The Generative Lexicon

The Generative Lexicon The Generative Lexicon James Pustejovsky The MIT Press Cambridge, Massachusetts London, Engl...

Author: James Pustejovsky

196 downloads 2274 Views 7MB 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

The Generative Lexicon

The Generative Lexicon

James Pustejovsky

The MIT Press Cambridge, Massachusetts London, England

Second printing, 1996 © Massachusetts Institute of Technology All rights reserved. No part of this book may be reproduced in any form by any electronic or mechanical means (including photocopying, recording, or information storage and retreival) without permission in writing of the publisher. This book was set in Computer Modern by author and printed and bound in the United States of America. Library of Congress Cataloging-in-Publication Data Pustejovsky, J. (James) The Generative Lexicon / James Pustejovsky. p. cm. Based on the author's unpublished manuscript, "Towards a Generative Lexicon." Includes bibliographical references (p. ) and indexes. ISBN 0-262-16158-3 1. Semantics. 2. Generative grammar 3. Computational linguistics. I. Title P325.P85 1995 401'.43-dc20 95-32875 CIP

Contents

Preface

ix

Acknowledgements

xi

1

Introduction

1

2

The Nature of Lexical Knowledge

5

2.1 Semantic Classes and Categorial Alternation 2.2 Interlexical Relations

8 23

3 3.1 3.2 3.3 3.4

The Logical Problem of Polysemy Varieties of Sense Extension Contrasitive Ambiguity Complemetary Polysemy An Elemetary Lexical Semantic Theory

27 27 29 31 33

4 4.1 4.2 4.3 4.4 4.5 4.6 4.7

Limitations of Sense Enumerative Lexicons The Goals of Lexical Semantic Theory The Creative Use of Words Permeability of Word Senses Difference in Syntactic Forms Semantic Expressiveness Generative Lexical Models Strong vs. Weak Compositionality

39 40 42 46 50 55 57 59

5 5.1 5.2 5.3 5.4 5.5

The Semantic Type System Levels of Representation Argument Structure Extended Event Structure Qualia Structure The Interaction of Semantic Levels

61 61 62 67 76 81

vi

Contents

6 Qualia Structure 6.1 Modes of Explanation 6.2 The Qualia Structure of Nominals

85 85 90

7 7.1 7.2 7.3 7.4

Generative Mechanisms in Semantics Coertion and Type Shifting Co-composition Selective Binding Semantic Selection

105 106 122 127 131

8 8.1 8.2 8.3 8.4 8.5 8.6

The Sematics of Nominals Basic Issues Nominals with Unified Types Nominals with Complex Types Propositions and Event Descriptions Varieties of Nominalization Lexicalisation of LCPs

141 141 142 149 157 165 177

9 9.1 9.2 9.3 9.4 9.5 9.6

The Lexical Semantics of Causation How Language Encodes Causation Causation and Uncaussativity Aspectual Causatives and Coertion Experiencer Predicates Modal Causatives Conclusion

183 183 188 198 208 215 218

10 Consequences of Generative Lexicon 10.1 Co-composition and Verbal Semantics 10.2 Stage-Level Predication 10.3 Further Applications of Coertion

221 221 225 230

Contents

vii

10.4 Lexical Inference and Rhetoric 10.5 Summary

236 238

Notes

241

Bibliography

269

Author Index

289

Subject Index

293

Preface

This book is basedon a larger unpublished manuscript, Towardsa Generative Lexicon, written between 1988- 1993. Many of the basic ideas for this manuscript are first explored in a general way in an article writ ten in 1987 with Bob Ingria , entitled "Active Objects in Syntax and Semantics." Since the original publication of the article "Generative Lexicon" in the journal Computational Linguistics in 1991, what had started merely as a critique of theories of lexical senseenumeration has developed into a fairly specific proposal for how to perform lexical semantic analysis. Substantial new material has emergedfrom cooperative work with several colleagues. This includes the original work done with Bran Boguraev and discussionwith Ted Briscoe and Ann Copestakein the context of their Acquilex researchfunded by Esprit in Europe. The work on unaccusativity stems from joint research with Federica Busa of Brandeis. Extensions and elaborations of the coercion analysis for aspectual predicates in French has been done in closecollaboration with Pierrette Bouillon of ISSCO and the University of Paris. Several chapters have been omitted for clarity of presentation and in some casesfor clarity of the content. Hence, some topics that were expected to be included have been deleted entirely. For example, two important themes in generative lexical studies (the role played by Lexical Inheritance Theory and the theory of co-specification) are not e~amined in any depth in this monograph; both of these areas have become too large to make only passingreferenceto , and I felt the discussionpossible in this monograph would do no justice to these issues. Regarding lexical inheritance, recent work in computational lexicography and semantics, much of it done in the context of 'the Acquilex project (and reported in Briscoe et al., 1993) has pointed to new and exciting directions for how lexicons should be organized globally. The promise of realizing a projective inheritance model, as suggestedin my 1991 article, awaits further investigation, although it is currently one of the topics of research at Brandeis in conjunction with Bran Boguraev at Apple. The secondmajor omission in this monograph is a comprehensivediscussion of co-specification and processesof selection. This has proved to be a central concern in the applied computational researchat Brandeis as well as the recent work on lexical acquisition and induction from corpora. In the present study, however, I have chosento concentrate on the core mechanismsinvolved in semantic selection as they relate to

x

Preface

syntactic expression . Because co-specification treats the subtle use and variation in selection below the level of conventional semantic and syntactic types , it is impossible to do it justice without extensive discussion . This can be found in Pustejovsky (forthcoming ) and to a certain extent in Boguraev and Pustejovsky ( 1996) . Finally , I should point out that many questions relating to natural language semantics are not investigated in any great detail here . In par ticular , issues surrounding quantification and genericity are only touched on briefly , if at all . Furthermore , details of several of the mechanisms of composition are to be found not here but in other works , including Pustejovsky ( 1995b) and Pustejovsky and Johnston (forthcoming ) . My aim in the current work has been to outline what I feel is the necessary infrastructure for a truly generative , highly distributed , and lexically based semantic theory for language .

James Pustejovsky Brandeis University

Acknowledgments

The ideas contained in this work have been greatly influenced by conversations and interactions with numerous friends and colleagues. Those who have contributed to the specifics of this work include : Noam Chom sky, Bob Ingria , Bran Boguraev , Ted Briscoe , Federica Busa , Patrick Saint -Dizier , Nicholas Asher , Ray JackendofI , Geoffrey Nunberg , Michael Johnston , Pierrette Bouillon , Scott Waterman , Sabine Bergler , Peter Anick , Paul Buitelaar , Evelyne Viegas , and David McDonald . Early encouragement from Dave Waltz , Jane Grimshaw , Jim Higgin botham , Remko Scha, and Edwin Williams helped shape the direction of the work , and their influence is seen throughout the pages of this book . Critical interactions with many colleagues have been useful in sharpening my proposal and the specific analyses. In particular , George Miller , Jerry Hobbs , Manfred Bierwisch , Jiirgen Kunze , Ewald Lang , Beth Levin , Barbara Grosz , Sergei Nirenburg , Aravind Joshi , Chuck Fillmore

, Rick

Alterman

, Yorick

Wilks

, David

Waltz

, and

Peter

Cariani

have provided various degrees of commentary and criticism , making , I believe , the resulting work much clearer than it might have been. Large portions of the material reported on in this book have been presented in front , of audiences from linguistics and computer science conferences and departments , including Brandeis , UMIST , MIT , University of Pennsylvania , Princeton , McGill , the University of Texas , Stuttgart , Humboldt University in Berlin , the CUNY Sentence Processing Confer ence in New York , IBM , SUNY Buffalo , Toronto , Toulouse , Cambridge University , University of Copenhagen , and Charles University in Prague . Many of the comments and questions from these interactions have found their way into the work presented here, and I would like to thank the participants of these talks for their critical and helpful remarks . I would like to also thank the following for carefully reading the final manuscript : Bob Ingria , Federica Busa , Ann Copestake , Henk Verkuyl , Ted Briscoe , Nicholas Asher , Alex Lascarides , Anne -Marie Mineur , Michael Johnston , and Marc Verhagen . I would like to give particular thanks to MIT Press, for their encour agement and editorial support of this project . To Harry Stanton , for his patience and faith , and to Amy Pierce , for her enormous help with guiding the book through the Press on time , and further expediting the publication . And finally , to Deborah Cantor -Adams , for her wonderful editorial help and support with the manuscript .

.. XII

Acknow ledgments

For moral support, there are really too many peopleto thank . I would like to thank my mother, Frances, for her constant energy and encouragement; Charlie and Helga Marqusee, for their wonderful creature comforts and Charlie's personal mail and delivery service; John Broglio, for lending me his apartment in Northampton to rewrite the manuscript; and to Brent Cochran, Peter Cariani, Bob Ingria , and Fede Busa, for their undeservedfaith in me. And finally and most significantly, to my family, Cathie, Zachary, and Sophia for always being there, and enduring the tedious hours and my testy moodiness. This work was supported in part by grants from The Advanced ResearchProject Agency (MDA904-91-C-9328); the National ScienceFoundation (IRI -93-14955); Digital Equipment Corporation ; and the National Institutes for Health. I would like to dedicate this book to the memory of my father, George Harry Pustejovsky.

At this point , as it often happensin philosophy, we suddenly realize that the path of inquiry we hoped to open is already marked by the footprints of Aristotle . Zeno Vendler Linguistics in Philosophy

1 This

Introduction

book

deals

semantics

of

with

words

,

compositionality

.

words

of

a

have

largely

with

features

Under

guage

the

.

for

an

In

have

of

study

I

as

general

goal

of

of

a

the

goals

we

,

then

will

see

the

form

,

and

ative

of

of

I

current

cal

.

associated

of

natural

lan

issues

basis

of

-

:

composition

review

-

on

distinctions

, "

.

distinction

plementary

As

I

accidently

carries

issues

to

represent

models

.

the

recursively

as

If

.

way

in

in

The

traditional

which

,

to

in

ar

-

categories

.

means

language

,

on

passive

changes

a

antonymy

or

itself

be

the

meanings

,

functors

can

view

ultimately

denote

the

affect

we

synonymy

active

change

address

assign

language

either

in

lexical

Hence

,

if

reevaluate

order

to

what

This

satisfy

chapter

several

former

distinct

,

senses

3

,

( 1964

basic

and

,

both

incorporating

on

however

is

and

ambiguity

view

Weinreich

The

representation

lexical

word

in

. !

.

-

adequately

will

depends

semantics

distinguishes

between

ambiguity

directly

ways

such

we

we

composi

by

meaning

several

compositionality

lexical

basic

argue

,

,

.

how

computational

techniques

if

of

and

that

lexical

to

composition

some

view

senses

in

as

that

theory

meaning

argue

compositionality

of

semantic

semantic

word

of

phenomena

then

behave

,

nature

First

the

the

will

word

able

categories

words

I

theory

being

for

lexical

that

But

creative

theory

as

,

.

Our

semantic

comprehensively

very

item

a

metonymy

basic

studied

the

tagged

;

of

concerns

.

accounting

and

denote

the

,

information

important

on

interaction

of

theory

been

.

can

the

linguists

generally

two

of

what

senses

theories

models

these

problem

of

,

guments

no

the

semantic

,

has

the

to

compositionality

expressions

what

problem

and

word

been

contexts

semantic

the

semantic

formal

address

novel

of

and

have

,

to

in

lexical

design

polysemy

view

words

relate

for

the

,

.

how

the

theoretical

set

senses

little

examine

they

accounting

issue

static

i . e

of

particular

.

this

the

of

study

and

a

in

,

the

Nevertheless

done

and

combination

morphological

word

.

,

Computational

as

,

items

use

in

is

.

different

evaluation

tionality

to

,

lexical

ty

and

lexicon

syntactic

view

semantics

semantics

denote

creative

ali

alone

the

semantics

.

both

treated

distinct

language

Lexical

language

this

with

natural

,

)

theoreti

" sense

the

basis

such

an

calls

homonymy

unrelated

present

in

of

finite

,

makes

and

,

meanings

-

feature

approach

contrastive

-

enumer

where

a

,

com

lexical

whereas

-

2

Chapter 1

the latter refers to logically related word senses of the same lexical item . I then turn to some further problems with the enumeration method for lexical description illustrated in chapter 3. It will be shown that the representations assumed by current theories are inadequate to account for the richness of natural language semantics . As I show in chapters 2 and 3, most of the careful representation

work has been done on verb classes(e.g., Levin, 1993). In fact, the semantic weight in both lexical and compositional terms usually falls on the verb . This has obvious consequences for how lexical ambiguity has been treated . In chapter 4, I discuss several devices which simplify our semantic description , but which fall outside the conception of enumera tive lexical semantics . Looking at these devices closely, we notice that they point to a very different view of lexical semantics and how word meanings

are combined .

Given the discussion in these chapters , the following conception of lexical semantic systems emerges. Under such a theory , a core set of word senses, typically with greater internal structure than is assumed in previous theories , is used to generate a larger set of word senses when individual lexical items are combined with others in phrases and clauses. I will

refer to such an organization

as a generative

lexicon , and the op -

erations which generate these "extended senses" as generative devices, including operations such as type coercion and co- composition . I discuss how this view supports an explanatory view of semantic modeling . I then examine the goals of linguistic theory in general and lexical semantics in particular . I argue that our framework of knowledge for lexical items must be guided by a concern for semanticality in addition to grammati cality . The model of semantic interpretation we construct should reflect the particular properties and difficulties of natural language , and not simply be an application of a ready -to -wear logical formalism to a new body of data . I will view natural languages as positioned on a hierarchy of semantic descriptions , characterized in terms of their underlying poly morphic generative power . I argue that natural languages fall within the weakly polymorphic languages, more expressive than monomorphic , but well below the power of unrestricted polymorphic languages . This par ticular characterization is rich enough to capture the behavior of logical polysemy as well as effects of co-compositionality . Next , in chapter 5, I outline the type system for our semantics . A generative theory of the lexicon includes multiple levels of representation for

3

Introduction

the different types of lexical information

needed. Among such levels are

Argument Structure (for the representation of adicity information for functional elements), Event Structure (for the representation of information

related

to Aktionsarten

and event type , in the sense of Vendler ,

1967, and related work), Qualia Structure (for the representation of the defining attributes

of an object , such as its constituent parts , purpose

and function , mode of creation, etc.), and Inheritance Structure (for the representation of the relation between the lexical item and others in the lexicon ) . Chapter 6 presents in more detail the structure of qualia , and the role they play in distributing the functional behavior of words and phrases in composition . Chapter 7 presents the application of the mechanisms outlined in chapters 5 and 6 to the polymorphic behavior of language . A variety of poly morphic types is studied and I consider what operations are needed to adequately account for the syntactic expressiveness of semantic types . In particular , I examine the role of coercion in the grammar as well as the need for other generative devices, such as selective binding and co-composition . There is no single form of polymorphism ; rather , poly semy and type ambiguity are a result of several semantic phenomena in specific interaction . Chapter 8 examines briefly what the consequences of qualia structure are for the semantics of nomina Is. Nouns can be formally characteriza .ble in terms of three dimensions of analysis , involving argument structure , event type , and qualia structure . An analysis of nominal polysemy is presented , making use of the type system outlined in the previous chapters , and explaining in more detail the distinction between unified types and dot objects . In the next two chapters , I outline some areas of grammar that can be greatly simplified if we apply to them principles of generative lexical analysis through the use of the generative devices and the type system presented in chapter 5. In particular , I treat argument selection as driven by semantic types , modulated by constraints on coercion rules , selective binding , and co-composition operations in the grammar . This approach will permit us to explain the polymorphic nature of verbs taking multiple syntactic types . In chapter 9, I discuss the role that qualia and event structure have in describing the way causal relations are lexicalized in

language. Specifically, I look at the semantics of causativejinchoative verbs , aspectual predicates , experiencer predicates , and modal causatives

4

Chapter 1

such as risk. Finally , I discuss how this view of lexical organization relates to current theories of metaphor and pragmatically-induced metonymy. I argue, on methodological grounds, for a strong distinction between commonsenseknowledge and lexical structure, although the issue is clearly an empirical one. The types of creative polysemy examined in this work exhibit a regularity and systematicity across languagesthat is absent from patterns of pragmatic senseextension or modes of metaphor.

2


Onlya fewyearsago,it wasconventional practice in boththeoretical andcomputational linguistics textbooks to coverall that needed to be saidregarding thelexiconin onequickchapter , beforegettingto the moreinteresting andsubstantive topicsof syntactic formandseman tic interpretation . Suchanimpoverished coverage todaywouldscarcely reflectthevibrancyof the fieldof lexicalresearch or the centralrole playedbylexicalknowledge in linguistic theoryandprocessing models . It is nowstandardly assumed bymostlinguistic frameworks (bothcom putational andtheoretical ) thatmuchofthestructural information of a sentence is bestencoded froma lexicalized perspective .! Themostpressing problems for lexicalsemantics , I believe , arethe following : (a) (b) (c) (d)

Explaining thepolymorphic natureof language ; Characterizing thesemanticality of naturallanguage utterances ; Capturing thecreative useof words in novelcontexts ; Developing a richer , co-compositional semantic representation .

I believe wehavereached aninteresting turningpointin research , where linguisticstudiescanbeinformed by computational toolsfor lexicol ogyaswellasanappreciation ofthecomputational complexity of large lexicaldatabases . Likewise , computational research canprofitfroman awareness ofthegrammatical andsyntactic distinctions oflexicalitems ; naturallanguage processing (NLP) systems mustaccount forthesedifferences intheirlexicons andgrammars . Thewedding ofthese disciplines issoimportant , in fact, thatI believe it willsoonbedifficultto carryout serious computational research in thefieldsoflinguistics andNLPwithoutthehelpof electronic dictionaries andcomputational lexicographic resources (cf. Zampolli andAtkins,1994 , Boguraev andBriscoe , 1988 ). Positioned at thecenterofthissynthesis is thestudyofwordmeaning , lexicalsemantics . Before addressing these questions , I wouldliketo discuss twoassump tionsthatwill figureprominently in mysuggestions fora lexicalseman ticsframework . Thefirstis that, withoutanappreciation of thesyntacticstructure of a language , thestudyof lexicalsemantics is bound to fail. Thereis nowayin whichmeaning canbecompletely divorced fromthestructure thatcarries it. Thisis animportant methodological

6

Chapter 2

point , since grammatical distinctions are a useful metric in evaluating competing semantic theories. The second point is that the meanings of words should somehowreflect the deeper conceptual structures in the cognitive system, and the domain it oper"ates in . This is tantamount to stating that the semantics of natural languageshould be the image of nonlinguistic conc--eptual organizi.ng principles, whatever their structure. Computational lexical semantics should be guided by the following principles. First , a clear notion of semantic well-formednesswill be necessary in order to characterize a theory of possible word meaning. This may entail abstracting the notion of lexical meaning away from other semantic influences. For instance, this might suggestthat discourseand pragmatic factors should be handled differently or separately from the semantic contributions of lexical items in composition.2 Although this is not a necessaryassumption and may in fact be wrong, it will help narrow our focus on what is important for lexical semantic descriptions. Secondly, lexical semantics must look for representations that are richer than thematic role descriptions (cf. Gruber, 1965, Fillmore , 1968). As argued in Levin and Rappaport (1986), named roles are useful at best for establishing fairly general mapping strategies to the syntactic structures in language. The distinctions possible with thematic roles are much too coarse-grained to provide a useful semantic interpretation of a sentence. What is needed, I will argue, is a principled method of lexical decomposition. This presupposes, if it is to work at all, (1) a rich, recursive theory of semantic composition, (2) the notion of semantic well-formednessmentioned above, and (3) an appeal to severallevels of interpretation in the semantics (cf. Scha, 1983). Thirdly , and related to the preceding point , lexical semantics must study all syntactic categories in order to characterize the semantics of natural language. That is, contrary to the recent trends in semantic representation, the lexicon must encodeinformation for categoriesother than verbs. Recent work has done much to clarify the nature of verb classesand the syntactic constructions that each allows (cf. Levin 1985, 1993). Yet it is not clear whether we are any closerto understanding the underlying nature of verb meaning, why the classesdevelop as they do, and what consequences these distinctions have for the rest of the lexicon and grammar. The curious thing is that there has been little attention paid to the other lexical categories (but cf. Miller and Johnson-Laird ,

7


1976, Miller and Fellbaum, 1991). That is, we have little insight into the semantic nature of adjectival predication , and even less into the semantics of nominals . Not until all major categories have been studied can we hope to arrive at a balanced understanding of the lexicon and the methods

of composition

.

Stepping back from the lexicon for a moment , let me say briefly what I think the position of lexical research should be within the larger semantic picture . Ever since the earliest attempts at real text understanding , a major problem has been that of controlling the inferences associated with the interpretation process. In other words , how deep or shallow is the understanding of a text ? What is the unit of well -formedness when doing natural language understanding ; the sentence, utterance , paragraph , or discourse ? There the

sentence

researchers

, these would

is no easy answer to this question terms agree

are not

even

formalizable

because , except in a way

that

for

most

Oll .

It is my opinion that the representation of the context of an utter ance should be viewed as involving many different generative factors that account for the way that language users create and manipulate the

context

under

constraints

, in order

to

be understood

. Within

such

a theory, where many separate semantic levels (e.g., lexical semantics, compositional semantics, discoursestructure , temporal structure) have independent interpretations , the global meaning of a "discourse" is a highly flexible and malleable structure that has no single interpretation . The individual sources of semantic knowledge compute local inferences

with a high degree of certainty (cf. Hobbs et al., 1988, and Charniak and Goldman, 1988). When integrated together, these inferencesmust be globally coherent , a state which is accomplished by processes of cooperation among separate semantic modules . The basic result of such a view is that semantic interpretation proceeds in a principled fashion , always aware of what the source of a particular inference is, and what the certainty of its value is. Such an approach allows the reasoning process to be both tractable and computationally efficient . The representation of lexical in a richer

semantics , therefore , should be seen as just one of many levels characterization

of contextual

structure

.3

Given what I have said , let us examine the questions presented above in more detail . First , let us turn to the issue of methodology . In this chapter , I shall review the most common methods used for semantic classification of lexical items ,. and characterize the richness of the problem

8

Chapter2

of

representing

It

lexical

is

the

the

goal

lexical

and

semantic

the

way

it

facilitates

Within

of

a

word

' s

egorial

or

meaning

,

pIe

,

the

individuals

in

all

or

viduals

the

,

of

applying

standard

texts

.

.

1

.

A

recently

of

words

.

cf

.

1972

linguistic

.

For

)

as

Harris

is

of

1951

nouns

generally

)

and

to

Schubert

,

a

1989

.

by

For

and

example

,

due

to

literature

mass

)

of

collocation

classes

book

least

-

procedural

accomplished

basis

,

as

indi

subsets

selectional

at

or

set

of

are

this

For

the

sets

selectional

the

.

such

over

,

,

out

distinctions

on

1955

to

operators

example

in

1993

)

English

(

example

e

in

,

methodology

classes

begun

(

for

type

the

and

semantic

work

in

,

different

due

Pelletier

developed

Levin

1974

-

between

pick

Logical

cat

/

the

con

-

partition

count

distinction

.

Alternations

into

nations

while

classes

,

.

,

behaves

relations

would

Conventionally

,

,

.

such

view

word

refer

as

distinguishes

into

this

a

operations

Because

Chomsky

partition

Verbal

,

.

On

category

woman

analysis

animacy

participate

ample

classes

(

book

,

words

)

the

women

further

selectional

1

1975

basis

in

aspect

how

viewed

theoretic

,

categories

and

Verkuyl

-

distributional

different

cf

are

set

Montague

,

these

involving

into

.

Woods

tests

dog

cf

of

noun

who

as

.

only

be

the

semantics

cooccurrence

in

.

lexical

members

2

cf

(

merely

fundamental

type

would

whereas

world

not

individual

words

most

semantic

elements

hate

viewed

world

(

broad

the

be

the

instructions

nouns

,

the

not

the

and

an

between

,

its

what

world

might

in

perhaps

love

in

and

so

also

individuals

the

(

but

on

Alternation

determines

verbs

syntactic

should

,

Categorial

semantics

information

theory

their

.

formal

is

type

syntactically

exam

of

of

a

relations

and

tradition

such

classify

of

sentence

mapping

Classes

the

per

semantic

this

adequately

predictive

,

items

the

to

classes

Furthermore

lexical

capture

Semantic

theory

into

.

should

.

semantic

language

of

which

1

information

lexical

a

expression

,

.

any

of

meanings

Rather

2

of

items

map

a

semantic

. g

.

is

,

common

the

outlines

in

order

,

the

grouping

the

broad

Project

of

sink

verbs

,

roll

,

into

and

break

meanings

patterns

that

)

and

classification

classify

the

syntactic

alternations

Lexicon

to

verbs

study

grammatical

MIT

a

,

for

to

recently

verb

For

ex

-

codified

argument

alter

semantically

all

.

-

unique

have

both

tran

-


9

sitive and intransitiveforms, wherethe lexicalsensesare relatedby the interpretivefeatureof causation . There are of course, numerousexampIes of intransitive verbs which have no zero-derived causative forms, e.g., arrive, die, fall (cf. Fillmore , 1968, Lakoff, 1970, Hale and Keyser, 1986, 1993, and Kunze, 1991): (1) a. The boat ~!!k in stormy weather. b. The plane ~ k the boat in stormy weather. (2) a. The ball ! Qllg4 down the hill .

b. Bill !:QIl~ the ball downthe hill. (3) a. The bottle .QrQ ~ suddenly .

b. MaryQ! Q~ the bottle suddenly . (4) a. The letter arrived on time . b. *The mailman arrived the letter on time . (5) a. My terminal gi g last night . b. *The storm .cti d my terminal last night . (6) a. The block tower full.. b. * Zachary full. the block tower. c. Zachary .fuil~ the block tower. While the sentencesin (4b)- (6b) are ungrammatical, they are certainly understandable. The lexical semantics should specify what it is that these two classesshare, such that they have grammatical intransitive forms, but equally important is the characterization of how they differ, such that the latter class permits no transitive form. Other useful alternation patterns include the conative, as illustrated below in (7)- (10):4 (7) a. Mary h.9-.t. the target . b. Mary ,!!.Qtat the target . (8) a. Mary scraped the window. b. Mary scraped at the window. (9) a. The cat touched my leg. b. *The cat touched at my leg.

10

Chapter 2

(10) a. Mary hQ .t. the arrow (at the target ). b. * Mary hQi at the arrow. The question is whether it is possible to identify the semantic discriminants leading to the distinct behavior of the transitive verbs above, while still explaining why (9b)- (10b) are ungrammatical.S Perhapseven more interesting is how the polysemy of those verbs taking multiple forms can be representedlexically. What the examples above clearly show is that participation in one grammatical alternation does not sufficiently determine the semantic class of the verb. In fact, even once a complete cataloguing of participa tion in alternation classesis achieved, we must ask ourselvesjust what we have accomplished. Descriptively, we may have achieveda great deal, in terms of how verbs behave according to semantically-labeled classes. But we must realize that explaining the behavior of a verb's semantic class can come only from acknowledgingthat the syntactic patterns in an alternation are not independent of the information carried by the arguments characterized in the very patterns themselves. In other words, the diversity of complement types that a verb or other category may take is in large part also determined by the semantics of the complements themselves. One of the methodological points I will argue is that alternation classifications do not constitute theory. Indeed, as Levin (1993) herself points out , the theoretical mechanismswhich give rise to the descriptive distribution of syntactic behavior are not transparent in the classesby themselves. Still another kind of syntactic diagnostic that seemsto have some theoretical utility is polyadicity more narrowly construed. As Bresnan (1982), Fillmore (1986), and Levin (1993) point out , there are not only argument changing alternations such as those discussedabove, but also argument dropping alternations as well. The rule of "indefinite NP deletion" is the term for the following alternation paradigm: ( 11 )

( 12 )

( 13 )

a . The

woman

~

her

b . The

woman

~

quickly

a . The

dog

devoured

meal

the

b . * The

dog

devoured

a . John

. dm ! ! k his beer

quickly

. cookie

.

. feverishly

.

.

11


b. JohngI~nk feverishly . (14) a. John~~! ~~ his beerfeverishly . b . * John gulped feverishly .

(15) a. Mary hummed a song while she walked. b . Mary hummed while she walked .

(16) a. Mary performed a song while she ate her dinner. b . * Mary performed while she ate her dinner . In the examples here one might attribute the possibility of object -drop to an aspectual difference between the verbs being contrasted . That

is, while eat denotes an activity of unbounded duration (at least lexically) , devour, one might argue, denotes a transition . Although devour is generally considered a manner specification of the verb eat, it carries a completive implicature that is absent from eat. Similar remarks hold for the other two pairs above: while drink is an activity , gulp carries the implicature of completive aspect ; and finally , while hum is an activity , perform has a completive aspect lexically .6 If this were a complete account of the above data , we might expect it to explain the patterns of deletion for the other cases of complement dropping . This would seem difficult for many of the verbs entering into

this alternation . For example, as Fillmore (1986) points out , there are cases where near synonyms seem to behave differently with respect to

licensing of complement-drop (cf. (17) and (18)).

(17) ( 18

As

a. Mary k! ~ to start her car in the morning. b. Mary .tI!~ in the morning.

)

we

all

a

.

Mary

b

.

*

can

see

In

,

no

to

ditransitive

.

start

Pustejovsky

in

one

drop

addition

( cf

to

attempted

complement

mented

below

attempted

Mary

her

the

semantic

cases

car

the

morning

.

.

parameter

will

be

sufficient

to

explain

. 7

transitive

- intransitive

- transitive

,

in

morning

polyadicity

shifts

1992

)

such

:

(19) a. John~~ ~ a book to Mary.

