Having outlined the basic mechanism of coercion under constraints, we can explain now why examples like (47) are ungrammatical. (47) a. *Marie a commence l'autoroute. (driving on ) *Mary began the highway. b. *Jean a commence le dictionnaire. (consulting) *John began the dictionary. c. *Jean a commence le sommet. (reaching) "'John began the top of the mountain. d. *Jean a commence la symphonie. (listening to ) "'John began the symphony.

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

l

lex commence le livre

James Pustejovsky and Pierrene Bouillon 1 5 3

e. *Jean commence le livre. (destroying) *John began the book £ *Jean commence le desen de Gobi. (going through ) *John began the Gobi desert.

sonate ARGSTR -

QUALIA =

[

[

- e : event . ARG2 - x : abstract_obJ performance-lcp ARGI

J

music (x ) TELIC = perform (e, w , x ) , listen (e', z , e) AGENT = compose (e", y, x ) FORM -

]

The first thing to notice is that the lexical item directly denotes an event, as well as an individual sortally restricted to music . As pointed out by psychologists such as Miller (I 99 I ), social artefacts are very different from simple physical anefacts, in that their function is defined in a more complex manner. For example, in defining the TELIC role for an event object such as symphony, one cannot ignore the role of the listener (the experiencer). That is, music is performed for an audience. This must be reflected in the qualia structure as a conjunction of relational values, i.e. perform and listen . Recovering the event listen in the metonymic reconstruction due to coercion without also recovering perform is similar to binding a variable with a partial value; that is, listen is a

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

The coercion in (47a) and (47b) is impossible for reasons already discussed; namely, if these sentences are examples of raising constructions, then they are ruled out according to our previous discussion. If they are control senses, then the metonymic reconstruction on the NP in each case produces a type (PROCEss) that is incoherent with that selected by the verb, namely an eventual function with a TRANSITION. In (47e), the coercion is impossible as the qualia does not stipulate how you destroy an object, but rather its bringing about or what you do with it. Similar remarks hold for (47c) and (47f). One might argue that (47d) should be possible with a control interpretation (the event in the complement is controllable and bounded), and yet coercion is not possible. This example points to a somewhat different phenomenon, one involving the semantics of the nominal itsel£ The qualia representation of symphonie appears to make reference to both an event and an individual reading. It suggests that the semantic distinction between objects such as books and tapes on the one hand, and symphonies and sonatas on the other is responsible for the unavailability of such coerced readings. The qualia structure for event-objects (using Dowty's 1979 terminology) such as sonate and symphonie can be schematically given as follows:

1 54

Aspectual Coercion and Logical Polysemy

dependent event while perform is independent, being projectable through coercion by itsel£ We can see how, with a raising sense, as in the sentence La symphonie a commence, nouns such as symphonie and sonate are directly selected by the verb 2 since they are able to denote events; hence, no coercion is involved (c£ (49)): 3

(49)

'Raising' (right-headed transition): s

[:

]

Ia symphonie TYPE event

(I s) a. b. ( I 6) a. b. ( I 7) a. b. c. I 8. a. b. c.

He finished reading the book. (VP[+PRG]) He finished the book. (NP) II a fini de lire son livre. (VP[+INF]) II a fini le livre. (NP) *I finished the symphony. (listening) I finished the chocolate. (eating) * I finished chocolate. (eating) *]'a fini la symphonie. J'ai fini le chocolat. *J'ai fini du chocolat.

The ungrammaticality of (I 7a) and ( I Sa, c) will follow from the same analysis given for begin above. Notice, however, that finish differs from commencer and begin in that it is not logically polysemous, having only a control sense. That is, the raising examples in (so) are ungrammatical. (so) a. *It has finished raining. b. *The sun has finished shining in my eyes. c. *The acid finished corroding the marble. This would suggest thatfinish is not lexically underspecified with respect to headedness, as is begin , but is already specified with a head. It is this lexical specification which gives rise to the control reading only. Interpreting Dowry's (I 979) analysis o(finish within our framework, we can analyze this verb as making reference to two events: ( I ) that subevent which brings about,

sine qua non , the culmination of the event as a whole; and (2) an

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

Let us now turn briefly to the semantics of the verbs finish and finir. These verb patterns are the same as commencer and begin with respect to coercion (c£ ( 1 s )-(I 8) repeated below),

------

--- -

- ---

James Pustejovsky and Pierrette Bouillon

--

1ss

assertion that the entire event has occurred. As Dowty makes clear, this presupposes that the event has a natural division into two subparts. This would seem to indicate that finish is an aspectualizer which type-shifts the complement event into an achievement (i.e. a right-headed transition). It does this, however, by preserving the integrity of the complement event, for notice how finish -sentences behave in many respects as both accomplishments and achievements: (51) a. Mary finished building the house in 3 months. b. Mary finished building the house at 3:oo pm today.

(52) V'e1'1e2 [cul(e1) = e2

++

-.3e [e2 :S e1 1\ e2 < e 1\ e :5 e1]]

we can build this relation directly into the event structure itself, in which case it would be a relation on event trees. This should essentially be a logical culmination relation between events; cula:(e1 , e2). Now we can express the semantics offinish as a right-headed transition, where the subevent standing in the culminating relation with the larger event is seen as the AGENTIVE of the overall aspectual event. Furthermore, the FORMAL or result of the aspectual event is the assertability of the entire transition, of which the AGENTIVE is a part. (53) h-A.ef:le13R3P (finish: ES = e1 cufa: e2 1\ FORMAL = (P(x)(e?)]] 1\ AGENTIVE = [R (e1, x)]] Consider now the sentences in (54) and (5 5), which appear to be raising constructions and are grammatical in French and marginally acceptable in English.

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

Hence, even lexical accomplishments (left-headed transitions) can be interpreted as achievements (right-headed transitions) when complements of finish . In order to capture this intuition while still satisfying Dowty's fundamental interpretation offinish , we define a general relation of logical culmination, cui, between an event and one of its subevenr:

( 54) a. ?The leaves have finished falling. b. ?The paint has finished drying.

( 5 5 ) a. Les feuilles finissent de tomber. b.

Le

peinture a fini de secher.

These data would suggest that raising construction is possible with finish and finir with some nominals. But in fact, the sentences in ( 54) and ( 5 5) are best analyzed as pseudo-control cases, and they are restricted to a certain well defmed class of norninals. In general, these verbs do not pass the standard rais ing tests, bur nouns such as paint and leaves are exceptions because they carry qualia information indicating a kind of 'autonomy of behavior' relative to

-,

1 56

Aspecrual Coercion and Logical Polysemy

certain predicates. Hence, paint , for example, is consmted as a pseudo-agent in the control relation because of this property; i.e. it can dry on its own. It is interesting to observe that even in English an intransitive consmtction is possible; namely, if the event nominal in subject position has an agentive component (c£ (s6b) and (s6c)), then a control interpretation is possible in what would otherwise appear to be an intransitive (i.e. raising) consmtction. We will refer to these as intransitive control consmtctions. (56) a. b. c. (57) a.

While classes and talks have an apparent agentivity and controllable component to them, parties are less controllable, resulting in the less acceptable (s6a). Since rain is completely uncontrollable, it is ungrammatical in an intransitive control construction with the verb finish (c£ (57a)). The verb stop , as we see below, allows a raising interpretation and permits the intransitive raising consmtction in (s7b). The semantics of arreter and stop are interesting because they have both control and raising senses, yet do not allow complement coercion at all (c£ ( I 9) and (2o) repeated below).

( I 9)

Madame arrete de verser le the. ?Madame is stopping steeping the tea. b. *Madame arrete le the. *Madame is stopping the tea. (2o) a. ]'arrete de lire le livre. ?I am stopping reading the book. b. *]'arrete le livre. *I am stopping the book. a.

Why then should coercion not be possible with what would appear to be a verbal form almost identical to that of commencer and begin ? The answer to this question emerges from a closer examination of the data. Observe that stop appears in the sentences in (s8) with a non-control consmtction, assuming the sense of 'prevent': (58) a. John stopped Mary from smoking in his house. b. Mary stopped the man from hitting her. c. John stopped the bomb from exploding. In fact, there is a kind of coercion possible in complement position with stop , essentially reconsmtcting an ellipsed predicate, as in (59).

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

??The party finishes at midnight. Class will finish at 2:oo pm. The talk will finish by noon. *The rain will finish by noon. b. The rain will stop by noon.

James Pustejovsky and Pierrette Bouillon

1s7

(59) a. John stopped the car. (from moving) b. The referee stopped the clock. (from moving) c. Mary stopped the record. (from playing/moving) While constructions such as (59) are possible in French, those in (5 8) are not. But there are data suggesting that arreter does allow non-control complementa tion, as in (6o), with the sense of empecher (prevent). (6o) a. ]'arrete la bombe avant qu'elle explose. I am stopping the bomb before it explodes. b. ]'arrete le moteur avant qu'il ne chauffe. I am stopping the car before it overheats.

7

CONCLUSION

In this paper, we have attempted to illustrate how coercion operations are constrained by typing judgements and the structure of lexical semantic representations. Constraints on generative operations such as coercion are an integral component of the approach to semantics we have presented. In the process of this discussion, we have reiterated the advantages of a generative lexicon in the context of the larger theoretical and methodological issues in lexical semantics. More specifically, we have shown how begin and commencer exhibit both raising and control behaviour, and that this is an instance of the larger alternation class between causative and inchoative verbs, itself an example of logical polysemy. We have further shown why coercion is possible only with the control sense of commencer and begin and illustrated both the type inference involved and the semantic derivation of these constructions. We have also examined the behavior of two other classes of aspectual verbs, arreter and stop and finir and finish , and have shown whyfinir and finish are unambiguous but do exhibit complement coercion, and do appear to allow raising construc tions. Finally, we explained why arreter and stop do not allow coercion, even though they have control readings. We believe that the advantages accom panying generative mechanisms and the characterization of languages as

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

What these data suggest is that the complement type of arreter and stop is not an eventual function, as with commencer and begin , but rather simply an event, where the type ofverb is (�. (e, ET)). That is, these verbs are not strict obligatory control verbs, such as try and begin , but impose 'available controller' binding, as with verbs such as want (c£ Chomsky 1 98 1 ; Dowry 1 98 5; and Farkas 1 988). Given that the complement of both arreter and stop is �. it is clear that coercion is not possible since this is not among the type aliases for the NP complements given in (r 9b) and (2ob).

I

s 8 Aspectual Coercion and Logical Polysemy

polymorphic in well-defined ways allow us to overcome the explanatory inadequacies inherent in word sense enumeration approaches to lexical semantics. Although some of the details of the analysis have been omitted in order to concentrate on the general strategy of lexical analysis employed here, we have hopefully made clearer what some of the specific theoretical advantages of this approach are. Acknowledgements We would like to thank Peter Bosch, Bob Ingria, Noam Chomsky, Federica Busa, Michael Johnston, Evelyne Viegas, Graham Russell, and two anonymous reviewers for their helpful comments and suggestions. All responsibility for errors are ofcourse our own. Received: I 8.07.94 Revised version received: 05.I2.94

Computer Science Department Brandeis University Waltham, MA 02254 USA [email protected] PIERRETTE BOUILLON

ISSCO University ofGeneva 54 route des Acacias Ch-t zz7 Geneva Switzerland [email protected]

NOTES

=-

I See, for example, the work reported in Buicelaar & Mineur (I 994), which attempts to unifY some of the notions from Generative Lexicon Theory wii:h Categorial Grammar. Busa & Dini ( I994) attempt to import the notions of coercion with qualia srructure (see below) into HPSG for the handling of control phenomena, while Bouillon & Viegas (I994) handle cross-linguistic pheno mena of adjective-noun collocations. Both Copestake (I993) and Sanfilippo (I 990) are also interesting in how they model the projection of lexical semantic information to the syntax.

2 Qualia structure can be seen as providing the 'modes of explanation' for a concept, as lexicalized in a particular word. This is not strictly rrue, as we shall see in Section 5 below, where we introduce the notion of an eventual function. 4 As one reviewer points out, we experience objects in any number of ways. That is, one need not read books in order to be bored by them. One can be bored by looking at them, shopping for them, writing them, or chinking about them. This is not in any way inconsistent with the GL approach. As discussed in Puste jovsky (I99Ia) and elaborated on in

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

JAMES PUSTEJOVSKY

James Pustejovsky and Pierrette Bouillon 1 5 9 Pustejovsky ( 1 995). the qualia determine two types ofinformation in the context of coercion: i. type and sort information which the qualia must satisfy; ii. specific qualia values which are the explanatory modes in understanding a word.

1I

I2

I3

14

1s

16

17 18

---

--

-- -- ---

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

For words such asfilm and book, the TELIC quale role value of watch and read respectively are not optional in any sense, but are part of the semantics of the words. When an NP enters into a coercive environment, such as here with experien cer verbs, the qualia values act only to determine the default assignment for how the type environment is recon srructed. Strictly speaking, in terms of type saris faction either TELIC or AGENTIVE is a possible target of the modification. Hence, it can also be interpreted as e.g. 'a book taking a long rime to write'. 6 We follow Landman (199 1 ) in making the distinction between the temporal relations in an event structure and the intensional relation between event parts. 7 Formally, the head is interpreted as a focus srructure over the domain ofevents. This approach is explored in Pustejovsky (1995)· 8 We will continue referring to the binding relation between the matrix subject and an argument position in the complement position as semantic control, although the analysis does not necessarily assume a syntactic reflex for this binding relatiotL 9 There are at least two other major properties of raising predicates that should be mentioned, but that are not as important to our discussion. These are: (a) the inheritance property, which ensures that any syntactic restrictions imposed by the embedded VP on the subject are inherited or reflected in the 'raised' position; and (b) the narrow scope interpretation of the raised NP relative to the raising verb. See Jacobson (1990) and

10

Di Sciullo & Williams (1987) for discus sion. Further discussion of the syntactic patterning associated with aspectual predicates in English can be found in Freed ( 1979) and Rudanko ( I 989). Other verbs in this natural class include grope for, reach for, and other conative verbs as well. On this view, Tense is analyzed as a function from events to propositions, viz of rype (ff', t). We simplify this analysis here for ease of presentation. In fact, an eventual function is a function from individuals to sets of events: ((o, r)r)D. With this typing, begin as a conrrol verb has the following type: ((e, (er, t)), ((e, (eT))). Furthermore, under this analysis, Tense is rreated as a generalized quanti fier: ((£0, t), t). For details see Pustejovsky (1995). We will also refer to the typing assign ments in (3 I ) as (e - eT) - (e - eT) and (ro - eT), respectively, for use in type inference trees below. We follow Asher & Morreau ( 199 1 ) and Asher & Lascarides ( 1 993) in the use of the defeasible conditional > for specifying default lexical inferences. Unless otherwise stated, the default event srructure (Es) associated with the qualia for a TRANSITION is: ES - [,T eP <« e 5]. Furthermore, the qualia structure inter prets the negation of the FORMAL value as holding in the AGENTIVE as well. See Pustej ovsky (1995) for discussion and justification of this move. Notice that the representation of expeti encer predicates given in ( 1 0) above is actually a specific form of this causative paradigm, where the relation in the AGENTIVE quale is sortally restricted to experiencing predicates. This particular observation was suggested by Robert Ingria (personal communica tion). We ignore for now the details of the lexical representation. Our concern in this paper is simply to illustrate the source

----

160 Aspectual Coercion and Logical Polysemy 22 See P!Jstejovsky (1993) for details on the control relation. We follow generally Klein & Sag (1 985) for how binding is achieved in £qui-constructions such as

begin. 23 Sentence (49) should be contrasted with sentences such as *The book began last week, which are ungrammatical. In this sentence, the subject NP is not an event denoting nominal such as symphony. The metonymic reconstruction resulting from coercion would make available a rype alias of an eventual function, (e, � rather than a simple event, 1!'. Hence, such cases are ruled out because of a type mismatch.