,

as

those

there

shown

are

well

in

( 19

- docu

) -

-

( 24

)

12

Chapter 2

b. *John gave abook . (20) a. John ~~~ a lecture to the academy. b. John gave a lecture. (21) a. John mailed a book to his brother. b. * John mailed a book. (22) a. John mailed a letter to his brother.

b. Johnmaileda letter .

(23) a. Bill showeda book to Mary. b. *Bill showeda book. (24) a. Bill showeda movieto the audience . b. Bill showeda movie. Thus , in certain cases, the otherwise obligatory expression of the goal argument is dropped and the verb becomes a simple transitive . What allows the alternation , I will argue in later chapters , is the interaction of the verbal semantics with semantic information itself .

from the complement

Such grammatical alternations can be used throughout the grammar of a language to make semantic distinctions on the basis of syntactic behavior . Using category and selectional information as well as gram matical alternation data , words can be grouped into semantic classes following more or less predictable syntactic behaviors . Nevertheless , it is still necessary to explain why these and just these grammatical forms are part of a certain alternation class. This is addressed in chapter 9 below . Finally , let us consider briefly one of the oldest semantic classifications for verbs , that of aspectual class or Aktionsarten . The essential idea behind this classification is that verbs and verb phrases differ in the kinds of eventualities in the world they denote . It is normally assumed that there are at least three aspectual types : state, activity , and event, where the last class is itself sometimes broken down into accomplishment , and achievement events .8 For example , the verb walk in sentence (25) denotes an activity of unspecified duration . That is, the sentence itself does not convey information regarding the temporal extent of the activ ity , although deictically it is an event in the past which did terminate .9

13


(25) a. Mary walked yesterday. b. Mary walked to her house yesterday. Such 1963

a ,

Vendler

Dowty and the

sentence ,

,

1979

drink

.

1967

) .

of

a

( 25b

her

logical

home

.

house

.

of

the

type

as

the

are

verbs

which

the

verbs

build

plishment

( 26

)

conveys however

seem

,

verbs

the

to

lexically

their

there

is

is

( 25a

when

over

to

1993

,

work

,

) ,

with

to

the

process

Mary

is

at

event

an

activity .

use

,

activity

accomplishment

transitive

logical

,

the

accomplishments

a

,

run

reference that

default

typical

Kenny

her

assert

an

.

1972 ,

as

explicit

does

denote

denote

, in

, sleep

terminated

activity

to

lexically

because

)

the said

Verkuyl

information

making

( 25b

,

( cf

are

Mary

not

activity

1978

same

that

,

seems

to

activity

,

is

an

of

whereby

destroy

denote

Mourelatos

activity

walk

and

a .

Maryhyilt Mary

) in

direct

the ( 26b

a

Creation mance

existence

of

) , the

,

For

example

, denote

culmination

to

.

there ,

accom

the

-

activity

or

As

permit

best

the

piano

the

sonata

in

in

( 27b

such ( 28

)

)

above

as

that

an in

both ,

a .

Mary

walked

b .

Mary

. Ql ! ilt

an

hour

to a

house

the

for is

i .e . ,

( 29

store in

in a

year

the

do

an .

)

accomplishments not

hour

.

.

testing

modification

so - called

lexical )

,

the

and

Perfor

-

accom

-

: 10

.

diagnostic

and

activities

act

is

) . minutes

classic

,

' s

table

. ( 27a structure

accomplishment

derived

while

usage

hours

a

accomplishments

complement

15

Mary

as

. of

the

in

, one

denotes

act

activity

( for

of

denotable

her

on

played

culmination

example

depending

E ! ~ ~ ~ ~

phrase

the

both

Mary

modification

)

the

play ) ,

is something of

Mary

verb

.

of

a .

adverbials

Notice

house

b .

or

poral

the

only

as ( 27b

illustrated

verb

table

consequence

are

such usage

)

. the

nonexistence

- verbs - verbs

plishment

house

destroyed

culmination

( 28

) ,

sentence

verb

b .

( 26a

such

,

to

.

while

( 27

said

1949

,

of

events

performed

is

Although

culmination

This

Just

,

constraint at

)

examples

duration

has

Ryle

Sentence

walking

In

,

( 25a

Other

additional

temporal

a

as

frame

whether by

tem

adverbials license

.

14

Chapter2

(29) a. * John gr2!!k in 20 minutes. b. *Mary worked in an hour. The

frame

adverbial

reference

to

and

( 29b

)

The

is

instantaneously

.

is

it

not

.

Of

)

( 31

,

John

gi

~

fuylli1

b

.

John

c

.

Mary

arrived

point

adverbial

a

.

She

b

.

The

.

J

the

specific

by

the

an

an

( 33

)

make

in

( 29a

)

~

!

state

thought

that

at

noon

)

,

of

,

( 30b

)

has

and

( 30c

an

)

- like

as

.

as

occurring

point

such

( cf

.

just

as

,

a

adverbials

achievement

at

3

Dowty

,

the

qual

-

pm

1979

is

) .

3

pm

.

.

is

not

restricted

verbs

at

10

3

am

show

at

seminar

pm

achievements

,

:

.

sonata

hour

: 00

to

below

: 00

the

his

4

of

achievement

.

channel

at

of

at

noon

2

.

: 30

pm

.

d

.

He

deijvereg

.

indicates

the

starting

time

of

an

event

of

some

.

lexical

could

in

as

not

( 32

in

)

,

properties

possibly

where

( 32a

)

specified

with

)

~

is

point

performed

apparently

lexically

( 32

phrase

that

change

( 30a

modification

~

a

sentences

accomplishment

~ ~

that

activity

verb

missing

something

an

wallet

at

- adverbial

activity

appear

his

lecture

sentences

the

in

but

pm

the

ames

are

factors

3

pianist

duration

What

or

is

change

by

at

swam

point

,

,

denotes

with

his

Here

verb

precondition

in

the

modification

sentence

.

c

a

results

where

one

a

examples

)

,

classification

example

that

,

the

the

state

that

but

gradual

a

course

as

event

,

For

a

Hence

suggestive

( 30

an

does

to

that

of

aspectual

achievement

ity

require

change

conventional

accomplishment

change

to

explicit

.

last

An

seems

an

we

to

an

of

be

see

a

either

a

"bare

( cf

.

a

.

Mary

~

cookies

b

.

Mary

~

a

a

.

Brown

and

cookie

Root

and

.

(

.

Inc

build

or

( 33b

activity

) )

of

( 32b

)

.

mass

term

,

. il1

thereby

assuming

)

the

runway

in

eat

above

)

QY

consider

Tehran

from

Similarly

mentioned

.

( accomplishment

.

in

affected

,

meaning

as

object

)

be

instance

the

accomplishment

plural

can

For

in

verb

( 33a

verb

.

shift

accomplishment

reading

the

lexical

.

,

can

15


b .

The

Brown

and

presence

ically

of

telic

1986

,

,

subject

patterns

aspectual

)

)

a .

* John

b .

* The

a .

John

b .

The

Finally

, )

cates

individual

the

as of

as

the

( 36

)

of ( 37

( 37

,

sick

)

)

.

,

a .

John

b .

Watching

c .

Bill

these clearly

a .

* Bill * John

,

pluralization

and

finish of

.

Normally

these

,

verbs

subjects

,

as

suggests

himself the

wiped

,

on

typically

TV

permit

tall

can

be

,

Properties - permanent

predicates

. II

himself

overweight intelligent

in

brandy

the with

( 36

) .

hungry us

our

- level

years the

. coffee

predicates

. Great

is

.

John

serving

individual

over

It

construction

sentences

made

, -

with

. non

-

re

identified

predicates

cheap

predi

intelligent

individual

resultative

the

before

stative

an

- level

the

that

~

as

with

stage of in

with

~ ~

and

Carlson

of

that

- level

forms

Following

kinds such

called

k

.

two

.

himself read

!

states

identified

shown

counter

constructions

ate

individual usually

as

.

properties

are

commercial

the

dog

lifetime

in ,

.

Predicates as

been

predicate

his

are

appears

dog

of

of

have

his

.

.

its

and

illustrates

b .

typ

Bach

complementation

plural

distinguish

thought

clean

drank

on ~

- level

These

, and

on

behavior

throughout

typically

culminating

None as

which

a .

.

~ gi

can

stage

be

individuals

class

begin

with

flea

fleas

the ) , we

and

directly ,

from from

complements

arrive

to

( 1989

might

hungry

states this

~ ~ ~ ! ~ ~

examine

less

a

to

began

- level

individual

such

( cf

) .

comes

predicates

.

of

process

resulting

as

Asia

interpretation

details

such

~ ~ ~ ! ~ ~

began

or

for shift

as

began

Kratzer

more

the

Southwest

.

guests

us

1989

aspectual

same

in

unbounded

predicates

the

began

overweight

,

an

an

grammatical

but

guest

let

and

,

Krifka

distinction

( 1977

tains

) ,

runways

to

predicates

not

( 34

~

shifts

event

achievement

are in

( 35

and

aspectual

achievements illustrated

( 34

,

Q , yil

object

)

of

of

with

and

1993

.

plural

indication

the

an

bare

Inc

completive

Verkuyl

Another of

a

( or

Root

Books

.

.

,

16

Chapter2

One final characteristic for distinguishing

activities from accomplish -

ments, known as the "imperfective paradox" (cf. Bach, 1986, Dowty, 1979) , involves the possible entailments from the progressiveaspect. To illustrate the nature of this paradox, consider the sentencesin (38). (38) a. John is running . (Therefore, John has run.) b. John is building a house. (*Therefore, John has built a house.) What

this

difference

in entailment

indicates

is whether

an action

is ho -

mogeneousin nature or has a culmination of some sort. Sentence(38a) is an activity

and entails

the statement

John has run . That

is , John has

already engagedin some running. Sentence(38b), on the other hand, does not allow the entailment John has built a house because building is not a homogeneous process, but rather culminates in a changed state , i .e., it is an accomplishment . Thus , if x is ,t > : b. every woman E e, t >, t >: APVx(woman(x) -t P (x)] c. John E e type-shifts to e, t >t >: AP[P (j )] Type shifting has been useful for bringing together two distinct approaches to the semantics of interrogatives, as proposed recently by Groenendijk and Stokhof (1989). Other applications have also been proposed, and are explored in Partee (1985) and Dowty (1988).1 In Pustejovsky (1993) a particular application of type-shifting is discussed, illustrated in (3) below. (3) John considersMary ~jQQ} . In (3), the type of the NP a fool is changedto that of a predicate, <e , t >. Thus, Mary (of type e) and the predicate can combine in the standard fashion. Following a suggestion in Partee (1985), we can represent the verb consider as selecting for an argument of a particular type, namely a predicative phrase.2 (4) a. John considersMary ~_fulli . b. a fool: AP3x (fool (x) 1\ P (x)], c. Mary : E e

e, t > ,t > ~

<e , t >

The type that consider selectsfor by default is <e , t >, as seenwith the infinitival complement in (5), and it is exactly this type of complement selection which illustrates type coercion, to which I return below.3 (5) John considers [MarY]e [to be a fool] <ett > 0 The attraction of type-shifting is that it will provide us with one of the tools neededfor expressingthe semantics-to-syntax mapping in the grammar; namely, it can capture the semantic relatednessbetween systematically ambiguouslexical items. For example, Klein and Sag (1985) show how the "raised" and "non-raised" forms with believe(cf. (6)), and Equi and non-Equi sentenceswith prefer (cf. (7)) can be accounted for (cf. also Gazdar, 1982). Their analysis in effect provides one solution to the polymorphic syntactic nature of these verbs.

108

Chapter7

(6) a. John believes Marytobehonest . b. John believes thatMaryishonest . (7) a. Maryprefers toprogram inCommonlisp . b. Maryprefers forherstudents toprogram inCommonlisp -. They suggest lexical entries for believe and prefer as shown in (8) and (9) , where IR and IE are type -shifting operators for Raising and Equi , respectively . (8) a. believe E <S J, t , the interpretation it assumesin this position is clearly that of a proposition. We saw in (34) above that there are at least two possible event interpretations associatedwith a noun such as book(and subsequentlywith the NP containing it ), but in fact there is neither an event reading for (32c) nor is there a "reconstructed proposition" from the qualia structure of the NP. How, then, doesthis reading comeabout? Recall from our discussionin chapter 6 and the qualia structure in (34) above that the semantics of the bookis more complex than standardly assumedfor a nominal term phrase. Namely, the noun bookis a complex (or dot) object, and its type is info .physobj . Recall that this is the result of the type constructor lcp repeated here applying to the two types info and physobj : (33)

a : info

a : physobj

lcp(a ) : info . physobj

The dot object is the logical pairing of the sensesdenoted by the individ ual types in the complex type. That is, following the treatment taken for relational nominals outlined in Pustejovsky (1989) and in Pustejovsky (1994), we assumethat nominals such as bookare a sort of container ,

Generative Mechanisms in Semantics

119

which are further specified as a relation between a physical object and the

textual

information

contained

within

it .

Along with the complex type , we need to define the operations over this object which effectively foreground a particular sense. The coercion operations projecting one type from the complex type are a special case

of type pumping (or projection), and can be simply defined as El and ~ 2 below . These two operations , together with the dot object itself will form the definition of the type cluster we called an lcp in the previous chapter .

(34) lcp = { 0"1 . 0"2, E1[0"1 . 0"2] : 0' 1, E2[0' 1 . 0'2] : 0' 2} Hence , the lcp for book provides for the following aliases, licensing the polysemous behavior of the lexical item demonstrated previously . Assuming

the operators

El

and E2 to be defined

as above , then the full

type cluster , first mentioned in 6.2, can be given as in (35c) , with the appropriate derivation .

(35) a. ~ 1[info 'physobj ] :info b. E2[info .physobj ]:physobj c. info .physobj _lcp = { info .physobj , info , physobj } Intuitively , then , a book always denotes , in part , textual information . Because of the subtyping relations between these two types and the type

proposition

(prop ):

book ~ text

~ prop

the NP the book is able to "stand in " for a full propositional expression . This is illustrated in the tree structure below , once again showing semantic

selection

:

(36)

the

120

Chapter 7

Given this discussion, we return to sentence (32c) and the coercion involving the NP, the book. The type derivation below shows both the application of the type pumping operator, E, and the subtyping relation , 8 . The lcp interpretation of book, and there exists a subtyping relation O :' ~ S' . 37 (

)

The

El

(

info

associated

(

38

.

physobj

)

semantic

)

a

b

c

d

.

Mary

:

info

,

6

[ info

interpretation

believes

the

.

believe

(

"' 8

.

believe

'

(

A8

.

believe

'

(

" ' the

(

E1

(

(

book

the

the

-

-

-

is

book

:

given

5

prop

]

below

:

info

in

-

(

38

)

t

prop

.

.

book

)

book

:

:

)

)

(

info

prop

Mary

)

)

(

Mary

)

)

(

Mary

=

)

>

= > -

)

What the composition in (38) illustrates is how the deep typing on the complement is satisfied by enriching the semanticsof the object description as well as adding functionality to the manner in which phrases compose (cf. Pustejovsky, 1994). Finally , consider how the typing requirement for a verb such as sell, selecting for a complement of type ind , is satisfied by the lcp of book(cf. (39)). (39) Mary sold the book to John .

In sentence (39), the type physobj is available by application of the coercion operator E2, and likewise, the type ind is available through subtyping with the coercion operator 8 : i.e., physobj :,0::; ind .13 (40) a. Mary sold the book to John. b. sell (John )(e (~ 2(the -book )))(Mary ) =} c. sell (John ) (e (the - book : physobj )) (Mary ) =} d. sell (John )(the -book : ind ) (Mary ) Now let us return to the more difficult caseof coercion involving the complement of believe, namely (32d) above, repeated below. (41) Mary believes John . (NP)

121


Unlike the case of an NP such as the book, where the type required by selection was part of the underlying semanticsof the expression, this is not the casein (41). What is similar , however, is the interpretation of the complement; namely, it refers to a proposition. Hence, under this view, there is coercion of an individual , John, to a propositional interpretation . In terms of typing , the shift simply involvestaking the type of an NP like John (i .e., a generalized quantifier, ind , prop > , prop and coercing it to a proposition (prop ).

(42) Type-shift:

ind , prop>,prop>

=>

prop

By itself, of course , this is both uninteresting and unintuitive . It is onething to simplyprovidethe appropriate type shiftingmechanisms to satisfythe typerequirements of a complement , suchasin (32c) and (32d). Thetask, however , is to explainwhysucha typeis available as an interpretationto beginwith. That is, why shouldJohnbe ableto metonymically projectto a propositional denotation , andindeed , what wouldthismean ? Intuitively,weunderstand thissentence to meanthat, whatJohnutters, typicallyor onthis oneoccasion , is believed by Mary. Thus, thereis an implicitrelationbetween Johnandthe propositionc/J expressed by him, conveyed by somemeansof communication R, andit is this proposition whichis beingjudged. Therelationhereis illustrated with an individualtype, indo (43) john ~ 33R[c/> 1\ [R(john,<jJ)])] The propositionct>is not simplyany proposition , but is construedas that whichis communicated by John. Our knowledge of humanagents as possiblespeakers or writersallowsus to infer (semantically ) that sucha relationasR in (43) exists, andthat it is sortallyrestrictedto communicative acts. Furtherdetails , however , on whatthis relationis seems out of the domainof linguisticknowledge , andproperlypart of defaultreasoning andabduction (cf. Lascarides andAsher,1993 , Hobbs , 1987 , and Hobbset al., 1988 ). Because the interpretationfrom the compositional semantics is conservative - by not completely specifying whatthe relationbetween thetwotypespropandind is- I will referto this particularapplication of typechange asminimalcoercion . That is, the typingenvironment is satisfied , but the semantics is not completely determined .14 Theassociated semantic derivationis below .15

122

Chapter7

( 44 )

As

a .

Mary believe

( Ap2 ( John

believe

( A3cjJ3R

[ / \

[ R ( John

, rjJ ) ] ) ] ) ( Mary

d .

believe

( A3 <j >3R

[ <j > A

[ R ( John

, cjJ) ] ] ) ( Mary

before

nicative

relation

information

,

but

say

radio

limits to

in

of

chapter

Co

In

section

we

begin

by

sense

( 45 )

( 47 )

( 48 )

Safire

is

.

For

most

example

likely

to not

lexical

inferences

,

an

-

, be -

you

is

how

with

commu semantic

that

illustrates also

of

, lexical

factors

example

only

type

something

, but

a

, the

verb

creation

bake

sense

. Q~ ~

. d the

potato

. d the

cake

verbs

in

heard believe

only

the

semantics issue

we

is return

.

Normally entries

are

Mary

~ !E ~i

the

table

b .

Mary

~ ! . E~ i

the

table

on have

function

verbs

a

application

to

meanings

dry

John

hammered

the

metal

John

hammered

the

metal

illustrate

, both

a

. this

change

of

) .

a .

Mary

.

between

a

presence

of

enter , 1995

.

. flat

the to

.

a .

car

one

al . , 1988

ambiguous

Rappaport

b .

the

two et

would and

a .

waxed

baking

, depending

Levin

involving describes

.

, lexicons ( cf .

of has

polysemy

, co - composition

.

( 46 ) - ( 49 )

reading

ballogical

than

( Atkins

Q~ ~

transition

ver

, more

polysemy

John

a

of

) . Briefly

, superficially the

below

cases , 1991a

studying

and

, the

( 46 )

those

John

lexical

other

William

a .

adjectival rate

possible

what

, however

believe

this

b .

and

not

commonsense

discuss

allows

( 45 )

Similarly ing

by to

knowledge

( Pustejovsky which

. In

supplied

What

linguistic

is

)

is specify

=?-

- composition

structure

state

.

subsequent

co - composition

point

This

)

10 .

7 .2

We

further

as .

believing

wrote

=?-

interpretation to

probably

, while

license

this

is is

he

)

complete

defined

rather

that

natural

is

Schorr

on

something

a

) ) ( Mary

operating R

Daniel

able

,

inferences

lieving

.

c .

mentioned

him

John

b .

subsequent

to

believes

.

both ) .

process a

forms

read

-

sepa

-

resultative as

123


b. Mary ~~ ~ the car clean.

(49) a. Mary ~ . b. Mary ~ herselfsick. In order to capture the logical polysemy in all these casesand obviate the need for multiple listings of words, Pustejovsky (1991a) proposed that the complements carry information which acts on the governing verb, essentially taking the verb as argument and shifting its event type. Here we will make this proposal more explicit and describe what mechanism makes such an operation available. Let us assumethat the lexical structure for a verb such as bakeis that given below.

(50)

.I

I wish to claim that there is only one sensefor bake, and that any other readings are derived through generativemechanismsin composition with its arguments. What needsto be explained, however, is why nouns such as cake, bread and cookie "shift" the meaning of the verb bake, while other nouns (such as potato and garlic) do not. Intuitively , we would like to capture the fact that the former objects are prototypically brought about by the activity they are in composition with , something that the qualia structure should be able to express. Assume that the qualia for cakemakes referenceto an AGENTIVEvalue of ;\y;\e3x[bake(e, y, x)]. cake

ARGSTR ( 51

=

)

[

D - ARGl1 ARG

QUALIA

=

TELIC AGENTIVE

y : mass _ ind x= : food

=

CaNST FORMAL

= = =

]

y x

eat

( e2 =

bake

,z ,x

)

_ act

( el

,w

,y )

Notice that the AGENTIVEmakes referenceto the very process within which it is embeddedin this phrase, a relation which Pustejovsky (1991a) called cospecification.

124

Chapter 7

The

semantics

First

for

ment

its

The

governing

(B )

The

complement

The

7 and

the

verb

VP

The

derived

The

conditions

below

also

Keenan

by

then

a

we

, we

is

plements by

can

structure

for

the

:

in

a

derived

sense

AGENTIVE

becomes

the

, 1985

roles

FORMAL

of

matcha - ~- - - -

role

for

the

word

is

polysemous

not part verb

the

by

The

the

of

{ 3 , of

[ Qi

==

"Y ] ]

.

in

of

and

( 52 )

.

QSaC

Rather of

polysemous )

associated

Q Sa

/ 3) =

the

a

to with

by

the

Q S /3,

/ 3 . 16

putting , in

virtue

this

such of of

certain basic

"Y

qualia

sense

by

because add

=

and

that

,

[ Qi

QSanQS

bake

-

value

two

creation

cake

struc

. . .

suggests ,

two

Quale

these

like

view

meaning

verb

of

For

qualia a

[ QS {3

of

verbs

. This

:

is

bound

b with

senses

the

unification stated

a , with there

unification

type

appears

representation

type , if

lower

NP

are

UNIFICATION

, then

qualia

, a ( , a ) is

qualia

) :

greatest

co - specifying .

:

the

unique

call apply

QUALIA

. . .

on

The

( i . e . , those

[ QSa

weight

in .

I will can

< a , b > , and

both

itself

operation

WITH

define

derive

artifact

structure

verb

;

results

/ 3 , respectively ( 3,

the

semantic

co - composition

lowing

the

;

operation

Faltz

Further

contributed an

and can

.

verb

an

type

QS

/ 3 , as

structures

bake

-

unification in

include

complement

complement quale

APPLICATION

QSa

QSanQS

being

verb

argu

) .

qualia

These

.

the

feature

cake

a .

into

AGENTIVE

( the in

its

of

for

QA

structures

this

and

, a , of

, the

the

from

and

cases

=

to

and

which

expressions

the

applies

qualia

tures

of

)

operations

object

type

constituents

verbal

results under

shared

Thus

, a values

results

FORMAL

FUNCTION

]] ,

( bake

several

the

.

sense

( cf .

( 52 )

of the

from

binds

qualia

co - specifies

, where

results

Secondly

both

bake

complement

entire

. of

is , QA

of

composition verb

cake

co - composition

aspects

(A )

a

bake

identity

That

of

reflects

the

bake

application

verb

the .

operation

that

some

the

by

argument

The VP

of

, licensed

and

VP

function

structure

occurs

(C )

the

, conventional

it

com

-

meaning has

the

fol

-


125

(53)

The result of co-composition is a semantic representation at the VP level that is identical in structure to the lexical form for a creation verb such as build. What this claims is that the creation senseof bakeembedsthe change-of-state reading within it by systematic rules of composition.17 The sensearises not by lexical enumeration, but generativelyin the semantics itself. As another brief example of how co-composition operates to give rise to "derived" senses, consider the lexical conflation examples first discussed in Talmy (1985), and characterized as lexical subordination in Levin and Rapoport (1988); namely, the systematic polysemy of the verb float and related verbs in the following construction: (54) a. The bottle is floating in the river. b. The bottle floated under the bridge. There is a systematic polysemy exhibited by these examples, where a processinterpretation for the verb float, as in (54a), is shifted to a tran sitional reading in sentence(54b). A partial lexical representation for the verb is given in (55).

(55)

In (54b), both manner and motion aspects of meaning are conflated into a new senseof float. In Pustejovsky (1991a) , it was suggestedthat


127

the lexicon. Once a mechanism is in place which admits of making use of information from complements and other non-functor elements in a phrase, it is interesting to imagine how the notion of co-composition can be extended to handle other phenomena. Three applications are ill ustrative to this point ; namely, manner co-composition, feature transcription , and light verb specification. I will present an analysis of one phenomenon related to the semanticsof light verbs in chapter 10, and discussseveral further applications of this mechanism involving the underspecification of semantic forms. The other phenomenamentioned here are discussed in Pustejovsky (forthcoming) .18

7.3

Selective Binding

Now we return to the problem of adjectival polysemy presentedearlier. In chanters 3 and 4,. I discussedthree types of polysemy with adjectival ~ modification , repeated below :

(59) a. We will need a ~ boat to get back in time . b. John is a ~ typist . c. Th .t drivers will be caught and ticketed. (60) a. John put on a ~ album during dinner. b. I 'd like a really bright bulb for my desk. c. Mary dumped the pasta into the boiling pot . (61) a. The man is ~ . b. John is a ~ man c. That was truly a ~

day (event, occasion).

These are interesting from our point of view becauseof the apparent noncompositional nature of the modification as well as the productivity of these constructions. Consider first the examplesin (59). Recall that the standard view on selectionwithin an SEL for these types of adjectives is to enumerate the senses. This requires that every finely-nuanced sense of fast must be typed specifically for the noun class or type it modifies. We observed, however, that the meaning of fast is determined largely by the semantics of the head it is in construction with . Following our discussionof qualia structure in 6.2, we can now give more substanceto this observation.

128

Chapter 7

In predicativeposition, adjectivessuchasfast are ambiguousaswell, beingableto modify both NPs (cf. (62)) aswell asVPs (cf. (63) below). (62) a. That was M ! You're back already? b. Your dog is ~ . (63) a. Emanuel Ax 's hand moved so ~ were a blur . b . Mary

Even

in

the

pretation

was

individual

of

the

interpretation while

dog . The

There are

modifier

two

are

itself

Consider

issues

how

of

fast

to

be

able

the

( 63 ) are

of

the

( 62 ) , however an

event

.

moving

car .

, the In

inter

being

gone

is predicated

standard

-

( 62a ) , the

of someone

when

both

do

we

such

at

Assume

as

in

, 1985 in

modifier

arrive

the

here

event

, of

predicative

context

or

such

events

depends

on

as fast

; and

the

( b ) the

semantics

of

). ( 5gb ) above would

the

: ( a ) adjectives

individuals

. Treating

give

of

qualia

is as follows

the

us Ax [typist

interpretation

that

typist

with

modify

sentence

intersective

?"

dealt

to

adjective

the

an

typing

nominal

event

in

( cf . Bartsch

again

as simply

at

to

of the

being

control

in

reference

property

, being of

head

But

readings

duration

sentences

two

interpretation

fast

the

maintain

.

polysemous

the

fu ! ; . to

requires

to

( 62b ) , the

interpretations

too

predicative

refers

in

the

driving

during the scherzo they

" John

structure

for

adjective

' ( x ) Afast is

a typist

a deverbal

fast ' ( x ) J. who

is

agentive

:

typist ( 64 )

If

fast

ARGSTR

=

QUALIA

=

is

an

position

that

namely

, that

( 65 )

The event

adjective

such as

predicate

is

an

allow

( e ,x ) ]

, then the

]

there

desired

is

no

standard

interpretation

mode

for

of

sentence

com

=

able

Ae [type

to

make

contained

in

interpretation

selective

binding

' (e, X)

fast

available the possible

.

1\

( e ) ] . . .]

a selective

qualia is

for

the

a generative

interpretation head

noun

of . 19

mechanism

-

( 5gb ) ,

( 65 ) .

expression

to

= x : human

[ TELIC FORMAL = type = X

would in

AX [ . . . Telic

makes refer

event

[ ARGI

an

What I will

129


(66) SELECTIVEBINDING: If a is of type , ~ is of type b , and the qualia structure of (3, QS {3, has quale , q of type a, then a ~ is of type b , where

[a{3] = (3 n a (q{3). The semantic device giving this interpretation

can be seen as treating

the adjective as a function and applying it to a particular quale within the N ' that it is in composition with . This same interpretive mechanism will now allow us to account

for the contextualized

senses for evaluative

adjectives such as good, discussed in chapters 2 and 3, and as used in

(67) below.2o (67) a good knife: a knife that cuts well Recall from chapter 6 that the qualia structure for knife is that given in (68) .

knife ARGSTR = [ ARG1 = x:tool ]

(68)

QUALIA= [ TELIC = cut FORMAL = x (e,x ,y) ] Sincegoodfunctions as an event predicate , the event description jn the TELIC quale interpretation given in ( 67 ) . Finally , consider the selective modification (69) below.

it is able to selectively of the

noun , resulting

within

the

NP

modify in the

in sentence

(69) a. John boughta long record. b. a long record: a recordwhoseplayingtime is long. Assuming

that

that

the

the

result

playing

adjective

the

qualia

structure

for

record

lon .g has an interpretation

is a selective

," as illustrated

interpretation

is that

as an event

over the TELIC

in ( 71 ) .

record = [ARGl =Y x:physobj (70) ARGSTR ARCl= .lnfo ] QUALIA = FORMAL = R(x.y) TELIC = play-Icp (e,x.y) ] [ info .physobj

given

event

in ( 70 ) and

predicate of " the

, then record

130

Chapter 7

(71)Ax [. . . Telic= [play(x)(e) Ae A long(e)] . . .J

Noticethat, because theoverall typeoftheNPdoesnotchangeas a resultoftheselective binding operation, sentences suchas (69)arenot

typingviolations.That is, whilethe verb buyselectsfor an individual for its internalargument,the adjectivelongselectsfor an event. Since

thesetypesarebothsatisfied, butat different levels inthecomposition,

the sentence is well-formed.

Whattheseadjectives demonstrate is not a violation or puzzlefor

coercionand selection;rather, they serveto illustratethe subselective

bindingproperties ofdifferent classesof adjectives, as modifying differentfacetsor qualiaofthehead,byvirtueoftheirtype.Modification by an adjective suchas long,fast,or bright,canbeseenaseventpredica-

tion, selectivelybindingthe appropriatequaleof the head. 2 As wesaw

above, theadjectives inthesecasesmodify a distinguished eventpredicate(i.e.,the TELIC quale)associated withthehead,readforbook, and

illuminate forbulb.Thus,a longbook isinterpreted asonetakinga long

timeto read,whilea brightbulbis a bulbwhichshinesbrightlywhen

illuminated.Theseadjectivescan be comparedto modifierssuchas ex-

pensive andopaque intheNPsin(72),bothofwhichreferto thephysical objectratherthanan activityor stateassociated withtheobject. (72) a. an expensive book b. an opaque bulb

Theseadjectives canbeseenasmodifying theFORMAL roleofthequalia

structure

for these nouns.

Thereis anotherinteresting phenomenon thatmightbeexplained by qualiaandselective binding mechanisms; namely, polysemous adjectives suchas oldandnewas usedin an oldfriendanda newneighbor, with thenon-intersective interpretation. Representationally, theadjective selectively modifies ina waysimilartofastandlonginouranalysis above. (73) a. an Qj j friend:(a friendfora longtime) b. {Formal [ AxBy = friend(x,y)A[Telic =

)teS{friend_state(eS,x,y) A long(eS)}} ...J Theadjective oldin thisexample is ambiguous between predication of theindividual andthatofthefriendship itself.Thedurative reading faciliated byselective bindingin (73a)doesnotseemto be available,


131

however , with non-relational nouns such as movie and house, as illus -

trated in (74): (74) a. * an Qld movie : (one that I have had for a long time ) b . * an Qld house: (one that I have had for a long time ) It is not clear whether this is a semantic or pragmatic distinction . While friendship is something that continues , and one can have any number of friendships , possession of a house is typically something that persists uniquely , in successive stages. Hence, the phrase in (74b) does not mean " a house that I 've had for a long time ," but rather assumes the sense of previously lived in or owned. Such an analysis , however , cannot explain why (74a) is not a possible interpretation for a noun such as movie .22 In either case, it is the prepositional phrase of mine or the genitive which brings out the internal (i .e., selective ) reading which is possible , namely , the sense of former . In this section , I have tried to illustrate some further enrichments to the mechanisms of semantic composition , in order to capture the creative use of adjectives in adjective -noun constructions . The rule of selective binding is, in fact , an instance of a more general mechanism facilitating the selection of substructures through a path of features in the semantic description of a phrase . This is developed more fully in Pustejovsky and Johnston (forthcoming ) .

7 .4

Semantic Selection

As stated in above chapter 4, the goal of lexical semantic theory is to provide the foundation on which projection to syntax can be accomplished. One of the avenuesexplored in this researchis to determine to what extent syntactic behavior can be seen as following from semantic selection, and what is ,due to constraints on syntactic form. It is not our aim to completely reduce syntactic selection to underlying semantic types, a rather misguided goal in itself, since the semantic types would simply increasein specialization to reflect distinct syntactic patterning . Rather, the goal is to see how the grammar is affected by a specific approach to modeling lexical semantic knowledge. Although many of the problems in mapping semantic forms to syntax are left unanswered here, it is necessaryto identify the scopeof the changesthat result from adopting a generative lexical approach to semantics.

132

Chapter 7

There are two ways in which a generative lexicon as outlined above affects the mapping from lexical semanticsto syntax: (A ) There is no one-to-one mapping from underlying semantic types to syntactic representations; rather, a syntactic phrase is only fully interpretable within the specific semantic context within which it is embedded; (B) Because the representation of semantic information in the qualia structure , argument structure , and extended event structure is richer than what conventional models associatewith a word, a more complex model of filtering and checking is necessaryfor restricting the output to actual syntactic form. 7 .4.1

Canonical Syntactic Forms

Given the previous discussionof the mechanicsand operation of coercion and other generative devices, what appearsobvious is that the semantic type associated with an argument for a verb allows for a multiplicity of syntactic expressions; this forces us to addressthe issue of what role semantic selection is playing in generalin the grammar. The view taken here is that the grammar associatesa canonical syntactic form with a semantic type, but there are in fact many possiblerealizations for a type due to the different possibilities available from generative devices such as coercion and co-composition. Our methodology here, however, is to first look for semantic distinguishability betweendistinct sets of syntactic realizations. If no empirical evidence presents itself suggesting that two lexical items are truly distinguished in their semantic types, then the residual difference between these items is attributable to syntactic distinctions rather than semantic typing . For an expression of a particular type, there is a unique syntactic realization, which is able to adequately expressthe semanticsof the expression, and which can be seen as the most direct realization of the type , what I will call the canonical syntactic form (csf). The alternative syntactic forms which are available to an expressionare licensed by virtue of principles of semantic "recoverability," such that the information conveyedby the expressionwhosecanonical form is 7ris also able to be recovered by generative devices in another syntactic realization, 1[' . The proposal put forth here can be seenas an elaboration and extension of the ideas discussedin Chomsky (1986), where the notion of canonical

:ve Mechanisms in Semantics Generati

133

syntactic realization is introduced. In previous chapters, I discussedhow semantic types can be expressed in the lexicalization process. The lexical conceptualparadigm (lcp) can be seenas the lexicalization of a number of distinct semantic types into one lexical form. The semantic type which results is a dot object, and the lcp itself is the type cluster of the individual types together with the dot object . 1~he syntactic realizations for a semantic type are in part determined by virtue of the semantic type. Hence the form will exhibit the behavior of each type it is composedof in addition to the unique behavior of the dot object. The set of phrasal alternations associated with a semantic type is similar in nature to an lcp, in that they are determined by virtue of the semantic type. Unlike the lcp, however, which can be seen as taking a number of types and fusing them into a lexical item , the mapping from a single semantic type to syntactic forms is a one-to-many relation. What determinesthis mapping is the manner in which the generative devices are allowed to transform one semantic type into ano,ther, under a set of specified constraints. Factors that are outside of the realm of the syntax-to-semantic mapping might restrict an otherwise legitimate set of mappings or syntactic possibilities. As we shall see in chapter 8, an lcp may not necessarily exist as a unique lexicalization, but may be distributed across several lexical entries. The lcp itself as a conceptual relation may exist, but logical polysemy in a single lexical item would not occur for that concept in this language. For example, while the noun newspaperis logically polysemousbetweenthe organization and the printed information-containing object , the noun book refers only to the latter , while the noun author makes referenceto the "producer" of the book. In what follows, I will discusstwo examples of how the semantics of a lexical item is able to determine the resulting syntactic forms possible for complements to that item . The casesI wish to examine include selectional distinctions betweenthe following two minimal pairs: (75) a. the verbs like and enjoy; b. the interrogative selecting verbs ask and wonder. What emergesfrom this discussionis that semantic selection can in fact be a good indicator of the syntactic behavior of a lexical item , but only when viewed together with the generative operations that connect the distinct syntactic forms for a particular semantic type.

134

Chapter 7

Let us assume that , for any semantic

type , 7 , there is a unique

canon -

ical syntactic form {csf} that expressesthis type as a syntactic object, Xi . Expressed as a function , csf can be viewed as that unique mapping from semantic types to syntactic forms , cst 7 - t Xi , such that :

(76) For every type 7 in the set of semantic types, there is a function, canonical syntactic form {cst} , such that CSf (7) = Xi , except for a . when

7 = T , or

b . when

7 = -1,

in which case csf is undefined .

A syntactic expression, yj , of type a (where csf( a) = yj ), is substitutable for the csf of a type 7 only if this type is fully recoverable from licensed

semantic

operations

on a .

For example , assume a verb is typed for selecting an argument of

type 7, where cst( 7) = xi . The direct realization of this type as Xi is well -formed , assuming surface constraints are satisfied in the syntax ;

that is, [V Xi . . .] is a legitimate structure. Now consider the same verb appearing together with a syntactic phrase yj , of type a2; i .e.,

[V yj . . .]. From what we have said above, this structure is well-formed only if 7 is fully recoverable from a2. This is illustrated in (77) below .

(77)

Canonical

Form X ?

The resulting set of syntactic forms associatedwith a particular semantic type will be called a phrasal paradigm.23 The operations ensuring recoverability on the semantic type in this case are exactly the generative devices introduced earlier in this chapter; namely, type coercion, selective binding , and co-composition. It should be pointed out that the overall enterprise of establishing a lexicon operating along the principles outlined in this work are to some extent independent of the successor failure of semantic selection determining syntactic behavior.

135


Let us turn now to specific examples illustrating

this mechanism , be-

ginning with the classic sub categorization differences between the verbs

enjoy and like. Observe in (78) and (79) how these verbs differ in their complement selection behavior :

(78) a. Mary likes to watch movies. b. Mary likes watching movies . c. Mary likes movies .

d. Mary likes (for) John to watch movies with her. e. Mary

likes that

John watches

movies with

her .

f . Mary likes it that John watches movies with her .

(79) a. Mary enjoys watching movies. b. Mary enjoys movies . The syntactic differences between these and other verbs are typically used as counterexamples to strong views of the semantic selection hy-

potheses(as held by Lakoff, 1971, Dixon , 1984, Wierzbicka, 1988, Chomsky, 1986). The discussionfrom the early literature on these differences were suggestive , but not conclusive of a real semantic type distinction . Thjs was due, in part , to the lack of a sufficiently rich type system that was both linguistically motivated as well as formally characterized . Viewed within the current enterprise , however , there are diagnostics in dicating that the semantics of these verbs are in fact distinguished by the type of complement each selects. Notice the distinction between these two verbs that emerges in the following discourse . Imagine that two roommates are discussing a chair in their apartment ; speaker A utters

sentence(80) to B , who in turn respondswith sentence(81). (80) I want to get rid of this chair. (81) But I ~

that chair!

The statement made by B in (81) is difficult for A to question, since the questioner is not privy to the actions . By uttering (81) , B towards the object , although now an alternate response to

attitudes of the hearer , only to his or her does not express any articulated attitude more explicit attitudes do exist . Consider (80) , that given in (82) :

(82) But I ~!~~~ that chair!

136

Chapter 7

Sentence (82) expressesa very different relation with respect to the complement. In this case, the proposition is readily verifiable, since it says of an object, that the speaker stands in a relation of performing some activity with the object; as such, (82) can be doubted, questioned, or denied, sincethe judgment relies on the person's observablebehaviors, entailed by the semanticsof the selectionalproperties of the verb. Hence, (83) is a legitimate responsefor A to make for (82) but not for (81). (83) But you never use it ! The distinction here is not just a pragmatic one, but is due to the underlyingly distinct types selectedfor by theseverbs. Namely, the verb enjoy selects an event function , while like selects for an attitude towards any type , which I represent simply as the top most type available to interpretation , i .e., T . This would include the semantic types event , prop , property , ind , and facti ve. These are shown with their respective canonical syntactic forms in (84).

(84)

VP[+ PRG ] NP S[+ INF] S[+ TNS] Becausethe verb like selectsfor T , it has no unique cst In fact, a unique csfexists only if there is a unique least upper bound in the type lattice . Notice from the figure that there is still only one canonical syntactic form given for the proposition even though full sentential infinitival and PRO-form infinitivals are often considered in free variation , and constrained due to binding constraints. What this states is that PRO interpretation might be usefully associatedwith a coerced form of the complement. In (85), the two verbs are compared, where their syntactic behaviors are associatedwith distinct semantic typings:


137

(85)

[ind]

What

this

straints

illustrates to

thing

give

to

a

forms

the

us

turn

the

semantic

of

lection

is of

properties

( 87 )

)

in

relation questions

John

asked

me

what

John

asked

me

the

wondered

a .

John * John

a .

John

b .

* John

didn

of

concealed of

ing

semantic

and

care

that

differences

case

care

verbs

The

( 87 ) - ( 88 )

the

by

between

strategy in

distinction

is these

the

employed

syntactic in

in

the

verbs

. 24

.

wa

care

to

~.

verbs

take

this

the

NP

by

Pesetsky

rather

than

complements

expressibility

selectional

.

.

attributed

two

in

the

.

was

temperature

and

se -

and

.

temperature

wonder

from

semantic

.

temperature

the

of

below

was

nature

interrogative

difference

temperature

temperature

interpretation assigned

distinction .

the

and

~ ~ at

care

of

complements

Observe

the

.

discussion

idea

.

the

't

the

question

absence

't

didn

) , the

exists

the

a

-

those

there

type

problem

interrogative

the

which

continuing

temperature

only

illustrating

( 86 )

what

wondered

for

con

important

and

semantic

( 1982

to

a .

inability

semantic

Pesetsky

all

verb

, the ) ,

under

. The

that

pair

, namely

between

a

base

( 1986

and

b .

( 88 )

a

1981

concealed

b .

their

Chomsky

illustrated

( 86 )

The

,

is

minimal

relation

In

discussed

problem

from

, operates

polymorphism

to

second

selection

( 1979

verbal

complements

our

example

discussion

coercion to

.

, for

of

as

now

complementation Grimshaw

coercion

previous

appear path

Let

how

appearance

from

legitimate

of

the

note

syntactic

is

S[+ INF] S[+ TNS]

NP

VP[+ PRG]

work should

of ,

ask

however be

( 1982 to

an

is

of

the

) to

the

underly

versus ,

first

form

-

wonder to

attributed

assume to

138

The

Chapter7

difference

would

appear

to

be that

ask selects

for

a true

inter -

rogative (interpreted as a set of propositions), while wonder selects for an attitude

towards

a set of propositions

. Even with

this discussion , it

might be argued that such a finely -articulated distinction between ask and wonder is not something that should be encoded in the semantics typing of the complements , since such a difference could not possibly be learned by the child in linguistic experience . Interestingly , such semantic

distinctions

seem to be more difficult

for children

to acquire , and

the failure to make these subtle semantic differences is accompanied by a systematic misuse of the complement patterns associated with these verbs (cf . Pinker , 1984, 1989) . If that is possible to maintain , then it is interesting to speculate that the semantic differences are motivated and furthermore obviate the case theory analysis for this example . The syntactic consequence of such a semantic split might be reflected in distinct tendencies towards transitivity , but that too would be as a result of the semantic What

typing . would

the semantic

types

of these two verbs

order to give rise to both the appropriately mentioned

above , as well as the syntactic

need to be , in

fine-tuned interpretations

forms possible

for their

com -

plement position ? Exactly that distinction made above; namely , wonder takes as its complement an attitude towards a question , while ask selects a question directly . Some useful work towards characterizing this

distinction has already been made by Groenendijk and Stokhof (1989), who distinguish

between know and wonder as extensional and inten -

sional complement interrogatives , respectively .25 Assume that semantic selection

assigns the internal

argument

of ask the type quest ion , which

is a shorthand for the type <s , t >, where an interrogative is seen as a function from worlds to answers. Following the observations made in Groenendijk and Stokhof (1989) and others , let the type selected for by wonder be , simply the intension of the type assigned to the complement of ask. This distinction turns out to have consequences for the syntactic complementation pattern for each verb , as we see below . The tree in (89) gives the two semantic types associated with the interrogative argument of the verbs ask and wonder along with their canonical

and non -canonical

syntactic

forms .

139


(89)

S[+ WH]

NP

It is possiblethat the subtleties of such distinctions are not really appreciated or recognizableuntil larger segments(or units) of analysis in the language are investigated, such as the way these verbs are embeddedin a particular discourse and context. This does not mean, however, that the information is not lexically encodedinto distinct semantic types for the complements. The NP complementation pattern for verbs like ask illustrates two aspect of the approach outlined here: A . the ability of the verb to coerceits complement; B . the ability of the complement to metonymically reconstruct the required coercing type from the semantic structure within the complement. Without detailing the mechanismsbehind these shifts in other classes, it is worth briefly looking at what the range of complementation patterns is for a small number of verbs, where the phrasal paradigm associated with each verb below is illustrative of a distinct semantic type

(90) LIKE-class: (like, hate, prefer) a. John would h~ Bill to leave. b. John h~ ~ (it ) that Bill left . c. John .h~ ~g to lose the game. d. John h~ .E?-dlosing the game. e. John .h~ ~ that he lost the game. (91) WAGER-class: ( wager, bet) a. * John wagered Bill to have left .

140

Chapter7

b . John wagered that Bill left . c. * John wagered to have left .

-class: (92) MAINTAIN a. * John maintainedBill to haveleft. b. John maintained(for a year) that Bill left. c. * John maintainedto haveleft.

(93) TRY-class: (try, attempt) a. * John Y !~ Bill to read the book. b. * John trig.Qthat Bill read the book. c. John k ! 4 to read the book. d. John t! !~ the book.

(94) REMEMBER -class: (remember, forget) a. John rememberedto lock the door. b. John rememberedthat he locked the door. c. John remembereglocking the door. d. John rememberegwhere he put the keys. e. John remember~Q his phone number. (95) PERSUADE -class: (PERSUADE , CONVINCE ) a. John convinced Mary to build a house. b. John convinced Mary that she should build a house. c. John convinced Mary that she had built a house. From our previous discussion, it follows that the broader the selectional possibilities for a verb, the more general is the semantic type associated with the complement. We saw above that the verb like selects for the type T , for which there is no specific canonical syntactic form (csJ) , but it is associatedwith the csfsfor its immediate subtypes; hence, indirectly , it behavesas though it has several csJs, along with the derived syntactic expressionswithin the phrasal paradigm for each type. The verbs .fo~ qet - ~ and rememberselect very generally for a factive inte~pretation of any type .

8

The Semantics of Nominals

8.1

Basic Issues

In this chapter , I give a general overview of theory can contribute towards a classification types . In studying the semantics of nominals major areas of concern from the perspective of

how a generative lexical of the different nominal , we can distinguish four a lexical semantic theory :

(1) The distinction in complement-taking behavior between nouns and verbs ;

(2) How nominalizations and event-denoting nominals are distinguished from their corresponding verbal representations and the events they denote : that is, what is the difference between an event represented as a sentence, and an event represented as an NP ? And likewise , for facts and propositions , represented as sentences and NPs .

(3) The representation of logical polysemy in nominals, such as window, record, book, and how these implicitly relational

nominals

relational nominals differ from

;

(4) How the semantics of nominals facilitates the richer compositional in terpretation required for characterizing natural language semantics as polymorphic ; that is, what allows for co-compositional interpreta tion in natural

language ?

The first two issues are of course related , and I will argue that they cannot be addressed independently of each other . Furthermore , a formal distinction

in how nominals

and verbs refer to event descriptions

is nec -

essary to show how the grammaticalization of events differs i 1} these two syntactic domains . This is presented in 8.3 and 8.4. The third point was addressed in part in chapter 6, but many questions remain regarding the descriptive power and the exact nature of dot objects and how argument and event p~rameters interact in the qualia expressions . These issues are discussed below in 8.1 and 8.2, where I distinguish between unified types and dotted types . The former

are a more formal

interpretation

of the or -

thogonal types of Pustejovsky and Boguraev (1993), incorporating the structures employed in Copestake (1992) and Copestake et at. (1993). The last point above was touched on briefly in chapter 7, and will be the subject of later discussion . This concerns the exact manner in which

142

Chapter 8

coercion and co-composition rules makes use of qualia-based information . The variation in the expression of complements between nouns and verbs has long been a major concern to linguists, and has motivated many of the shifts in the theory of grammar. We will not be able to address this issue until we have explored in somewhat more detail the semantics of nominals from a GL perspective. As mentioned in chapter 6, for any category we can potentially distinguish three distinct dimensionsalong which the elementsof that category can be analyzed semantically. With respect to nouns, the interpretation can vary according to the three dimensions below: (A ) ARGUMENTSTRUCTURE : How many arguments the nominal takes; what they are typed as; whether they are simple, unified, or complex types. (B) EVENT STRUCTURE : What events the nominal refers to , both explicitly and implicitly . (C) QUALIA STRUCTURE : What the basic predicative force of the nominal is, and what relational information is associatedwith the nominal, both explicitly and implicitly . The manner in which these three representational levels can help us classify the semantics and associatedbehavior of the different nominal types will becomeclear in the subsequentsections.

8.2

N ominals with Unified Types

For purposes of lexical representation, it is often necessaryto allow a lexical item , making referenceto a type in a lattice , to be able to inherit from multiple parents. For some conceptual structures and their associatedlexicalizations, we will also needto employ a schemeallowing multiple inheritance. This has been explored in a generative lexicon to some extent in Pustejovsky and Boguraev (1993) and more generally in Copestake (1992) and Copestake et ale (1993). Previous attempts at structuring conceptual hierarchies (whether explicitly language related or not) have made heavy use of multiple inheritance, as systems have to grapple with accounting for the fact that , according to particular lexical-conceptual projections, biased by a variety of context factors,

TheSemantics of Nominals

143

different aspects of objects become more or less prominent as context varies . Thus , as illustrated below , a "book " is _a "information ," as well as a "physical _object ; a "dictionary " is _a "physical _object ," as well as "reference" ; a "car " is _a both "vehicle " and an "artifact ," and so forth . The conventional view on these sorts of inheritance relations is shown in the figure below .

information

.._-----~ -~ -artifact -

reference

Figure 8.1 Conventional Representation of Inheritance Relations

Still , as descriptive as such relations may appear, models like these suffer from a very limited notion of lexical structure; one particular consequence of this is the ambiguity of class membership (or, in our termi nology, 'hidden' lexical ambiguity). Thus, even though elaborate mechanisms have been proposed to control and limit the flow of information along the generalization/ specialization links, there has been no theory to either ( a) explain how to assign structure to lexical items, or (b) specify lexical relations between lexical items in terms of links between only certain aspectsof their respective lexical structures. The approach presented here, with its several distinct levels of semantic description, and in particular the qualia structure , are relevant to just this issue. On this view, a lexical item inherits information according to the qualia structure it carries. In this way, the different sensesfor words can be rooted into suitable, but orthogonal lattices. To illustrate this point , consider the two is _a relations below, and the differencesin what relations the objects enter into .

144

Chapter 8

play is_a book read

dictionary

ok

is_a book

no

buy

ok

ok

consult

no

ok

begin

ok

no

Figure 8.2 Table of is - a relations

This table illustrates a serious problem with most current inheritance systems for lexical knowledge. Namely, although it might seemreasonable to think of both plays and dictionaries as "books," they behavevery differently in terms of how they are selectedby different relations. This suggeststhat a single lattice for inheritance is inadequate for capturing the different dimensionsof meaning for lexical items. In Pustejovsky and Boguraev (1993), a proposal is made in regards to the structure of Lexical Inheritance Theory, and the need for typed inheritance for lexical information . The proposal, although not fully formalized there, was to posit a separate lattice per role in the qualia structure . Briefly, inheritance through qualia amounts to the following relations for this example: (1) a. book is _formal phys-object b . book is _telic information c. d. e. f. g. h.

book is _agent information dictionary is _formal book dictionary is _telic reference dictionary is _agent compiled -material play is _agent literature play is _telic book

The different inheritance structures just mentioned can be illustrated by the diagram below .

145

The Semantics of N ominals

reference

F

T compiled_matter

A A

The qualia roles differentiate the lattice structures resulting in a typed inheritance , excluding the unwanted inferences listed above, and allow -

ing only the desired ones.l Let us now integrate

this perspective

into the typing

system outlined

in previous chapters. In Carpenter (1992), type unification is defined as the unique greatest lower bound , glb , for two types al and 0"2, i .e., 0"1n0"2. This will allow us to define those objects which are non -polysemous , yet do involve the logical conjunction of their types . For this reason , I will refer to these objects as unified types. To illustrate

the formal

properties

of nominals

with

unified

types ,

consider the following examples . A type lattice provides the essential characteristics of a concept as a type in a hierarchy . Orthogonal inher itance arises not from the multiple assignment of essential typing of an object but from constraints on functional information about the object

146

Chapter 8

.I

suppliedby the qualia. The type hierarchyshouldbe so constrained as to allow only orthogonaltypes to be unifiable (i.e., distinct qualia types); in other words, differentqualiamayunify to form a unifiedtype, if the qualia unificationis well-formed. For example, giventwo orthogonaltypes a and T with qualia structures (2a) and (2b), (2) a. u = [0- . . . [QT = a ]] b. T = [r . . . [QA = {3 ]] they may unify, anT , to form the unifiedtype a_T, with the resulting qualiastructuregivenin (3):

TELIC =~={3 (3) [o QUALIA = AGENTIVE -_r [ r\I ] ]

The creation of a concept (i.e., type) that refers both to a physical object and a proposition, however, is not a possible unified type, since the FORMALqualia values for the two concepts are not unifiable. The ability to construct types that are otherwise outside of the set of unifiable types is something enabled by the construction of dot objects with the lcp type constructor, mentioned in 6.2, and discussedbelow. Let us look at a few examplesof unified types to illustrate what role the qualia are playing in supplying orthogonal dimensionsof an object's denotation. Consider first the nominal food. As mentioned in chapter 6, the TELIC role specifies for a given concept its use or function ; for food this value is the activity of eating. The construction of the type associatedwith food arises from a qualia-based constraint on the type physobj that it be edible. food

(4)

ARGSTR= U LIA = Q A

ARGI = x :physobj ] TELIC = eat FORMAL = X(eP,y,x ) ]

Similarly , the semantics for an artifactual object states minimally that it is something that was created , made , manufactured , or brought about by some human activity . The value of the AGENTIVE role for the type physobj indicates just this relation to the object , as shown in (5) , where top is the top -level type in the lattice , as defined above in Section 7.4.1.

147


(5)

artifact ARGSTR= QUALIA =

D-ARGl y:human ] [ ARG 1== x:top = make [ TELIC FORMAL = X (eT,y,x) ]

This representation says nothing about the nature of the object except that it is human derived. That is, as an artifact , it could be a social construct, a verbal act, a physical object, and so on. Given the underspecified representation in (5) regarding what an artifact is, as opposed to how it came about, we can further restrict the type which partici pates in the qualia relation, effectively unifying a subtype with the qualia structure for the concept artifact . For example, to represent artifacts that are physical objects, we would like a type which combinesboth the FORMALvalue of a physobj and the AGENTIVEvalue of an artifact . Becausewe are treating the argument and the qualia relation both as types to be unified, it is possible to view the resulting lexical representation as a unified type, effectively arriving at a type structure similar to Copestake's (1992) treatment of greatest lower bound types. The resulting qualia structure is shown in (6) below.

(6)

y:human :physobj ] :~make (eT ,y,x)]

Another example of how the types instantiated as qualia structures can be unified to form new unified types is illustrated with the concepts of tool and artifactual tool . As discussedin chapter 6, a tool is simply defined as an object which has a TELICvalue indicating a purpose telic rather than a direct telic. That is, the predicative argument is the first argument in the TE~IC relation . Hence, for a word a , defining a TELIC value essentially classifiesit as something that has a use;

(7)

a ARGSTR =[ARG 1=x:top ] QUALIA =[FORMAL TELlO =R =(e x,x,y)]

Here again, this sayssimply that the object is usablein a certain capacity (i .e., TELIC), and saysnothing about how it cameabout (i.e, AGENTIVE ),

148

or

Chapter8

what

a

its

basic

naturally

device to

used

be

type

is , i . e . , its

occurring as

both

a tool

an

artifact

object of

persuasion

artifact

_ tool

FORMAL

being

and

, which

is

. a

role

used

as

.

tool

Restricting

tool

,

represented

the

however

as

Hence

a

in

,

, it ,

or

could to

nature

creates

a of

a

refer

to

rhetorical this

unified

object type

,

(8 ) .

a:

ARGSTR (

=

) QUALIA

x : top = y : human

TELIC

=

R ( e ,x ,y )

[ AGENTIVE FORMAL

Nouns

such

type

unification

be

as

knife

, discussed

very

specified

= =x

clearly

locally

for

in .

knife

make

6 .3

since

( eT

above

Observe ,

]

that lexical

,y ,x ) ]

, illustrate the

this

AGENTIVE

inheritance

operation value through

need the

.

not

= 1

=

8

of

[ ARGI D - ARG

FORMAL type

value artifact

will

bind

_ tool

.

the

AGENTIVE

to

the

value

inherited

from

the

(9)

ARGSTR

=

QUALIA

=

.

knife

D-ARGl y :physobj [ ARG 1 = = x :artifact _tool ] = cut [ TELIC FORMAL = X(e,x ,y ) ]

A look at the type lattice which results from such constructions illus trates the process of type unification , where unlabeled edges denote the FORMAL Quale. (10)

.

.

.

~

T entity , / /" 2 : " 1 Q ~ abstrac ~phys arti ~ tool _ o bj ~ . / ~ QAnQT .

.