RE FERE N C E S Asher, N . & Lascarides, A. ( 1 993), 'Temporal interpretation, discourse relations and commonsense entailment', Linguistics and Philosophy, 16 , 437-94· Asher, N. & Morreau, M. ( 1 99 1 ), 'Common sense entailment a modal theory of nonmonotonic reasoning', in Proceedings to

the 1 2th International Joint Conference on Artificial Intelligence, Sydney, Australia, August 1 99 1 . Beierle, C., Hedstiick, U., Pletat, U., Schmitt, P. H., & Siekmann, J. (1 992), 'An order sorted logic for knowledge representation systems', Artiflciallntelligence , 55, 1 49-9 1 . Benthem, J . van. (1 983), The Logic of Time, Reidel, Dordrecht. Bouillon, P. & Viegas, E. (1 994), 'A semi polymorphic approach to the interpreta tion of adjectival constructions: a cross-linguistic perspective', Proceedings of Euralex 94 , Amsterdam. Buitelaar, P. & Mineur, A.-M. (1 994), 'Com positionaliry and coercion in caregorial grammar', in Proceedings ofNinth Amsterdam Colloquium , 1 4- 1 7 December 1 993, Amsterdam. Busa, F. & Dini, L. (1 994), 'Generative opera-

tions in a constraint-based grammar', in

Proceedings of KONVENS-94 , Vienna, Austria, 28-30 September 1 994. Chomsky, N. ( 1 9 8 1 ), Lectures on Government and Binding, Foris,.Dordrecht. Copesrake, A. ( 1993), 'Defaults in the LKB', in T. Briscoe & A. Copestake (eds), Default Inheritance in the Lexicon , Cambridge University Press, Cambridge. Copestake, A. & Briscoe, E. ( 1992), 'Lexical operations in a unification-based frame work', in J. Pustejovsky & S. Bergler (eds),

Lexical Semantics and Knowledge Represen tation , Springer Verlag, New York. Di Sciullo, A.-M. & Williams, E. (1 987), On the Definition of a Word, MIT Press, Cambridge, MA. Dowry, D. R. ( 1 98 s), 'On some recent analyses of control', Linguistics and Philo sophy, 8, 1-4 1 . Dowry, D . R. (1 979), Word Meaning and Montague Grammar, Reidel, Dordrecht. Farkas, D. F. (1 988), 'On obligatory control', in Linguistics and Philosophy, 1 1 , 1 , 27-58. Freed, A. F. (1 979), The Semantics of English Aspectual Complementation , Reidel, Dor d recht.

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

of the polysemy and how the rwo senses are logically related. 1 9 C£ Pustejovsky (1 994a, 1 994b) for dis cussiOn. 20 See Gunther (1 992) for explanation of formal mechanisms of type inference within the ). -calculus, and Morrill ( 1 992), Copestake (1993), and Pustejovsky ( 1994b) for irs application to lexical representation. 2 1 For the purpose of the present discussion, we will ignore the type distinction berween individuals, e, and generalized quantifiers ((e,t),t). We assume, however, that they are related by a type shifting operation as discussed in Partee ( 1 992).

James Pustejovsky and Pierrette Bouillon I 6 I

.

.

conjunction and rype ambiguiry', in Mean

in� Use, and Interpretation of Language,

R.

Bauerle, Schwarze, & A. von Stechow (eds), Walter de Gruyter, Berlin. Perlmutter, D. (I 970), 'On the two verbs begin ', in R. Jacobs & P. Rosenbaum (eds),

Readings in English Transformational Gram mar, Ginn, Waltham, M.A. Pollard, C. & Sag, I. (I987), Information-Based Syntax and Semantics, CSLI Lecrure Notes

Number I 3, Stanford, CA. Pustejovsky, J. (I988), 'The geometry of events', in Carol Tenny (ed.), Studies in Generative Approaches to Aspect, Lexicon Project Working Papers 24, MIT, Cam bridge, MA Pustejovsky, J. (199 I a), 'The generative lexi con', Computational Linguistics , 17, 4· Pustejovsky, J. (I 99 I b), 'The syntax of event structure', Cognition , 41, 47-8 1 . Pustejovsky, J . (I993), 'Type coercion and lexical selection', in J. Pustejovsky (ed.), Semantics and the Lexicon , Kluwer Aca demic Publishers, Dordrecht. Pustejovsky,]. (I 994a), 'Linguistic constraints on rype coercion', in P. Saint-Dizier & E. Viegas (eds), Computational Lexical Seman tics, Cambridge Universiry Press, Cam bridge. Pustejovsky, J. (I 994b), 'Semantic ryping and degrees of polymorphism', in Marrin-Vide (ed.), Current Issues in Mathematical Linguis tics, Elsevier, Holland. Pustejovsky,]. (I99S). The Generative Lexicon , MIT Press, Cambridge, MA. Pustejovsky, J. & Anick, P. (I988), 'On the semantic interpretation of nominals', in .

Proceedings oft zth International Conference on Computational Linguistics, Budapest, Hun gary, 5 I 8-23. Pustejovsky, J. & Boguraev, B. (I993). 'Lexical knowledge representation and narural language processing', Artificial Intelligence, 63, I 93-223. Pustejovsky,]. & Busa, F. (I 994), 'Unaccusari viry and event composition', in P.-M. Bertinetto (ed.), Approaches to Tense and Aspect, Elsevier, Amsterdam. Rudanko, J. (I 989), Complementation and Cose

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

Godard, D. & Jayez, J. (I993), 'Towards a proper rreatment of coercion phenomena', in Proceedings ofthe 1 993 European ACL . Gunter, C. (I 992), Semantics of Programming Languages, MIT Press, Cambridge, MA Grimshaw,]. (I 990), Argument Structure, MIT Press, Cambridge, MA Jacobson, P. (I990). 'Raising as function composition', in Linguistics and Philosophy, 13, 423-76. Kamp, H. ( I 98o), 'Some remarks on the logic ofchanger: Part I' in C. Rohrer (ed.), Time, Tense, and Quantifiers, Niemeyer, Tiibingen. Klein, E. & Sag, I. (I98 s). 'Type-driven translation', in Linguistics and Philosophy, 8, I 63-202. Krifka, M. (I 992), 'Themarics relation as links between nominal reference and temporal consrirurion', in I. A. Sag & A. Szalolcsi (eds), Lexical Matters, CSLI Lecrure Notes Number 24, Stanford, CA. Lamiroy, B. (I 987), 'The complementation of aspecrual verbs in French', Language , 63, 2. Landman, F. (I99 I), Structures for Semantics, Kluwer, Amsterdam. Levin, B. (I 99 3), Lexical Organization ofEnglish Verbs, Universiry of Chicago Press, Chicago, IL. Miller, G. (I99 1 ), The Science of Words, Scientific American Press, New York. Moens, M. (I 987), 'Tense, aspect and temporal reference', Ph.D. thesis, Edin burgh. Moens, M. & Steedman, M. I 98 8, 'Temporal oncology and temporal reference', Compu tational Linguistics, 14, I 5-28. Morrill, G. (I992), Type-Logical Grammar, Onderzoeksinstiruut voor Taal en Spraak, Utrecht. Nunberg, G. (I 979). 'The non-uniqueness of semantic solutions: polysemy', Linguistics and Philosophy, 3, I43-84. Partee, B. (I 992), 'Syntactic categories and semantic rype', in M. Rosner & R. Johnson (eds), Computational Linguistics and Formal Semantics , Cambridge Universiry Press, Cambridge. Partee, B. & Rooth, M. (I98 3). 'Generalized

162

Aspecrual Coercion and Logical Polysemy

Grammar, State University of New York Press, Albany, NY. Sanfilippo, A. (1 990), 'Grammatical relations, thematic roles, and verb semantics,' Ph.D. thesis, Centre for Cognitive Science, University of Edinburgh.

Zaenen, A. (1993), 'Unaccusativity in Dutch: integrating syntax and lexical semantics', in J. Pustejovsky (ed.), Semantics and the Lexicon , Kluwer Academic Publishers, Dordrecht.

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

journal ofSemantics

1 2: 163-209

©Oxford Universiry Press 1 995

A Typology and Discourse Semantics for Motion Verbs and Spatial PPs in French N I C H O L A S A S H E R and P I E R R E S A B L A Y R O L L E S

IRIT-LRC, University Paul Saba tier

Abstract

1

I NTRODUCTION

I n this paper we offer a semantic study of motion verbs and motion verb com plexes determined by motion verbs and spatial prepositional phrase adjuncts. We propose a classification of motion verbs and of motion verb complexes. Unlike other semantic or syntactic studies, we build up the spatiotemporal semantic properties of motion verb complexes compositionally, on the basis of the semantic properties of the verbs, their arguments and adjuncts. Motion verbs convey much more information than just a simple and strict spatiotemporal description of motion activities (as we will see shortly). The immense richness of motion verbs is probably at the origin of the difficulties encountered by all those (including us) who have attempted to study them. The organization of this paper is as follows. First, we set out in Section 2 an ontology of eventualities and spatiotemporal extensions, which furnishes a philosophical background against which to analyze motion describing expres Sions. Then, in Section 3 we present the typology for motion verbs in French. We have considered motion verbs without any adjuncts, in their atemporal form and independently of any context in order to extract their, and only their, intrinsic semantics. To be sure, even this requires certain ontological assumptions about processes, objects, boundaries, and the like. We detail these in Section 2. We have also complemented our purely linguistic justifications for this typology with some considerations that stem from compositional

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

In this paper we offer a semantic study of motion verbs and motion verb complexes " determined by motion verbs and spatial prepositional phrase adjuncts. We propose a classification of motion verbs and of motion verb complexes. Unlike other semantic or syntac tic studies, we build up the spatiotemporal semantic properties of motion verb complexes compositionally, on the basis of the semantic properties of the verbs, their arguments and adjuncts. We show how to combine this lexical information with discourse information to determine the spatiotemporal structure of texts and to help with lexical disambiguation.

1 64 ·

A

Typology and Discourse Semantics for Motion Verbs and Spatial PPs in French

o

o

motion verbs have in themselves an intrinisic spatial and temporal semantics; the spatial and temporal semantics of a motion verb complex is a specifiable function of the spatial and temporal semantics of each component used in this complex (for example, Laur 1 99 1 has studied the combination of displacement verbs and spatial prepositions in French). We can quickly illustrate these two points with the following examples:

1.

(a) entrer

to go in

(b) sortir

2.

to go out (c) s'approcher to approach (a) sortir du jardin to go out ofthe garden

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

preoccupations; the verb classes of our typology are organized according to a certain class of semantic properties so that we can monotonically specify possible (and linguistically accepted) paths of transition from one class to another when we add to the verb, for example, a PP adjunct. Further, in this way, motion complexes can be classified using the same classes as for motion verbs. We give rules of combination by means of which we can derive the properties and classification of a motion verb complex from the properties of the verb, its arguments, and its PP adjuncts. So we distinguish motion verbs and motion complexes. But since both levels use the same kind of classification, the transition to one {the motion verb level) to another {the motion complex level) is clearly and logically defined (unlike some other semantic studies to be discussed in Section s). In Section 4 we consider in greater detail one of the classes found in Section J . J , namely the class of verbs of change of location. We refine this class with more precise spatial criteria. This allows us, in Section 4·5· to give compositional rules for those motion verbs and the spatial prepositional phrases they can be combined with. With these rules, we then obtain a classification ofmotion verb complexes. This spatiotemporal knowledge about motion verb complexes can then be used in the understanding processes of the spatiotemporal structure of texts, c£ Section 4·7· In Section s we present and discuss the main ways people have tackled the difficult problem of the study of motion verbs/motion verb complexes, and propose a general comparative table of all these works including ours. Finally, we give in Section 6 some evidence in favor of our typology and how it has been used for several other detailed investigations on motion. Our typology is based on two principles.

Nicholas Asher and Pierre Sablayrolles

165

(b) sortir par le jardin to go out by thegarden (c) sortir dans le jardin to go out into the garden (d) s'eloigner dans Ia rue togo away down the street

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

The first point is exemplified by comparing verbs like those in ( 1 a) and ( I b) or in ( I a) and (rc). Verbs like the one in ( I a) require the moving entity to be outside some location before the process and inside after. In contrast, verbs like the one in ( 1 b) require the moving entity to be inside some location before the process begins and outside after the process ends. Verbs like the one in ( I c) require the moving entity to be outside some location both before and after the process; but they also require that the position after the process has to be closer to this location than the position before the process. The second point can be illustrated by contrasting the complexes given in (2a), (2b), and (2c). In these three examples, the same motion verb sortir and the same location le jardin are used. These complexes differ only in the preposition used to link the location in the PP to the motion verb. In (2a), the complex requires that the moving entity inside the garden before the process and outside after; in (2b), the complex requires that the moving entity be inside some loca tion before the process, outside it (i.e. inside another location) after the process, and that it must pass through the garden during its moving process; in (2c), the complex requires that the moving entity be inside some location, which is not the garden, before the process and inside the garden after it. In (2a), (2b), and (2c), we have in fact changed just one component of the complex and we have seen that the resulting meaning has completely changed. But all the components can also change. Compare for example (2a) and (2d). We need to be able to calculate the meanings of such complexes in a rule governed way. Motion verbs and motion verb complexes are two different things (see, for instance, Sablayrolles I 992b or I 993) that should not be indiscriminately mixed together in a typology. But if we consider them separately, we of course need something to connect them-if they are obviously related. This is the aim of the compositional rules which compute of the semantics of a complex form one of each component. The computation of the spatiotemporal properties of motion complexes is similar to the calculation of aspect, as studied in Verkuyl (I 972), Smith ( I 991), and Krifka ( I 987).

166 A Typology and Discourse Semantics for Motion Verbs and Spatial PPs in French

2 E VE N T U A L I T I E S A N D S P A T I O TEMPORAL EXTE N S I O N S

•

=-

•

•

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

We are interested here in a typology of motion as it is expressed in natural language. Motion is a property of objects (i.e. objects move), but motions themselves are a type of event. Motions are changes of some sort of spatial posi tion over time. To explore motion, we must first make clear our ontological assumptions concerning eventualities and spatiotemporal extensions. Eventualities are concrete but complex objects. They may have both objects as constituents and also other eventualities. Every concrete entity, we suggest along with Vieu ( 1 99 1 ), has a spatiotemporal trajectory, which we will denote with the aid of a function STrej. STret{x) is the spatiotemporal extent ofx. In DR theoretic terms (Discourse Representation Theory or DRT (Kamp 1 979), STref(x) is a new discourse entity designating the trajectory described by x all along its 'life'. STref may be additionally parametrized with the aid of a temporal variable: STref(x, e) (or STret{x, t)) denotes the 'temporal slice' of STref(x) whose time matches the time of the event e (or the time t), if e (or t) is temporally included in STref(x); otherwise it is not defined. Spatiotemporal relations between these referents, such as inclusion (P(x, y)), overlap (O(x, y)), or contact (EC(x, y)), are axiomatized in a theory of space-time, derived from Clarke's calculus of individuals (Clarke 1 98 1 , 1 98 s ), and which is based on mereology (Lesniewski 1927-3 1 ). This theory, fully presented in Aurnague ( 1 991), Vieu ( 1 99 1), and Sablayrolles ( 1 99 1 a), is used in Aurnague & Vieu ( 1 993) for representing the geometrical aspects of the lexical semantics of static relations such as etre dans to be in ' etre sur to be on ' etre devant to be infront of, and internal localization nouns such as le haut the top , le dessus the top surface , le coin the corner . . . It also can be used to encode various facts about the geometrical and topological properties of objects and processes or, rather, their STref projections (c£ Vieu 1 99 1 ). We further suppose that all eventualities have a beginning or source, and end or goal, and a middle or path. Thus we will define Source(e) as a 'location' whose spatioremporal referent contains or is in contact with STref(e,lni�e)), and similarly for Goal(e). Finally, Path(e) will be a set of sequences of'locations' 1;. A path is a set of sequences because we need to allow for descriptions at different granularities, and this is necessary to handle the discourse relation Elaboration correctly. For each sequence of locations (!1 , , In) we have P(STrej(e), 'L;STrej(/;)), and for each i, EC(STrej(l;), STrej(/;�,)), i.e. adjacent locations in the sequence are externally connected. Source(e) and Goal(e) are not sets of 'locations', however. They are often lexicalized, and rhus uniquely identified within the discourse by a particular discourse referent introduced by

Nicholas Asher and Pierre Sablayrolles

167

the noun phrase; in other cases they are anaphorically identified, and in the remaining cases their existence is assumed through the introduction of discourse referents. As a consequence, we do not have (I) but only (2) and (3) since we should allow for refinements of the path within Source(e) or Goal(e).