.

ph ys _ai~t_ rt t1 oo

149


8.3

Nominals with Complex Types

Having discussedthe propertiesof unified types, we turn now to the more complexstructure associatedwith dot objects, first discussedin chapters6 and 7. With the help of dotted types and qualia structure, we cananalyzethe inherentlyrelationalaspectof certainnominalswithout having to treat them as overtly relationalin denotation. The dot object is a type whichnecessarily incorporatesthe meanings(i.e., types) of its simpletypes into the complexobject. The examplesI haveused throughoutthis work haveincludednounssuchas book , door, and novel. The mannerin which suchnounsarerelationalis expressed in the FOR MALquale, asoutlinedin 6.2. For the noun book , for example,a relation specifyinga particularcontainmentrelationbetweenan individualphysical object andinformationis explicitly part of the makeup of the noun'8 FORMAL role, as repeatedbelow: a (11)

ARGSTR = ARG 1 = X:Tl ARG2 Y:T2] QUALIA =

FORMAL = R(y,x) T1. T2-1CP ]

The predicativeproperty of book , as discussed in chapter6, is givenby its type as a dot object. Hence, we canview the FORMAL Qualehereas the heador predicativequaliarole in this structure, givingthe equivalent representationin (12): (12) [FORMAL = x.y : R(x, y)] This is, in turn, translatableasthe expressionin (13) below: (13) Ax.y3R[book(x.y) /\ FORMAL = R(x, y) . . .] Notice that we have constructed a type x .y in (13) above without concern for how it is related by subtyping in the lattice . Because of the qualia -based relation between the types in the FORMAL role of the object , there is no way of exhaustively defining the behavior or characteristics of the dot type in terms of the type lattice alone . We can , however , provide for a partial characterization of the type for purposes of type selection

150

Chapter 8

within

the grammar

object

a . {3 to be selected

.

Namely , we need a mechanism in an environment

of its dot elements , a or {3. I will call this particular Object Sub typing , e ., and define it below . ( 14)

for allowing

less specific

than

an

either

coercion

rule Dot

0' 1 . 0' 2, which

is a dot

DOT OBJECT SUBTYPING : , : 0" 1 . 0"2,

e ' [0"1 ~ 7 , 0"2 ~ 7 ] : 0' 1 . 0' 2 - t 7

8 -[0"1 ~ 7 , 0' 2 ~ 7 ] (, ) : 7 This

states that , given an expression

, of type

object , there is a subtyping relation possible between the dot object and a type 7 , just in case 7 is the least upper bound of both of the dot elements , 0"1 and 0"2; coercion furthermore in an environment normally typed for 7 . To illustrate

the utility

lectional

properties

towards

a dot object

what

specific

allows the dot object

of this operation

of the verb

to pass

, we need only recall the se-

like . By expressing

a specific

such as a book , we do not commit

attitude

ourselves

aspect of the book causes me to like it . It might

be its " bookness ," independent

of what

information

it contains

to

simply ( a dot

element ) , what it looks like ( another dot element ) , what I can do to it (a true coercion ) , and so on . In fact , the generic nature of the statement in ( 15a ) suggests that process of reading terpretation . ( 15)

( 16 )

it cannot

or writing

, since that

a . Mary

likes the book .

b . Mary

liked the book .

a . ?Mary b . Mary

be associated would

why a generic interpretation telic

an iterative

book , she enjoyed

event

reading

it ) . As argued

The application

of the dot object

of book is illustrated material

a generic

in -

paradigm

with

Mary

read that

discussion , enjoy seman -

and coerces its complement subtyping

in ( 18 ) below . Assume

," associated

for ( 16a ) is only possible

(i .e., every time

in previous

tically selects for an event function appropriate type is not present .

conceptual

not entail

single

enjoys the book .

This also explains

" printed

any specific

enjoyed the book .

with

lexical

with

, if the

rule above to the type that

book is a species of

the type print

-D1atter . Then , the

for this dot object

is given in ( 17 ) below .

151


(17) printJIlatter _lcp = {physobj.info , physobj, info } (18) a. e [physobj~ T] :physobj~ T b. e [info ~ T] :info ~ T c. e .[physobj~ T ,info ~ T] :physobj.info ~ T Hence , the type restriction for the complement position of like is satisfied by direct subtyping , as illustrated in (19) . ( 19) S / / / / " " " '" [human ] VP I a

/ V

-/

/ /"-.." " ~"" [T] e ' :physobj . info ~ T

like

the Jook

Now let us turn to how nominals denoting dot objects are different from relational nominals , such as father and sister . In chapter 6, I discussed simple and complex typed nominals , where the form of the qualia structure determined the general class of nominal a lexical item denotes . Hence , along the FORMAL dimension alone , we can distinguish between simple natural kind terms such as rock and man , as well as simple typed artifacts , such as knife , and complex typed nominals such as book and record. In spite of this difference in typing , however , all these noun classes are predicative nominals . The manner in which predicative nouns differ from nominals that are overtly relational in nature , such as brother , sister , and neighbor , must still be addressed. As discussed above in chapter 2, a relational noun denotes a set of individuals standing in relation to at least one other individual in a specific way.

(20) a. John's neighbor is visiting . b . Mary 's brothers came home together . In (2Gb) , Mary stands in a particular relation to the head brother , which is given directly in the semantics of the noun 's FORMAL role :

152

Chapter 8

(21)

The conditionsunder which the argumenty is existentiallyclosedor bound to a constantare discussedin Eschenbach (1993) and elsewhere (cf. Bierwisch,1983). The important point to noticeabout the represen tation above, however , is how the FORMAL Qualediffersfrom the forms we haveexaminedthus far.2 That is, unlikea nounsuchas cakeor rock, with simpletyping and a FORMAL valueno differentfrom the typing on the argument, relationalnounshavea "relationalrepresentation " in the FORMAL Quale . However , they must be distinguishedfrom the complex typed nominalssuchas book , whichalsocarry a relationalFORMAL value (cf. (3)):

(22)

ARGI= X:/ l ARGSTR = ARG2= Y:/ 2 a: ] ' 1. ' 2_lcp QUALIA = FORMAL = P{y,x)

]

The distinction betweenrelationalnominalsand nounssuchas bookis due to the latter being typed as dot objects. That is, although the complexobject denotedby bookis partially definedby a relation in the FORMALquale, its actual type is quite distinct from a simple typed relationalnominalsuchas brother. The ability of sucha nounto assume onecomplexpredicativeinterpretation, or oneof two simplepredicative interpretationsis what the lcp accomplishes (with the help of coercion operators). I return to this point in chapters9 and 10, whendiscussing constraintson coercion. It is perhapsworthwhile at this point to recall what the linguistic motivation is for the existenceof dot objects. There are, I believe, two major reasonsfor admitting suchobjectsinto our semantics : (i) SEMANTIC MOTIVATIONS : The knowledgewe haveof the concepts associatedwith doors, windows, books, computerprograms, etc. is not characterizable asthe conjunctionof simpletypes (or properties)

153


in a conventional type hierarchy . The predicates and relations for the lexical

item associated

with

such a concept

are characteristic

of that

concept alone . For example , the concepts of "reading " and "writing " are not conceivable

activity

without

the existence

of the concept to which the

is applied , i .e., the dot object of printed material .

(ii ) LEXICAL MOTIVATIONS : The dot object captures a specific type of logical polysemy , one that is not necessarily associated with true complement coercion . There is strong cross-linguistic evidence suggesting that the way such concepts are lexicalized is systematic and predictable . So far we have focussed our attention on what the dot object is not : it is not a unified

type , created

from the meet n on types ; nor is it a

standard generalization (or join ) on types, in the conventional senseof this operation (cf . Morrill , 1994) . I have characterized it as a Cartesian type product of n types , with a particularly retricted interpretation . The product 71 x 72, of types 71 and 72, each denoting sets, is the ordered pair

< t1 , t2 > , where

t1 E 71, t2 E 7"'2. But

the pairing

alone does

not adequately determine the semantics of the dot object ; rather , the relation , R , which

of the definition

structures

the component

types must be seen as part

of the semantics for the lcp type constructor

itself .

That is, in order for the dot object 71 . 72 to be well -formed , there must exist

a relation

R , that

" structures

"

the

elements

of 71 and

72 ; i .e . ,

R (t1, t2)' For nouns such as bookand record, the relation R is a subtype of "containment ," while for partially event-denoting nouns such as lunch

and sonata, the relation is more complex (cf. 8.5 below). For now, we might view the set of relations, { Ri } , as specializedtype product operators , where the specific relation is built into the construc tor

itself :

(23) { Ri } == oR! , R2, oo. , Rn Then , for a dot object nominal such as lunch , which is polysemous between "event " and " food " interpretations (cf . 8.5) , the appropriate dot

object is that shown in (24). (24 ) lunch : event , lunch : food lCPR!(lunch) : event . food The common noun lunch would then have the following interpretation :

154

Chapter 8

(25 ) Ax .y3Rl [lunch (x : event .y : food ) 1\ FORMAL = Rl ( X , y ) . . .J With

abstraction

eralize

over tuples

the semantics

phically

such as that

of determiners

to any tuple

type . This is discussed

( 1995b ) , and in some respects , is similar of parametric polymorphisms . Given turn

a mechanism

for creating

to distinguishing

related

lexical

different

- matter

_lcp

to apply

more fully

to Morrill

types such as dot objects , let us now the nouns

- matter

for semantically

book and newspaper , which

. This type has the lcp referred

= { physobj

.info

, physobj

, info

tion

that

the physical

one generally

disbelieves

in ( 27 ) , where New York ( 27 )

object

that

to

}

As one would expect with this type , there is a logical polysemy paper between

polymor -

in Pustejovsky

's ( 1994 ) treatment

classes of polysemies

items . Consider

both refer to the type print above , repeated here ; ( 26 ) print

in ( 25 ) , we will need to gen -

and quantifiers

for news -

one purchases , and the informa -

or believes . Consider

Times is a subtype

of newspaper

the sentences .

a. Eno the cat is sitting on yesterday 's New York Time ~. b . Yesterday 's New York Times really got me upset .

In addition

to this

complex

of senses , represented

by the dot object

shown above , the noun newspaper refers logically also to the organization which publishes it , a sense which is unavailable to the noun book . ( 28 )

a. The newspaper

has just

b . The newspaper is filing for wire tapping . ( 29 )

suit

against

the federal

government

a . * The book has raised the price of paperbacks . b . * The author

While

fired its sports editor .

is suing the book for breach of contract .

newspaper carries a sense corresponding

publishes

the paper , the noun

to the organization

which

book is unable to denote the publisher

of

the book in these contexts . A concept such as that denoted by the noun newspaper is really a construction from types , one of which is itself a dot object .


155

(30) newspaper_lcp = {print - matter .organization , print - matter , organization } As discussedin 8.6 below, whether a languageactually lexicalizesthe dot object representedabove in (30) by a single lexical item or not, the concept exists potentially with this structure in the semantic representation for the language. We discusssplit lexicalization below in 8.6. From these considerations, the observation here is that dot objects are constructed in a pair-wise recursive fashion, as illustrated in (31) and (32) below. There may, in fact, turn out to be instances of dot objects that are constructed from three dot elements or more, but this is question open to further investigation. (31) (32)

a : 0"1 a : 0"2 lcP1(a ) : 0"1 . 0"2 /3 : 0"3

/3 : 0"1 . 0"2

Icp2({3) : 0"3 . (0"1 . 0"2)

The lattice structure

relating the type associated with newspaper to

physob info organiza ~t m ~ atte '~ ".newsp . /(.)book

that for print - matter and book is shown in (33). (33)

The way in which a noun such as newspaperdenotes, however, is quite different from that of book, due to the type of lcp it belongs to . That is, the qualia structure for newspaperis a specialization of the lcp known as product .producer , where the AGENTIVEquale makes referenceto the dot element denoting the producer, and the FORMALquale refers to the product ; the dot object itself does not appear in the qualia except to define the type itself:

156

Chapter 8

ARGI ARG2

(34) -

Hence, although newspaper is logically polysemous , it can not denote the complete dot object , as can book. Rather , one sense or the other is available

for interpretation

, but not both .3

Given this type of representational strategy , one might ask what the appropriate

distinction

is between nominals

such as book and novel . Al -

though book may refer to novel in many contexts , the latter is obviously more informative and less extensive in its selectional distribution . They are both dot objects , and in fact , novel is arguably a subtype of the dot object book. But this subtyping alone does not inform us of what distinguishes these concepts ; namely , while a book is possibly an informa tion holder of potentially any type of information , a novel is restricted to a specific structure and form of the information , e.g., a narrative . The

relation

between

these

nominals

can

therefore

be characterized

as

the specialization of one dot element in the dot object . That is, for book denoting type physobj .info , the dot object associated with novel is physobj .narrati ve , where narrati ve ~ info . This is illustrated in

(35).

(35)

TheS_ emantics of Nominals

157

In this section , I have tried to better characterize the behavior and properties of the complex types known as dot objects . There are surely many questions that remain unanswered , but it is clear that the nature of the denotation relation for dot objects is distinct from the unified types presented in 8.2.

8.4

Propositions

and Event Descriptions

In order to understand the position taken here with regard to event descriptions and norninalizations, it is necessaryto digress briefly about our assumptions regarding the treatment of tense, and the distinction between propositions and events in semantics. I will make some very particular assumptionsabout the relation between events, propositions, and facts. An untensed utterance such as Mary build a housewill denote a set of events, and becomesa proposition only by virtue of being "tensed" (cf. Verkuyl , 1993). A fact is a proposition generally judged to be known as true . An unsaturated proposition, typically denoting a propositional function , will be replaced by an event function , to be defined below. The way in which a sentencedenotes events is distinct from the way NPs are event-denoting. To make this clearer, let us compare the referenceto the event of Mary 's arriving , expressedas first a sentencein (36a) and as an NP in (36b). (36) a. Mary arrived at 3 pm. b. Mary 's arrival at 3 pm. Although both (36a) and (36b) ostensibly may refer to the same event, they do not expressthe same content about this event, since (36a) is an assertion of an event having occurred while (36b) denotesan event with out an assertion that it has occurred. In order to make this distinction clearer, we will distinguish the way in which eventsare quantified as NPs or sentences. In (36a), the event is tensed and thereby interpreted as a proposition. I will argue that only through tensing of an event-selecting predicate can the event in (36b) contribute propositional information ; (37) Mary 's arrival was at 3 pm. Our assumptions regarding the connection between propositions and events is essentially the following: both untensed sentencesand NPs

158

Chapter 8

may denote event descriptions , but the manner with which they denote is quite different , distinguished by their types . A proposition is seen as the result of applying tense to an event description . To this end , I assume that tense acts as a generalized quantifier over event descriptions , and behaves in many respects similarly to SPEC in an N P structure . The proper name of an individual object is typically construed as denoting a type e directly . Let the simplest cases of event reference also be an individual , eO ' . Both are types that refer to individuated

objects ,

either spatial (e) or spatia-temporal (eO ' ). Consider the proper named event The Vietnam War or Vietnam as in (38) below. (38) We will never forget Vietnam . Just as the proper name Nixon is typed e, and yet is understood to refer to a particular

;'"1dividual , there is no reason for us to treat

Vietnam

any

differently , in terms of presupposition or existential closure due to the name alone . Consider the following two sentences.

(39) a. I read a book about ~ !~Q!! . b . I read

a book

about

The

Vietnam

War .

Since books can be about anything , the preposition about obviously selects

for a type

" general

that

individual

is illustrated

covers both

e and eO".

Let

us call this

type

a

," g , where e ~ g , eO' ~ g . The way about predicates

in (40) below .

(40) '\g'\x [about(x , g)] (V ietnameO ") The

type

for a quantifier

such as every or a is now represented

as

shown in (41), where 9 is the general individual type: (41)

g ,t >,

g , <eU, t

, <eU, t

For example , the quantifier a will have the following denotation , where : F is a variable

of type , and

( 1993

-

ex -

to

this

is

a substantive to

aspect

( cf .

) .

tense

anch

Verkuyl

other

, 1980

that

which

purposes

how

is

proposition

rest

expository

, 1976

relation

( 46 )

the

is

predicate

tense

primary

a language

, in

sor

, < eO' , t

in

, i .e . , V P ' ( N

that

relation is

the

English

Assume are

, the

of

, with

- denoting

[V P ( e , x ) ] )

role

aspect

anchors

through for

which

fact ,

system

in

meant

the

verb

quantifier

AxAe

are

Comrie

subject

generalized

=

that

"

its

a

) expresses

=

/ aspect

" tense

to

a

expression of

) ; ( 45b

[V P ]

In

The

description

examine

P

event

eO' , < eO' , t

of .

[ V P ]

.

and

interpretation

a .

us

remain

and

.

projects

applies

languages

marker

, < eO' , t

P / ( V PI

redefined

reading

tion

it

will

- denoting

b .

some

tense

the

))

quantifier

individual

functional

the

let

tional in

what which

, which for

( 45 )

generalized for

it , i . e . , N

variable

( e2 , el

: F eO" , respectively

VP

to

) 1\ Feu

e , < eO' , t

and

analyze

lower

applies

a

variables

resulting

dard

( el

notating

of

given

( x ) 1\ : Fe ( e , x ) ]

[ war

specified

pressions

in

(woman

an and

AE3i3e

event

Kamp

as is

a

function

itself

of

within and

[a ( i , n ) A

over

type an

event

descriptions

eO" , t > , t > . The interval

Reyle

( 1993

anch

( i , e ) 1\

structure )

( cf .

(e)]

also

, as En

,

anchoring explored

~]dExE3Y = [pd] (tg)

162

Chapter 8

NP . That

is , both untensed

sentences and common

event nouns denote

event descriptions . As mentioned earlier , Tense functions of S , just as Det is the specifier of N P .

as the specifier

Now let us see how NPs refer to events in the simplest complicated

modes of reference will be considered

section . Consider a simple the sentence in ( 56 ) . ( 56 ) A war occurred

event denoting

case . More

below and in the next

nominal

such as war , as in

last year .

I will assume that an NP may denote an event in one of two ways ; either it denotes an individual proper name event , such as Thanksgiving or Mardi Gras , of type eO " ; or it denotes a quantified expression of an event description , such as every war , which shown in (57 ) below .

is of type

eO" , t > , <eO" , t

, as

(57 ) [ every war ] == A: FeUAe13e2 [war ( e2) /\ : Feu(el , e2) ] The qualia

structure

for war is shown in (58 ) , where I assume that

logical agent of the activity Krifka , 1989 ) .

is the join of the two parties

the

( cf . Link , 1983 ,

war ARGSTR = [ D- ARG2 ARGI = yx :ind ] EVENTSTR = [ HEAD el El = el=:process ]

( 58 )

QUALIA = [ AGENTIVE = war _act (el ,XEBY) ] The basic predicative individuals

force of the noun is to denote an event between two

, both of which are represented

as default

ar ~uments , and are

optionally expressed in the syntax . They must be logically represented , however , because of the interpretation necessary for NPs such as those in ( 59 ) .4 ( 59 )

a. the war between the U .S. and Vietnam b . France 's war with Russia c . the American

There

are , furthermore

war with , other

Vietnam

uses of nouns such as war where there is

an extend .ed sense of attack or assault , such as the war on drugs and the war on poverty .

The Semantics of Nominals

163

by Aver b such as occur, selecting an event ar gument, is represented the following expression :

(60) APeCTPeCT (Ae2Ael[OCCUr (el, e2)]) The

semantic

tree

given

descri

selection in

ption

on

( 61 ) ,

the

where

selected

by

subject

we the

have

verb

for

this

simplified

verb the

is

similar

typing

for

to the

the event

.

( 61 ) [ prop

/

/

/

]

/ - " --"

", , -

Tense

S

I [

event

[ < event

, prop

> , prop

>] /

/

[ event

, prop -' ~

> ]

" " '~

,

]

VP I

a

V I occur

The of

event

description

tensing

, which

however

,

is

is

the

proposition

( 62 )

in

the

a .

A

~ i ~ e2 [i

:::::; n

First

,

( e , el

a

is

by

only

verbal

sentential

anchored

by

predication

event

virtue

. The

description

result

,

, namely

a

1\ anch

::3i ::3e2 [ i

:::::; n A anch

3i

~

::3e2 [ i

A

n

section

man

however with

the

section

( i , e2 ) 1\ & ( e2 ) ] ( Ae3el

[ war

( el

) A

)]])

c .

structed next

position

available

with

b .

below

( 63 )

only as

subject

.

next

peated

itself

same

[ occur

In

the

1\ anch

, we

will

( i , e ) A Ae3 ( i , e2 ) A

return

::3el ::3el

to

[ 'war

[ war

the

( el ( el

) 1\ [ occur

) 1\

example

[ occur

in

( e3 , el ( e2 , el

( 4 7b

)

) ] ] ( e2 ) ] ) ]]]

above

,

re -

:

's

arrival

,

we

the

occurred

will lcp

.

need type

yesterday

to

.

examine

constructor

how into

complex dot

objects

events , the

are subject

con

of

164

Chapter 8

Finally , let us revisit the semantics of true complement coercion in volving event selection , in order to examine the role played by the mechanisms introduced here for tense binding into the complement . Recall

from 7.4, that sentencessuch as (64) require coercion on the complement to reconstruct the type selected for by the verb , as well as to provide an appropriate interpretation of the complement .

(64) John enjoyed War and Peace. The complement of enjoy is an event description , <e , E>. (65) S

/ / '/ [ human ]

'-'--"-"" ' "", -", VP

I

/ / / --'-"""" "",

Q

V

[ <e , E >]

AxAe[read(e,x,war_and_peace)] .

enJoy

[

I

War

e , E > , <e ,

ana

Peace

]

The formal derivationassociatedwith this coercionis shownbelow. ( 66 )

a . John

enjoyed [enjoy

' ( QT

c . AeT

[enjoy

' ( AX , e [read

e . John f . AeT

this

propositions

) ) ( John

) ( eT ) ] =>-

( WP

) ( x ) ( e ) ] ) ( John

{ AX , eT [enjoY

' ( Ae [read

[enjoY

section are

, I

' ( 3e [read

have

related

event

descriptions

as the

relation

briefly

.

to of

( WP

these

analysis to

and

) ( x * ) ( e ) ] ) ( x * ) ( eT ) ] } =>) ( e ) ] ) ( John

a

view of

have propositions

of tense Pustejovsky

) ( eT ) ] =>-

) ( x ) ( e ) ] ( x * ) ) ( x * ) ( eT ) ] } =>-

treatment

questions

events

issues

( WP

) ( John

the

Many

the

( WP

presented

through

between

consequences discussion

.

' ( AX , e [read

over

further

( WP

Peace

{ AX , eT [enjoy

such

I defer

and

b . AeT

d . John

In

War

) ( eT ) ]

of

how

tense

been

events as

left to

facts

as a generalized ( forthcoming

a

and

function

unanswered , as well quantifier ).

, as .

165


8.5

Varieties of Nominalization

In Davidson (1967) and Parsons (1981), the connection between nominalizations and events was explored , and the motivation for adding an event variable to verbs as well as to nouns such as burning and arrival was supported by inference patterns that mirror the sentential struc tures associated with such nouns . For example , consider the pair of

sentencesin (67), where in (67a), two sentencesare connectedby a temporal connective, while in (67b) a nominalization is predicated in a way that results in a reading similar to (67a) . (67) a. When wood burns, it requires oxygen. b . The burning of wood requires oxygen . This pair illustrates similar information

how ing - nominalizations and sentences contribute for purposes of inferences . Davidson 's point was to

elevate the event to a first -order object in the logic , such that quantifi cation over such objects would be possible with event nominals such as the ing- phrase in (67b) above just as it is with sentences as in (67a) .

Higginbotham (1983, 1985), Verkuyl (1990, 1993) and Grimshaw (1990) extend this notion to the role that event quantification has on the syntax and examine

a broad

function

event quantification . One of the remaining difficult

of nominalization

types with

respect

to

questions in event semantics is to for -

mally characterize the linguistic devices which facilitate an interpreta tion of these sentences as making similar if not identical causal state -

ments. To illustrate this point , consider the sentencesin (68). (68) a. Charcoal requires oxygen to burn . b . The burning

of charcoal

: equir ~~ oxygen .

c. When charcoal burns, it Eequire~ oxygen. d . Every

burning

of charcoal

requires

oxygen .

There are several things to point out about these sentences. First , the

quantified expressionin (68d) is simply a more explicit wording of what is implicitly stated in (68b) and (68c). Secondly, the statement in (68a) brings out the functional dependency between the subject phrase char coal and the rationale clause to burn ; namely , that only together with the rationale clause does the subject satisfy the event selecting property of the verb require .

166

Chapter 8

Let

me

require

is

subject

illustrate polymorphic

( cf .

infinitival

) ) ,

( cf .

a .

Making

b .

The

c .

For

the

and

dry

point in

( 69a

VP

(6 9 )

this

or

( 69b

that

it

selects

a

split

allows

dive

further either

data for

.

an

construction

requires

require divers

-

some

Observe

event

with

how

directly

as

control

of

an

) ) .

this divers

with

to

suits

having

both make

both

wet

and

this

dive

wet

dry

and

suits

to

requires

dry

suits

make

.

this

( having

)

dive

both

.

wet

.

d . Making this dive requires the divers to have both wet and dry suits . Notice

how

the

constructions

( 70

)

( 71

)

While

to

a

.

*

b

.

The

synonym

.

*

.

Animals

need

c

.

*

can

states

of

temporal

overlap

need

)

affairs

g

! ! ~ ~

both

,

however

,

seen

does

not

allow

both

,

as

resulting

discussed

~

more

reference

in

a

in

~ . Q

like

.

and

to

dry

suits

make

this

.

dive

.

.

animals

event

wet

suits

.

breath

patterning

making

dry

oxygen

to

another

as

both

and

! ! ~ ~

of

is

to

having

wet

oxygen

breathing

event

be

~ gg

breathing

! ! ~

Every

verb

dive

n

Animals

an

sitions

this

divers

a

the

( 72

Making

b

relate

or

near

.

a

The

oxygen

want

verb

to

a

meaning

chapter

VeVxVl: loc [burn (e, x) /\ at (e, x , l ) oxygen(y)]

- verb

in

, 5

these

require

the

seems

generic

propo

quantification

like

5

.

over

following

,

-

events

where

0

is

:

- t 3e

' [ o

( e

, e

' )

1\

at

( e

' ,

y

,

1 )

1 \

I amnot concerned in this section, however , with justifying the relevance of events - which we havediscussed previouslyin any case-- as muchas establishingthe way that linguistic items makereferenceto them. To fully understandthe contributionmadeby nominalsto causalinference , we must first explainthe distinctionsin event-denotingtypes. Wesawfrom Section8.5 that simpleevent-denotingnominalssuchas war havea fairly direct representation in a qualia-basedstructure. What would the qualia structureand event-denotingpropertiesfor a nominal suchas burningbe, suchthat the readingsin (68) could be correlated ?

167


Notice

that

this

event

variable

is not

in both

the

( 73 ) VeVx [burning Regardless simply it

of

is telling

The

the

the

, that

verbal

expressions

of

semantics

reference

statement verb

appears

( x , y ) ]]

, this

require

of the

- like

, as in ( 73 ) .

( y ) 1\ require

this

of the

an individual

a Davidsonian

move

and

sentences

what

in

in subject

( 69 )

position

that

is required

by the

verb

is to

of ing - nominals

of

event the

description

such

as burning

should

make

argument

as the

, as the

verb , while

ungrammatical

still

ref -

requiring

forms

in

( 74 )

.6

( 74 )

a . * The

burning

b . * Burning The

event

tice

that

cated

placing

.

same

expression

suggest

underlying

underlying

representation the

and

behavior

even when

description

to

merely

expressiveness

polymorphic

us about

( as in ( 69b ) ) , the

erence

nominal

overall

the

above . Namely

by

( e , x ) 1\ wood ( x ) - 4 jy [oxygen

the

ignores

an event

answered

description the

requires

requires

associated

AGENTIVE

by the

oxygen

oxygen with

argument

argument

structure

.

. burning

is illustrated

, x , is necessarily assignment

in ( 75 ) . No -

expressed

, as indi -

.

burning ARGSTR = ( 75 )

EVENTSTR

emphasis

the

initial

out

by

Asher

aspect

of the

a. b.

suffix form

event In

= burn

that

being

referred

this

is due to the

verbal

system

in .

above

seems to be on

to , an observation

appears

for left - headed

- ing , resulting

in the

_act ( el ,x ) ]

of the sentences

fact , this

, at least

( 1993 ) claims

nominalization

( 76 )

]

( 1993 ) .

in general

progressive

]

in the interpretation

Asher

nominals

=

= x :phYSobj

HEAD el [ El = el= :process

= [ AGENTIVE

QUALIA The

[ ARGI

to

be

events , as we shall aspectualizing

an interpretation

pointed

a property

of

ing -

see below

nature similar

of the to

the

.

168

Chapter8

Extending

to

Asher

agentive

such

as

baker

subevents

R

,

( el

of

' s

,

e2

)

,

el

and

e2

.

are

coercing

the

Space

preparatory

.

way

all

-

,

an

aspectualizing

.

much

Thus

as

event

,

the

the

' nominal

-

verbs

and

veT

nominalizers

,

by

where

a

kind

.

The

is

simply

of

can

complete

,

,

and

such

cut

are

also

over

completed

the

not

as

quit

operations

the

)

to

you

a

similar

denote

( 76b

aspectualizer

stop

perform

and

that

verbs

as

and

discussed

reference

fact

the

-

launching

denotes

abort

not

the

com

party

)

forcing

associated

to

as

( 76a

already

b

to

and

,

the

,

both

,

such

seems

R

of

headedness

such

denoted

The

seem

,

pair

by

the

examination

in

event

causatives

nominal

,

derived

events

NP

.

a

denotes

as

event

form

as

,

ing

other

the

the

suffixes

as

to

nominal

such

this

,

predicates

events

-

simple

that

well

ing

launching

argument

as

be

similar

interpretation

the

- or

seen

.

both

the

nominalizer

relation

is

the

that

-

and

eo

temporal

nominals

suggest

the

,

5

that

changing

of

event

this

ing

- er

would

,

chapter

to

the

the

an

fact

aspectualized

this

on

for

a

In

argue

might

suffix

'

be

might

,

terminate

,

nominal

)

event

to

at

is

identical

one

Following

ing

in

phase

)

, e2

might

predicates

( 76b

short

a

,

That

through

( ~

polysemous

Shuttle

the

the

in

as

Indeed

say

event

R

compare

with

process

. ,

view

well

below

i . e

might

suffixation

related

encountered

event

as

.