'
�

3x E Path (e), x - (Source (e), . . ., Goal(e)) (/1 , , /n) .... (O (STrej( /1), STrej(Source (e))) !\ O (STref(ln), STrej(Goal (e))))) •

•

•

(3)

{y : 3x E Path (e) 3 /1 /k, 3/m , . . . In ((x (/1, , fk) U y U (/m, . . . , /n)) !\ O (STref( /1 ), STrej(Source (e))) !\ . . . !\ O(STrej(lk ), STref(Source (e))) A O(STref(lm ), STrej(Goal(e))) A . . . A O (STref(ln), STrej(Goal (e)))) SIP(e)

�

•

�

•

•

•

•

•

(4)

As we shall see, the notions of Source, Goal and SIP may be focalized by motion verbs, prepositions and motion complexes in French; they are important types of entities in natural language metaphysics. Our study also requires us to make explicit our assumptions concerning the semantics of motion expressions. We will follow roughly the approach of Davidson (I 967) (an approach which is also used in DRT), on which eventualities, along with other objects, satisfY predicates derived from natural language verbs. In OR-theoretic terms eventuality discourse referents will be arguments of conditions introduced by the verbs, and in fact the eventuality discourse referents themselves are introduced by the inflection node in the syntactic tree. Such eventuality discourse referents are also arguments to various adverbial phrases, and, in particular, spatial prepositional phrases. One of our tasks in this paper is to classify the various eventuality discourse referents that might be understood as representing movements and to determine those inferences that allow us to specifY the spatiotemporal position of the eventuality as a whole or of its parts or actors with respect to various reference locations given in a text (see Borillo & Sablayrolles 1 993 for a discussion). In chis paper we will present both a typology and rules for drawing inferences about locational structure. Our ultimate aim is to use this lexical research to drive discourse reasoning about spatiotemporal structure (see, for instance, Asher et a/ . I 993). But our task here is to make clear what inferences are allowed by the lexical content of the verbal predicate on an eventuality and with the other

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

In what follows we shall employ a subsequence of the path of an eventuality-the strict internalpath ofe or SIP(e), for each eventuality e . SIP(e) is defined the set Y of sequences y such that for each element y in Y there is an element x in Path (e) such that y contains no locations chat overlap the Goal(e) or Source(e). In symbols, where u is the concatenation operation:

1 68 A Typology and

Discourse Semantics for Motion Verbs and Spatial PPs in French

modifiers. We will restrict this study to verb complexes and ignore verbal nominalization and adjectival phrases, among others. To simplify matters further, we will consider only transitive motion verbs and verb complexes with only one PP adjunct, though our theory straightforwardly extends to arbitrary spatial PP adjuncts as long as these are of different types of the sort that we will define below (see Sablayrolles I 99 I b for a discussion on this point). We postpone the treatment of multiple spatial PPs of the same type in a single verbal complex for a future occasion.

3 THE T Y P O L O G Y

What are locations?

One concept that plays a central role in a typology of motion is the concept of 'location'. Despite first impressions, it is not easy to give a definition oflocation. Boons (I 98 s ) and Laur ( I 99 I) use the concept in a rough and ready way, without any analysis. But this leads to difficulties, especially at the level of inferences, where we need to specify the spatiotemporal position of one eventu ality with respect to some reference location, as shown in Sablayrolles I 992a. Boons and Laur define their displacement verbs as 'verbs absolutely requiring a change of location of a body, which suffers no modification of its shape nor of its substance during the process' (Boons 1 98 5). This definition is imprecise due to the non-specified nature of their concept of 'location', and it leads them to conflate verbs such as entrer to get in and courir to run , for example. Let us com pare ( 3) and (4). 3- Demainj'entrerai a la cuisine, sous un pretexte quelconque (Martin du Gard) Tomorrow, I will enter the kitchen under some pretext 4· Les joueurs courent sur le terrain de football The players run on the soccerfield S· Le gardien de but s'appuie contre le montane de ses buts The goalkeeper leans against his goalpost Both entrer in (3) and courir in (4) are displacement verbs (in the terrns of Boons {I 98 5) and Laur { I 99 I ). Both imply a change of'location' in their sense of the moving entity(ies). The PPs in (3) and (4) give us a 'location': Ia cuisine the kitchen , and le terrain de football the soccerfield , respectively. Knowing that in both sentences there is a change of'location', they would infer that in (3) there is a change oflocation with respect to Ia cuisine, which is absolutely true, and that in (4) there is a change oflocation with respect to le terrain de football, which is definitely false. Indeed, in ( J ), the author is located, before the process, outside the kitchen,

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

3. 1

Nicholas Asher and Pierre Sablayrolles

1 69

3 .2

Locations, positions, and postures

In order to analyze the concepts of location, position, and posture more fully, something has to be said concerning the relations between the locations, positions, and postures and what is introduced by the PP in the sentence. In the case of a complex which exemplifies a change oflocation, as in ( 3), what is introduced by the PP is definitely a location. This location can, as Laur ( 1 99 1 ) has shown, have an initial polarity (i.e. be considered as the Source(e)) or a final polarity (i.e. be considered as the Goal(e)). In the case ofa change ofposition, what is introduced by the PP is a location too. This location is such that the process takes place inside it; it can be con ceived, as we said earlier, as a background location. In such cases, as the name indicates, there is no change oflocation (the moving entity stays during the whole process inside this location), but rather a change ofposition. This position (c£ Definition 2 below) cannot be lexicalized and indeed cannot be expressed at all except with the help ofdeictic devices. Consequently, the PP always introduces a background location with such verbs. The positions referred to when we talk about change of position are generally either conceptualized by the hearer/ reader ofthe sentence as spatial sub-parts ofthe background location, with the presupposition that the moving entity does not stay inside the same sub-part during the whole process. This conceptualization is derived from contextual or

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

and will be located, after the process, inside the kitchen; there truly is a change of'location'. We can naturally talk about a change oflocation here. 1 But in (4), the players were located, before the process, on the soccerfield, and will still be located, after the process, on the soccerfield . However, ifwe are in the context of a soccer game, they have surely passed several times from one sub part of the soccerfield to another, even if they always stay on the soccerfield . So there are some changes of spatial properties, but they are not changes of location. Compare, for example, (s). In the case of (4), we will speak not of change of location but of change cifposition 2 and consider the soccer field as a background location. In the case of ( s ), we will speak of a change cifposture .3 Although it is probably possible to continue to refine the granularity of analysis and look at ever finer detailed movements, our study stops with changes of posture. The study of motion takes place always relative to objects that move, and for practically all purposes of which we can conceive, change of posture of an object of the right sort of granularity suffices to capture the nature of movement as expressed in natural language. Additionally, we suspect that further refinements of motion will lead to a disintegration of the concept-we will end up with a Heraclitean universe in which all things change. In the next section we analyze the concepts of location, position and posture more fully.

1 70 A

Typology and Discourse Semantics for Motion Verbs and Spatial PPs in French

world knowledge {i.e. directly from the world, or the previous discourse, or from both). In the case of a change of posture, what is introduced by the PP is again a location. In such cases, as the name indicates, there is no change of location or position {the moving entity stays within this location and at the same position for the whole process), but rather a change of posture. The process then consists in realizing a special, physical relation between one or more parts of the moving entity and the location introduced by the PP. For example, in {6), it is a matter of realizing a physical contact between the knees4 and the bench; in (7) of passing the head and the shoulders through the window.

In all cases, PPs used in combination with motion verbs always introduce a location. We now propose the following analyses for our concepts of location (c£ Definition r), position (c£ Definition 2), and posture (c£ Definition 3).

Definition 1 A location5 is analysed as a portion ofspace6 which can be 'designated' in natura / language, and with which is associated afunctionality. This location can be recognized as one in which certain types ofactivity takeplace .7 Locations aregenerally lexicalized by means ofa rea / lexical item (e.g. Ia cuisine the kitchen in (3)) or a construction using a preposition and a real lexical item (eg. derriere Ia maison behind the house). They can also be expressed with deictic constructions like here or where John was 5 minutes ago. Definition 2 A position8 is analyzed as a portion of suiface, without any functionality or real associated lexical item, 9 and only geometrically defined by the pragmatic shape (cf Definition 4) associated with a given entity. A position thus is necessarily dependent on the entity usedfor its definition. Definition 3 A posture10 is analyzed as a special way to be inside one's pragmatic shape, with which is associated a certainfunctionality (cf Definition 4). Postures are always postures ofan entity and they are defined by the relations between the parts of that entity. Postures are lexicalized byparticipialforms ofverbs ofchanges ofposture e.g. ass is sitting down or by adjectives like de bout standing up. Definition 4 By pragmatic shape we mean the 3-D portion of space fully occupied by the entity, plus that space that would be occupied by the entity x were it to undergo a change ofposture :rr, where :rr is constrained such that if:rr occursfrom t to t' and posture(x, t) - posture{.'<, t'), then STrej(x, t) is at least roughly the same as STre(x, t ').

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

6. S'agenouiller sur le bane To kneel down on the bench 7· Se pencher par le fenetre To lean out ofthe window

Nicholas Asher and Pierre Sablayrolles 1 7 1

We are now able to define precisely the generic concepts ofSource, Goal and Path given in Section 2, with each of our location, position, and posture. Source(e), lnit-position(e) and Init-posture(e) are a location, position, and posture, respectively, whose spatiotemporal referent contains or is in contact with STref(e,lnit(e)); we define Goal(e), Final-position(e), and Final-posture(e) analgously. Finally, Path(e) will be a set of sequences of locations and Path position(e) and Path-posture(e) will be a sequence of positions and postures, respectively. Similarly for SIP(e), SIP-position(e), SIP-posture(e). 3·3

Thefour subclasses ofmotion verbs

8. Les joueurs courent sur le terrain de football The players run on the soccerfield 9· A Los Angeles, Ia plupart des jogueurs courent sur place In LA, most ofthejoggers run on the spot Io. Pauline se deplace en bus au centre ville Pauline takes the bus in the centre ofthe city I I . *Pauline se deplace en bus sur place *Pauline moves around the bus while it is stationary

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

With the definitions oflocation, position, and posture given in Section 3.2, we can now talk about class of motion verbs of change of location, change of position, and change of posture. Let us have a closer look at the verbs ofchange of position. Compare, for example, (4), which we give below again as (8) and (9). Compare also (ro) and ( I I ). Some of these verbs (like courir to run , danser to dance , voler tofly . . .) can be combined with sur place on the spot . Others (like se de placer to move around, circuler to circulate, descendre to go down , s'elever togo up . . .) cannot. This shows that the latter always imply a change of position, while the former only suggest it. The latter class (e.g. se deplacer to move around) will keep the name of verbs of change ofposition. We will call verbs like courir to run verbs of inertial change of position. We also distinguish verbs of inertial change of position from verbs of change of posture. Verbs of inertial change of position typically describe a change of position (8), although this is not a strict entailment (9). Nevertheless, where it is a strict entailment, the change oflocation concerns the entity in its entirety. On the contrary, verbs of change of posture usually, as their name suggests, describe a change of posture ( 1 2). They can also describe some kind of change of position, but in this case the change always concerns only parts of the entity ( I 3). It never concerns the whole entity. Otherwise this would imply that ( I 3) should have a meaning like that of ( I 4)!

1 72

A Typology and Discourse Semantics for Motion Verbs and Spatial PPs in French

I 2. Gaby se penche pour caresser son chien Gaby leans over to caress her dog I 3· Gaby se penche par le fenerre Gaby leans out ofthe window I 4. Gaby passe par/tombe de Ia fenerre Gabygoes out through/falls out ofthe window

·

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

We thus arrive at four classes of motion verbs. We give both a discursive and a logical definition below (Definitions s.I-8. I and 5.2-8.2) for each of these classes. To present our axioms, we need a formal language. We have already made use of a predicate calculus language for describing our spatiotemporal relations. First we extend this language with a conditional operator >, where A > B means 'if A then normally B' and represents a defeasible or generic rule. The semantics of this language is given in Asher & Morreau (I 991) and allows us to encode defaults about lexical entries. We will shortly be reasoning about the types and properties of eventualities that verbs and verb complexes introduce within our semantic representations. As we intend ultimately to talk of discourse structure using the framework of Asher et a / . ( I 993), our semantic representations for clauses will be DRSs and hence the representational elements that verbs will introduce will be predicative DRSs that abstract over discourse referents. The types of eventualities that we shall discuss are in fact types of discourse referents. Predicative DRSs are functions from tuples of discourse referents of the appropriate type to conditions in a DRS. Our axioms will then apply these predicative DRSs to a sequence of discourse referents of particular types that occur as arguments and spell out the consequences in terms of spatiotemporal conditions on those discourse referents. We will also generalize on the types of conditions; thus, we will have axioms that apply conditions of a certain type derived from those verbs and prepositions. These axioms should be understood as rules for expanding a representational structure-either a DRS or a structure that in the DRS construction procedure we favor will eventually yield a DRS when combined with other elements. This is similar to the approach taken by Kamp & RoGdeutscher in a number of papers (see e.g. Kamp & RoBdeutscher I 992). We will assume the existence of certain functional dependencies between eventualities and other discourse entities, e.g. Source(e), SIP(e), Goal(e), Mobile(e)-the moving entity-and Lre�e)-the location of reference. 1 1 We will forgo the box notation of DRT, since for us nothing hangs on it at the level of the spatiotemporal information we address here, though we shall construct DRSs when we apply our theory. 1 2 We give below the properties for each group of verb.

Nicholas Asher and P�erre Sablayrolles

1 73

Definition 5 Verbs ofchange oflocation 1 . A verb of change oflocation, entails that the moving entity changes location during the process ex: entrer to enter, arriver to arrive, atterrir to land, s'approcher to approach . . . 2. change-oflocation (e) -+ SIP(e) # {(Source(e)))

One may also wish to require that for changes of position, the moving entity at least by default stays in the same location during the whole process, but we will not do this here. So it is possible that a change of position is also a change of location. Definition 7 Verbs ofinertial change of position 1 . By defoult a verb of inertial change of position implies a change ofposition for the moving entity ex: courir to run, danser to dance, voler tofly . . . 2. inertial-change-ofposition (e) > Path-position(e) # {(!nit-position (e))} Definition 8 Verbs ofchange of posture 1 . Verbs of change of posture entails that the moving entity stays inside the same location and at the same position during the whole process, but also changes of posture during the process ex: se pencher to lean over/towards(forward!back, s'asseoir to sit down, se baisser to bend down . . . 2. change-ofposture(e) -+ Path-posture (e) # {(Init-posture (e))) change-of-posture (e) - Path-position (e) = {(!nit-position (e))} Note that a change of posture is not any change of posture, but just one that is expressed by a certain verb. Changes of posture in our sense, those that are expressed by verbs in French (and we suppose for other languages too), do not entail changes of position-and, hence, changes oflocation. 3 ·4

Embedding classes

For each subclass of motion verb defined in Section J.J, we associate truth conditional properties expressed in our formal language. These properties

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

Definition 6 Verbs of change of position 1 . A verb ofchange of position entails that the movement entity changes position during the process ex: se deplacer to move around, circular to circulate, descendre to go down, s'elever to go up . . . 2. change-ofposition (e) - Path-position (e) # {(!nit-position (e))}

1 74 A Typology and Discourse Semantics for Motion Verbs and Spatial PPs in French

.