,

we

by

agentive

,

alternative

plete

event

initial

as

phase

of

the

.

As

plausible

scrutiny

as

from

.

Right

.

as

There

the

ing

that

be

relations

( ei

an

less

-

process

,

e2

,

,

where

operation

which

given

of

*

than

are

-

appear

)

and

to

withstand

arrive

of

the

the

have

die

double

- headed

one

to

;

of

suffix

are

unheaded

-

.

an

event

would

be

. ,

,

event

the

or

i . e

Hence

- headed

,

of

ill

is

.

accept

,

.

,

)

structure

application

left

result

said

and

less

transitions

destruction

nominalization

event

a

the

as

headedness

in

much

- headed

as

such

we

as

are

left

nominals

what

indicates

the

already

not

achievements

ing

result

resulting

acceptable

does

,

( such

the

subevent

) )

it

processes

nominalizations

point

since

,

. ,

are

of

ion

first

,

the

( R

than

transitions

nominals

head

( i .e

nominals

with

- headed

-

seems

:

interpretation

is

Regarding

ing

-

no

there

analysis

observations

transitions

ing

is

as

right

this

two

- headed

able

B

advisor

the

we

An

A

and

then

operation

,

created

externalization

war

observation

nominals

expect

- formed

to

ing

( R

we

might

as

effectively

( el

,

e2

structure

to

would

event

.

) )

expect

- denoting

This

seems

= =


169

to be supportedby the contrastingbehaviorof arriving and arrival, as illustrated in (77)- (79) below. For somereason, the ing-form is grammatical only whensomereferenceis alsomadeto the culminatingstate of the event, as in (77b).7 (77) a. *The arriving of Johnwasgreetedwith mixedreactions. b. ?The arriving of John late wasmet with mixed reactions. (78) a. The arrival of Johnwasgreetedwith mixedreactions. - --- arrival -~~b. The of John late wasmet with mixedreactions.

(79) a. John 's arrivalwasgraceful . b. Thearrivalof thetrainis expected for3:00pm. Regarding not

the

actually

most

- ion

( 80 )

second

point

polysemous nominalizations

a . * The b . The

made between such

destroying destruction

( of was

above

, notice

that

and

result

process as

the

destruction

city

are

readings

are

, as

.

) was

widespread

ing - nominals

widespread

.

.

(81) a. *Theconstructing (ofthehouse ) hasadequate stability . b. Theconstruction hasadequate stability . Thewayin whichthe lexicalstructurereflectsthis singlesense , making reference onlyto the initial event,is illustratedin the representation in (82) of the semantics for the ing-nominalformconstructing . ing

(82)

El = el:process D-El = e2:state RESTR =

,

reveals

]

the

resulting

that

a

interpreta

the

manner

The

!

have

\

,

factive

to

sense

-

reading

similar

factive

1993

)

tried

.

in

true

is

comple

context

-

for

this

the

(

f

)

!

,

but

one

\

f

is

=

depart

that

(

this

arising

e

,

is

mary

)

not

through

a

]

reading

coercion

(

cf

.

10

to

the

fact

however

,

,

as

,

semantics

captures

,

109

' s

factives

in

anch

to

' s

,

lexicon

processes

\

Zucchi

section

(

]

,

.

n

NP

and

this

mously

in

in

110

the

of

description

point

in

Mary

treatment

above

: : ;

important

erative

give

a

general

polysemy

perspective

inherent

inability

of

other

on

in

many

nominals

how

a

gen

-

nominalization

to

behave

polyse

-

.

.

In

6

Lexicalization

this

and

section

,

icalization

I

theory

semantics

.

explored

,

sions

the

(

[ i

inherent

In

)

event

in

3i3e3f

Asher

1993

discussed

given

)

The

8

the

coercion

NP

(

shown

onto

ment

(

' s

of

fact

, <e, b . begin as a Raising verb : <eO ", >

204

Chapter9

Given that event-headednessacts to foreground or 'focus' a single quale of the verbal semantic representation, the effect of heading the final event from the lexical structure in (45), i .e., the FORMALrole, correspondsto the raising interpretation ; what is assertedis simply the initiation of an event, without explicit referenceto causal preconditions of the event. Consider now the semanticsof the control interpretation of begin, and how it interacts with coercion on the complement.

(47) Mary begana book. Assume

the

qualia

structure

for

book

to

be

as

given

in

(

48

)

.

book

ARGSTR

(

48

=

ARG2 ARG

[

)

QUALIA

=

= =

1

info

y x

= =

AGENT

heading

begin

,

trol

relation

(

49

the

we

initial

arrive

at

is

expressed

event

a

shadowing

.

the

~ (

of

y

e

, x , w

write

(

lexical

subject

]

lcp

hold

read

=

in

as

physobJ info

_

FORMAL TELlO

By

: :

. physobj

eT

) , x

, v

)

, x

)

representation

the

associated

culminating

and

event

event

function

,

while

complement

with

the

con

-

.

)

Following the discussionin chapter 7, the qualia can also be seenas type pumping operations, giving rise to new types for coercive environments. Thus for example, the type available to an expressiona with quale Qi of type T, can be seenas allowing the following type inference:

(50) a : a E9 Qi [a, T] : a -::;. T Qi [a , T] (a ) : T

205

The Lexical Semantics of Causation

This says that , given an expression a of type a , there is a coercion possible between 0" and T, which changes the type of a in this com position from a to T. We will illustrate the application of this coercion operation

below, as used in the beginexample above in (47). Becauseof the qualia structure of the complement inherited from that of its head book, the typing requirements specified by the governing verb begin are satisfied just in case the type of the NP is coerced to an event function .

As illustrated below in (51), 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

beg 'iln . (51)

:

the

(e

This corresponds

VP

.

C -) -

- 1- c, 1 ; ;;

within

11

derivation

- 1-

where

the

TELIC b

the

"'.(e~') .eQT --.".- :e~

role k

has been ,

selected

,

.

'

(e - 1- c,C -) wffi a 00 :e begin the book : -

to the selection

tree structure

shown below .

(52)

This states that the TELICrole of the book, AXAeT[read(eT, x , a-book)], is available as an alias for shifting the type of the NP . This metonymically reconstructed type , <e , ' >, is identical to that selected by the verb begin in complement position . After the coercion operation , the derivation proceeds as if a full predicate were present in the syntax .l0 The result of heading the final event in begin gives rise to either direct

event selection, as in (53a) below, or a raising construction, as in (53b). (53) a. The party began at midnight . b. Snow began to fall .

206

Chapter9

In (53a), the typing of the predicateis satisfieddirectly by the eventdescription in subjectposition (ignoringsometechnicaldetails). For the constructionin (53b), I assumethat raisingis accomplished by function composition(FC), in the mannerof Jacobson(1990). In particular, the raisingverb begin , of type <eO " , >, imposesthe type eO " on its complement. Assumingthe VP to fall, for example, in (53b), is <e, E>, then FC (begin,to fall ) returnsthe expressionAP[begin(jall (P)]. Finally, let me turn briefly to the semanticsof the verb finish and how the semantictype selectedby this verb corresponds to a different canonicalform as well as the associatedphrasalparadigm. There are two things to noteaboutfinish. Noticefirst that finish differsfrom begin in that it is not logicallypolysemous , havingonly a controlsense . That is, the raisingexamplesin (54) are ungrammatical . (54) a. *It hasfinishedraining. b. *The sun hasfinishedshiningin my eyes. c. *The acid finishedcorrodingthe marble. This would suggestthat finish is not lexically underspecifiedwith respectto headedness , as is begin, but is alreadyspecifiedwith a head. It is this lexicalspecificationthat givesriseto the controlreadingonly.II Considernow the sentences in (55), which appearto be raising constructions, contradictingthe claim madeaboveabout finish beinglexically specifiedas a controlverb. (55) a. The leaveshavefinishedfalling. b. The paint hasfinisheddrying. Thesedata would suggestthat a raising constructionis possiblewith finish with somenominals. But the sentences in (55) are best analyzed aspseudo -controlcases , and they arerestrictedto a certainwell-defined classofnominals. In general,theseverbsdo not passthe standardraising tests, but nounssuchas paint and leavesare exceptionsbecausethey carry qualia information indicating a kind of "autonomyof behavior" relativeto certainpredicates . Hence,paint, for example,is construedas a pseudo -agentin the control relation becauseof this property; i.e., it can dry on its own. It is interestingto observethat anothertype of intransitiveconstruction is possiblewith finish; namely, if the eventnominalin subject position has an agentivecomponent(cf. (56b) and (56c)), then a control

TheLexicalSemantics of Causation

interpretation

207

is possible in what would otherwise appear to be an in-

transitive (i.e., raising) construction. Pustejovsky and Bouillon (1995) refer

to these

as intransitive

control

constructions

.

(56) a. ??The party finishes at midnight . b . Class

will

finish

at 2 :00 pm .

c. The talk will finish by noon .

(57) a. *The rain will finish by noon. b . The rain will stop by noon . While

classes and talks have an apparent agentivity

and controllable

component to them , parties are less controllable , resulting in the less

acceptable (56a). Since rain is completely uncontrollable, it is ungrammatical in an intransitive

control construction with the verb finish (cf.

(57a)) . The verb stop, however , allows a raising interpretation and per mits the intransitive raising construction in (57b) . This verb is inter esting because it has both control

and raising

senses , yet does not allow

complement coercion at all . Observe that stop appears in the sentences

in (58) with a non-control construction, assumingthe senseof "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 reconstructing an ellipsed predicate , as in (59) . (59) a. John stopped the car . (from moving )

b. The referee stopped the clock. (from moving) c. Mary stopped the record. (from playing/ moving) What these data suggest is that the complement type of stop is not an event function , as with begin, but rather simply an event , where the type of the verb is <eO ' , <e , . 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 (cf. Chomsky, 1981, Dowty, 1985, and Farkas , 1988) . It is worth noting that , with the complement of stop specified as eO ' , it is clear that coercion is not possible since this is not among the type aliases for the NP complements (cf . Pustejovsky and

Bouillon , 1995 for discussion).12

208

9.4

Chapter 9

Experiencer

Predicates

Giventhe generalstatement of causation presented above , it is interest ing to speculate on whatroleverbalqualiamightplayin otherlexical causatives . Forexample , ashasbeenlongnotedandaswasmentioned in chapter2, the verbkill allowsfor a broadrangeof possible subject types, aslongasthe causalrelationbetween the argumentin subject positionandthe resultingstateis coherent in someway. (60) a. Johnkill. .dMary. b. Thegunk!~ g Mary. c. Thestormkill~ Mary. d. Thewark!ll~ Mary. e. John's shootingMarykilk.dher. In Pustejovsky (1991a ), it wassuggested that kill selects for aneventand coerces its subjectto thistype, in a mannersimilarto theinterpretations of enjoyand begin . Thisis essentially correct , but the detailsaremore complicated , andin fact, moreinteresting . Assume that the lexicalrepresentation for the verbkill is asgivenin (61), wherethe initial subevent is headed . kill

(61)

El = el:prOCess E2= e2:state EVENTSTR = RESTR = - B The assumed belief is that the speaker and hearer share ellipsed propo sitions

as common

belief . The speaker , by using such a rhetorical

form ,

is establishing coherencebetween (i .e., coercing) the propositions for the hearer

.

If the ellipsed proposition is the minor premise , then an enthymemic abduction results . This is illustrated by the following argument . A . Every American loves baseball . B . Therefore

, John

loves

baseball

.

C. [John is American.] [A 0 ' --t B '] !\ B = > 0 The assumed belief here is that the speaker and hearer know that the ellipsed proposition is true . The speaker in this case is using a non-analytic proposition as the major premise . This is the coercive assumption from which , with the common belief of C , the syllogism has rhetorical force . Finally , if the ellipsed proposition is the conclusion , then a simple enthymemic deduction results , shown below . A . All politicians are corrupt . B . Berlusconi is a politician .

C. [Berlusconi is corrupt .] [A Bf ~ Of] !\ B ~

0

238

Chapter10

Common knowledgeprovides the valid rule of syllogistic reasoning. But the speakerdoesnot completethe syllogism, leaving the inferenceas new and topical information conveyedor projected by the discourse. The effect of coercion and co-composition, when projected to discourselevel interpretations , could be viewed as a type of lexical enthemymic inference. Consider the example below, where the qualia contribute information that can be used in the discoursein a method similar to the inferencesdiscussedabove. A. B. C. C'.

Steven Kingbegan a newnovel . [Steven Kingis a writer.] AGENTIVE (novel ) = AzAx .YAeT [write(eT,z,x.y)] Steven Kingbegan to writea newnovel .

As argued in chapter 7, coercion requires that the verb begin selects an event function as its complement, but the interpretation is not fully determined by the lexical semantics alone. The knowledge of what the type of the NP Steven King is, and what qualia values mayor may not contribute to further specifying this interpretation , can be seen as discourse inferences that are biased by the compositional semantics of the sentence. Similar remarks hold for data mentioned in chapter 4 above, where the full interpretation of the type required by the coercion on the complement is logically separate from the type coercion itself. (37) a. Most commercial pilots ~~~ E Kennedy to Logan. b. Most commercial pilots prefer New York to Boston. These comments are merely suggestive, but establishing the contribution of lexical semantics to pragmatic and contextual inferences is an important area of research.

10.6

Summary

In this book, I have presented a framework for the semantic analysis of natural language, motivated by the following theoretical and computational concerns: (1) Explaining the interpretation of words in context;

Consequencesof a Generative Lexicon

239

(2)

for wordsfromfinite Deriving a potentiallyinfinitenumberof senses resources;

(3)

Accounting for the systematic relatedness between word senses in a formal and predictable way ;

(4)

Characterizing natural languages in terms of their polymorphic prop erties ; more specifically , studying what types of polymorphisms exist and how productive they are in natural language ;

(5) Providing

a semantic representation that is applicable cross-categori ally , and is not restricted to verb semantics alone .

By structuring the representation of words in a manner completely par allel to that of syntactic objects , we obviate the need to postulate lexical primitives , since the natural language vocabulary is itself the data set of primitives . Just as the generativity of syntactic structures is the major concern of syntactic theory , accounting for the generativity of senses in a language is the primary goal of lexical and compositional semantics . This , in my view , characterizes what a generative lexicon should be. There are obviously many questions that have been left unanswered in the current work . For example , issues relating to parsing , details of interpretation concerning quantification , and connections to pragmatic and commonsense inferencing mechanisms have not been seriously dealt with here , although they have contributed to the design of the frame work . I have attempted to present a model of semantics for natural language that provides the necessary lexical resources to perform these tasks , while simplifying and , I hope , clarifying the role of the lexicon in linguistic theory .

Notes

Chapter 1.

1

Weinreich (1964) , in making this distinction , was critiquing the somewhat arbi trary

manner

with

which

the

Webster

' s Third

Dictionary

handled

lexical

items

with multiple word senses. The observation shows an insight into the problem of distinguishing accidental sharing of lexical forms from linguistically motivated sense relations , that others did not generally recognize .

Chapter 1.

2

Perhaps the most notable exception to this claim or movement would be Fill more 's recent move towards the encoding of larger units of phrasal structure ,

known as constructions (cf. Fillmore , 1985) . Goldberg (1994) extends this perspective to a fairly broad range of phenomena , showing how constructions can adequately account for cases of polyvalency and polysemy . In some sense , this is not contradictory to my claim above , since larger structural units are essentially

lexicalized rather than subject to general rules of composition (cf. chapter 8 below) .

2.

This is still a contentious

point

and is an issue that is not at all resolved in the

community . Hobbs (1987) and Wilensky ( 1986,1991) , for example, question the distinction between commonsense knowledge and lexical knowledge . Even more fundamentally , some previous research in NLP and computational linguistics ,

such as Wilks (1975) and Schank (1975) saw no utility in distinguishing knowl edge of the world and knowledge

of how we talk about the world . More recently ,

Wilks ' work (cf. Wilks et al., 1988,1993) has pointed to examining how faithfully lexical knowledge

represents our knowledge

of the world , as encoded in machine -

readable formats such as dictionaries , e.g., LDOCE ( cf. Procter , 1978) . Schank has not recently addressed the issue of knowledge representation as it relates to language . Nevertheless , I will suggest below that there are good reasons , both methodological and empirical , for establishing just such a division . Pustejovsky and Bergler ( 1992 ) and Pustejovsky (1993 ) contain a good survey on how this

issue is addressed by the community . Saint-Dizier and Viegas (1995) compare some recent approaches

to addressing

this distinction

more formally . Cf . also

Atkins and Zampolli ( 1994) for approaches to integration of lexicographic and computational

methods .

3. The context within which an utterance is situated must reflect the state of the discourse, as many have recently argued, for example, Heim ( 1982) , Kamp and Reyle ( 1993) , Asher (1993), and recent trends within situation semantics. Just as important , however, is some notion of the purpose of the utterance . Grosz and Sidner (1986) , Pollack (1986) , Cohen, Morgan , and Pollack ( 1990) , and others have rightly stressed the importance of goals in laying the contextual groundwork for the communicative act . Thus , in some sense, the rhetorical structure of the discourse defines what the speaker wants to accomplish (cf. Hovy , 1993) . I will return to this point in chapter 10. 4. See Dixon (1991) and Levin (1993) for discussion. 5. The generalization , as Levin (1993) seesit , bears on verbs denoting both contact and motion (or perhaps more correctly , motion towards contact ) . Notice , however, that (9b) above becomes much better if we contextualize the sense with adjunct modification :

242

Notesto pages1023

(i)?Underthetable,thecatkepttouching at mylegwithits frontpaw. The issueof sensecontextualization willfiguremoreprominently later in our

discussion of co-compositionin chapter 7.

6. Aswewillseebelow,however, eventhisdistinction requiresmodification, since allthreeofthelexically marked activity verbscanphrasally orsententially appear

in transitional(i.e.,telic) contexts.Seechapter9 belowfor further discussion. 7. I arguein Pustejovsky(1995c)that the aspectualdistinctionin the indefiniteNP deletioncasesis in fact sufficientto explainwhichverbslicensethe alternation.

Forthe broaderclassof complement-drop phenomena, however, including the infinitival complement casesin (17)and(18)above,moreisinvolved thansimple

aspect.

8. On details of this, see Dowty (1979).

9. It is usefulhereto pointoutthe distinction betweentemporalculmination, where an eventof whateversort simplystops,and logicalculmination, wherebysomethingis fulfilledor finishedas a resultofthe activity(cf.alsoParsons,1990and Tenny, 1992).

10. Theissueofvariableaspectuality isdiscussed insomedetailin chapter9 interms of polysemy and event coercion.

11. Althoughthere are severaldiagnostics for indicatingmembership in oneor the otherclass,manyof themare simplyinadequate to accordwithspeakerintu-

itions. For example,both there-insertionand progressivetests seem to favor stage-levelpredicatesoverindividual-level ones,but there are manycounterexamplesto the generalization. I discussthe generalissueofstage-level predication below in chapter 10.

12. For crystallographers, of course,thereis a verydifferentexperiencewithsand as a material. They mightverywellreferto individualgrainsof sand as sand.

Theclassification ofa materialorthingin theworldis obviously drivenbyhow

we perceiveit and constructthe realityfor that object,givena certain modeof measurement.See Modrak (1987)and Hacking(1983). 13. I return to a discussionof this issuelater in chapter 7. 14. Chomsky(1957)seesthis as a clearmismatchbetweenthe structurethat Markov

modelsassign to suchsentences andthe underlying representations necessary to them.

understand

15. Anypropertyis of coursepossibleas the complement of easyor difficult when overtcomplementation is used,as illustratedin (i) and (ii): (i) a. That exam was diffiylt to grade.

b. The examhas beeneasyto write,but its goingto be verydifficultfor the students

to take.

(ii) This highwaywillbe easyforthe workcrewto pavein the summer.

Defaultinterpretationsof ellipsedexpressions willalwaysgivewayto overtlin-

guistic expressions.

16. Thisdefinition ignoresthe issueof opacityandnon-substitutability in opaque contexts.Katz (1972)has a usefuldiscussionon the generalnature of semantic

relations, whileBierwisch andSchreuder (1992) discuss theissuesofrepresentationandaccessforthedifferent semantic relations between lexicalitems.Apresjan (1973a)examines the linguistic contextsforsynonymy in particular,and howit isstructurally identified. Sparck Jones(1986) presents a helpful studyof criteriaforsemantic classification andtheidentification oftermsassynonyms.

243

Notes to pages 24- 43

17 .

The issue of presupposition and how lexical meaning relates to projection cannot be given a full discussion in this work . I will return , however, to these issues briefly in the context of factivity in chapter 7. Soames (1989) presents a helpful guide to most of the recent approaches to the treatment of presupposition in nat ural language.

Chapter 1.

2. 3.

3

Hirst treats taxiedasambiguousin this sentenceaswell. Although the uninflected form taxi is certainly ambiguous,the past tensemarkerhereactsto disambiguate the form, thereforewe neednot considerit for disambiguationpurposes. Seeoriginal articles by Katz (1965), Vendler (1963), on the useof "good." But we return to this in the next chapter. In

some

sense

proposed

,

interesting

a

the

raise

is

clear

in

Fodor

that

have

for

to

biguate

.

item

SYNSET

it

,

the

set

of

extensionally

i

)

CAT

(

ii

=

count

)

=

This

6.

of

1980

and

Chapter 1.

exact

most

Pollack

in

et

al

of

(

2

. ,

word

,

,

approach

properly

disam

1990a

to

)

.

On

information

)

)

namely

synonyms

;

,

i

. e

with

will

for

.

. ,

example

-

aid

this

in

view

also

those

,

store

a

words

This

that

would

help

in

:

(

1987

,

1988

)

.

and

)

,

to

Bookman

representations

Sag

(

1994

selection

is

do

reported

on

Memory

-

(

1994

in

Based

vary

several

Retrieval

not

trivial

,

varies

same

is

recently

.

performed

the

Strategies

and

)

are

selection

acknowledge

senses

,

lexical

such

and

techniques

Hirst

like

and

for

search

contrastive

,

-

Pollard

mechanisms

sense

Passing

1987

above

Aspects

of

researchers

of

,

given

,

-

1988

noun

work

word

biguation

Marker

sets

the

sur

,

raised

to

GENUS

with

Not

enumerative

distribution

line

with

HPSG

the

,

sense

Miller

.

border

starts

the

such

system

.

)

]

in

how

had

features

and

)

Mehler

-

of

noun

-

=

approach

Although

cf

associated

line2

codified

(

CAT

1986

that

message

count

SYNSET

5.

item

is

]

CAT

[

-

=

the

)

It

. "

using

complementary

of

critic

selectional

suggested

to

synonyms

sense

SYNSET linel

[

lexical

addition

(

to

and

Theory

,

1963

con

identical

Pinker

Fodor

Semantic

of

have

in

the

-

(

.

disambiguation

McClellan

in

Katz

system

Fodor

forms

sense

papers

ambiguities

a

in

appears

distinguishing

(

,

the

and

almost

and

.

in

rich

of

would

word

senses

strongest

researchers

senses

lexical

lexical

Explorations

fairly

Some

disambiguating

cf

contrastive

a

senses

the

some

employ

(

Katz

ambiguous

Rumelhart

work

that

"

.

since

representing

at

of

( cf

this

essay

much

for

attempts

theory

of

his

that

system

model

issues

,

changed

connectionist

critics

same

Weinreich

It

-

the

the

' t

same

structure

Katz

,

will

early

feature

early

prisingly

a

haven

the

that

structed

4.

things

essentially

basic

quite

and

bit

details

system

strategy

a

for

,

of

(

such

cf

to

disam

course

implementations

techniques

the

from

,

.

-

from

Hendler

as

Waltz

)

4

The context-dependenceof the predicatesmentionedabovewas noticed by Aristotle in the Categoriesand the NicomacheanEthics, where he speaksof the particular aspectof an individual to which is attributed a quality such as goodness; for examplea goodlyre-player or a goodsculptorspeaksto the function of

244


particular ness

aspect

; for

that

in

, 1967

)

. of

chapter

Examples

,

.

These

Street

who

to

a

of in

polysemy chapter

)

and 10 .

and

the

it

or

the

.

with

strategy

meaning

as

.

of

Vendler

adjectival

in

I

good

function

sculpting

interacts

general of

such to

Aristotle

return

to

's

this

in

( 1964 extensions

by

psychological

( 1974

the

matrix

I

( ii )

know

I

The

how

don

't

matrix

just

as

)

* 1 don

Whatever Van

tall

't

filters Valin

member

,

and

the

examines

the

this

question

See

Verma

and

constructions

Morrill

and

positionality

as

the

constituent

tion

, we

See

Pustejovsky

do

.

rich

senses

and

( 1973b

, 1981

study

of

Nunberg

are

and

the

's

) , of

Viegas

( 1994

and

Pustejovsky of

on

to

) ,

proposals

in

classes

perspective

how

pragmatics

recent

explored

related

and

adjectives

the )

,

polysemy

and

Bouillon

.

from

" questions

"

whetherwh

on

the

basis

of

- complementation and

with

tests .

questions

Ob

-

disappears

address )

)

for

make

a

degree

of

the

sentence

dimension further

filter

the

analysis

factive

in

the

verb

of

complementation

a

related of

verb

( 1990

of

the

extensive

,

) , in

survey

a

of

( 1993

and

the

)

syntac

somewhat

understand

) .

re and

Jackendoff

intend

verb

( iii

patterns

manner the

Wheeler

an

the

for

In

to

interpretation

-

related .

I

return

experiencer

to

subject

.

to

this

appears

properties

)

complement

possible .

it , while

( 1991

)

interesting

structure

languages

( 1990

also an

verb

different

relates

question

exclamation

its

the

selectional 7

indirect

reading

Mohanan

( 1994b

a

exlamations

provide

conceptual

structure not

tall

.

it

is

word

general

to

the ) :

below

Carpenter

cor

( BCET

Dictionary

Apresjan

examined "

between

chapter

in

) ,

to

Bouillon

( only

( iii

)

of

associated

in

is

( 1993

underlying

the

) , Longman

related

- linguistic

are

blocks

exclamation

relation

, discusses

cross

between

is .

in

semantics

realizations

manner

he

he

Wilkins

underlying

)

this

sensitivity

( i )

( ii )

that

the

and

tall

factive

know

various Text

is .

in

the

from

English

( ii ) .

he

how

blocks

a

, see

in

negation

it

, for

, and

in

know

in on

" exclamations

negation

taken of

return

adjectives

adjectives

ambiguity

of

( 1979 we

details

adjectives

, polarity

are

( RD

( 1983

, and

Also

) distinguishes

negation that

(i)

these

issue

) , Nunberg

of

) .

,

Digest

Bierwisch

further

( 1995a

book Collection

general

are

For

the

) , Readers

the

sense

) .

of

) .

properties ( 1994

) , where

Elliott

( WSJ

, Katz

( 1995

with

9 .

on

Birmingham

addressed

Pustejovsky

serve

8 .

lyre

how

modes

quality

speaks

the

and

at

a

sculptor

playing

expands

( LDOCE

demonstrated

tic

)

attributed

good

at

remainder the

semantically

subselectional

the

is

a

function

looking

the :

, 1989

extent

involving

7 .

in are

have

lesser

The

and

them

( 1993

6 .

i .e . ,

of

( 1975

English

Others

( iii

or

,

role

by

Journal

represent

see

which

- player

capacity the

Moravcsik

here

sources

Wall

5 .

that

understanding

Contemporary

4 .

to

lyre

6 .

pus

3 .

individual

good

discusses

modification theory

an

a

individual

( 1963

2 .

of

example

distinction of

between

how .

strongly

Although

weak the

this

here

, but

discussion

.

see

rather

is

and

composition an

important

Pustejovsky

strong

com

-

respects distinc

-

( 1994

) .

Notesto pages61- 67

245

Chapter 5

1.

As

an

operation

forming and

2.

on

a Wegner

Both

,

Levin

involving

separates

arguments

The

in

tempted

to

although

it

though of

the ,

Levin

the

and

phrase

Restrictions

type a

su

( i )

on

as

in

the

an

btyping

Mary

~

b

John

dyed

c .

Zac

For

the

the

incorporated

~

colored

verbs

~

dye

of

( 1990

( 1981

)

01

and

On

this

) ( e , x

formal

( i ) , AYAxAe

individual

and the

, in

have :::;

the

2 ,

to of

the

Baker

' s

operations

,

is

For

a

come

from

some

verbs

where

the

predicate the

as

both

( al

-

nature

hammer

and adjectival

the

complement the

is verb

the

adjunct

restrict

For

- like ) ,

the

the

.

process

such by

be ,

semantically

discussion

verbs

might

example

construction are

for

one

For

not

the

contributed

of

closer

semantic

.

7

.

obey

) .

that

there

is

,

dyes

a ,

shadow and

resultative

logically

restricted

:

.

These

seem

,

associated

respectively :

to

be in

white

~

the

event

paint

a

_ color

Pustejovsky

, the

( 1991b

, is

LFrom

between

representation

where

with .

positioned

discussed

Davidsonian

expression

,

relations

resultatives

purely

argument colors

following

color

( e , x

, y

added

the

verb

as

an

) .

build

argument

) ]

captures ) ,

02

the ( e , y

) ,

relational

( e )

event logic

/ \

01

( e , X

essentially

of

roughly of

is .

structure

corresponding

interpretations

the

verb

to

incorporation

arguments

chapter

is

the

emphatic

a

those

:

[ building

view

,

paints

green

from

.

color

we

following

[ build

,

,

-

.

green

and

re

.

of

directly

AYAXAe

functions

,

within

the

relation

Parsons

,

,

denote the

dye

and

specifically

would

dragon

view

_ color

resultatives

More

white

,

is

construction

argument

purple

1995b

into

type

shadow

semantics point

pure

,

to

,

-

further

many

of

chapter

the

and

notice

house

jeans

the

paint

selectional purple

a

the

his

)

fashion

to

~

in

predicate

But

tighter

relation

a . .

even

.

type

entering

state

this

verb

trans

Cardelli

appear

constructions in

verb

.

distinction

verb

athough

shadow

verbs

resulting

This

proposal

,

the

certain

verbs

from

the

theory

,

and

( 1995

differ of

well

all

make

.

discusses

' s

)

Pustejovsky

the

)

the

who

Wunderlich

briefly of

that

can

. as

discussed

as

( cf

arguments ) ,

into

of

terms

shadow

( 1988

default

Rappaport

semantics

predicate itself

in

appear

arguments

wipe

( i )

,

would

see

by

,

class

.

from

( 1987

classification

construction

homogeneous

( cf

of

the

of

directly

lexicon

seen

.

.

fairly

( 1984 arguments

directly

's

be types

) .

preposition

Baker

can

Marantz

internal

German

is

perspective

rethink

1987

and

Wunderlich

in than

generative

the

)

expression

in

translate

a

resultative

the

processes

from

a

coercion polymorphic

,

1995

indirect

from

with

operations

,

type

with

O - assignment

{ } - role

here

,

one

Benthem

versus

described

presented

available

van

( 1988

receiving

derivational

Viewed

A - calculus into

and

a

behavior

syntactic

to

Klein

involved of

analysis

5.

,

direct

operations type

a

Rappaport

assigned

within

4.

1985

finements

the

within language

and

arguments

3.

types

monomorphic

)

/ \

these

02

to

roles

as

the O - roles

in

Dowty

verb

through as

a

used ( 1989

by ,

set

Chomsky

1991

) .

( e , V ) ]

identified

by

predication

,

as

with

any

other

of

246

6 .


This

is

( 1989

the

view

) , and

espoused 7 .

here

Landman van

only of

9 .

in

a

verbs

allow

(i )

a

Mary

( ii )

~

Mary

tion

11 .

~ ~

( cf .

Jackendoff

,

lexical

or

differ

to

might

be

example

,

it

,

also and

which

strategy

for

dictionaries

.

For

the

( cf .

Pollard

of

variation

fully

alternation

in

more

fully

both

structures

define

overlap

define

in

" exhaustive

given

a set

planning

of

- Ginet

adverb

1987

a

( 1982

, such

part the

as

of

the

subject

read

in

of

events

.

events

for

) , some

ad

-

( i) :

event

in

( ii )

- oriented

,

, ,

a

above

interpreta

-

1994

I

chapter

broadly

a

in

10

I

am

( 1993

below

closest .

) . to

that

and

The

of

the

kinds

are

presented

of

HPSG aspects

addressed of

here

etc

listing

certain

treatment

of ,

structured

style

the

general

many

These

) .

activity use

dictionaries

ignoring ) .

X quote

the

with

a

It

For

particular

our

have

as

or

to

music

notation ,

( forthcoming is

, we ,

. .

Malcolm

some

that

along

construed

Sanfilippo

)

,

and

however of

reference

Hence

feature

Furthermore

Johnston ( 1993

.

case

in

listing

how

world

passage

however

dictionaries

of

the

,

written

make

a

of

maintain ( the

8 , of

visual

adopt ) .

discussed

Sanfilippo

also

reflection distinctions

particular

also

chapter

a

ontology to

a

are

more

our

dictionary

compilation ,

even

in difficult

structure

i .e . ,

only

informative

( e . g . , for

in

the

course are

objects

novel

argue

to

is .

of

these

must

I

and

in

to

dictionaries

exposition

as

in

's

.

gives

are

to

lists

,

the

are ,

dictionaries

Pustejovsky

given this

.

Lexikon some sort

Sag

the

McConnell

reference

of

to

more

term

of and

typology

- scope

AGENTIVE

novels

purpose

are on

and

distinctions

that

the

word of

syntactic

more

discuss

argued

e .g . ,

The German of information

that wide

refer

constructing ,

)

on

possible

refers

orderings

event

classification

) , and

dictionary

new

differences

these

therefore

,

.

Nevertheless

certainly

be

to

differs

structure

Benthem

us

.

) system

event

to

allows

events

) .

.

characterizes

term

12

1972

that

( 1979

of authors

.

while

sortal

above

might

way

the

is

this

collocational

argued

mentioned

interface

observes

semantically

contribute

's 's

suggestion

,

) , Descles

.

~

)

reading

these

van

restrictions

interpretation

! ~

( 1985

's

by

Kamp Kamp

, while

possible the

( 1985

departed

manner

These

the

explore

of

- scope

departed

In

Landman

Higginbotham

the

they

It

richness

wide

Higginbotham gives

)

language

in

how

.

, furthermore

what

( 1993

the

out

shared

of

( 1978

decomposition

.

natural

of

pointed

be

relations

This

) , Mourelatos

structural

events

follow

.

) illustrates

Pulman

Discussion As

well

for

primitive We

of

discussion

logic

are

( 1976

view necessarily

useful )

.

as

, 1984

and

discourse

10 .

"

) , Comrie

The

not

inclusion

of

( 1983

Crouch

a

precedence

part

Allen

) .

( 1983

overlap

temporal

overlap

( 1973

( 1990 , would

's

and

contain

Dahl

) provides

Benthem

precedence

terms

by

, however

( 1991

from

8 .

taken

Guentcheva

,

verbal and

we

.

Notesto pages85- 87

Chapter

247

6

1. Much of the perspective taken here has been influenced directly or indirectly by readings of Aristotle 's Organon and subsequent commentaries on it . 2. Hobbs' (1987) "systems," within the model of commonsense reasoning he assumes, are good approximations for the type of understanding needed to categorize and name things in the world . Nirenburg 's (1989a,1989b) notion of "mi crotheory " is a rich model of what language maps into after superficial linguistic interpretation has occurred . Miller 's ideas regarding the structure of the mental lexicon are also interesting in this respect, since the qualia can be viewed as "distinguished " links between lexical concepts in the lexicon . 3. As in Moravcsik 's (1990) treatment of Aristotle 's aitiae - - the modes of explanation mentioned above - the goal is to explore a sound cognitive view on semantics, largely based on the constructivist assumptions in Aristotle 's metaphysics. Qualia structure , as interpreted in a generative lexicon , is an integrated part of a larger methodological shift from conventional approaches to lexical semantics. The qualia provide the foundation for an inherently polymorphic approach to semantics. But they are only one tool for the characterization of natural language as appropriately polymorphic . What gives the qualia their true generative power is an explicit statement of how they are typed , i.e., what the expression in the quale role actually refers, as well as reference to explicit mechanisms of enriched composition ) such as coercion, selective binding , and co-composition . to be discussed in chapter 7. Hence, as presented in the literature , there is some divergence between our use of qualia structure and Moravcsik 's use of aitiae . GL makes explicit just what the generative devices are that allow qualia to provide different ways of explaining what an object is. What these approaches share is the goal of looking beyond conventional truth -theoretic models of semantics to a model of meaning which admits of a multiplicity of reference in linguistics expressions. For further exposition of this position , see Pustejovsky (forthcoming ) . There is possibly some similarity between aspects of qualia theory and the ontological theory of moments put forth by Smith and Mulligan (1983) . They argue for an ontology consisting of "dependent parts " of objects , which are not prop erly characterizable in standard mereological terms . Many of these "parts " are characteristic properties of an object which might be viewed as qualia associated with it . Another theoretical similarity exists between GL 's use of qualia structure and the notion of "lexical function " as employed by Mel ' cuk and his colleagues (cf. Mel 'cuk (1973a,1988a) , within the Meaning-Text Model . Heylen (1995) , in fact , compares qualia structure and lexical functions as descriptive mechanisms for accounting for semantic relatedness in potential , as well as actual , collocational usage. Finally , some of the proposals given in Bes and Lecomte (1995) are much in line with the methodology here, concerning what level of lexical knowledge is both appropriate and necessary for linguistic theory . 4. It might be argued that the subject always stands in a containment relation to the substance denoted by the complement NP. That is, yeast is contained in beer, tomatoes are contained in the pasta sauce, and so on. There are some interesting counter -examples, however, to this generalization , as (i) illustrates . (i) Methylene Chloride is used in decaffeinated coff~~. This example suggests that the construction is actually making reference to some process involved in bringing about the substance, as opposed to the substance itself , since the chemical in subject position is not present in the resulting liquid .

248


5.Another The notion of defeasibility here isthat asused insentences Lascarides et ale (1994 ). way of viewing the distinction between these is that the space of interpretations available torestricted the NPs New York and Boston is that available tothe more sortally airport proper names inlarger (4a ).than 6.qualia InPustejovsky and Anick (1988 ),the senses are "split "role according toits separate assignments . Under this analysis , the FORMAL takes as value the Figure of a physical object , while the CaNST role assumes the Invert -Figure value ofan aperture .The polysemy was seen as resulting from the foregroundin or backgrounding of a nominal ' s qualia . That is , in (9a ) paint applies tothe FORMAL role of the door , while in (9b ), through will apply to the CaNST interpretation ofthe same NP .The problem ,however ,to with sense splitting inside the qualia structure for a lexical item is that it fails capture the selectional behavior ofthe item for allcontexts . The interpretation of complex such as these given (1994 )overcomes these difficulties ,types as we see below infirst chapter 8.inPustejovsky 7.regularly Pustejovsky and Boguraev (1994 )show how these logical polysemies are infact encoded indictionary definitions for these words . 8.See LeiB (1991 )and Gunter (1992 )for anexplanation offormal mechanisms of type inference within the A -calculus , and Copestake (1993 ), Morrill (1994 ), Buitelaar and Mineur (1994 ) for its application to lexical representation . Shieber (1992 )discusses general issues oftype inference for natural language processing . It should be pointed out that the details of the typing assumptions discussed in those works differ in ways that are beyond the scope of the present discussion . For further details of the typing rules and the use of typed feature structures within generative lexicons ,see Pustejovsky and Johnston (forthcoming ). 9.munication Following a),suggestion by Manfred Krifka and Nicholas Asher (personal com the specific relation that structures the types can be built into the type structure itself .This isessentially the analysis pursued below in8.3. 10 . For expository purposes ,Iignore the other qualia roles until the discussion below . 11 . AGENTIVE Inprevious treatments of the semantics for dotted types ,Iassociated aunique and TELIC role value for each argument in the argument structure (cf.Pustejovsky (1994 ),Viegas and Bouillon (1994 )).for This seemed appropriate since the information for both modes of presentation an object must be ac counted for . For example , on this view , for a two -argument dotted type nominal such asbook ,there is amode ofexplanation for both the coming about of the information and the coming about of the physical object as well . That is , the qualia structure was given as follows : book =yx:phys _ob ]~iOn ARGSTR =[ARG2 ARGI :infOrma ]container information -phys _ obj -lcp QUALIA = TELIC FORM = hold (Ytx ) ==read (P(T ,w,v ,x),x), puhlish AGENT write (T,z:printer ,y)

The problem with this interpretation isthat the predicates that are sui generis to entities that are both physical objects and information bearing objects are actually mistyped here ; that is , both read and write are only understandable in terms of a complex typing system , such as that outlined in the previous chapter . The objects selected for byapredicate such as read and the predicates made

Notesto pages100- 108

249

reference to by objects such as books are the characteristic functions for those entities . 12. Aristotle Metaphysics , Physics II , and Moravcsik (1975). 13. These strategies replace the mapping rules discussed in Pustejovsky (1991b) , where it was less clear what type of semantic representation the mapping to syntax actually allows. These were given as follows : (A ) The semantic participant involved in a predicate opposition is mapped onto the "internal " argument position of the lexical structure . All transitions involve a predicate opposition of some sort . Independent of this principle , Pustejovsky (1988) and Grimshaw (1990) argue that the notions of causer and agent are associated with the initial subevent of an event structure . (B ) The agentive participant in the initial subevent or event structure is mapped onto the external argument position of the lexical structure . The result of both principles A and B applying in the structures above is not general enough, however, to account for the mapping to passive constructions as well as to experiencer constructions (cf. chapter 9 for discussion) .

14. This proposal should be compared to the lexical mapping theory , as outlined in Bresnan and Kanerva (1989) and Bresnan and Moshi (1990) , where semantc arguments are related to their morphosyntactic expressions by means of specific syntactic functions . Along with these principles , they assume a universal hierarchy of thematic roles, as given below: (i) ag > ben > goal > inst > patient > locative Cf . also Alsina and Mchombo (1990) for further discussion. Chapter

7

1. Briefly, we can summarizesomeof the type shifting operationsthat have been suggestedin the literature (cf. Partee (1985), Chierchia(1984), Hendriks (1987), Dowty (1988). (

i

)

Type

a

Shifting

.

b

.

c

.

d

g

2.

3 .

4 .

:

.

iota

pred

.

=

lower

:

:

general

> "

j

P

x

AX

~

P

=

~

lifting

)

]

lift

( j

)

=

[ P

n

u

( j

[ X

LP

>

:

[ P

(

~

P

:

:

>

lower

:

nom

.

functions

j

ident

.

e

f

lift

(

j

X

)

)

=

j

]

]

P

x

:

a

~

AX

[ X

(

a

)

]

There is a long controversyover how to best analyzethe infinitival in such constructions; namely, as either a VP of some sort or as a full sentencewith a PRO-subject. In order to focuson issueof the coercionphenomenathemselves , I will not discussthis issuehere. Cf. Chomsky, 1981and Pollard and Sag, 1994. As we seein 7.4 below, this is in agreementwith the generalstrategy outlined in Chomsky (1986) moving towards constraints making referenceto semantic selectionrather than syntactic configurationsdirectly. Seebelow for discussion . For details seeKlein and Sag (1985) and Gazdar et ale (1985).

250

Notesto pages109-110

5. One recent proposalwhich would disagreewith this conclusionis Chierchiaand Turner (1989), wherepropertiesare allowedas arguments. 6. If we return to the type selectionof considerin (3) above, we can seeanother way of solving this type clash; namely, to allow the NP to shift in denotation accordingto the referencetype ladder. Followingthe proposalsjust discussed , however, there is another possibility; to shift the type of the verb itself, so that it would accept an NP as type e,t >,t >, without the NP shifting at all. Assumethat this could be accomplishedby a type-shifting operator, f sc, (for small clause), which would act to relate the two structures in (ia) and (ib). (i) a. John considers Mary to be an honest person. b. John considers Mary an honest person. Thus , f sc exhibits the shifting shown in (ii ) below. (ii ) a. consider E: ,t >

The meaning (ii ) :

=>-

postulate

<e , t > associated with this type shifting

rule is given below in

(ii ) VPD [pCP ) +-)0 Ax3P [P (P ) (x )]] The result of applying this operator to an NP is effectively to create a metonymic extension of the NP meaning . In the case of the NP John , for example , the operator produces the following : (iii ) p (J\ P [P (j )] ) => J\x3Q [Q (J\P [P (j )] ) (x )] This operator , together with the independently able to generate an " underspecified " translation

motivated Equi operator , fE , is of (iv ) , as shown in (v ) below .

(iv) MarywantsJohn. (v) a. fE (want')("'p(j ))(m) ~ b. fE (want')("'AX3P[P(j )(x)])(m) => c. m{Ax[want'("'Ax3P[P(j )(x)](x*))(x*)]} => d. m{Ax[want'("'3P[P(j )(x*)])(x*)]} ~ e. want'("'3P[P(j )(m)])(m) The expression in (ve) leaves the relation between Mary and John underspecified . This actually happens to be an advantage rather than a problem , since lexical information from the particular elements in composition with the verb may act to specify this relation in unique ways.

252

10

.


Pollard

and

be

Sag

( 1994

considered

In

a

particular

below

)

)

( i )

from

subject

was

John

promised

and be

order

to

in

Sag

show

GL

,

at

but

.

In

an

general

interesting

we

have

eliminates .

subject

of

( i )

a

the

denied

his

He

denied

having

similar

.

I

began

to

b

.

I

began

the

calls

the

need

of

)

a

.

b

I .

1

to

began

of

In

chapter

In

particular

the

type

.

.

aspectual

For

this

I

discuss

,

I

the not

the

, ,

complement

sense

for

many

" zeroing

as

is

studied

promise

cases

.

of

transformations a

cause in

been

type

treats

specific

" to controller

on

what surface

transformation

long

as

the

responsible

"

meaning

for

the

that

can

zeroing

be

of

the

:

( iia

)

for

" appropriateness

,"

Harris

.

.

variable

the

order

X

zeroing

to

apt

and

admits

transformation

prevent

Xap

that

are

from

applying

the

unclear

in

and

( iiia

) ,

for

rule

, t

be >

, t

> .

of

type

apply given

as

( 1989

to

)

to follows

lexical

generalized ,

where

Verkuyl

'

,

'

the

language

from is

NP

Harris

basic

an

and

-

the

insights

can

.

through and

the

the

type

newspaper

issue

( 1993

)

types

this

section

lattice

.

formally

in

,

quantification at

in

some

length

.

we ,

of discuss

will

not

since

concern

we

are

ourselves

here

interested

.

quantifiers S

,

novel

that

.

semantics

alone

Notice

appropriateness

complernent

inheritance

complement

in

behavior

Hence

ignoring

and

verbal

verbal

the

for

from

been

examples the

.

.

of

framework

structures

have

problem

question with

differs

qualia

I

Kritka

of

this

the

associated

related book

,

,

lexical

issues

their

subsequent

revisits

representation

richer

how

exposition

semantic

should

now

this

some

aspects

never lexicon

values

a

.

.

work

qualia

and

all

in

The

later

consequences

and

book

generative

examine

of

further

e

)

into

shift

.

within

position

the

can

in

the

over

but

hierarchy

primarily

15

the

,

purposes

with

a

coercion

8

complement

14

coerced

have

have

conditions

licenses

book

' of

by of

buy

Harris

view

preserved

For

of

context

,

the

directly

operation

.

book

it

study

,

point

swered

13

is to

.

which

began *

Unfortunately

.

to

( 1965

below

from

appropriatedness

further

)

.

semantic

Harris

different

the

)

.

derived

book

( ii

.

party

data

be

example )

.

in

.

under be

These

sequence

( ib

shift

.

( iii

12

but

can

for

slept slept

read

under

example

be

)

a

Harris

the

)

( iib

a

( ii rule

introduces

in

having

manner that

conditions

in

NP

He

)

,

in

lexical

terms

Harris of

complement

b

.

in )

transformation

a .

suggests

( ii

coercions

members

This

,

party the

underlying

article

properly here

controller

the

attend a

one

the

attend

.

of

particular

secondary

preserved

In

In

attend

would

virtue

type .

to

employ

methodology

in

to

allowed

relevant

considered

to

interpretation

by

and

forms

be

can examined

.

allowed

They

correct

coerced

syntactic

to

that being

involved

allowed be

coercion

phenomenon

- control

be to

."

complement

"

object

to

how

the

the

systematically

11

to

attend

of general

" coercion

Sandy

to

arrive

the

promised

allowed

type

more

- control

Sandy

to

a the

discuss

b

Pollard x

discuss of

they

a . .

)

subset

abbreviates

as the

well

. type

The t

type ,

and

derivation NP

'

the

type

253

Notes to pages 123127

believe : s + (NP + s) (i)

believe(p 2[NP,

John:NP p2[NP,S]:NP S p2[NP,SJ(John):S

S](John)) : NP

16. Insomesense,thiscanbeseenassimilarto thegoalsofthelexicalsubordination

operation in LevinandRapoport(1988).In fact,co-composition provides an elegant solution tothepolysemy involving resultative verbssuchashammer and wipe,aswellaswithverbssuchas rattleinrattledownthehill(cf.Jackendoff, 1990 and Levin, 1993). See chapter 10 for discussion.

17.Asweshallseein chapter10,lightverbspecification characterizes the wayin whicha specificsensein contextof certainverbsis determined by the complement.In manyofthesecases,the verbactsonlyas a generalfunction over qualia-based information fromthecomplement; e.g.,theverbsopen,close,break,

andfix.Thisisformally distinct fromthetypesofconstructions discussed inDi Sciullo and Rosen (1990) and related work, however.

Two other casesdiscussedin Pustejovsky(forthcoming)includemanner co-

composition andfeaturetranscription. Theformerdetermines howanargument to the verbmayitselfspecifythe mannerin whichthat objectis actedupon. Thereare twobasicsubtypesconsidered: (a) wherethe complement specifies manner, e.g.,theverbstryandsample; and(b)wherethesubjectspecifies manner,e.g.,theverbsbreakandothercausatives, aswellaspsychverbs. Featuretranscriptioninvolvesthoseconstructions wherea modifierto an ex-

pression contributes information to further specify a semantic feature ofthat expression that wasleftunspecified. Forexample, adjectives suchas pregnant, whenapplied to termsnotsensitive togender, suchasprofessor, willspecify the expression forthatfeature.Weinreich (1972), in hiscritique oftheKatzand Fodormodelof semanticdescription,calledsuchconstructionsfeature transfers.

Whatis interesting abouttheseoperations is that, althoughtheyare similar to coercions in somerespect,theyarenot typechanging operations; rather,

theymake useofqualia-based information tofurther specify theverbssensein context.

18.Thistypeofmodification mightat firstseemsimilar to theanaphoric island effectsdiscussedin Postal (1971),illustratedin (i) below.

(i)*Johnis anorphanandhemisses them verymuch.

Obviously, in (i)theindexed pronoun isunable to refertotheputative implicit reference to (Johns)parents,andthesentence is ungrammatical underthisinterpretation. Theanaphoric binding in (i)is qualitatively different fromthe selective binding operations above,however. Notice that an adjective suchas fastorgood isabletoidentify partofthequalia structure oftheexpression it is incomposition with.Making anexpression available formodification isnotthe sameas elevating it to a positionof antecedenthood forsubsequent anaphora, asin(i)above. Wecanthinkofthequalia withtheirvalues asa moreexplicit statement ofthetypeofthe nounitself.In thissense,selective bindingallows anadjective to modify a typefragment oftheexpression. Suchaninterpretation would prevent theprojection ofimplicit features expressing aspects oftheFORMALroleforthenounorphan.Notice that,although anaphoric islandeffects are preserved, eventpredicate selection isinfactpossible withorphan, asillustrated in (ii):

(ii) a recent orphan

254

Notes to pages 128134

Thissupportstheviewthat theseareverydifferent phenomena. 19.It mightbeargued thatthemodification byadjectives suchasgood asillustrated

in (67)above is notlicensed bythetypewithin thequalia, butratherbythe qualia themselves. Thatis,therulemightbetterbestatedintermsofqualia

selection:

(i) SELECTIVE BINDING (byquale):

Ifa isq,of type , /3is typeb,andthequalia structure of3,,QS quale, then cx/3 is of type b,of where = /3 fl a(q, ).

has

Although it mayappearthatspecific qualiaareselected for,in general, such statements canberecastintermsoftypes.Iftheformulation ofselective binding in (66)is correct, thenwewould predictthat anyqualesatisfying thetype required bytheadjective wouldbe available forselective interpretation. This

doesseemto be the case,as illustratedin (ii) below. (ii) a goodknife:a knifethat is well-made; (ii) Thats a goodknife,but it doesnt cut verywell.

Ifsuchdistinctions aregrammatical, which theyappear tobe,thentheadjective

is ableto selectthe AGENTIVE qualeofthe noun,supporting the viewthat a type rather than a quale is selected.

20. Asstatedabove, strictlyspeaking themodification isintermsoftypesatisfaction andnota particularqualerole.Hence,eitherTELICor AGENTIVE wouldbe a

possible target of the modification.

21.Onecanimagine someuseofhouse ina sublanguage where thingsaredifferent, forexample, Victorian orEdwardian England, wherea family mighthaveany number ofhouses.Blandings Castle, forexample mightbe an oldhouseof mine,stillin mypossession, whileI mighttypicallylivein a newerhouse.The wordhouse wouldnt havechanged significantly inthissubculture, butenough to reconceptualize it as an objectwitha differentintension.Theissueof how muchofthissemantic shiftiscultural andwhatislexical isbeyond thescope of thepresentinvestigation. Mymethodological assumption throughout hasbeen toattribute specific grammatical effects topredominantly linguistic distinctions, whichbytheirverynature,revealtheparticular conceptualizations oftheworld constructed by a word or phrase.

22. Forsomestates,beingmodified byan adjective suchas oldwillreferto relations

that arepossibly extensionally equivalent in duration to theinterpretation of interpretation forthisreason (i.e.,thereisnoHeisoldasmybrotherreading). The relationshipof brotherhoodor sisterhooddoesnot allowthis. theobjectitself.Forexample, an oldbrother soundsparticularly oddunderthis

23. Kiparsky andKiparsky (1971)attemptto relatethe syntactic behavior ofcomplementsto the semanticsof the predicate,the complement, and the relation

between them.Grimshaw (1979) distinguishes thebehavior ofsyntactic cornplementation in English in termsofsemantic selection overthreecomplement

types,propositions, interrogatives, andexclamations (cf. Bresnan, 1972).As beenpursuedby functionalists, andis centralto the goalsof RoleandRefer-

VanValinandWilkins(1993)pointout,the traditionofsemanticselection has

enceGrammar (RRG) theorists (cf.VanValin, 1993 andthepapers therein).


255

Wierzbicka (1980,1982,1988) develops an increasingly rich system of semantic selection , with many interesting insights regarding the role of semantic categories for syntactic form . Yet the goal of semantic theory in this respect should be to capture the most elegant generalizations concerning syntactic expressiveness from semantic selection, and not to completely recreate the richness of the syntactic descriptive system within semantics. The observations by Grimshaw (1982) and Jacobson (1992) concerning the limitations of semantic selection are worth noting , and there is no reason to think that syntax is completely parasitic on semantic types .

24. That something besides case assignment is involved can be seen from sentences such as the following : (i) John asked me about my name. (ii ) John asked me my name. In sentence (i) , the preposition is not just case marking the complement but shifting the interpretation of the predicate as well .

25. Groenendijk and Stokhof (1989) discuss the scope distinctions allowed by these verbs in sentences like (i ) and (ii ) below. (i) Peter knows who John loves or who Mary loves. (ii ) Peter ~ o!!ders who John loves or who Mary loves. The sentence in (ii ) is ambiguous where the disjunction may take wide or narrow scope

relative

to wonder

.

Chapter 8 1.

In Pustejovsky (1991) and Pustejovsky and Boguraev (1993) , a distinction is made between fixed and projective inheritance . Following Touretzky (1986) , the fixed inheritance are concepts

structure

in our

model

of a lexical item is defined as follows , where Q and P of lexical

network

. Then

:

(i) DEFINITION : A sequence< Ql , PI , " " Pn > is an inheritance path, which can be read as the conjunction of ordered pairs { < x I , Yi > 11~ i ~ n } . Furthermore

, from

this the set of concepts

that

lie on an inheritance

path

is

defined , as distinguished by a particular quale role (e.g., TELlO vs. FORMAL ) . This is called the conclusion

space for a given quale , q'

(ii ) DEFINITION : The conclusion space of a set of sequences, q, is the set of all pairs < Q , P > such that a sequence < Q , . . . , P > appears in q , where q is one of the qualia for the concept Q .

(iii ) DEFINITION

: The complete conclusion space is the set of all conclusion

spaces defined for each quale for a concept : = q'i ' Finally , by adopting

Touretzky

's operator

I nh - where , for every set of sequences

S, Inh (S) denotes the set of values inheritable from S- the lattice structures shown

above

for

book can be differentiated

as follows

:

256


Let [a ] cfI stand for the denotation of Q: with respect to a model of inheritance over the set of sequences, 4>. Then ,

[book]

Mary

's

the

purchase

knows

the

coercion

to and

awaits

Then

, t

.

of

these

phenomena

( el

, y

explored the

, x

) ] ]

in

Busa

semantics

of within

explanation

for

4 .

explains

It

is

which a

the

useful

typed

classes

the

can

an

well

of .

embedded

inter

and

reveal

presumably

- formed

on ,

as

also

constraints

be

in

argument

understand would

syntactic

to

-

interpretation that

for know

on

below

taking

assume

type verb

par lexical

available

as

-

think

generative

propositional

we

inter

to

noun

Under

simple

) ,

The

an

subcategorizations

remarks

apply

further

be

coercion

which

of

,

course

.

the of

car

the

.

car

.

.

in

can

car

.

a

completely

unconstrained

unexplained research

effects

a

buying

answer

leaves

syntactic

coerCIon

> .

bought

knows

how

study

."

a

bought

knows

problem

, t

, < e

it

be

( 1989

Mary

* John

for

above

larger

chapter

multiple

allowing

Mary

* John

,

as

would

sentences

d

John

the

further

the

c . .

is

[ R

.

John

overgenerates

is

Stokhof

some all

John

Thus

It

without

=

made

as in

extension

verbs

and as

b

.

invoked

,

for

these also

s

AGENTIVE

approaches

mentioned

coercion

while

be

a .

e .

ing

,

,

a

and

is ,

account is

typed

expect

are

( i )

be would

typed

coercion

of

would

Groenendijk

used

,

A

SLNs within

" metonymic

type

rogatives

type

terms

That

) ]

.

particular

.