1 S· Courir sur place To run on the spot 16. Courir autour du stade To run around the stadium 17. Courir vers Ia maison To run towards the house

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

reflect the minimal intrinsic core meaning common to all the verbs belonging to the class. The three first subclasses are not exclusive; when we take each class to be defined by the eventualities they describe, they are in fact strictly ordered by the subset relation. That is, the spatial properties of the minimal core meaning of verbs of change of location are more constrained than those of verbs of change of position, which, in their turn, are more constrained than those of verbs of inertial change of position. Our typology is based on intuition and our definitions of location and position: one cannot change location if one does not change position during the process of changing of location; and one cannot change position if one does not inertially change position during the process. Our taxonomy allows for a monotonic calculation of constraints for a verb complex. The combination of a verb and its PP or other arguments (direct object, for example) can only add to the spatiotemporal properties that affect the Source, Goal and Path of the targets' posture, position, or location. There is an evident advantage in doing this. Non-monotonic constraint propagation would require us to assume the existence of something like an intensional operator that might block the truth conditional content of the verb itself-in the sense that Believe (¢ ) fails to entail ¢ -or to rephrase all our constraints at the default level. The latter makes lexical semantics very complex (for a discussion see Lascarides & Asher 1 993), and we do not see our way through. The former seems semantically unmotivated. With monotonic constraint propagation, we need resort to neither alternative.U Take, as an illustration, the following verbs: courir to run , voyager to travel , and entrer to enter, which are verbs of inertial change of position, of change of posi tion, and ofchange oflocation, respectively. The composition of to run with some other elements can either result in a complex having the same constraints (such as in ( 1 s)), or more constraints (such as in ( 16)), or even more constraints (such as in (17)). In the case of the verb to travel, the compositionality can either result in a complex having the same constraints on the movements of the moving entity (such as in (r 8}}, or more constraints (such as in (19)), but never fewer constraints (as (2o) shows). In the case of the verb to enter , which already belongs to the class placing the most constraints on the movement of the moving entity (ofthe three classes considered here), only a complex having the same constraints can result from compositional calculation (c£ for example (2 1 ))

Nicholas Asher and Pierre Sablayrolles

175

8. Voyager en Europe To travel inside Europe 19. Voyager de Toulouse a Austin, Texas To travelfrom Toulouse to Austin, Texas 20. *Voyager sur place *To travel in place 2 1 . Entrer dans une maison To enter a house 22. S'asseoir sur une chaise To sit on a chair 23. S'asseoir par terre To sit down 1

4 F R O M LEX I C A L S E M A N T I C S TO D I S C O U RSE U NDERSTANDI N G 4.1

The verbs ofchange oflocation

In this section, we focus on verbs of change of location and ignore three other groups we have defined in Section 3·3· Verbs of change of location are propitious for a detailed analysis of how the space is organized in and around the location with respect to which the displacement takes sense. They describe displacement going from the outside of a location to the inside, or the reverse displacement. Nevertheless, a coarse organization of the space into only two 'zones', the interior and the exterior of the location, quickly proves insufficient. Compare, for example, (24) with (25) and (26) with (27). 24. Abby est sortie de la maison Abby has gone out ofthe house 25. Abby est partie de la maison Abby hasgone awayfrom the house 26. L'avion a atterri sur la piste 4 The plane has touched down on runway 4 27. L'avoin s'est approche de la piste 4 The plane has approached runway 4

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

The composition ofverbs ofchange of posture with any PP (cf (22) and (23)) will always result in a complex having the same constraints as those the verb introduces. If one does not change position or posture during the process described by a verb or verbal complex, then this 'process' can only be a static one.

1 76

A Typology and Discourse Semantics for Motion Verbs and Spatial PPs in French

In both

(24)

and

(25), Abby has

gone from the inside of

the house

to the

outside ofit. But if sortir in (24) only describes this sort of displacement, partir in

(25) forces Abby to continue her displacement till she is away from the house, at a sufficient distance. Under a certain critical distance, the verb partir could not be used. For example, ifAbby took two minutes to talk in the garden with her neighbour Peggy, and if someone asked AI, who is inside the house 'Abby est

elle partie?' 'Has Abby gone away?', AI will surely answer no, if he can hear her

talking or see her through a window. This critical distance is a function of many

parameters-not only the size of the moving entity and of the location, but also the presence and the size of the other referents (locations and/or objects) in the

difficult to circumscribe precisely this critical distance. But we do believe it is pan and parcel of our natural language metaphysics and classification of movements. We believe that the contextual factors relevant to describing this critical distance must be handled non-monotonically. But we shall do no more here than assume that there is some contextually given and pragmatically constrained Ininimal distance. With

(26)

and

(27),

the plane stays during its whole motion outside the

location which here is runway 4· Nevertheless, the displacement is not the same in the two sentences. In (26), the plane touches down and thus finally comes in contact with the runway, whereas in touching it.

(27)

it comes near the runway without

In order to be able to take into account and to represent formally these differences, we propose seven relations (cf Section

4.2).

Four of these relations

encoding how the space is structured by verbs of change oflocation are sirnilar to the ones presented byJackendoff& Landau ( 1 992) which encode the space as it is structured by spatial prepositions. For example, the preposition describes the interior of the reference object;

against

inside

describes an external

contact with the reference object; near describes an exterior close by with respect to the reference object; finally for describes a far away exterior with respect to the reference object. As we have come to these distinctions by examining motion verbs, we conclude that language structures space in the same way whatever son of lexical items (motion verbs or spatial prepositions) we examine. We talk about (spatiaVspatiotemporal) relations and not about (spatiaVspatiotemporal) zones1• below for the following reason. Consider, for example,15 our plane and our runway in the sentences (26) and (27). If we want to define a zone of contact between the plane and the runway, we have to construct this zone as the union of all the spatial positions the plane can occupy so long as it is, in one way or another, in contact with the runway. This will result in a zone whose height is equal to the longest dimension of the plane (and

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

scene, the perceptual (visual, auditory . . .) capabilities in the case of human beings, and other elements of the context. The critical distance will be different ifwe know that Abby intends to go away by foot or by car, for example. It is very

Nicholas Asher and Pierre Sablayrolles

1 77

so a tone determined by imagining the plane in a vertical position and its nose in contact with the runway). It is then easy to place in such a zone the plane in its canonical position at, let's say, one metre up from the ground. So, the plane is inside the zone of contact but definitely not in a relation of contact with the runway. In conclusion, it would be false to say that the plane is in contact with the runway just because it belongs to the zone of contact. Belonging to the zone of contact is a necessary but not sufficient condition to be in contact. 4.2

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

We introduce our relations discursively (Definitions 9. 1-1 5.1) and offer a formal definition (Definitions 9.2- 1 5.2). In the logical formulation, we make use of the mereology of Lesniewski ( 1 92 7-3 1 ), formalized and extended with topological structure by Clarke (198 1 , 1 98 5). 1 6 In Arnague & Vieu ( 1 993), the reader can find an extension of Clarke's topological mereology and an application to the formal representation of the semantics of space and space time in natural language. Definition 9 x is in a relation ofinner-halo with y iff 1 . x is strictly in thefunctionafl1 inside ofy, i.e. without sharing any ofits points with the fro ntier ofy 2. inner-halo {x, y) =def PP(x,fint(y, x}} Definition I O x is in a relation ofcontact with y: 1 . x is in thefunctional outside ofy and is weakly connected toy, i.e. x andy are in contact but share no points (Aurnague and Vieu talk about a 'natural contact' in Aurnague & Vieu (1993) 2. contact (x, y) =tkf PP(x,fext(y, x)) A weak- contact(x, y) Definition I I x is in a relation of outer-halo with y: 1 . x is in thefunctional outside ofy but at a distance lesser than the critical distance, i.e. x is 'in proximity' ofy 2. outer-halo {x, y, C) =def PP(x, prox{y, x, C)) Definition 1 2 x is in a relation ofouter-most with y: 1 . x is in theJunctional outside ofy and at a distancegreater than the critical distance, i.e. x is not 'in proximity' ofy 2. outer-most (x, y, C) =d�J PP(x,fext(y, x}} A -. PP(x, prox{y, x, C)) Definition I3 x is in a relation ofinner-transit with y: 1. x shares at least one ofits points with thefrontier ofy 2. inner-transit(x, y) =dif EC(x, y) --

--1

178

A Typology and Discourse Semantics for Motion Verbs and Spatial PPs in French

Definition 1 4 x is in a relation ofcontact-transit with y: 1 . x shares at least one of its points with the frontier delimiting the foct to be in contact with y and thefoct to be in an outer-halo relation with y 2. contact-transit (x, y) =J4 contact (x, y) 1\ outer-halo(x, y) Definition 1 5 x is in a relation ofouter-transit with y: 1 . x shares at least one ofitspoints with thefrontier delimiting thefoct to be in a relation of outer-halo with y and thefoct to be in a relation ofouter-most with y 2. outer-transit(x, y) =dtf outer-halo(x, y) 1\ outer-most(x, y)

With our relations defined in Section 4.2, we have at our disposal seven generic locations linked to a given referent (see Figure I ). Talk about locations, however, suggests zones and not relations. We have shown why such zones cannot be defined simply relative to the reference location but rather must be defined relative to the referent, the target, and some relation of spatial inclusion.18 These zones must obey certain constraints given by the relations we have introduced. But we can still define the relevant zones by the following axiom scheme I . Let R be any one of our seven generic relations, and Z - R the corresponding zone. Axiom

1

R (x , y) - P (x , Z - R (x, y )) 'proximity limits ' - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - · ' . ' .

Z-contact-transit Z-contact

Reference Location Z-outer-most

Z-inner-halo

Z-inner-transit Z-outer-halo •-

- - -----�

�---- - - - - J .

Z-outer-transit

Figure 1 The seven generic locations

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

4·3 A subclassificationfor verbs ofchange oflocation

Nicholas Asher and Pierre Sablayrolles

1 79

Note the use of an implication and not an equivalence relation in this axiom scheme. Now for the definition of the relevant R-zone, we appeal to a notion of physical possibility that is quite restricted; we are interested in the possible positions ofx with respect to y without any deformation or change in the shapes or volumes of x and y. We also suppose that the reference location y then does change. Axiom 2

Z - R (x , y) � U {z : <>(z

�

STref(x ) 1\ R (x , y ))}

1. Name of the group: a'approcher to approach Logical definition: Approach(e) -+ {P(Source(e),Z-outer-most(cible(e),Lref(e))) 1\ P(SIP(e),Z-outer-transit(cible(e),Lref(e))) 1\ P(Goal(e),Z-outer-halo(cible(e),Lref(e)))} Lexical entries: 6 Examples: s' avancer to moveforward ; accou rir to rush up Explanation: Such verbs describe a motion going from a far away outside to a near outside of a location of reference. 2. Name of the group: Arriver to arrive Logical definition: Arrive(e) -+ [P(Source(e).z-outer-mos«cible(e),Lreqe))) 1\ P(SIP(e).z-outer-halo(cible(e),Lreqe))) 1\ P(Goal(e),Z-inner-halo(cible(e),Lref(e)))} Lexical entries: 2 3 Examples: aller to go ; venir to come Explanation: Such verbs describe a motion going from a far away outside to the inside of a location of reference, via a near outside of this location.

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

Our axiom scheme and definition of zones yields the picture of their interrelation as depicted in Figure I. In this section we will concentrate on the verbs of change oflocation. The zones we have so constructed can be considered as locations; they have functionality and they can, even if rarely for some, be lexicalized in different ways. We are then able to subclassify these verbs on the basis of which locations the moving entity is inside, at the beginning, the middle, and the end of its motion, respectively. Not all the possibilities are lexicalized in French. We have listed 2 1 6 lexical entries for French intransitive verbs of change of location, which can be dispatched in ten groups. Nine of them correspond to a change of location in the pure sense of location. The tenth one is a little more special and we will return to it at the end of this section.

1 80

3·

4·

A Typology and Discourse Semantics for Motion Verbs and Spatial PPs in French

s. Name of the group: S'eloigner to distance oneselffrom Logical definition: Distance-from(e) -+ {P(Source(e),.Z-outer-halo(cible(e),Lref{e))) I\ P(SIP(e),Z-outer-transi«cible(e),Lref(e))) I\ P(Goal(e),Z-outer-mos«cible(e),Lref(e)))} Lexical entries: 8 Examples: s' isoler to become isolated ; se reculer to move back Explanation: Such verbs describe a motion going from a near outside to a far away outside of a location of reference. 6. Name of the group: Partir to leave Logical definition: Leave(e) -+ {P(Source(e),.Z-inner-halo(cible(e),Lref(e))) I\ P(SIP(e),.Z-outer-halo(cible(e),Lref{e))) I\ P(Goal(e),.Z-outer-mos«cible(e),Lref{e)))J Lexical entries: 44 Examples: s'en aller to go away; deserter to desert Explanation: Such verbs describe a motion going from the inside to a far away outside of a location of reference, via a near outside of this location. ·

7· Name of the group: Sortir to go out Logical definition: Go-ou«e) -+ {P(Source(e),.Z-inner-halo(cible(e),Lref{e))) I\ P(SIP(e),.Z-inner-transi«cible(e),Lref{e))) I\ P(Goal(e),.Z-outer-halo(cible(e),Lref{e)))J

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

Name of the group: Entrer to enter Logical definition: Entrer(e) {P(Source(e),.Z-outer-halo(cible(e),Lref(e))) I\ P(SIP(e),.Z-inner-transi«cible(e),Lref{e))) I\ P(Goal(e),.Z-inner-halo(cible(e),Lref{e)))J Lexical entries: 4 3 Examples: s'embarquer to board; pemetrer to penetrate Explanation: Such verbs describe a motion of going from a near outside to the inside of a location of reference, crossing its 'frontier'. Name of the group: Se poser to alight ; to land Logical definition: Land(e) -+ {P(Source(e),Z-outer-halo(cible(e),Lref{e))) I\ P(SIP(e),Z-contact-transi«cible(e),Lref(e))) I\ P(Goal(e),Z-contac«cible(e),Lref(e)))} Lexical entries: 44 Examples: se jucher to perch ;1 se suspendre to hang Explanation: Such verbs describe a motion going from a near outside of a location of reference to an external contact with this location of reference.

Nicholas Asher and Pierre Sablayrolles

I8I

Lexical entries: I 6 Examples: debarquer to land ; jaillir to spring (up) Explanation: Such verbs describe a motion going from the inside to a near outside of a location of reference, crossing its 'frontier'. 8.

9· Name of the group: Passer (par) to go through ; to cross Logical definition: Cross(e) -+ {P(Source(e),Z-outer-halo(cible(e),Lref(e))) 1\ P(SIP(e),Z-inner-halo(cible(e),Lref(e))) 1\ P(Goal(e),Z-outer-halo(cible(e),Lref(e)))} Lexical entries: 4 Examples: couper to cross; repasser (par) to pass by again Explanation: Such verbs describe a motion going from a near outside to a location of reference, entering the location, crossing it, and going outside to a near outside of this location. 1 0.

Name of the group: Devier to deviate Logical definition: Deviate(e) -+ [P(Source(e),Z-inner-halo(cible(e),Lref(e))) 1\ P(SIP(e),Z-inner-transmit(cible(e),Lref(e))) 1\ P(Goal(e),2-outer-halo(cible(e),Lref(e)))} Lexical entries: 1 7 Examples: devier to deviate; bifurquer to turn off Explanation: Such verbs describe a motion going from the inside to a near outside of an ideal trajectory.

Now, as promised, some words about the tenth group, i.e. the group of verbs such as Devier. In contrast with the other groups for which there is a change of location in the pure definition of the term location, here the location considered is a little special. The location does not consist in a 'normal' portion ofspace, but rather in an ideal trajectory, or ideal path which might be captured by means of aXioms using >. This is reminiscent of the English progressive in which we similarly appeal to 'ideal' or 'inertial' paths. With the progressive,

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

Name of the group: Decoller to take off Logical definition: Take-off(e) -+ [P(Source(e),2-contact(cible(e),Lref(e))) 1\ P(SIP(e),Z-contact-transit(cible(e),Lref(e))) 1\ P(Goal(e),2-outer-halo(cible(e),Lref(e)))} Lexical entries: I I Examples: se decoller to come unstuck ; deconnecter to disconnect Explanation: Such verbs describe a motion going from an external contact with a location of reference to a near outside of this location.

1 82 A Typology and

Discourse Semantics for Motion Verbs and Spatial PPs in French

such paths can be naturally captured in a non-monotonic formalism {Asher 1 992a).19 We propose the schema depicted in Figure 2 for our ten groups of verbs of change oflocation. 'proximity limits ' � - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - ' '

Passes

Entrer to enter

Z-outer-most

S'eloigner to distance oneselffrom

Lref

·

·

·

�

·

·

·

Partir to leave

Sortir

to go out

Z-outer-halo

'ideal trajectory ' �

Devier to deviate , I

Z-inner-halo

Z-outer-most

Figure 2 The 1 0 groups of verbs of change oflocation

As Tal my { 1 98 3) noted, motion complexes, and for us verbs and prepositions, can focus on particular spatial portions of the STref of a motion. We will make use of (especially in Section 4·5) the Initial, Median, and Final polarities defined in Boons ( 1 98 5) and Laur ( 1 99 1 ). Items with an initial polarity focus on the source of the motion, items with a median polarity on the SIP, and items with a final polarity on the goal of the motion. Our groups S'eloigner, Partir, Sortir, Decoller, and Devier have an Initial polarity. Our groups S'approcher, Arriver, Entrer, and Se poser have a Final polarity. Our group Passer has a Median polarity.

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

S'approcher to approach . . . ----i>+- . . . . : Arrive to arrive

Nicholas Asher and Pierre Sablayrolles 1 8 3 4·4 A

classificationfor spatial prepositions

Prepositions

inner-halo

contact

outer-balp

outer-most

Positional

chez; dans at; in

sur; conrre on ; against

sous; derriere; a below; behind ; at

loin de Jar awayfrom

Initial directional

de chez from -'s

de sur from on to

de derriere from behind

de dehors from the outside

Medial positional

par through

au fil de

le long de along

au-deJa de beyond

jusque dans up to the inside of

jusque sur up on to

vers towards

pour for

Final positional

FigUre 3

Classification of French spatial prepositions

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

As we have said from the outset, we are interested in the compositional determination of the spatiotemporal properties of motion complexes consisting of a verb of change of location and a spatial preposition. In order to formulate rules of combination (c£ Section 4.5), we briefly present in this section a classification for French spatial prepositions. It is based on the classification proposed in Laur ( 1 991), which we have refined using the spatial relations defined in Section 4.2. Under the term 'preposition', following Laur ( 1 991), we consider simple prepositions (such as dans in ) and complex prepositions, i.e. prepositional phrases (such as en face de infront of). We separate prepositions into two main groups: the positional pre positions, which just describe a relation of localization (such as dans in ), and the directional prepositions, which in addition to a relation of localization also suggest a direction of motion (such as de from , which has an Initial polarity, par through , which has a Median polarity, or jusqu'a to , which has a Final polarity). We set out this classification in Figure 3· In all, 1 99 French prepositions have been listed and classified. We nevertheless have to remark that such a classifica tion using the spatial relations dedicated to CoL verbs is not as refined as we would like it to be. For example, sous below and derriere behind could be differentiated considering the intrinsic orientation of the entity; others require a distinction using the vertical dimension (i.e. the axis of gravity). At present, we are investigating such refinements for motion verbs in a decompositional semantic study (Asher & Sablayrolles, forthcoming) and we intend to generalize them for spatial prepositions.

1 84

A Typology and Discourse Semantics for Motion Verbs and Spatial PPs in French 4· 5

In providing rules for lexical · entries, we must extend our formal language. Roughly, we will be going from a syntactic parse of a motion complex to the semantic conditions introduced in the DRS. We avoid any issues here about incremental parsing, and in any case we will need information both about the verb and its arguments to introduce the appropriate semantic conditions into the DRSs. To expand our formal language, we will use ¢ as a variable for motion complexes from which we may isolate the verb and propositional phrase adjuncts; though the fragment we will codify here only makes use of one prepositional phrase adjunct, our language will be general enough to handle more prepositional phrases. v will range over verbs in these complexes while n will range over prepositions. Polarity will be used to denote the polarity of the verb (Polarity (v)) or of the preposition (Polarity(n)); IZone(v, mobile(e),l), MZone(v,mobile(e),l) and FZone(v,mobile(e),l) will denote the kind of initial, median, and final zone intrinsically implied by the verb, the mobile, i.e. the moving entity, and a location, which may depend on e, the eventuality that it is an argument to the motion verb of the motion complex. Prep Zone(n,mobile(e),l) will denote the kind of zone intrinsically implied by the preposition and its arguments; Ns(n) will denote the location introduced by the noun phrase in the PP of the motion complex; the question mark (?) will denote a location that will have to be matched with some location given by the discourse (in previous or following sentences) by anaphora resolution in DRT (Asher & Wada 1 989). Finally the DR-theoretic lexical entry for an item v will be denoted ii. Figure 4 illustrates, for each of our ten groups of verbs of change of loca tion, the polarity Polarity (v) and the zones in which the moving entity is located at the initial (variable IZone (v,mobile(e),l)) median (variable MZone(v,mobile(e)), and final (variable FZone(v,mobile(e),l)) stage of the groups displacement, respectively.20 These zones will be related to the Source( e), SIP( e) and Goal(e) by axioms later on in this Section. We show in Figure s. for each group of directional prepositions the polarity (Polarity(n)) and the spatial 'zone' it suggests (PrepZone(n,mobile(e),l)). This 'zone' (c£ remark made in Section 4.3) is a function of the moving entity, the location introduced by the preposition and the preposition itself Finally, we show in Figure 6, for each group of positional prepositions the spatial 'zone' (PrepZone(.n,mobile(e),l)) it suggests. Now we propose the following compositional rules (under the form of logical axioms), with some discursive explanations. In Axioms I I and I 2, > denotes the weak conditional of DICE used in Asher et al. ( 1 993), Lascarides & Asher ( 1 993), and Asher ( I 992b). As in the treatment of 'Discourse in Commonsense Entailment' or DICE developed in Lascarides & Asher ( 1 993),

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

1-

Compositional rules

Nicholas Asher and Pierre Sablayrolles 1 8 s

Groups of verbs of change of location Decoller Sortir Partir S'eloigner Devier

MZone

FZone

Z-contact Z-inner-halo Z-inner-halo Z-outer-halo Z-inner-halo Z-outer-most Z-outer-most Z-outer-halo Z-outer-halo Z-outer-halo

Z-contact-transit Z-inner-transit Z-outer-halo Z-outer-transit Z-inner-transit Z-outer-transit Z-outer-halo Z-inner-transit Z-contact-transit Z-inner-halo

Z-outer-halo Z-outer-halo Z-outer-most Z-outer-most Z-outer-halo Z-outer-halo Z-inner-halo Z-inner-halo Z-contact Z-outer-halo

I

F F F F M

Data for French verbs of change of location

Figure 4

Directional prepositions

Polarity

PrepZone

I F F F F M M M M

Z-inner-halo Z-contact Z-outer-halo Z-outer-most Z-inner-halo Z-contact Z-outer-halo Z-outer-most Z-inner-halo Z-contact Z-outer-halo Z-outer-most

de chez de sur de derriere de dehors jusque dans jusque sur vers pour par au 6l de le long de au-dela de Figure 5

Data for French directional prepositions

Positional preposition

PrepZone

chez; dans sur; contre sous; derriere loin de

Z-inner-halo Z-contact Z-outer-halo Z-outer-most

Figure 6

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

S'approcher Arriver Entrer Se poser Passer

!Zone

Polarity

Data for French positional prepositions

--

-- ---

-- -

- - - - -- - ----

1

1 86

A Typology and Discourse Semantics for Motion Verbs and Spatial PPs in French

Asher ( r 993), and Lascarides & Asher ( r 99I), we will later label consituents of a discourse representation as having certain properties, and we will reason about lexical properties within Commonsense Entailment. ·

Axiom 3

( Verb (v, � ) A Change-oflocation (v) A ii(e)) .... P(Source(e), IZone (v, Mobile(e), Lref(e))) Axiom 4

( Verb (v, � ) A Change-oflocation (v) A v(e)) .... P(SIP(e), MZone(v, Mobile(e), Lref(e))) ( Verb ( v, � ) A Change-oflocation ( v) A v( e)) .... P(Goal(e), FZone(v, Mobile(e), Lref(e)))

Axioms 3, 4 and 5 say that.the Source, Path, and Goal of a motion are always a part of a zone given by the verb via !Zone, MZone and FZone, and processed with respect to the reference location Lre£ Axiom 6

( Verb (v, � ) A Change-oflocation (v) A v (e) A Preposition (n , � ) A Directional_Prep (n) A Polarity (v) = Polarity(n)) .... Lref(e) = PrepZone (n, Mobile(e), strej(Ns (n))) Axiom 6 says that when the preposition is directional with the same polarity as the verb (e.g. sortir de Ia maison to go out ofthe house) the reference location (Lref(e)) is a zone given via PrepZone (Z-inner-halo, in our example) and processed with respect to the Ns (Ia maison in our example). Compare that, for example, with sortir de derriere Ia maison to go out ofthe back ofthe house , where the reference location (Lre�e)) cannot directly correspond to the location introduced by the noun phrase (here Ia maison), but has to be built using the PrepZone (here Z-outer-halo for the preposition de derriere) to give the zone of the back ofthe house . Axiom 7

( Verb ( v , � ) A Change-of-location ( v) A ii (e) A Preposition (:rc , � ) A Directional_Prep (n) A Polarity(v) � Polarity (n) A Polarity(n) = I) .... (P(Source(e), PrepZone(n, Mobile(e), strej(Ns (n)))) A Lref(e) = ?)

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

Axiom s

Nicholas Asher and Pierre Sablayrolles

187

Axiom 8

-+

[ Verb ( v, ¢ ) 1\ Change-of-location ( v) 1\ ii( e) 1\ Preposition (n, ¢) 1\ Directional_Prep (:rr ) 1\ Polarity( v) � Polarity(n) 1\ Polarity(n) - M) (P(SIP (e), PrepZone(:rr , Mobile(e ), strej(Ns(:rr )))) 1\ Lref(e) - ?)

Axiom 9

-+

[ Verb ( v, ¢ ) 1\ Change-of-location ( v ) 1\ ii (e) 1\ Preposition (n, ¢ ) 1\ Directional_Prep ( n) 1\ Polarity (v) � Polarity(n) 1\ Polarity(n) - F) (P( Goal (e), PrepZone(n) , Mobile(e ) strej(Ns(:rr )))) 1\ Lref(e) - ?) ,

Axiom I O

-+

[ Verb ( v , ¢ ) 1\ Change-oflocation ( v) 1\ ii (e) 1\ Preposition (n , ¢ ) 1\ Positional_Prep (n) 1\ Polarity ( v) - I) (P( Goal (e), PrepZone(n) , Mobile (e ), strej(Ns (:rr )))) 1\ Lref(e) - ?)

Axiom 1 0 says that when the preposition is positional and the verb Initial (e.g. sortir dans togo out in/into ) the Goal is a part of a zone given via PrepZone, and processed with respect to Ns; the reference location remains unmatched. Axiom I I

[ Verb ( v , ¢ ) 1\ Change-oflocation ( v ) 1\ ii( e) 1\ Preposition (n, ¢) 1\ Positional_Prep (n) 1\ Polarity( v ) - M 1\ :J ((Info (a , {3) 1\ ( r , a , {3) 1\ Lref(e) - PrepZone(n , Mobile(e), strej(Ns (n)))), R (a , {3) 1\ R � Narration 1\ R � Result) 1\ --. :J ((Info (a , {3) 1\ ( r , a , {J)), R (a , b ) 1\ R � Narration 1\ R � Result)) > Lref(e) - PrepZone(n , Mobile (e), stref(Ns (n)))

Axiom I I encodes a complex interaction between lexical inference and discourse context. It is of the sort we have found common in other areas (Lascarides & Asher I 993). When the preposition is positional and the verb medial (e.g. courir dans to run in ), and if the fact that the reference location is the zone given via PrepZone and processed with respect to Ns enables us to deduce

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

Axioms 7, 8, and 9 say that when the preposition is directional with a polarity different from the verb (which is Initial (e.g. sortir par to go out through ), Medial or Final) the Source, Path, or Goal, respectively, is a part of a zone given via PrepZone, and processed with respect to the Ns; the reference location remains unmatched.

1 8 8 A Typology and Discourse Semantics for Motion Verbs and Spatial PPs in French

in CE a discourse relation R, other than Narration or Result, that we cannot deduce without this information, then we infer that the location of reference is so determined. :l encodes in the object language of CE the non-monotonic consequence relation F· Axiom 12

Axiom I 2 is similar to Axiom I I. When the preposition is positional and the verb Medial (e.g. courir dans to run into ), and if the fact that the Goal is a part of a zone given via PrepZone and processed with respect to Ns enables us to deduce in CE a discourse relation R, other than Narration or Result, that we cannot deduce without this information, then we infer that the Goal is so determined. The reference location remains unmatched. Axiom 13

( Verb ( v, �) 1\ Change-oflocation ( v) 1\ ii (e) 1\ Preposition (:rc , �) 1\ Positional_Prep (:rc) 1\ Polarity(v) = F) -+ Lref(e) = PrepZone (:rc , Mobile (e), stref(Ns (:rc))) Axiom I 3 says that when the preposition is positional and the verb Final (e.g. entrer dans togo into ) the reference location is the zone given via PrepZone, and processed with respect to the Ns. Having detailed in axiomatic form the lexical entries for verb classes and classes of prepositions and their combination, we propose to illustrate their application with the following example: (28), (29) and (3o). 28. Leticia est arrivee par le jardin Leticia has a"ived by the garden 29. Eva est sortie de derriere Ia maison Eva has gone out via the back ofthe house 30. Ines est passee sous le porche Ines has passed under the porch We will derive DRSs for these sentences, so we now briefly mention our views on the construction of DRSs and the syntax/semantics interface. We

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

( Verb (v, � ) 1\ Change-oflocation (v) 1\ ii(e) 1\ Preposition (:rc, �) 1\ Positional_Prep (:rc) 1\ Polarity ( v) = M 1\ :J ((Info (a . {J) 1\ (T, a , {J) 1\ Lref(e) = ? 1\ P(Goal(e), PrepZone (:rc, Mobile (e), stref(Ns (:rc))))), R (a , fJ) 1\ R i' Na"ation 1\ R i' Result) 1\ -. ((Info (a , {J) 1\ (T, a , {J)), R (a , fJ) 1\ R i' Na"ation 1\ R i' Result)} > (P(Goal(e), PrepZone(:rc, Mobile(e), stref(Ns(:rc)))) 1\ Lref(e) = ?)