, x

and

Italian

questions

coercion

Following

so

) ,

in this

interrogative

we

and

concealed

examples

an

English

( e2

embedded

framework

alternation

ticipate

these

is

( 1995a of

[P

ILNs

analysis

lexicon

pretation

=

between

Her

nominalizations

FORMAL

why ,

and

a

( ic

)

solution

contribute

to

and

fashion ( id

)

should

are prove

limiting

the

in

these

ungrammatical

cases .

helpful

in

application

This

determin of

-

semantic

.

perhaps

worth

shifting

,

in the

sentence .

each verb

pointed

mentioning out is

by embedded

another Hobbs

et within

al

interesting .

( 1987a an

case , 1987b

interpretation

) .

of

coercion

Notice predicted

how

, the by

involving

sort

complement the

typing

of

(ii) a. Thatchervetoegthe channeltunnel. (theproposal ) b. The boardvetoedan increase in payfor the teachers . (theproposal ) (iii) The organizers havebookedElla into Symphony Hall. (theconcert ) (iv) Hersecretaryhasscheduled Johnfor nextweek. (theappointment ) That is, in (ii) the complements are modallysubordinated within the type of proposal, whichis the type selectedby the verb veto. This producesa weak intensionalcontextfor the complement , sinceit is intensionally"wrapped " by the denotationof anothersort. A similarexplanation accounts for the apparent metonymies in (iii) and (iv)~where , in fact, strict typing is beingobeyed , and sort coercionis providingfor the correctinterpretation(cf. Pustejovsky (1995c ) for furtherdiscussion ). 6. Li (1994 ) discusses spatialprepositions andtheirinterpretation in German , within Lang's (1989 ) theoryof spatialrelationsand Bierwisch 's modelof conceptual structures(cf. alsoBierwischandLang, 1989 ).

267

N dtes to pages 233- 235

7.

One

lexical

ations is

knowledge

is that

of

entries

,

lexicon

words

is

.

The

classes

From

the

of

,

the

relation to assuming

for

a

example

,

generic

is

a .

.

,

{

animal

{

artifact

a

,

,

are

for

The

of divide

a

example

,

typically

attribute to

the

organized

,

,

verbs

relations

.

entails

He

which

( 1991

issue

The

my

( 1988

Lakoff

characterize

them

uniquely

,

a

different

The

the

of

the

it a

of

this

of

ascriptive

.

The

former

modifies

ascribes

class

principles

.

noun

heavy antonym

,

set

nonascriptive

.

value

For

of

adjective

,

the light

,

adjectives

.

which

do way

,

not

the

a

see

set

of

lexical

is

formal

)

snoring

semantics

lexical

( 1990a

musical

entailment He

in

relating

Miller

a

.

since

theories

device ,

predicate ,

musical

sleep

to

formal

WordNet

directly example

is

entails similar

not For

that

to

lexically is

.

one

according

snore

organization

has

more

not

.

( cf

.

Fass ( 1987

so

most

chapter 1988 ) ,

and

ignored

are

, 1993 Sowa

,

to of

no

in

this

patterns

easy

forms

7 ) , ,

been systematic

always

that

( cf

largely

on

assumes

) ,

:

items

and

.

Miller

For and

) .

are

inferencing

to

below

}

of

music

as

correspond listed

}

follows

Thus

structured

on

a

}

condition

unspecified

for

,

each

.

words

some

verb

information

attention

coercion

the

This

metaphor

phenomena

literature of

of

in

postulates

and

Fellbaum

focused

.

meaning

details

"

are ,

.

those

used

WordNet

sleeping

adopt

more

it

,

is

principles

to

example

is

weight

include

instrument

in For

i .e .

,

}

package .

element primes

}

flora

and

package

different

" pertain an

the

noun ,

very

but is

Finally

The

the

adjectives

noun

,

predicates

in

attribute

along

instrument

8.

to

same

Nonascriptive the

heavy

weight

et

is

Rather

}

ascriptive

attribute

adjective

of

refers

bipolar

:

topmost

are

phenomenon

attributes

WordN

classes

which .

actually

object

which

or

the

fields

{ natural

. .

,

semantic

{ natural

features

in

two

of

the

topical relations

relation

These

{ process

function

adjectives

into

the

by ' s

in into

hyperspaces

as

set .

{ state

concept

organization

involves

is

noun

interesting

languages

these

{ plant

}

distinguished

a

Adjectives

}

}

further

example

three

entailment

entity a

of

}

{ food

various

with

some

knowledge

,

function

leaving

organized

hyponymy

into

}

,

WordNet

and

,

is are

{ person

{ cognition

,

representation

}

{ attribute

lexical

In

modifiers

n - dimensional

hierarchy

activity

.

WordNet

of as

separate

fauna

,

Nouns

means

nouns

,

different

explicitly

conventional

example

action

verbs

, .

structure

For

, 000

synonyms

-

relation

.

knowledge

the

forms

fields

{ act

Nouns

the of

partitions

which

semantic

is

consider lexical

64

of

represented

structured

hierarchical et

,

by

are

relation

single

WordN

concept

distinct

-

nouns

principles

,

nouns

the

:

categories

organized

hand

between

equivalent than

are

other

sets

adjectives

three

psychological

fundamental over

, 000

not

and

organization

verbs

on

,

these

while

,

The

lexical

structures

hierarchies Adjectives

actually

verbs

the

39

categories are

,

to

where contains

about

four

nouns

according ) ,

currently

words

view it

designed

, 1990b

into

into

:

is

1990a

WordNet

divided

words

that

,

organized

function

of

way

. are

that

Miller

synonymy

which

the

base

( WordNet

and ( 1984

delineate

,

metonymy

different

, from

Martin , 1993

,

also

,

and

I

polysemy of

as

, 1992 argue

,

much in

types

of

the

and it

. of

the

context

metaphorical

Brugman that

instead

metonymy

and

discussed

) .

have

and

course

certain 1990

)

work

of

is

difficult

Lakoff to

268


maintain any clear separation between lexical and linguistic knowledge, on the one hand , and general conceptual knowledge on the other , based on the modes of sense extension associated with the metaphorical usage of lexical items . This is similar to the position held by Nunberg (1995) as discussed above, but Nunberg would presumably still distinguish between lexically and non-lexically based inferences.

Bibliography

Abraham , W . 1986 . " Unaccusatives manistischen

in German ," Groningen

Abusch . D . 1985 . On Verbs and Time , Doctoral sachusetts

Arbeiten

zur Ger -

28 : 1- 72 . , Amherst

dissertation

, University

of Mas -

, MA .

Ackerman , F . 1992 . " Complex

Predicates

and Morphological

Relatedness : Loca -

tive Alternation in Hungarian ," in I . Sag and A . Szabolcsi (eds.) , Lexical Mat ters , CSLI Lecture Notes , Stanford , distributed by U . of Chicago Press . Ackerman , F . 1993 . " Entailments of Predicates and the Encoding of Causees ," Linguistic Inquiry 24 :535 - 546 . Allen , J . 1983 . " Maintaining Knowledge About Temporal Intervals ," Communi : cations of the A CM 26 :832- 843 . Allen ~ J . 1984 . "Towards a General Theory of Action and Time ," Artificial Intel ligence 23 : 123- 154 . Alsina , A . 1992 . " On the Argument Structure of Causatives ," Linguistic Inquiry 23 : 517 - 555 .

Alsina , A . and S. Mchombo . 1990 . "The Syntax of Applicatives in Chichewa ," Natural Language and Linguistic Theory 8 :493 - 506 . Amsler , R . A . 1980 . The Structure of the Merriam - Webster Pocket Dictionary , Doctoral dissertation , University of Texas , Austin , TX . Anderson

, M . 1979 . Noun

Connecticut

, Storrs

Phrase

Structure

, Doctoral

dissertation

, University

of

.

Anderson , M . 1984 . "Prenominal Genitive NPs ," The Linguistic Review 3 :1- 24 . Anick , P . and S. Bergler . 1992 . " Lexical Structures for Linguistic Inference ," in J . Pustejovsky and S. Bergler (eds .) , Lexical Semantics and Know ledge Repere sentation , Springer Verlag , Berlin . Anick , P . and J . Pustejovsky . 1990 . " An Application of Lexical Semantics to Knowledge Acquisition from Corpora ," in Proceedings of 13th International Conference on Computational Linguistics , Helsinki , Finland , 1990 , pp .7- 12.

Apresjan , J. D . 1973a. "Synonymy and Synonyms," in F . Kiefer (ed.) , 173- 199. Apresjan , J . D . 1973b . " Regular Polysemy ," Linguistics 142 :5- 32 . Apresjan , J . D . 1981 . " Semantic Amalgamation Rules for Russian Verbs ," in P .

Jacobson and H . L . Krag (eds.) , The Slavic Verb, Rosenkilde and Bagger, Copenhagen , 7- 13. Asher , N . and M . Morreau . 1991 . " Common Sense Entailment : A Modal Theory of Nonmonotonic Reasoning ," in Proceedings to the 12th International Joint Conference on Artificial Intelligence , Sydney , Australia . Atkins , B . T . 1987 . " Semantic ID Tags : Corpus Evidence for Dictionary Senses," in Proceedings of 3rd Annual Conference at University of Waterloo , Center for the

New

OED

, pgs . 17 - 36 .

Atkins , B . T . 1991 . " Building a Lexicon : Reconciling Anisomorphic Sense Differ entiations in Machine - Readable Dictionaries ," International Journal of Lexi cography .

Atkins , B . T and B . Levin . 1991. "Admitting Impediments ," in Zernik (ed.) , Lexi cal Acquisition NJ

: Using On -Line Resources to Build

a Lexicon , LEA , Hillsdale ,

.

Atkins , B . T ., J . Kegl and B . Levin . 1986 "Explicit and Implicit Information in Dictionaries ," Advances in Lexicology , Proceedings of the Second Conference of the Centre for the New OED , University of Waterloo , Waterloo , Ontario . Atkins , B . T ., J . Kegl and B . Levin . 1988 . " Anatomy of a Verb Entry : From Linguistic Theory to Lexicographic Practice ," International Journal of Lexi cography 1:84- 126 . Austin , J . L . 1962 . How to Do Things with Words , Harvard University Press , Cambridge , MA .

270

Bibliography

Bach, E. 1981 . "On Time, Tense , andAspect: an Essayin EnglishMetaphysics ," in P. Cole(ed.), RadicalPragmatics , AcademicPress , NewYork, 63-81. Bach, E. 1986 . "TheAlgebraof Events ," in LinguisticsandPhilosophy 9:5-16. Baker, C.L. 1969 . "Concealed Questions ," paperdeliveredat LSA. Baker, M. 1988 . Incorporation : A Theoryof Grammatical FunctionChanging , Universityof ChicagoPress , Chicago . Bar-Hillel, Y. 1953 . "A Quasi -arithmeticalNotationfor SyntacticDescription ," Language 29:47-58. Bartsch, R. 1985 . "The Structureof Word Meanings : Polysemy , Metaphor , Metonymy "in F. LandmanandF. Veltman(eds.) Varietiesof FormalSeman tics, Foris, Dordrecht . Beard, R. 1991 . "Decompositional Composition : TheSemantics of ScopeAmbiguitiesand'Bracketing Paradoxes '," in NaturalLanguage andLinguisticTheory 9:195- 229. Beckwith , R., C. Fellbaum , D. Gross , andG. Miller. 1989 . "WordNet : A Lexical DatabaseOrganized on Psycholinguistic Principles ," in U. Zernik(ed.), Proceedings of theFirst InternationalWorkshop on LexicalAcquisition , at IJCAI, Detroit. Beckwith , R. and G. A. Miller. 1990 . "Implementing a LexicalNetwork ," InternationalJournalof Lexicography 3:302-312. Beierle , C., U. Hedtsttick , U. Pletat, P. H. Schmitt, andJ. Siekmann . 1992 . "An Order-SortedLogicfor knowledge Representation Systems ," Artificial Intelli.qence , 55:149-191. Belletti, A. and L. Rizzi. 1988 . "Psych -verbsand Theta-Theory," NaturalLanguageandLinguisticTheory6:291-352. Benthem , J. van. 1983 . TheLogicof Time, KluwerAcademicPublishers , Dordrecht. Benthem , J. van. 1991 . Language in Action: Categories ,Lambda , andDynamic Logic, Studiesin Logic and the Foundations of Mathematics , Volume130, Noth-Holland,Amsterdam , alsopublishedin 1994 , by MIT Press , Cambridge . Bergler , S. 1991 . ('Thesemantics of collocational patternsfor reportingverbs," in Proceedings of theFifth Conference of theEuropean Chapterof theAssociation for Computational Linguistics , Berlin, April 9-11. Bergler,S. 1992 . EvidentialAnalysisof Reported Speech . PhDdissertation , BrandeisUniversity . Bes, G. and A. Lecomte . 1995 . "SemanticFeaturesin a GenericLexicon ," in P. Saint-DizierandE. Viegas(eds)., Computational LexicalSemantics , CambridgeUniversityPress . Bierwisch , M. 1983 . "Semantische undkonzeptuelle Reprasentationen lexikalischer Einheiten ," in R. RuzickaandW. Motsch(eds)., Untersuchungen zur Seman tik, Berlin, Akademische -Verlag. Bierwisch , M. andE. Lang. 1989 . Dimensional Adjectives . Grammatical Structure and Conceptual Interpretation , SpringerVerlag,Berlin. Bierwisch , M. andR. Schreuder . 1992 . '(FromConcepts to lexicalItems," Cognition 42:23-60. Bobrow, D. G. and T. Winograd . 1977 . "An Overviewof KRL, a Knowledge Representation Language ," CognitiveScience , 1:3-46. Boguraev , B. 1991 . "Buildinga Lexicon : The Contributionof Computers ," in B. Boguraev (ed.), Specialissueoncomputational lexicons , InternationalJournal of Lexicography 4(3). Boguraev , B. andT. Briscoe , eds. 1988 . Computational Lexicography for Natural Language Processing , Harlow,Essex , Longman . Boguraev , B. and J. Pustejovsky . 1990 . "LexicalAmbiguityand The Role of Knowledge Representation in LexiconDesign ," Proceedings of COLING-1990 ,

271

Bibliography

Helsinki. Boguraev, B. and J. Pustejovsky. 1994. "A Richer Characterizationof Dictionary Entries" in B. Atkins and A. Zampolli (eds.), Automating the Lexicon, Oxford University Press. Boguraev, B. and J. Pustejovsky, eds. 1996. CorpusProcessingfor Lexical Acquisition, MIT Press, Cambridge, MA . Bolinger, D. 1967. "Adjective Comparison: A SemanticScale," Journal of English Linguistics 1:2-10. Bolinger

,

D

.

1968

Bolinger

,

D

.

1971a

,

D

Inquiry Bolinger

.

,

Bouillon

, tation ings

P

.

,

.

.

:

through

Viegas

.

Adjectival

P

of

Nominal

Overloading

Trajectories

Structures in

a

study

1994

Knowledge

,

Kluwer

, "

the

Glossa

2 : 119

Progressive

of

the

.

" A

94 1995

.

,

Semi

,"

Verb

" The

:

A

a

- 127

.

Lan

-

Linguistic

' Remind

' , "

Dynamic

Publishers

- Polymorphic

:

Amsterdam

Space

Academic

Constructions

Euralex

, .

E

the

Cross

,

Framework

Boston

Approach - linguistic

to

. the

Perspective

Interpre ,"

-

Proceed

-

.

Semantics

of

Adjectival

Modification

,"

ms

.

ISSCO

,

.

Brachman

by

.

and

of

Geneva

Brachman

1994

Meaning

on

.

Comprehension

of

Bouillon

N

.

the

Note

" Semantic

- 547

Machine

and

Further .

.

: 522 L

" A

- 586

1971b

47

Bookman

" Entailment .

2 : 584

guage

for

.

R

.

Findler

J ,

. ed

Computers ,

1979

.

. ,

Associative

, R

.

J .

and

" On

Academic J

the

Epistemelogical

Press .

Schmolze

Status

Networks , .

: New

1985

York .

"

An

of

Semantic

Representation

and

Networks Use

of

,"

in

Knowledge

. Overview

of

the

KL

- ONE

Knowledge

Representation System," Cognitive Science 9.2. Bresnan , J., ed. 1972. Theory of Complementation in English Syntax, Ph .D . Dissertation , MIT . Bresnan , J., ed. 1982. The Mental Representation of Grammatical Relations , MIT Press, Cambridge , MA . Bresnan , J. and J. Kanerva . 1989. "Locative Inversion in Chichewa," Linguistic Inquiry 20:1-50. Bresnan , J. and L . Moshi . 1990. "Object Asymmetries in Comparative Bantu Syntax ," Linguistic Inquiry 21:147- 185. Bresnan , J. and A . Zaenen. 1990. "Deep Unaccusativity in LFG ," in K . Dziwirek et al., eds., Grammatical Relations : A Cross- Theoretical Perspective, CSLI , Stanford and U . of Chicago, Chicago. Briscoe , E ., A . Copestake, and B . Boguraev . 1990. "Enjoy the Paper: Lexical Semantics via Lexicology ," Proceedings of 13th International Conference on Computational Linguistics , Helsinki , Finland , pp . 42- 47. Briscoe ,T ., V . de Paiva , and A . Copestake (eds.) . 1993. Inheritance , Defaults , and the Lexicon, Cambridge University Press, Cambridge . Brugman , C. 1981. The Story of 'Over ': Polysemy, Semantics, and the Structure of the Lexicon , Master 's Thesis, University of California , Berkeley. Brugman , C. and G. Lakoff . 1988. "Cognitive Topology and Lexical Networks ," in S. I . Small , G. W . Cottrell , and M . K . Tanenhaus, eds., Lexical Ambiguity Resolution , Morgan Kaufmann , San Mateo , CA ., p. 477-508. Buitelaar , P., and A .-M . Mineur . 1994. "Compositionality and Coercion in Categorial Grammar ," in Proceedings of Ninth Amsterdam Colloquium , December 14- 17, 1993, Amsterdam . Burge , T . 1972. "Truth and Mass Terms," Journal of Philosophy 69, 263-282. Burzio , L . 1986. Italian Syntax: A Government and Binding Approach , Reidel , Dordrecht . Cardelli , L . 1988. "A Semantics of Multiple Inheritance ," InfoTm (l,tion and Computation 76:138- 164.

272

Bibliography

Cardelli , L . and P . Wegner . 1985 . " On Understanding Types , Data Abstraction , and Polymorphism ," ACM Computing Surveys , 17:471 - 522 . Carlson , G . 1977 . Reference to Kinds in English , PhD . dissertation , University of Massachusetts

, Amherst

.

Carlson , G . 1984 . "Thematic Roles and Their Role in Semantic Interpretation ," Linguistics 22 :259- 279 . Carlson , G . 1989 . " On the Semantic Composition of English Generic Sentences ," in

G . Chierchiat B . Partee and R. Turner (eds.), Properties , Types and Meaning , II

t Kluwer

, Dordrecht

, 167 - 192 .

Carnap , R . 1956 . Meaning and Necessity , University of Chicago Press , Chicago . Carpenter , B . 1992 . "Typed Feature Structures ," Computational Linguistics , 18 :2. Carrier , J . and J . H . Randall . 1992 . "The Argument Structure and Syntactic Struc ture of Resultatives ," Linguistic Inquiry 23 :173- 234 . Carter t R . 1988a . On Linking : Papers by Richard Carter , B . Levin and C . Tenny ., eds ., MIT Lexicon Project Working Papers 25 , Center for Cognitive Science , MIT , Cambridge , MA . Carter , R . 1988 . " Compositionality and Polysemy ," in B . Levin and C . TennYt (eds .) , On Linkin .fJ: Papers by Richard Carter , MIT Lexicon Proiect Working Papers no . 25 , Dept . of Linguistics and Philosophy , MIT , Cambridge . Cattell , R . 1984 . Composite Predicates in English , Syntax and Semantics 17, Aca demic

Press

, New

York

, NY

.

Centineo , G . 1986 . ~'A Lexical Theory of Auxiliary Selection in Italian ," Davis Working Papers in Linguistics , 1: 1-35 . Charniak , E . and R . Goldman . 1988 . "A Logic for Semantic Interpretation ," in Proceedings of the 26th Annual Meeting of the Association for Computational Linguistics , Buffalo . Chierchia

, G . 1982 . " Bare

Plurals

, Mass

Nouns

, and Nominalization

,"

WCCFL

1,

243 - 255 .

Chierchia , G . 1984 . Topics in the Syntax and Semantics of Infinitives and Gerunds , Doctoral dissertation , University of Massachusetts , Amherst , MA . Chierchia , G . 1989 . " A Semantics for Unaccusatives and its Syntactic Conse quences ," unpublished ms ., Cornell University , Ithaca , NY . Chierchia

, G . 1989 . " Structured

Meanings

, Thematic

Roles , and

Control

," in G .

Chierchia , B . Partee , and R . Turner , eds ., Properties , Types , and Meaning , vol . 2 , Kluwer

Academic

Publishers

, Dordrecht

, The

Netherlands

, 131 - 166 .

Chierchia , G . and S. McConnell -Ginet . 1990 . Meaning and Grammar : An Intro duction to Semantics , MIT Press , Cambridge . Chomsky , N . 1955 . The Logical Structure of Linguistic Theory , University of Chicago Press , first published 1975 . Chomsky , N . 1957 . Syntactic Structures , Mouton , The Hague . Chomsky , N . 1965 . Aspects of the Theory of Syntax , MIT Press , Cambridge . Chomsky , N . 1970 . "Remarks on Nominalization ," in R . Jacobs and P .S. Rosen baum (eds .) , English 7' ranformational Grammar , Ginn , Waltham , MA . Chomsky , N . 1972 . Studies on Semantics in Generative Grammar , Mouton , The Hague . Chomsky , N . 1973 . " Conditions on Transformations ," in S. Anderson and P .

Kiparsky , (eds.) , A Festschrift for Morris Halle , Holt , Rinehart , and Winston , New

York

.

Chomsky , N . 1975 . Reflections on Language , Pantheon , New York . Chomsky , N . 1981 . Lectures on Government and Binding , Faris Publications drecht

Chomsky , N . 1986 . Knowledge New

, Dor -

. York

of Language , its Nature , Origin , and Use, Praeger ,

.

Chomsky , N . 1994a . Language and Thought , Moyer Bell , Wakefield , R .I .

Bibliography

273

Chomsky , N . 1994b . " Language as a Natural Object ," Jacobsen Lecture , Univer sity College , London , May 23 , 1994 . Cohen , P.R ., J . Morgan , and M . Pollack , eds . 1990 . Intentions in Communication , MIT Press , Cambridge , MA . Collins , A . and M . Quillian . 1969 . " Retrieval Time from Semantic Memory ," Jour nal of Verbal Learning and Verbal Behavior , 9 :240- 247 . Comrie , B . 1976 . Aspect , Cambridge University Press , Cambridge . Comrie , B . 1980 . Tense , Cambridge University Press , Cambridge . Comrie , B . and S. Thompson . 1985 . " Causative Verb Formation and Other Verb deriving Morphology ," in T . Shopen (ed .) , Language typology and syntactic description .' Grammatical categories and the lexicon , Cambridge University Press , Cambridge . Conestake .. A . 1991 . " Defaults in the LRL ." in T . Briscoe , A . Copestake , and V . de Paiva , eds ., Proceedings of the Acquilex Workshop on Default Inheritance in the Lexicon , University of Cambridge Computer Laboratory Technical Report No . 238 , 1991 . Copestake , A . 1992 . The Representation of Lexical Semantic Information , CSRP 280 , University of Sussex . Copestake , A . 1995 . "The Representation of group denoting nouns in a lexical knowledge base ," in P . Saint - Dizier and E . Viegas , eds ., Computational Lexical Semantics , Cambridge University Press . Copestake , A . and T . Briscoe 1992 . " Lexical Operations in a Unification -based Framework ," in J . Pustejovsky and S. Bergler , eds ., Lexical Semantics and Knowledge Reperesentation , Springer Verlag , Berlin . Copestake , A . and T . Briscoe 1995 . " Semi -Productive Polysemy and Sense Exten sion ," Journal of Semantics 12. Copestake , A ., A . Sanfilippo , T . Briscoe , and V . de Paiva . 1993 . "TheACQUILEX LKB : An Introduction ." in T . Briscoe , V . de Paiva1 and A . Copestake (eds .) , Inheritance , Defaults , and the Lexicon , Cambridge University Press , Cam bridge . Croft , W .A . 1990 . " A Conceptual Framework for Grammatical Categories : A Taxonomy of Propositional Acts .," Journal of Semantics 7 :245 - 79 . Croft , W . 1991 . Typology and Universals , Cambridge University Press . Croft , W . 1991b . Categories and Relations in Syntax : The Clause -Level Organi zation of Information , University of Chicago Press . Crouch , R . and S. Pulman . 1993 . "Time and Modality in a Natural Language Interface to a Planning System ," Artificial Intelligence 63 :265- 304 . Cruse , D . A . 1986 . Lexical Semantics , Cambridge University Press . Cruse , D . A . 1995 . "Polysemy and related phenomena from a cognitive linguistic viewpoint ," in P . Saint - Dizier and E . Viegas (eds .) , Computational Lexical Se~ antics , Cambridge University Press . Dahl , O . 1981 . " On the Definition of the Telic -Atelic {Bounded -Nonbounded . Dis tinction ," Syntax and Semantics : Tense and Aspect 14, p . 79- 90 . Davidson , D . 1967 . "The Logical Form of Action Sentences ," in N . Rescher , ed ., The Logic of Decision and Action , Pittsburgh University Press , Pittsburgh . Davidson \ D . and G . Harman , eds ., 1972 . Semantics of Natural Language , Reidel , Dordrecht . Decles , J .- P . 1989 . "State , Event , Process , and Topology ," General Linguistics 29 :159- 200 . de Paiva , V . 1993 . " Types and Constraints in the LKB ," in T . Briscoe , V . de Paiva , and A . Copestake (eds .) , Inheritance , Defaults , and the Lexicon , Cambridge University Press , Cambridge . Diesing , M . 1992 . Indefinites , MIT Press , Cambridge .

274

Bibliography

Dini L . and F . Busa . 1994 . " Generative Operations in a Constraint -based Gram mar ," in Proceedings of KONVENS -94 , Vienna , Austria , September 28- 30 . Di Sciullo , A . M . and S. T . Rosen 1990 . " Light and Semi - Light Verb Construc tions ," in K . Dziwirek , P . Farrell and E . Mejfas - Bikandi , eds ., Grammatical Relations : A Cross - Theoretical Perspective , Center for the Study of Language and Information , Stanford University , Stanford , CA , 109- 125 . DiScuillo , A . M . and E . Williams 1987 . On the Definition of Word , MIT Press , Cambridge . Dixon , R . M . W . 1984 . " The Semantic Basis of Syntactic Properties ," BLS 10 ) 583 - 595 . Dixon , R . M . W . 1991 . A New Approach to English Grammar , on Semantic Prin ciples , Oxford University Press . Dolling , J . 1992 . " Flexible Interpretationen durch Sortenverschiebung ," in I . Zim mermann and A . Strigen (eds .) , Fugungspotenzen , Berlin , Akademie Verlag . Dor , D . 1992 . "Towards a Semantic Account of Concealed Questions ," ms . Stan ford University . Dowty , D .R . 1976 . " Montague Grammar and the Lexical Decomposition of Causative Verbs ," in B . Partee , ed ., Montague Grammar , Academic Press , New York , NY , 201-246 . Dowty , D . R . 1979 . Word Meaning and Montague Holland .

Gmmmar , D . Reidel , Dordrecht ,

Dowty , D .R . 1982 . " Grammatical Relations in Montague son and G .K . Pullum , eds ., The Nature of Syntactic Dordrecht . 79- 130 .

Grammar ," . in P . Jacob Representation , Reidel ,

Dowty , D . R . 1985 . " On Some Recent Analyses of Control ," Linguistics losophy 8 :1-41 . Dowty . D . R . 1989 . " On the Semantic Content of the Notion - ~ - ~- ' Thematic - -- -----~--1 G ~ 'Chierchia , B . Partee , and R . Turner , eds . Properties , Types , and Vol . 2, Semantic Issues , Dordrecht . Dowty , D . R . 1991 . " Thematic Proto -Roles and Argument Selection ," 67 :547 - 619 .

and Phi R.ol - -- --~' ." in Meaning ---, Language

Eberle , K . 1988 . " Partial Orderings and Aktionsarten in Discourse Representation Theory ," in Proceedings of COLING - 88 , Budapest , 1988 , pg . 160- 165 . Elliott , D . E . 1974 . " Towards a Grammar of Exclamations ," Foundations of Lan guage 11. En ~, M . 1981 . Tense without Scope, Ph .D . Dissertation , University of Wisconsin , Madison . En ~, M . 1987 . "Anchoring Conditions for Tense ," Linguistic Inquiry 18:633 - 658 . Eschenbach , C . 1993 . " Semantics of Number ," Journal of Semantics , 10:1-23 . Evans , R . and G . Gerald . 1989 . " Inference in DATR ," in the Proceedings of the Fourth European ACL Conference , April 10- 12 , 1989 , Manchester , England . Evans , R . and G . Gerald . 1990 . "The DATR papers : February 1990 ." Cognitive Science Research Paper CSRP 139 , School of Cognitive and Computing Science , University of Sussex , Brighton , England . Evens , M . 1987 . Relational Models of the Lexicon , Cambridge University Press , Cambridge . Farkas , D . F . 1988 . " On obligatory 58 .

Control ," in Linguistics

and Philosophy , 11:27-

Fass , D . 1988a . "An Account of Coherence , Semantic Relations , Metonymy , and Lexical Ambiguity Resolution ," in S.l . Small ~ G .W . Cottrell ~ and M .K . Tanen haus , eds . Lexical Ambiguity Resolution , Morgan Kaufmann , San Mateo , CA . Fass , D . 1993 . " Lexical Semantic Constraints ," in J . Pustejovsky (ed .) Semantics and the Lexicon , Kluwer Academic Publishers , Dordrecht . Fauconnier , G . 1985 . Mental

Spaces, MIT

Press , Cambridge .