Nicholas Asher and Pierre Sablayrolles 1 89

APAQ

X

P(x) Q(x) This structure says that the discourse referent introduced by the partial DRS above can fill in the argument position in a predicative DRS, which is introduced, for example, by the common noun jardin: AU

jardin(u) They combine to yield the following partial DRS for the noun phrase le jardin: AQ

X

jardin(x) Q(x) Before turning to the first example we should stress again that we take Mobile , Goal, Source , and SIP to pick out semantic arguments of a motion verb; we will use functions, e.g. mobile(e), goal(e), to refer to them. Now, as we have seen, sometimes other elements of the motion complex specifY what these various semantic arguments are. We are skeptical that syntax alone can determine which part of the motion complex specifies which argument in every case, especially with regard to spatial prepositional phrases. Certain lexical conditions or elements in the discourse context (some of which we have

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

employ the 'bottom-up' DRS construction procedure from Asher ( 1 993), in which the function argument structure of the syntax can be followed, if we ignore the problem of scopes of noun phrases and of anaphora resolution. With this approach, determiners are relations between dynamic properties, noun phrases a property of dynamic properties, while common noun phrases and intransitive verb phrases introduce dynamic properties. The account resembles Montague's intensional logic, but instead of formulas of intensional logic we use OR-theoretic notation to denote dynamic properties and propositions. A determiner, semantically, introduces a partial DRS, which when combined with the two dynamic properties yields the semantic value of a DRS (a set of embedding functions from worlds into the domains of those worlds); for instance, le introduces the following structure:

1 90 A Typology and Discourse Semantics for Morion Verbs and Spatial PPs in French

codified in the rules above) determine whether, for instance, the spatial prepo sitional phrase determines the goal or not. With this in mind let us consider example (28) within the bottom-up DRS construction procedure. We take a syntax of the form shown in Figure 7 with a VP internal subject (though this is not essential). Now we simply go back up the tree using the expansions given above of the conditions given by the lexical

� .� IP

I'

VP

A �estrA

•e

PP

VP

V'

Leticia

V

P

arrivee

par

I

Figure 7

P'

DP

I

I

ADP

I

D.

le jardin

The bottom-up DRS construction procedure

elements. In (28) arriver implies that Polarity(v) - F 1\ !Zone - Z-outer-most 1\ MZone Z-outer-halo 1\ FZone - Z-inner-halp . Then, using Axioms 3, 4, and 5 , we obtain the following predicative DRS from the verb: =

A.e.?.x P(Source(e), Z-outer-most(STref(mobile(e)),Lref(e))) P(SIP(e), Z-outer-most(STref(mobile(e)),Lre�e))) P(Goal(e), Z-inner-halo(stref(mobile(e)),Lre�e))) mobile(e) - x Then after conversion with the partial DRS derived from the DP Leticia, we following structure:

have the

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

I

Nicholas Asher and Pierre Sablayrolles 1 9 1

ux Leticia(x) mobile(e) - x P(Source(e), Z-outer-most(x,Lre�e))) P(SIP(e), Z-outer-halo(x,Lre�e))) P(Goal(e), Z-inner-halo(x,Lre�e)))

A.vA.PA.e'

u u = Z-inner-halo(STre�mobile(e ')),v) P(e ') Lre�e ') = ?

which, when combined with the translation of le jardin, yields:

A.PA.e'

uv u = Z-inner-halo(STre�mobile(e ')),STre�v)) P(e ') jardin(v)

Since the structure above has a lambda abstract for a property variable, it may combine with the property from the VP, given above, to give us: ux Leticia(x) mobile(e') = x P(Source(e'), Z-outer-most(x,Lre�e'))) P(SIP(e'), Z-outer-halo(x,Lre�e')))) P(Goal(e'), Z-inner-halo(x,Lre�e'))) u - Z-inner-halo(STre�mobile(e')),STre�v)) jardin(v)

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

In (28), we also have the directional preposition par which implies that Polarity(n) = M 1\ PrepZone = Z-inner-halo . Because of the directional preposition and Polarity(v) "' Polarity(n) and Polarity(n) = M, we can apply Axiom 8, which gives us: P(SIP(e), Z-inner-halo (stref(mobile(e)), strej(Ns(n)))) 1\ Lref(e) = ?. We will exploit the function argument structure of the syntax and propose the following translation of the preposition in this case:

1 92

A Typology and Discourse Semantics for Motion Verbs and Spatial PPs in French

Now we apply the partial DRS (givenjust below) derived from the inflection node in the syntactic tree that gives us the temporal conditions and introduces an eventuality filling in the event argument place in the partial DRS given just above.

euxn Leticia(x) mobil(e) = x P(Source(e), Z-ou ter-most(x,Lre�e))) P(SIP(e), Z-outer-most-halo(x,Lref(e))) P(Goal(e), Z-inner-halo(x,Lre�e))) u - Z-inner-hal(STre�mobile(e)),STref(v)) jardin(v) e
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

This is then the completed DRS for (28). Note that Lref(e) has not been determined and so musr be determined anaphorically.

Nicholas Asher and Pierre Sablayrolles 193

A.PA.e'

ux u - Z-outer-halo(STref(mobile(e)),STref{x)) Lref{e) = u maison(x)

When this is combined with the predicative DRS derived from the VP and the information from the inflection node in the syntactic tree, we obtain: evuxn

This says that Eva is in a zone of inner-halo, at the beginning of her displacement, with respect to a zone defined as the outer-halo21 of the house, in a zone of inner-transit, at the middle of her displacement, and in a zone of outer-halo, at the end, always with respect to the same zone. The last example shows us how our nonmonotonic axioms work in the framework. In (3o), passer implies that Polarity (v) = M /1. !Zone = Z-outer-halo /1. MZone = Z-inner-halo /1. FZone = Z-outer-halo . We have the positional preposition so us, which implies that PrepZone = Z-outer-halo . Because of the positional preposition and Polarity( v) - M, we have the choice of the two non monotonic Axioms I I and I 2, which give us, depending on the discourse context, two different translations for the preposition: Lref( e) = Z-outer halo (stref(mobile(e )), strej) (Ns (n ))), or P( Goal(e) Z-outer-halo (stref(mobile (e )), stref(Ns(n)))) /1. Lref(e) = ?. Using the construction procedure we have seen operating on the previous two examples together with Axioms 3, 4 and 5, we arrive at the following two DRSs:

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

Eva(v) P(Source(e), Z-inner-halo(STref(v),u)) P(SIP(e), Z-inner-transi�STref(v),u)) P(Goal(e), Z-outer-halo(STref(v),u)) mobile(e) = v u = z-outer-halo(STref(v),STref(x)) Lref(e) = u maison(x) e
1 94 A Typology and Discourse Semantics for Motion Verbs and Spatial PPs in French

evuxn Ines(v) porche(x) P(Source(e), Z-outer-halo(STreqv), u)) P(SIP(e), Z-inner-halo(STreqv),u)) P(Goal(e), Z-outer-halo(STreqv),u)) mobile(e) = v u = z-outer-halo(STref(v),STref(x)) Lreqe) - u e
evxn Ines(v) P(Source(e), Z-outer-halo(STreqv),Lref(e))) P(SIP(e), Z-inner-halo(STref(v),Lref(e))) P(Goal(e), Z-outer-halo(STref(v),Lref(e))) P(Goal(e), Z-outer-halo(STref(v), STref(x))) mobile(e) v Lref - ? porche(x) e
We thus obtain two interpretations for (3o), both possible in French: the first one, the median interpretation, where Ines is in front of the porche, goes through it, and reaches the other side, the second one, the final interpretation, where the location sous le porche is the ending point of her displacement. 4.6

Linear and nonlinear motions in expresions

Along any of the dimensions of a change of location, a change of position or a change of posture, we can distinguish motions that are 'linear' from those that are 'nonlinear' or circular. Roughly, circular motions are those in which there is a sequence of locations in the path of the eventuality such that the sequence contains a repeated element. But there are various kinds oflinear and nonlinear motion and these motions must be calculated with respect to the types of verb and prepositions if any in the adjunct. Luckily, we can use the same definitions of linearity for

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

or:

Nicholas Asher and Pierre Sablayrolles

195

prepostnons, verbs and verb complexes. Below we illustrate the relevant definitions of linear and nonlinear motion just for motions that are changes of location. The definitions for changes of position or posture are entirely analogous. De6.nition 1 6 A change of location verbal complex or verb, or a preposition � is strongly nonlinear iff� (e) ..... Source(e) Goal(e) �

De6.nition 17 A change of location verbal complex or verb, or a preposition � is weakly nonlinear iff � (e) -+3x(x E SIP(e) 1\ 3/;3� (I; occurs in the sequence x 1\ � occurs in the sequence x 1\ -.i j 1\ I l)J �

�

De6.nition 19 A change of location verbal complex or verb, or a preposition � is weakly linear iff�(e) ..... Source(e) i' Goal(e) As an example of a nonlinear preposition, we have autour de round . Weakly linear prepositions are jusqu'a to , de puis from ; for verbs we have weakly linear verbs like partir to leave, quitter to leave , aller to go , venir to come , while weakly nonlinear verbs are boucler to loop and graviter to orbit. There are no prepositions that are strongly linear, and for strongly linear verbs we have only potential candidates like tomber tofoil or manter togo up , and even with those it is not clear that their paths never cross. One can only get the effect of a strongly nonlinear motion in a verbal complex like John went to Paris and then returned , which we would in fact analyze as two motions. Strong linearity also is only found in complex expressions that are rather standardized as in the airplane climbed continuously at 1 ooofi/min for 30 seconds. If one sees this in an NTSB accident report, one would conclude that the airplane's path was strongly linear in this case. Some verbs of change of position like graviter to orbit are nonlinear though most others are weakly linear like the displacement verbs. Some verbs of inertial change of place like courir to run do not make any demands on the locational path of the eventuality they describe-they are neither linear or nonlinear. Just as we saw that one could calculate the position for the target compositionally on the basis of the properties of the verbs and prepositions, we can also easily calculate the properties of linearity of verb complexes based on the properties of the prepositions and constituent verbs. Roughly, the calculation is simply a matter of unification. If a verb or preposition is linear it may not combine with something with a nonlinear value. Again we can use the notation here of a first order language in which we reason declaratively about constraints on constituents. Because there are few if any strongly linear motion

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

De6.nition 18 A change of location verbal complex or verb, or a preposition � is strongly linear iff � (e) ..... -.3x (., E SIP(e) 1\ 31; 3� (I; occurs in the sequence x 1\ � occurs in the sequnce x 1\ i i' j 1\ I i' �))

1 96 A Typology

and Discourse Semantics for Motion Verbs and Spatial PPs in French

4.7

Discourse Structure and Discourse Semantics

We are now in a posicion to combine our lexical informacion with discourse informacion. We follow the studies ofAsher (1 992b) and Asher eta/. ( 1 993). We will work within the framework of Segmented Discourse Representation Theory (Asher 1 993). SORT consists in a nontrivial extension of Kamp's Discourse Representation Theory in which the dynamic propositions (DRSs) that can be derived from each clause in a discourse are related to each other by means of one or more discourse relations. The resulting structure is known as an SDRS (segmented DRS); it is recursively defined and may have SDRSs as constituents. SDRSs are constructed through a bottom-up procedure that is described at length in Asher (1993) and investigated for spacial applications in Asher et a/. ( 1 993). Here we will only look at simple two-sentence discourses to see how our more detailed lexical information interacts with the discourse structure. Asher et a/ . ( I 99 3) have at least preliminary results for more complex texts. To do this, we must introduce a crucial element that links the lexical semantics of the motion verb complexes with the discourse structure. This is the notion of a constituent salient location or CSL for each moving actor in the movement described. Each will have as well a constituent salient posicion and posture but here we concentrate only on the constituent salient location or CSL for the one target in our motion verb complexes. Tense affects the CSL; here we consider the CSL only for sentences in the simple past (passe simple or

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

verbs or prepositions we look only at the weakly linear cases. After each axiom we offer as example a verb complex governed by it. We give axioms that cover all the change of location verbs and chan�e of place verbs. Axiom 14 Non-linear{Verb) 1\ Weakly-Linear{Preposition) ... .l Example: "graviterjusqu'a Mars to orbit to Mars Axiom 15 Non-linear{Preposition) 1\ Linear{Verb) -+ .l Example: "aller autour du Gers to go round the Gers Axiom 16 Weakly-linear{Preposition) 1\ Weakly-linear{Verb) ... Weakly linear{Verb-complex} Example: allerjusqu'a Paris to go to Paris Axiom 17 Non-linear{Preposition) 1\ Non-linear(Verb} -+ Non-linear(Verb complex} Example: graviter autour d'une planete to orbit round a planet

Nicholas Asher and Pierre Sablayrolles

1 97

passe compose with the passe simple sense); for a more complete smdy see Asher & Bras (I 994, forthcoming). The CSL in our restricted srudy is the spatiotemporal location of the target once the action has been completed. Thus, we can formulate Axiom 1 8 relative to our verb complex types where the verb has the passe simple tense (PS). Axiom 18 PS(a ) - CSL (a )

=

Goal (ea)

Axiom 19 Narration (a , {J) - (Target(ea) = Target(ep) - Source (ep) = CSL (a )) Narration thus links the actions described in two constiruents together. With backgrounding, the location of the background location forms a setting for the action that is foregrounded. Axiom 20 Background (a , fJ) - O(CSL ({J),CSL (a )) Axiom 21 (Background (a , fJ) /\ Motion-Verb (a )) > P(CSL (a ),CSL ({J)) With elaboration, the CSL of the constituent that is elaborated contains the CSL of the elaborating constiruent, while with precondition and explanation the element that is the explanans or the precondition is part of a causal sequence of eventualities itself defining an event leading to the CSL of the constituent that is explained or that has the precondition. With result, we have the inverse relation between constiruents as with explanation. Axiom 22 (Elaboration (a , {J) /\ -.3yNarration (y, fJ)) > Source(ep) Source(ea) Axiom 23 Elaboration (a , fJ) - Vx E Path (ep)3y e Path (ea)Subsequence(x, y) Axiom 24 Precondition (a , fJ) - 3e3x E Path ( e )31;( I; occurs in the sequence x /\ P(STref(ep).Stref(/;)) 1\ goal(e) CSL (a )) Axiom 25 Explanation (a , {J) - 3e3x E Path (e)Vy E Path (ep)(Subsequence(y, x) /\ goal(e) = source(ea)) Axiom 26 Explanation (a , fJ) > goal(ep) source(ea) Axiom 27 Result(a , fJ) - 3e3y E Path (e)Vx E Path (ea)(Subsequence (y, x) /\ goal(e) source(ep)) Axiom 28 Result( a , fJ) > goal(ea) source(ep) =

=

=

=

=

Now we look at examples of Discourse Srrucrure Calculation. We first begin with some examples of Narration. Consider the following examples:

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

The axioms governing the various zones and the reference point will help us siruate the CSL relative to locations mentioned in the constiruent a . On the discourse side, we have several discourse relations that can affect the relationships of the CSLs of one constiment with another. Axioms I 9-28 come again from the srudies of Asher (I992b) and Asher et al . (I993), in which the discourse relations examined were Narration, Background, Precondition, Elaboration, Explanation, and Result. Their definitions and semantics can be found in Asher et a/. ( I 99 3) and Asher ( 1 99 3 ).

198 A Typology and Discourse Semantics for Motion Verbs and Spatial PPs in French

3 I . Pierre est aile dans le jardin. II a danse. Pierre isgone into thegarden. He danced. 32· Pierre est aile dans le jardin. Il s'est assis. Pierre isgone into the garden. He sat down. 3 3· Pierre est aile dans le jardin. Puis il a couru jusqu'a la fac. Pierre is gone into thegarden. He then ran to the Fac.

�

34· Jean a couru dans le jardin. II a vu le chat a travers la fenetre et a voulu l'attraper. John ran into thegarden. He saw the cat through the window and wanted to catch it. 3 5· Jean a couru dans le jardin. Il a voulu s'entrainer pour la competition. John ran in thegarden. He wanted to ilainfor the competition. In both cases (34) and (3 s), we conclude Explanation, rather than Narration because of a particular juxtaposition of event types: an action by the target,

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

The deduction of which discourse relation relates the two constituents takes place as in Asher et al. { I 993) within a nonmonotonic framework, which was first elaborated in Lascarides & Asher (I991 , I 993). Narration is the discourse relation that we infer by default between constituents when no other information overrides this default. This is the case here. In example (3 I ), the first clause has a motion verb complex that fits the antecedent of Axiom I 3, which makes the garden the location of reference. Thus, we have, through Axiom s. that the goal of the action is situated in the garden, and we know, through Axiom I 8, that the CSL is in the garden too. Axiom I 3 also disambiguates dans for us, since it in effect makes the preposition directional, specifying the goal. Since Narration holds between the two constituents, we can apply Axiom I 9 and arrive at CSL(a) - source(ep)· But since the verb in f3 is not a verb of change oflocation, we conclude that the source(ep) goal(ep). i.e. they occur in the same location, which is specified by the previous clause. This entails that he danced in the garden, which is what is desired. The same holds for the second discourse (32), except that we know that s'asseoir to sit down is a change of posture verb; if we assume that changes of posture and changes of position do not change location, then Axiom I 9 for Narration yields that he sat down in thegarden . Finally, the sequence of motion verbs in the third discourse (3 3), assuming once again that Narration holds between the two constituents, shows us that the source of the running is in the garden, but with the help of Axiom I 7 we know that the goal of the running is not just a simple inertial change of position but a displacement to another location- the Fac . Narration is not the only discourse relation that can be inferred in discourse contexts, of course. The following pair shows how we can disambiguate prepositions using the discourse context and particular discourse relations.

Nicholas Asher and Pierre Sablayrolles

1 99

followed by a statement about the target's desires. We believe it to be a general default pattern that when such event types are joined together in the discourse that the intended or preferred discourse relation is Explanation. We might express this as: Axiom 29

[( T, a , {3) 1\ Action-by-x(a ) 1\ volition-of-x-expressed (/3)] > Explanation (a , {3)

Axiom 30 (r, a , {3) > Narration (a , {3) Explanation and Narration are incompatible relations because they imply contradictory temporal information; but the more specific default about Explanation overrides the more general one about Narration and so we conclude Explanation in DICE (see Lascarides & Asher I 993 and Asher et al. I 993 for details). Notice, however, that we have two different inferences about dans in these examples. How is this possible? We have in the first sentence of(34) and (35) a verb of inertial change of location (meaning that it is highly likely that the target will change position from the pre to the post state of running but it is not obligatory). Such verbs may combine with either the locational in or directional into sense of dans, as we see and as our axioms predict. What does disambiguate dans is the second sentence, which supplies an explanation. Let us consider (34) first. The second sentence of (34) is itself complex; the first clause gives a cause of and results in the volition in the second clause. Now volitional expressions about x described in a constituent f3 are typically taken in the absence of other information to contribute explanatory causes of actions by x described in a if the two are related in the discourse and further a must be part of a plan to get the goal described in f3 . That is, we have the rule: Axiom 31

{3) 1\ volition-by-x(f3) 1\ action-by-x(a ) 1\ Cons (Post-state( e13),Pre-state ( ea))] > {Explanation (a , {3) 1\ a part-oj-a-plan-that-leads-to-goal-ofvolition ({3)) [(r,

a,

These notions involving plans can be made formally precise, but the inference is quite complicated. Only if a is a displacement from inside the salon to outside the salon and the enclosure (house or apartment) can it be taken

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

where > is the weak conditional of DICE used in Asher et a/ ( I 99 3), Lascarides & Asher ( I 993), and Asher ( I 992b). We will revise this rule shortly, but already this approximation serves to indicate how inferences in DICE exploit defaults. The antecedent of this conditional entails the more general antecedent to the default for Narration given below:

200

A Typology and Discourse Semantics for Motion Verbs and Spatial PPs in French

s

S U RVEY O F THE L I TERATURE

In this last section, we offer a discussion of the literature on the subject and a comparison with our own work. Two different approaches can be found in the literature: a syntactic one (followed among others by Wunderlich, Maienborn, Guillet, etc.), and a semantic one (followed among others by Hays, Dervillez-Bastuji, Lamiroy, Boons, Laur, etc.). These approaches are quite different but they also share some common points. s.I

Syntactic approaches

Wunderlich ( 1 991) distinguishes between spatial verbs (as in (36)), which are subcategorized for a spati:1l PP,22 and non-spatial verbs, which are not. He classifies spatial verbs into three categories: those which require a locative PP (c£ (36), (37), and (3 8)) which he calls verbs oflocation; those which require only directional prepositions (c£ (39) and (4o)) which he calls verbs of placement; and those which can accept all dynamic prepositions (c£ (41)) which he calls verbs of movement. He consequently groups together as verbs oflocation all verbs for which there is no change oflocation, that is, all our verbs of change and inertial change of position, of change of posture and all the static verbs (like those in(36)). This group is very large and quite hetero geneous. In contrast, his verbs ofplacement are very few and represent those verbs of change oflocation whose polarity is final and which require an agent to realize the change of location of the object. Finally, his class of verbs of movement is not very precisely defined and seems only to be used for all the verbs which cannot be placed into one of the first two classes. They seem to

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

to be part of a plan of catching the cat. For in order to catch the cat one has to go where the cat is, and the cat is, as is stated in fJ, outside the enclosure while the target Jean is inside the enclosure, something which we know from the first clause of the second sentence from the semantics of a travers through . By Explanation and the axioms governing the CSL, we conclude that the source of courir to run must be inside the enclosure, and in attempting to make the action of running part of the plan to get the cat, we infer that the goal of the running is the garden and thus disambiguate the preposition dans. The same explanation goes for (3 5). The second sentence of the second discourse gives us again an explanation in terms of plans. But a quite different plan is involved. Again depending on extralinguistic information about plans and also about intentional states, we can obtain different semantic effects.

Nicholas Asher and Pierre Sablayrolles

20 1

correspond to our verbs of change of location except those which are verbs of placement, but we also find in the examples he gives verbs like wand ern to hike which are verbs of change of position for us. The syntactic criteria used here appear not to be completely adequate for the kind of classification we need. 36. 37· 38. 39· 39· 40.

The ball is under the tree She plays under the tree The ball is rolling under the tree I in the room The ball is rolling under the tree I into the room He puts the ball under the tree He went into the kitchen Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

Maienbom ( I992) distinguishes between Positionsverben (which corres pond to (part of) the verbs oflocation of Wunderlich), and Bewegungverben (which correspond both to the verbs ofplacement and the verbs of movement of Wunderlich). Her subclassification is also based on a difference between locative and directional PPs. She nevertheless does not differentiate verbs which accept only directional (final) prepositions from those which accept all dynamic prepositions. She notes rather that locative PPs denote the property of being localized with respect to a 'path'. This classification is closer to ours than Wunderlich's. Bewegungsverben are our verbs of change of location; Positionsverben are our verbs of change and inertial change of position and change of posture.23 She also considers differentiating between 'region' and 'path' as important for the notion of localization. Guillet ( I 990) prefers to talk about locative verbs, which are those verbs which accept a locative 'compliment'. By locative 'complement', Guillet means a 'complement' answering the question ou? where?' but it is not a circumstantial 'complement' of location (the French complement circonstanciel de lieu ). He neither refines his class of locative verbs nor gives a precise definition ofit, but we can nevertheless see through his work that he distinguishes verbs which, for him, do not clearly denote a change of location (e.g. braquer to tum (the wheel) , remuer to move I tofidget . . .). With this consideration, we can say that all these verbs are our verbs of change of posture, and the remaining ones, his 'locative verbs', our verbs of change of location and change of inertial change of position. These syntactic approaches then do not furnish completely adequate and precise criteria for a classification of motion verbs. Nor, unfortunately, do the syntactic approaches cover all the verbs we would like. As Wunderlich remarks himself ( I 991 ), his definitions work well only for simple verbs in English or German, such as sitzen to sit , stellen to put, werfen to throw , gehen to walk . But there are many other verbs which already contain the locative predicate.24 He

202

A Typology and

Discourse Semantics for Motion Verbs and Spatial PPs in french

5 .2

Semantic approaches

Hays ( I 989), following Levelt et a/ . ( I 978), makes a distinction between trans position verbs and non-transposition verbs, in the specific domain of the VITRA project. The latter include concepts such as 'rotate', 'tremble', and 'sway'. Her transposition verbs in fact cover our verbs of change of location and change and inertial change of position; her non-transpositional verbs are our verbs of change of posture. She also makes a second level subclassifica tion by refining her transposition verb class, using the two following criteria. The first one consists of a distinction between a change and an absence of change over some dimension (i.e. in speed or direction, e.g.). The second one consists of a distinction between concepts referring to continuous change (e.g. walking) and those that refer to discrete state changes (e.g. stopping , starting , turning). This kind of criterion differs from any of those we have presented in this paper. We nevertheless agree with Hays that refinement of the classification in terms of meaning components (c£ Levin & Rappaport-Hovav I 992) is essential. But this has to remain as a second level classification. That is, both levels have to be accessible, depending on which kind of application it is used for. In other work, not presented here, we have circumscribed 3 s different meaning components concerning French verbs of change oflocation (c£ Asher & Sablayrolles I 994, in preparation), which lead to a very fine-grained classification (the largest class has I I lexical items, and most less than 3). This kind of classification is perhaps needed in applications such as automatic translation, for example, but does not help us understand basic inferences concerning movement, which is our task here.

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

gives the examples of the English verb to enter, the German25 verb bereisen to travel in , the Spanish verbs entrar to move into , salir to move out , pasar to move by. We can also add the French verb quitter to leave. It is true that some of them allow the expression of the relatum of the location by a direct object, which is evidence the syntactic approach; but not all of them, unfortunately. We remark that those verbs which run into trouble with a syntactic approach are easily recognized as 'motion verbs' with a semantic approach. Other very problematical verbs, such as chercher to lookfor or sonner to ring , are clearly rejected by a syntactic approach, but could eventually be recognized by a semantic approach, if there is some need of them for some particular application. We do not wish, with this negative conclusion, to imply that a syntactic approach for motion verbs classification is ineffectual. On the contrary, we think that, although this kind of approach alone is not sufficient, it can introduce some important elements that corroborate semantical criteria built on classifications.

Nicholas Asher and Pierre Sablayrolles

203

Dervillez-Bastuji ( 1 982) makes use of two kinds of criterion: the change versus the non-change oflocation, and the orientation. Motion does not always denote an orientation; the location of reference could be the 'background' location in the process, in contrast with the initial and the final locations for displacement. Motion only specifies the 'manner' in which the process is realized. A displacement is always oriented. The motion, concrete or metaphoric, is a necessary condition for a change of location, bur in no case a sufficient condition. Ir is why so many linguists and also sometimes psycholinguisrs, forgetting rhis consideration of change vs. non-change of rhe referential system, have pur rogerher as 'motion verbs' verbs such as remuer to move/tofidget , danser to dance, se promener to take a walk and verbs such as venir to come, partir to leave or entrer to get into . (Dervillez-Basruji

19�2)

The displacement verbs of Dervillez-Bastuji cover then our verbs of change of location, her agitation verbs our verbs of change of posture. Dervillez-Bastuji does not distinguish between our two classes of change of position verbs. She uses a directional criterion to make distinctions, which we do not. Verbs like longer to go along denote oriented-path directed motion, and then are displacement verbs; verbs like errer to wander denote free motion, and then are agitation verbs. Lamiroy ( 1 987) uses the concept of change of position in space. She defines three classes of motion verbs. The first class contains verbs of direction, which are polarized and relic verbs always having a starring point or a final point for the motion. They use in the passe compose the auxiliary etre.26 Examples are etrer toget into , arriver to arrive , sortir to get out , partir to leave . They correspond to our verbs of change of location. The second class contains verbs of displacement. They either denote a way of moving (e.g. marcher to walk , trotter to trot , deambuler to stroll about), or a speed of moving (e.g. se precipiter to dash at/in/out , foncer to rush ), or a direction of moving (e.g. se diriger to navigate/to make one's way (towards) , obliquer to take an oblique direction ). They use in the passe compose the auxiliary avoir. They correspond to both our verbs of change and inertial change of position. The third class contains verbs of movement of the body, which denote neither a direction nor a displacement from one location to another, but rather a change of posture or state. They correspond to our verbs of change of posture. Boons (1985, 1987) is the most careful of these authors. He thinks that the class of'motion verbs' is one of the most curious and interesting lexical class of

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

Every non-oriented motion is equivalent wirh a stare. This srare can manifest itself as rest or as agitation. To rhis rest or rhis free motion is opposed displacement, which is considered as an oriented-path directed motion defined by irs direction and, in a second rime, by irs meaning (Dervillez-Basruji 1 982)

204

A Typology and Discourse Semantics for Motion Verbs and Spatial PPs in French

6 CONCLUSION We have presented a typology for motion verbs and motion verb complexes which enables us to draw a detailed spatiotemporal semantics of motion expressions. We have also shown that a formalization of the calculus for our compositional semantics can be integrated in a Discourse Representation

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

the language. But he nevertheless realizes that this class is completely vague. Such verbs can denote a proper displacement of a body as well as displacements of one part with respect to other parts of a body. He then prefers to focus on the first interpretation of these verbs and avoids precise delimitation of the general class of the so-called 'motion verbs'. He considers, as does Dervillez-Bastuji ( I 982), the distinction between the change and the non-change of location as a crucial one. This (theoretically) leads him to the distinction between motion verbs and displacement verbs. Our verbs of change of location clearly belong to his motion verbs and our verbs of change of posture to his displacement verbs. But for our verbs of change and inertial change of position, the situation is absolutely unclear. If we refer to the definitions given by Boons, they must belong to the class of his motion verbs. But in fact, through the examples and the lists he has given, they appear in his class of displacement verbs. Laur (I 99 I) uses the concept of change. But unlike Lamiroy who is talking about only change of position in space, she is talking about change of state, shape, posture, or position of animate or inanimate bodies. She has also raised the ambiguity present in the work ofBoons by subdividing the displacement verbs into three subclasses with respect to a criterion of polarity (already introduced by Boons), that is the differentiation between verbs with an initial, a medial or a final polarity. Such verbs are naturally called initial verbs, median verbs, and final verbs. She has this time produced a list of verbs which accord with the definitions used. Initial and final verbs are our verbs of change oflocation. Median verbs are our verbs of change and inertial change of position. Figure 8 summarizes all the comparative works mentioned in this section. When we look at this it seems that the class of'motion verbs' is a really complex class, especially because of the different terminologies and the different ways and 'tools' used by each author to tackle this difficult problem. But in presenting the different authors' works by way of confrontation and superposition, we hope we have shown that the classifications are in fact compatible (as is clearly depicted in Figure 8). The fact that these works have been achieved on the basis of different kinds of criteria, such as syntactical, semantical, and conceptual criteria, and that they have arrived at compatible classifications, gives these classifications a very strong validity.

Nicholas Asher and Pierre Sablayrolles

Asher & Sablayrolles

Wunderlich Maienbom

verbs of change of location

verbs of placement

verbs of change of position

verbs of movement

verbs of inertial change of position

205

verbs of change of posture

verbs of location Posi tionsverben

Bewegungverben

Guillet

Verbes locatifs

other verbs

Hays

Transposition verbs

non transposition verbs Agitation verbs

Displacement verbs

Lamiroy

verbs of direction

Boons

displacement verbs

+-

Laur

Initial and final verbs

Figure 8

verbs of movement of the body

verbs of displacement (in the facts) AND (by definition)

motion verbs --+

Medial verbs

motion verbs

Comparison of classifications of motion verbs

framework (SORT). Our lexical information combined with discourse information allows for the understanding processes of the spatiotemporal structures of texts and for lexical disambiguation, which are essential when calculating the localization of eventualities denoting motion with respect to various locations given in the text. This work, presented in this paper, also forms the basis of other detailed investigations on motion in progress at the LRC (Language, Reasoning, Computation), in Toulouse, detailed below: o

o

e

a finer-grained typology of verbs of change of location, based on the one presented here, and using in addition a 3 5 meaning components for further categorization; an aspectual study of verbs of change of location, for which the seven spatial (spatiotemporal) relations used here for the typology lead to more detailed notions of Aktionsart; linguistic comparative studies realized on Basque (M. Aurnague) and Japanese Q. Saito) languages which corroborate our own results in French.

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

Dervillez Bastuji

206

A Typology and

Discourse Semantics for Motion Verbs and Spatial PPs in French

The approach and the methodology we have followed for this work suggest considerable promise and we are presently continuing our investigation on this basis. NICHOLAS ASHER

Received: 1 0.0 1.94 Revised version received: 29.07.94

Centerfor Cognitive Science University ofTexas CRG 22o Austin, TX 787 1 2 USA e-mail:nasher®Jeeves.la.utexas.edu PIERRE SABLAYROLLES

France

e-mail:sablay@ iritfr

NOTES To be more precise, we should talk about a change of relation between the whole moving entity and the location. 2 Rather, we should talk about a change of relation between the position defined by the moving entity and some referential fixed parr of the background location (such as a border). More precisely, we should talk about a change of relation between some part of the moving entity and the location. 4 The relevant parts of the moving entity which are concerned by the relation are not directly expressed in the sentence but they can generally be inferred from lexical informations derived from the motion verb. 5 In our French terminology: un lieu. 6 This portion of space can be a bit of ground, such as in the square or the lawn , or the space occupied and defined by an object, the inside ofwhich is a cavity, such as in the house , or not a cavity, such as in 1

the bench .

7 For example, a kitchen is normally a place one recognizes as being one where people can cook and eat, a street one on which cars and people move, a house a place in which people live . . . In contrast, positions (see Definition 2) have no such functionality. An unmarked part of the street has no associated activity and is not recognizable as such. It can only be defined deictically (which may involve referring to the position of some object by ostension or by the use of coordinates). 8 In our French terminology: un emplace ment. 9 They can only be lexicalized with deictic constructions like here, this position , the

position Mary occupies now . . . 10 In our French terminology: une posture. 1 1 Note that Talmy also has the notion of a location of reference. He also considers secondary locations of reference, which could prove useful to us. But we will forgo these here. 1 2 This is of course no longer the case once

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

IRIT Universite Paul Sabatier 1 1 8 Route de Narbonne F-3 1 o62 Toulouse Cedex

Nicholas Asher and Pierre Sablayrolles

o O O

o

o o

PP(x, y): x is a property part of y EC(x, y): x is externally connected to y f-int(y, x): rerum the functional interior of y with respect to x f-ex«y, x): return the functional exterior of y with respect to x weak-contac«x, y): x is in 'natural contact' with y prox(y, x, C): rerum a 'proximity zone' by deciding if x is 'in proximity' of y. This kind of decision needs informa tion about the context (C). Under the name 'context' we group together things like what we know from the

discourse, from world knowledge, about the nature of x and y and about the presence of other individuals around. In the rest of the paper, we will not write out this third argument C in order to simplify the notations. We use the functional interior defined by Aurnague and Vieu ( 1 993 ). 1 8 We use here the mereological relation P(x, y): x is a part of y. I 9 These ideal paths have the same proper ties as 'normal' locations plus some others (like, for example, a direction of motion), about which we will say nothing further in this paper. 2.0 The data contained in Figure 4 and the following Figures s and 6 are also written under the form of logical axioms. Never theless, we prefer, for the sake of clarity, to present these results in the form of tables. 2.I The computation makes use of the moving entity, but has also to make use of the geometrical semantics of the preposi tion. here de derriere is different from de devant, for example. This in fact concerns a semantical study on geometrical aspects of spatial prepositions we will touch upon in this paper. 2.2. i.e. they require the incorporation of a spatial PP. 23 Indeed, it appears that the Position sverben of Maienborn do not cover static verbs, in contrast to the verbs of location of Wunderlich. 24 As an incorporated directional or path oriented location. 25 A locative preposition is often incorpo rated in the German prefix verbs. 2.6 However, some of these verbs can use in the passe c0mpose the auxiliary avoir, in an accusative and agenrive form. For example, J'ai sorti le chien I have put the

·

I7

dog outside .

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

we begin to look at more complex dis courses with quantification and negation; there the srrucrure of the OS representa tions becomes very important for the analysis of anaphora, on which our work on spatiotemporal structure in discourse depends. 1 3 This classification, though developed principally with an eye to preserving monotonic constraint propagation, nevertheless also has a linguistic basis. Work in progress at Toulouse, carried out by Laure Sarda, shows that this distinc tion IS also triggered by syntactical grounds. Indeed, transmve French motion verbs, having a medial polarity, belong only to the class ofverbs of change of position, while intransitive French motion verbs, also having a medial polarity, cover both the classes of verbs of change of position and of inertial change of position. I4 As do, for instance, Jackendoff & Landau ( 1992), where places and relationships between figure and ground are defined as regions, often occupied by landmarks or reference objects and in which the figure object is located. I s We thank Laure Vieu for this example. I6 We use the following mereological and topological relations in our definitions:

207

208 A Typology and Discourse Semantics for Motion Verbs and Spatial PPs in French

REFERE N CE S Boons, J. P. (I 987), 'La notion si:mantique de deplacement dans une classification syn taxique des verbes locatifs', Langue Franfaise, 76, 5-40, December. Borillo, M. & P. Sablayrolles {I993), 'The semantics of motion verbs in French', Pro

ceedings ofthe 13th International Conference on Natural Language Processing ofAvignon , 24-

28 May I 993, Avignon, France. Clarke, B. L. (I 98 I ), 'A calculus of individuals based on "Connection"', Notre Dame journal ofFormal Logic, 2.2./3, 204- I 8, ]uly I 98 I . RFIA . Asher, N. & M. Morreau ( I 99 I ), 'Common Clarke, B. L. (I98 5), 'Individuals and points', Notre Dame Journal of Formal Logic, 26/I , sense Entailment: a modal theory of non 6 I -75, ]anuary I 985. monotonic reasoning', in Proceedings of Davidson, D. (I 967), 'The logical form of I]CAI-91 . action sentences', in N. Rescher (ed.), The Asher, N. & P. Sablayrolles ( I 994), 'Decom positional semantics for motion verbs in Logic of Decision and Action , University of French', forthcoming. Pittsburgh Press, Pittsburgh, PA. Asher, N., M. Aumague, M. Bras & L. Vieu Dervillez-Bastuji, J. (I982), 'Srrucrure des relations spatiales dans quelques langues ( I99 3), 'Space, rime and discourse', Proceed ings of the Dagstuhl Seminar on Universals in naturelles: introduction a une thi:orie si:mantique', Library Droz, Geneva-Paris. the Lexicon and Proceedings ofthe Workshop on Temporal and Spatial Reasoning, IJCAI. Guillet, A. ( I 990), 'The classification des Asher, N. & H. Wada (I 989), 'A computa verbes nansitifs locatifs', Ph.D. thesis, tional account of syntactic, semantic and Universiti: Paris 7, LADL, I 990. discourse principles for anaphora resolu Hays, E. M ( I989), 'On defining motion tion',journal ofSemantics, 6: 309-344. verbs and spatial prepositions', Bericht B61, SFB 314: Vl7RA , Fachbereich I o Aurnague, M. {I99 I ), 'Contribution a !'etude de Ia semantique formelle de l'espace er du Informatik IV, Saarbriicken, Germany, October I989. raisonnemenr spatial: Ia localisation interne en fran�ais, si:mantique et srruc J;;ckendoff, R. & B. Landau (I 992), ' "What" tures inferentielles', Ph.D. thesis, Univer and "Where" in spatial language and spatial cognition', BBS Report, Cambridge site Paul Sabatier, Toulouse, France, February. University Press, Cambridge. Aurnague, M. & L. Vieu (I 993), 'Toward a Kamp, H. { I979), 'Events, instants and formal representation of space in lan temporal reference', in R. Bauerle, U. Egli & A. von Stechow (eds), Semantics from guage: a commonsense reasonmg Different Points of View, Springer Verlag, approach', in Proceedings ofiJCAI-93 . Berlin, 376-4I 7. Boons, J. P. ( 1 98 5), 'Pri:liminaires a Ia classi fication des verbes locatifs: les comple Kamp, H. & A. Rolkleutscher (I 992), 'Remarks on lexical srructure, DRS men ts de lieu, leurs criteres, leurs valeurs consrruction and lexically driven infer aspecruelles', Linguisticae Investigationes, 9/ ences', Sprachtheoretische Grundlagen des 2, I 9 5-267.

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

Asher, N. (I 992a), 'A rruth conditional, default semantics for the progressive', Lin guistics and Philosophy, no. I 5· Asher, N. (I 992b), 'Spatiotemporal srructure in texts', in S. Akama (ed.), Festschriftfor II Onishi, Oxford University Press, Oxford. Asher, N. ( I993), Reference to Abstract Objects in Discourse , Kluwer Academic Publishers, Dordrecht. Asher, N. & M. Bras ( I 9!l4), 'Le raisonnement non monocone dans le consrruction e Ia structure temporelle de textes en fran�ais',

Nicholas Asher and Pierre Sablayrolles 209

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

Computerlinguistik , Arbeitspapiere . des ment er conttibution a !'etude de quelques Sonderforschungsbereichs 340. inferences naturelles', DEA Report, Uni K.rifka, M. (I 987), 'Nominal reference and versity Paul Sabatier, Toulouse, France, temporal constitution towards a semantics June I99 I . of quantity', FNS-Bericht 1 7 , Universicit Sablayrolles, P. ( 1 99 1 b), 'Semantique spatio Tiibingen. temporelle du deplacement en fran�ais: Lamiroy, B. (I986), 'Les verbes de mouve analyse er representation', Les Cahiers de Grammaire, 16, I I 9-59· ment, emplois figures et extensions meta Sablayrolles, P. (I 992a), 'How motion can be phoriques', LAnguefranfaise , 76, 4I-58. used for formalizing spatial and spatio Lascarides, A. & N. Asher ( I 99I), 'Discourse relations and defeasible knowledge', Pro remporal semantics of natural languages', Proceedings of Workshop Connecting Cognitive ceedings ofthe Meeting ofthe ACL , Berkeley, CA, 5 5-<>2. Theories with Formal Representations, ECAI Lascarides, A. & N. Asher (I993). 'Temporal I 992. Also published in S. Pribbenow & C. interpretation discourse relations, and Schlieder (eds), Kognitionswissenschafi Ham burg , February I 994· commonsense entailment', Linguistics and Sablayrolles, P. (I 992b), 'Spatia-temporal Philosophy, 16, 437-93 · semantics in natural languages: the case of Laur, D. ( I 99 I), 'Semantique du deplacement et de Ia localisation en fran�ais: une erude motion', Proceedings of the 4th Internal des verbes, des prepositions et de leurs Workshop on Time, Space and Movement, Chateau de Bonas, France, 4-8 September relations dans Ia phrase simple', Ph.D. thesis, Universire Toulouse le Mirail, Tou I 992. M. Aumague, A. Borillo, M. Borillo, M. Bras (eds), published in I 993· louse, France, June I 99 1 . Lesniewski, S. (I 927-3 I), '0 podsrawach Sablayrolles, P. (I993). 'A rwo-level semantics matematyki', in Przeglad filozoficzny, no. for French expressions of motion', Research 30-40. Partial translation by G. Kalinowski Report DFKI RR-93- 1 2 , Saarbrucken, in S. Lesniewski (I989). Germany, March I993 · Lesniewski, S. (I 989). Sur les Fondements de Ia Smith, C. S. (I99I), The Parameter of Aspect , mathematique, Hermes, Paris. Kluwer Academic Publishers, Dordrecht. Levelr, W. J. M., R. Schreuder & E. Hoen Talmy, L. (I 982), 'How language structures kamp (I 978), 'Structure and use of verbs of space', in Pick & Acredolo (eds), Spatial Orientation: Theory, Research and Applica motion', in R. N. Campbell & P. T. Smith (eds), Recent Advances in the Psychology of tion, Plenum Publishing Corporation, New York. LAnguage, Plenum Press, London, I 3 7-<>2. Levin, B. & M. Rappaport-Hovav (I 992), 'The Verkuyl, H. (I 972), 'On the compositional lexical semantics of verbs of motion: the narure of the aspects', Foundations ofLAn perspective from unaccusativity', in I. M. guage Supplementary Series, Reidel, Dor Roca (ed.), Thematic Structure: Its Role in drechr, No. I 5. I 972. Grammar, Foris Publications, Berlin, New Vieu, Laure (I99I), 'Semantique des relations York. spatiales er inferences spatio-remporelles: Maienborn, C. (I 992), 'Kompakte Strukturen: une conttibution a !'etude des structures de l'espace en Langage Nature!', Ph.D. Direktionale PPn und nichr-lokale thesis, Universite Paul Sabatier, Toulouse, Verben', in P. Bosch & P. Gersrl (eds), Sprachtheoretische Grundlagen fur die Com France, April I 991 . puterlinguistik, DFG Sonderforschungsbe Wunderlich, D. (I99I), 'How d o preposi tional phrases fir into compositional reich 340, Stuttgart, Berichr Nr. 30, December I 992, I 3 I -45· syntax and semantics?' Linguistics, 29, 59I 62 I . Sablayrolles, P. (I99Ia), 'Representation de Ia semantique spatio-remporelle du deplace-

Journal ofSmumtics

1 2: 2 1 1-2 1 2

© Oxford University Press 1 99)

Book Review

Stechow & Dieter Wunderlich (eds), Semantik/Semantics: Ein internationales Handbuch der zeitgenossischen Forschung. An International Handbook ofContemporary Research. Walter de Gruyter, Berlin/New York, 1 99 1 .

Arnim von

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

The title is bilingual, and the Preface is also printed in English and in German. Twenty-eight of the Handbook articles are in German; the remaining thirteen are in English-one wonders why. Of the many possible explanations only one seemed plausible in the end. And even this one does not make me happy. Could it really be that the budget did not allow for the translation of the last ten articles into German? Or was it for the international flair? Without doubt, several of the articles printed in German were not written in German, but were translated. And several of the articles printed in English are by authors whose native language is German and they could probably have been persuaded to write in German. The Preface offers no explanation. It was written by the editors, and the bilingual character of the book is unlikely to be part of the editors' conception of the book but just a result oflack of conception on the part of the publishers. It's a great shame. Handbooks are not cheap and are not meant to be (this one costs DM63o). Nor are they put together easily. And they are potentially very useful. But this one-for obvious reasons-cannot be recommended to anybody who does not read German. Now suppose you do read German. Is it a good book? Yes, unquestionably so. A major factor is the selection of authors. Almost all of them have played an important role in the development of natural language semantics. They are, in order of appearance: John Lyons, Max Cresswell, Dieter Wunderlich, Gisbert Fanselow, Peter Staudacher, Arnim von Stechow, Ede Zimmermann, Ulrike Hass-Spohn, Manfred Pinkal, Gunther Grewendorf, Dietmar Zaefferer, Pieter Seuren, Andreas Kaemmerling, Rainer Bauerle, Jean-Yves Lerner, Greg Carlson, Manfred Krifka, Godehard Link. Veronika Ehrich, Jan van Eijck, Irene Heim, Tanya Reinhard, Peter Pause, Joachim Jacobs, Ewald Lang, Kjell Saeb0, Ekkehard Konig, Angelika Kratzer, Cornelia Hamann, Ewan Klein, and Cathrine Fabricius-Hausen. It may be felt that an unproportionate number of the authors are German. Even if this is so, it does not show in the contents of the Handbook. The adjective international in the subtitle is certainly correct in the sense that the Handbook does reflect international developments.

21 2

Book Review

IBM Scientific Centre Heidelberg Germany

=-

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

I have not read the Handbook from cover to cover, but I have tested it, as it were, under real-life conditions for about a year: I have had it on my desk and whenever I was wondering about standard references for a particular topic, or wanted to make sure that I was not missing an important development in an area I am less familiar with, the Handbook was of help. It does not just provide you with ideas that are generally accepted, but the authors of the various articles-some more than others-use their own judgement as to what is and what is not an important development in the field. This may lead to an unbalanced representation occasionally, but I have not discovered any cases where at least references could not be found to relevant work that otherwise remained under-represented. What more could one ask? Now this holds for what may be labelled 'truth-conditional semantics' of natural language, which after all is the core of serious developments in the semantics of natural language in the last couple of decades. The book will be less helpful for anyone with stronger interests in psycholinguistics, com putational linguistics, or, for that matter, applications of semantics in the theory and interpretation of literature-unless, of course, someone from these fields wants to find out what happened in core semantics and what they can learn from it. The latter is exactly what the Handbook is for and where it has its function. Since there is no recent comprehensive introductory textbook for the semantics of natural language, the Handbook, or parts of it, will certainly come in very handy for teaching purposes as well. Of course, students won't buy it. They will copy the relevant pages and the publisher will not get a penny. Given their obvious and total obliviousness with respect to the market, I'd say: serves them right.

No title

No title

No title

No title

No title

No title

No title

No title

No title

No title

No title

No title

No title

No title

No title

No title

No title

No title

No title

No title

No title

No title

No title

No title

No title

No title

No title

No title

No title

No title

Recommend Documents