Bibliography

275

Fellbaum , C. 1990. "English Verbs as a Semantic Net ," International Journal of Lexicography, 3:278-301. Fillmore , C. 1968. "The Case for Case," in Universals in Linguistic Theory, E. Bach and R . Harms , eds.,. New York , Holt , Rinehart , and Winston . Fillmore , C. 1971. "Types of Lexical Information ," in D . Steinberg and L . Jakobovits , eds., 370-392. Fi11mor~., C. 1977a. "Tonics in Lexical Semantics," in R. W . Cole, ed., Current Issues in Linguistic Theory, Indiana University Press, Bloomington . Fillmore , C. 1977b. "The Case for Case Re-opened," in P. Cole, ed., Syntax and Semantics, Vol. 8, Academic Press, New York . Fillmore , C. 1986. "Pragmatically Controlled Zero Anaphora ," BLS 12, 95- 107. Fillmore , C. 1988. "The Mechanisms of Construction Grammar ," BLS 14, 35-55. Fillmore , C. and B . T . Atkins . 1991. "Toward a Frame-Based Lexicon : The Semantics of Risk ," paper delivered at ACL , Berkeley, CA . Fillmore , C. and B . T . Atkins . 1994. "Starting where the dictionaries stop : the Challend of corpus lexicography ," in B . T . Atkins and A . Zampolli , eds. Computational Approaches to the Lexicon, Oxford University Press, Oxford . Fillmore ., C. and D . T . Langendoen, eds., 1971. Studies in Linguistic Semantics, Holt , Rinehart and Winston , New York , NY . Fisher , C., H . Gleitman , and L . R . Gleitman 1991. "On the Semantic Content of Subcategorization Frames," Cognitive Psychology 23. Flickinger , D . C. Pollard , and T . Wasow 1985. "Structure -Sharing in Lexical Representation ," Proceedings of 23rd Annual Meeting of the ACL , Chicago, IL , pp .262- 267. Fodor , J. 1975. The Language of Thought, Harvard University Press, Cambridge . Foley, - W .. and R . Van Valin 1984. Functional Syntax and Universal Grammar , Camrbidge University Press, Cambridge . Freed, A . F . 1979. The Semantics of English Aspectual Complementation , Dor drecht : Reidel . Gawron , M . 1983. Lexical Semantics and the Semantics of Complementation , Ph .D . dissertation , University of California at Berkeley. Gazdar , G. , E . Klein , G . Pullum , and I . Sag. 1985. Generalized Phrase Structure Grammar , Harvard University Press. Gentner , D . and I . M . France 1988. "The Verb Mutability Effect : Studies of the Combinatorial Semantics," in S. I . Small , G. W . Cottrell , and M . K . Tanenhaus, eds., Lexical Ambiguity Resolution , Morgan Kaufmann , San Mateo , CA . Gleitman , L . 1990. "The Structural Sources of Verb Meaning ," Language Acqui sition , 1, 3-55. Godard , D . and J. Jayez. 1993. "Towards a proper treatment of Coercion Phenomena," in Proceeding of the 1993 European ACL . Goldberg , A . 1992. Argument Structure Constructions . Ph .D . Dissertation , Uni versity of California , Berkeley. Goldberg , A . 1995. Constructions : A Construction Grammar Approach to Argu ment Structure , University of Chicago Press, Chicago. Goodman , N . 1951. The Structure of Appearance , Reidel Publishing , Dordrecht , The Netherlands . Green, G. 1974. Semantics and Syntactic Regularity , Indiana University Press, Bloomington . Grice , H . P. 1971. "Meaning ," in D . Steinberg and L . Jacobovits , eds., Semantics : An Interdisciplinary Reader in Philosophy , Linguistics , and Psychology, Cambridge , Cambridge University Press. Grimshaw , J. 1979. "Complement Selection and the Lexicon " Linguistic Inquiry 10:279-326.

276

Bibliography

Grimshaw, J. 1982. "Subcategorizationand GrammaticalRelations," in A. Zaenen (ed.), Subjectsand other Subjects : Proceedingsof the Harvard Conferenceon GrammaticalRelations, Bloomington, Indiana Linguistics Club, p. 35-55. Grimshaw, J. 1990. Argument Structure, MIT Press, Cambridge, MA . Grimshaw, J. and S. Vikner 1993. "Obligatory Adjuncts and the Structure of Events," in E. Reuland and W. Abraham (eds.), Knowledgeand Language , Vol. II , Kluwer AcademicPublishers, Dordrecht, p. 143-155. Grinder, J. 1970. "SuperEqui-NP Deletion," CLS 6. Groenendijk, J., and M. Stokhof. 1989. "Type-Shifting Rulesand the Semanticsof Interrogatives" , in G. Chierchia, B. Partee, and R. Turner (eds.), Properties, Tpyes, and Meaning, Dordrecht, Reidel. Gropen, J., S. Pinker, M. Hollander, and R. Goldberg 1991. "Affectednessand Direct Objects: the Role of Lexical Semanticsin the Acquisition of Verb Argument Structure," Cognition41:153-195. Gross, D. and K . Miller 1990. "Adjectivesin WordNet," International Journal of Lexicography , 3, pp. 265-277. Grosz, B. and C. Sidner. 1986. "Attention, Intentions, and the Structure of Discourse," ComputationalLinguistics 12:175-204. Gruber, J. S. 1967. "Lookand See," Language43, 937-947. Gruber, J. S. 1976. Lexical Structuresin Syntax and Semantics, North-Holland, Amsterdam. Guentcheva , Z. 1990. "L'oppostion perfectifjimperfectif et la notion d'acheve ment," in J. Fontanille (ed.), Le discours aspectualise , Collection Nouveaux Actes Semiotiques , Benjamins, Paris. Gunter, C. 1992. Semanticsof ProgrammingLanguages , MIT Press, Cambridge. Hacking, I. 1983. Representingand intervening: introductory topics in the philosophy of natural science, CambridgeUniversity Press, New York. Hale, K . and S. J. Keyser. 1986. {'SomeTransitivity Alternations in English," Lexicon Project Working Papers7, Centerfor CognitiveScience , MIT , Cambridge, MA . Hale, K . and S. J. Keyser. 1987. "A View from the Middle," Lexicon Project Working Papers 10, Center for Cognitive Science , MIT , Cambridge, MA . Hale, K .L. and S. J. Keyser 1992. "The Syntactic Character of Thematic Structure," in I . M. Roca, ed., Thematic Structure: Its Role in Grammar, Foris, Berlin, 107-143. Harris, Z. 1951. Methodsin Structural Linguistics, University of ChicagoPress, Chicago. Harris, Z. 1952. "DiscourseAnalysis" Language28:130. Harris, Z. 1957. "Co-Occurrenceand Transformationin Linguistic Structure" Lan.qua.Qe33:283-340. Harris, Z. 1965. "TransformationalTheory," Language41:363-401. Hayes, P. 1979. "Naive PhysicsManifesto," in D. Mitchie (ed.), Expert Systemsin the Micro-electronicAge, Edinburgh University Press, Edinburgh, Scotland. Heim, I . 1979. "ConcealedQuestions", in R. Bauerle, U. Egli, and A. von Stechow (eds.), Semanticsfrom Different Points of View, Springer, Berlin. Hendriks, H. 1987. "Type Changein Semantics : The Scopeof Quantification and Coordination," in E. Klein and J. van Benthem (eds.), Categories , Polymorphism, and Unification, Centrefor CognitiveScience , University of Edinburgh, and Institute for Language, Logic and Information, University of Amsterdam. Heylen, D. 1995. "Lexical Functionsand KnowledgeRepresentation ," in P. SaintDizier and E. Viegas(eds.), ComputationalLexicalSemantics,CambridgeUniversity Press. Herskovits, A. 1987. Languageand Spatial Cognition, CambridgeUniversity Press, Cambridge, UK.

277

Bibliography

Herskovits , A . 1988. "Spatial Expressions and the Plasticity of Meaning ," in B . Rudzka- Ostyn (ed.) , Topics in Cognitive Grammar , John Benjamins , Amster dam Herzo

, 271 .e; ,

- 297

.

O . and

grating

C .- R .

CL

Higginbotham native

to

and

AI

, J .

1983

.

Situation

Higginbotham

,

quiry

14 : 395

J .

1983

- 420

.

Higginbotham

, J . 1985 ,

12 : 465

- 517

J .

Hirst

English ,.

G .

Hirst

1987

,

G .

.

1988

( eds

gies

,

Hobbs

1987

.

Meaning

:

An

:

Extensional

Philosophy and

80

Nominals

,

Inte

-

- 127

-

.

Linguistic

16 : 547

Linguistics

Alter

100

,"

Inquiry ,"

for

- 593

In

-

.

and

an

,"

, T

Philosophy

Computationally , G .W

S . I . Small

of

NP

Reference

Ambiguity

,

of , W

Cam

-

Word

Meaning

, and

,"

,"

. Ringle

,"

K

. Laws

.

Computational

. Edwards ,"

M

. Tanen

Mateo

Discourse

and

.K

-

, CA

in

, eds

.

Strate

-

. , Hills

-

in

Proceedings

of

.

, D . Edwards

Base

M

, San

in

. Lehnert

and

, D

Kaufmann

Coherence

Spreading

, and

.

Mexico

Semantics

with

. Cottrell

, Morgan

Knowledge

. Davies

Resolution

Ambiguity

Associates

. Davies

and

.

in

Processing

Lexical , T

the

Understanding

, New

Aktionarten

University

Resolution

Knowledge

, and

K

Artificial

1987a

.

" Commonsense

Linguistics . Laws

13 .

. 1987b

Intelligence

.

" The

Center

,

TACITUS SRI

Interna

-

. . Stickel

, P . Martin

Intelligence ,

T

.

Structure

1992 .'

, L . R . 1980

Horn

,

L .

R

.

Chicago

.

, D . Edwards : 69 - 142

" Aspect

Role

.

Theta

Grammar

Natural

.

" Projection

Theory

, Foris

and

A

. 1993

. " Interpretation

as

Abduction

,"

.

and

in

" Affixation

1989

the

,"

, Berlin

in ,

Unaccusative

History

of

I .

145

M

.

Roca

- 174

( ed

Hypothesis

Negation

!

Event

Structure

.) ,

Thematic

. ,"

University

CLS of

16 , 134

- 146

Chicago

.

Press

,

.

, A

. van

.

Everaert

1995

, and

Lawrence , E .

1993

.

,"

Artificial

, R . and

R .

Associates

,"

Project

J . in

Jackendoff

147

Tenny

, M

.

Insertion

,

- 385

using

Discourse

Structure

Re

-

.

Bake

Revisited

,"

International

Journal

of

1990 .,

2 ,

.

The Center

" Active

Objects

MIT

in

Parsing for

Syntax

Volume

Cognitive

,

Science

,

Semantics

1988 ,

,

- 1989

MIT

,

,

and

Parsing

Cambridge

,

, R .

1983

,

R .

1985a

,

R .

and

. Semantics .

and

" Multiple

Linguistic

1985b Inquiry

Interpretation

in

Generative

Grammar

, MIT

Press

,

.

Jackendoff

Linguistic

ed

. Semanic

, MA

Language

. ,

Papers

J ackendoff

J ackendoff

Lexical

.

, R . 1972

Cambridge

P . Coopmans

, N .J .

: 341

" English

in

and

.

Pustejovsky

C .

- 169

.

,"

Specification

Generation 63

1989 - 346

Working ,

Discourse

' cuk

2 : 325

on

. ) , Lexical

, Rillsdale

Intelligence

I . Mel

and

( eds

" Automated

Lexicography

Parsing

based

J . Grimshaw

Erlbaum

lations

,

63

.

Its

Horn

MA

Linguistic

of

and

Lexical

Cruces

. Croft

, J ., M

Ingria

LILOG

.

Words

" World

and

, J ., W

Hoekstra

Iison

,"

State

, Cambridge

Erlbaum

. Croft

Artificial

Hovy

,

Semantics

Language

Commonsense

Rout

in

Reports of

Binding

Interpretation -

. " Towards

, J ., W

Hobbs

,

Ohio

Ambiguity

- 3 , Las

tional

Form

,

Polaroid

Metaphysics Hobbs

Understanding .

Journal

Semantics

" Resolving

Natural

J .

Text

Perceptual

" Elucidations

Press

, N J , Lawrence

TINLAP

.

, Berlin

The

Compositional

. ) , Lexical

for

" On .

1991

of ,"

" Logical

Semantic

.

, J . 1982

dale

Logic

Dissertation

and

haus

Hobbs

. A

University

Activation

Hobbs

.

1989

. PhD

bridge

" The

.

.

- Verlag

.

, E . 1985

in

, eds

Semantics

Higginbotham

Hinrichs

Rollinger , Springer

.

Cognition

Theory

3 : 271

" Believing 16 : 445

, MIT

Subcategorization

and - 460

.

Press and

- 295

Intending

, Cambridge the

O - Criterion

, MA ,"

. Natural

. :

Two

Sides

of

the

Same

Coin

,"

278

Bibliography

Jackendoff, R. 1985c. " Multiple

Subcategorization and the Theta - Criterion : The Case of Climb ," Natural Language and Linguistic Theory 3 '271- 295 . Jackendoff , R . 1987 . "The Status of Thematic Relations in Linguistic Theory ," Linguistic Inquiry 18:369 -411 . J ackendoff , R . 1990 . Semantic Structures , MIT Press , Cambridge , MA . Jackendoff , R .S. 1992a . " Babe Ruth Homered His Way into the Hearts of Amer ica ," in T . Stowell and E . Wehrli (eds .) , Syntax and the lexicon , Academic Press , San Diego , 155- 178 . Jackendoff , R . 1992b . " Mme . Tussaud Meets the Binding Theory ," Natural Lan guage and Linguistic Theory 21 :427 -455 . Jackendoff , R . 1993 . "The Conceptual Structure of Intendingand Volitional Ac -

tion ," in H . Campos and P. Kempchinsky (eds.) , Festschrift for Carlos Otero, Georgetown University Press , Georgetown . Jacobs , R . and P . Rosenbaum , eds . 1967 . Readings in English Grammar

, Ginn

, Waltham

, MA

Transformational

.

Jakobson , R . 1970 . Recent Developments in Linguistic Science , Perenial Press . Jacobson , P. 1990 . " Raising as Function Composition ," in Linguistics and Philos ophy 13:423 -476 . Jacobson , P . 1992 . " The Lexical Entailment Theory of Control and the Tough Construction ," in I .A . Sag and A . Szabolcsi , (eds .) , Lexical Matters , CSLI , Stanford

, 269 - 299 .

Jespersen , O . 1924 . The Philosophy of Grammar , Norton , New York . Johnston , M . 1995 . " Semantic Underspecification and Lexical Types : Capturing Polysemy without Lexical Rules ," in Proceedings of ACQUILEX Workshop on Lexical Rules , August 9- 11, 1995 , Cambridgeshire . Johnston , M ., B . Boguraev , and J . Pustejovsky . 1995 . "The Acquisition and In terpretation of Technical Complex Nominals ," in Proceedings of AAAI Sym posium on Lexical Acquisition , March 29- 31 , Stanford . Kamp , H . 1979 . " Some Remarks on the Logic of Change : Part 1" in C . Rohrer ,

(ed.) , Time , Tense, and Quantifiers , Tlibingen , Niemeyer. Kamp , H . and U . Reyle . 1993 . From Discourse lishers

, Dordrecht

to Logic , Kluwer

Academic

Pub -

.

Kamp , H . and A . RoBdeutscher . 1992 . " Remarks on Lexical Structure , DRS Construction , and Lexically Driven Inferences ," Arbeitspapiere des Sonder forschungsbereichs 340 Sprachtheoretische Grundlagen fur die Computerlin guistik , Nr . 21 , Stuttgart . Karttunen , L . 1971 . " Implicative Verbs " . Language 47 :340 -58 . Karttunen , L . 1974 . " Presupposition and Linguistic Context " . Theoretical Lin guistics 1:181-93 . Katz , J . 1964 . " Semantic Theory and the Meaning of ' Good ' ," Journal of Philos ophy61 :739- 766 . Katz , J . 1972 . Semantic Theory , Harper and Row , New York . Katz , J . and J . Fodor . 1963 . "The Structure of a Semantic Theory ," Language 39 : 170 - 210 .

Kautz , H . 1987 . A Formal Theory of Plan Recognition versity of Rochester . Keenan , E . and L . Faltz . 1985 . Boolean Semantics for Publishing , Dordrecht . Kenny , A . 1963 . Action , Emotion , and Will , Routledge Keyser , S. J . and T . Roeper . 1984 . " On the Middle and English ," Linguistic Inquiry 15:381-416 . Keyser , S. J . and T . Roeper . 1992 . " Re : The Abstract guistic Inquiry 23 :89- 125 .

, Ph .D . Dissertation Natural

, Uni -

Language , Reidel

and Kegan Paul , London . Ergative Constructions in Clitic

"

Hypothesis .

Lin -

279

Bib Ii 0gr aphy

Kiefer , F . 1992 . " Hungarian

Derivational

Morphology

, Semantic

Complexity

, and

Semantic Markedness," in I .A . Sag and A . Szabolcsi, (eds.), Lexical Matters , CSLI

, Stanford

, 183 - 208 .

Kilgarriff , A . 1995 . " Inheriting Polysemy ," in P . Saint -Dizier and E . Viegas (eds .) , Computational Lexical Semantics , Cambridge University Press , Cambridge . Kiparsky , P . and C . Kiparsky . 1971 . "Fact ," in D . Steinberg and L . Jakobovitz ,

(eds.) , Semantics, Cambridge University Press, Cambridge , 345-369. Klavans , J ., M . Chodorow , N . Wacholder . 1990 . " From Dictionary to Knowledge Base via Taxonomy ," Proceedings of the 6th Conf . UW Centre for the New OED , Waterloo

, 110 - 132 .

Klein , E . and I . Sag . 1985 . " Type - Driven Translation

," Linguistics

and Philosophy

8 : 163 - 202 .

Klein , E . and J . van Benthem . 1987 . Categories , Polymorphism , and Unification , Centre for Cognitive Science , University of Edinburgh , and Institute for Lan guage , Logic and Information , University of Amsterdam . Kowalski , R . and M . Sergot . 1986 . "A Logic -based Calculus of Events ," New Gen eration Computing 4 :67-95 . Kratzer , A . 1988 . Amherst .

" Stage - level

vs . Individual

- level

Predicates

,"

ms . UMASS

,

Kratzer , A . 1989 . " An Investigation into the Lumps of Thought ," Linguistics and Philosophy , 12:607- 653 . Krifka , M . 1989 . " Nominal Reference , Temporal Constitution , and Quantification in

Event

Semantics

,"

in

R . Bartsch

, J . van

Ebthe

, and

P . van

Emde

Boas

(eds.), Semantics and Contextual Expressions, Foris, Dordrecht , pg. 75- 115. Krifka

, M . 1992 . " Thematic

Relations

as Links

between

Nominal

Reference

and

Temporal Constitution ," in I . Sag and A . Szabolcsi, (eds.), Lexical Matters , CSLI Lecture Notes , University of Chicago Press , Chicago , Ill . Kunze , J . 1991 . Kasusrelationen und semantische Emphase , Akademie Berlin

Verlag ,

.

Ladusaw , W . and D . Dowty 1988 . "Toward a Nongrammatical Account of The matic Roles ," in W . Wilkins , (ed .) , 62- 73 . Lakoff , G . 1968 . " Instrumental Adverbs and the Concept of Deep Structure ," Foundations of Language , 4 :4- 29 . Lakoff G . 1970 . Irregularity in Syntax . Holt , Rinehart , and Winston . Lakoff G . 1971 . " On Generative Semantics ," in Semantics : An Interdisciplinary Reader , D . Steinberg and L . Jakobovits , eds ., Cambridge University Press . Lakoff

, G . 1972 . " Linguistics

and

Natural

Logic

," in D . Davidson

and

G . H . Har -

mon , eds ., The Semantics of Natural Language , Reidel , Dordrecht , 545 -665 . Lakoff G . 1987 . Women , Fire , and Dangerous Objects , University of Chicago Press , Chicago , 1987 . Lambek , J . 1958 . " The Mathematics matica ~ Monthlv 65 . 154 - 170 . also

of Sentence Structure ," American in W . Buskowski , W . Marciszewski

Mathe , and J .

van Benthem (eds.) , 1988, Categorial Grammar , John Benjamins , Amsterdam , 153 - 172 .

Lambek , J. 1961. "The the Calculus of Syntactic Types ," in R . Jakobson (ed.) , Structure of Language and its Mathematical Aspects , Proceedings of the Sym posia in Applied Mathematics XII , American Mathematical Society , 166- 178 Lamiroy , B . 1987 . " Les Verbes de Mouvement Emplois Figures et Extensions Metaphoriques ," Langue Fran ~aise 26 :41-58 . Lamiroy , B . 1987 . " The complementation of aspectual verbs in French ," in Lan guage

, 63 .

Landman , F . 1991 . Structures drecht

.

for Semantics , Kluwer

Academic

Publishers , Dor -

280

Bibliography

Lang , E . 1989 . " The Semantics

of Dimensional

Designation

of Spatial

Objects ,"

in M . Bierwisch and E . Lang (eds.) , Dimensional Adjectives : Grammatical Structures and Conceptual Interpretation , Springer Verlag , Berlin . Lang , E . 1991 . " The LILOC Ontology from a Linguistic Point of View ," in O .

Herzog and C.-R . Rollinger (eds.) , Text Understanding in LILOC : Integrating GL and AI , Springer -Verlag , Berlin . Langacker , R .W . 1991 . Foundations of Cognitive Grammar : Descriptive Applica tion , Vol . 2 , Stanford University Press , Stanford , CA . Lascarides A . and N . Asher . 1993 . " Temporal Interpretation , Discourse Relations and Commonsense Entailment ," in Linguistics and Philosophy 16:437 -494 . Lascarides , A ., E .J . Briscoe , N . Asher and A . Copestake . 1994 . "Persistent associative default unification ," Lin .quistics and Philosophy 17. Lasnik , H . and R , Fiengo . 1974 . " Complement Object Deletion ," Linguistic In Quiry

5 :535 - 571 .

Lebeaux , D . 1986 . " The Interpretation of Derived Nominals ," CLS fJfJ, Part 1, Papers from the General Session , 231- 247 . Lees , R . B . 1966 . The Grammar of English Nominalizations , Mouton , The Hague . Leifi , H . 1991 . " Polymorphic Constructs in Natural and Programming Languages ," Semantics of Programming Languages , Springer , Berlin . Lenders , W . 1989 . " Computergestiitzte Verfahren zur semantischen Beschreibung yon Spracher ," in Handbuch Computerlinguistik , Berlin , De Gruyter , pg . 231244 .

Levi , J . N . 1978 . The Syntax Press

, New

York

and Semantics

Levin , B ., ed . 1985 . Lexical Semantics 1 , MIT

of Complex

Nominals , Academic

.

in Review , Lexicon Project

Working

Papers

.

Levin , B . 1985. "Introduction ," in B . Levin , (ed.) , Lexical Semantics in Review, Lexicon Project Working Papers 1, Center for Cognitive Science , MIT , 1-62 . Levin , B . 1993 . Towards a Lexical Organization of English Verbs , University of Chicago Press , Chicago . Levin , B . and T . R . Rapoport 1988 . " Lexical Subordination ," Proceedings of CLS 24 , 275 - 289 , 1988 .

Levin , B . and M . Rappaport . 1986 . " The Formation of Adjectival Passives ," Lin guistic Inquiry , 17.4 . Levin , B . and M . Rappaport . 1988 . " On the Nature of Unaccusativity ," in Pro ceedings of NELS 1988 . Levin , B . and M . Rappaport Hovav . 1991 . " Wiping the Slate Clean : A Lexical Semantic Exploration ," Cognition 41 :123- 151. Levin , B . and M . Rappaport Hovav . 1995 . Unaccusatives : At the Syntax -Lexical Semantics Interface , MIT Press , Cambridge , MA . Levin , L . 1986 . Operations on Lexical Forms . Ph .D . dissertation , MIT , Cambridge . Li , J . 1994 . Riiumliche Relationen und Objektwissen : am Beispiel 'an ' und 'bei ', Gunter Narr , Tiibingen . Link , G . 1983 . " The Logical Analysis of Plurals and Mass Terms : A Lattice -

Theoretic Approach ," in r . Bauerle, C. Schwarze, and A . von Stechow, (eds.) , Meaning , Use, and Interpretation of Language , Mouton , Berlin . Lloyd , G . E . R . 1968 . Aristotle : The Growth and Structure of his Thought , Cam bridge University Press . LoCascio , V . and C . Vet , eds . 1986 . Temporal Structure in Sentence and Discourse , Foris

, Dordrecht

.

Lytinen , S. L . 1988 . " Are Vague Words Ambiguous ?" in S. I . Small , G . W . Cottrell , and M . K . Tanenhaus , eds ., Lexical Ambiguity Resolution , Morgan Kaufmann , San

Mateo

, CA

.

281

Bibliography

McArthur , T . 1981. LongmanLexicon of ContemporaryEnglish, Longman, Harlow, Essex. McCawlev. J. D. 1968b. "The Role of Semanticsin a Grammar," in E. Bach and R. T . Harms (eds.), Universalsin Linguistic Theory, Holt, Rinehart, and Winston, New York, NY. McCawley, J. D. 1975. "Lexicographyand the Count-Mass Distinction," BLS 1, 314-321. Reprinted in J.D. McCawley(1979), p. 165-173. McCawley, N. 1976. "On ExperiencerCausatives ," in M. Shibatani (ed.), 181-203. McCawley, J. 1979. Adverbs, Vowels , and other Objectsof Wonder, University of ChicagoPress, Chicago. McClure, W . 1990. "A Lexical Semantic Explanation for UnaccusativeMis.matches," in K . Dziwirek, P. Farrel, and E. Mejias-Bikandi (eds.), Grammatical Relations: A Cross- TheoreticalPerspective , CSLI, Stanford. McConnell-Ginet. S. 1982. "Adverbsand Lo.gicalForm: A Linguistically Realistic Theory ," Language 58 : 144- 184 . McKeon , R ., ed . 1968 . The Basic Works of Aristotle , Random Marantz

. A .P . 1984 . On

the Nature

of Grammatical

Relations

House , New York . , MIT

Press , Cam -

bridge , MA . Martin ~ J . 1990 . A Computational Model of Metaphor Interpretation , Academic Press , Cambridge , MA . Martin , J . 1992 . " Conventional Metaphor and the Lexicon ," in J . Pustejovsky and S. Bergler (eds .) , Lexical Semantics and Knowledge Reperesentation , Springer Verlag , Berlin . Mel 'cuk , I . A . 1973a . " Lexical Functions in Lexicographic Description ," BLS 8 , 427 - 444

.

Mel 'cuk , I . A . 1988a . " Semantic Description of Lexical Units in an Explanatory Combinatorial Dictionary : Basic Principles and Heuristic Criteria ," Interna tional Journal of Lexicography 1:165- 188 . Mel ' cuk , I . A . 1988b . Dependency Syntax , SUNY Press . Albany , New York . Miller , G . 1985 . "Dictionaries of the Mind " in Proceedings of the 23rd Annual Meeting of the Association for Computational Linguistics , Chicago . Miller , G . 1989 . " Contexts of Antonymous Adjectives ," Applied Psycholinguistics 10 : 357 - 375 .

Miller , G . 1990a . " Nouns in WordNet : A Lexical tional Journal of Lexicography 3:245- 264 . Miller

, G . , ed . 1990b . " WordNet

: An

On - Line

Inheritance

Lexical

Database

Journal of Lexicography 3 :4 . Miller , G . 1991 . The Science of Words , Scientific

American

Miller

and

, G . , R . Beckwith

, C , Fellbaum

, D . Gross

System ," Interna ," International

Press , New York .

K . J . Miller

. 1990 . " Intro

-

duction to WordNet : An On - line Lexical Database ," International Journal of Lexicography 3 :235 - 244 . Miller , G . and C . Fellbaum 1991 . " Semantic Networks of English ," Cognition 41 : 197 - 229 .

Miller , G . and P . Johnson - Laird . 1976 . Language and Perception , Belknap , Har vard University Press , Cambridge , MA . Mineur , A .- M . and P . Buitelaar . 1995 . "A Compositional Treatment of Polysemous

Arguments in Categorial Grammar ," in S. Peters and K . van Deemter (eds.) , , . . .

Ambiguity and Underspecification ~ CSLI , Stanford University . Minsky , M . 1975 . "A Framework for Representing Knowledge " in P . Winston , ed ., The Psychology of Computer Vision , McGraw - Hill , New York . Mitchell , J . C . 1988 . " Polymorphic Type Inference and Containment ," Information and Computation 76 . Mitchell , J . C . 1991 . "Type Inference with Simple Subtypes ," Journal of Func tional Programming 1:245 - 285 .

282

Bibliography

Modrak, D. 1987. Aristotle: the Power of Perception, University of ChicagoPress, Chicago. Moens, M. 1987. Tense, Aspectand TemporalReference . Ph.D. Dissertation, Edinburgh. Moens, M. and M. Steedman . 1988."TemporalOntologyand TemporalReference ," ComputationalLinguistics 14:15-28. Montague, R. 1970. "Universal Grammar," Theoria 36:373-398. Reprinted in Thomason1974, pp. 222-246. Montague, R. 1973. "The Proper in OrdinarY - Treatment of Quantification . --------'" -En-glish," in K . Hintikka, J. Moravcsikand P. Suppes(eds.), Approachesto Natural Language , Dordrecht, Kluwer, pp. 221-242. Reprinted in Thomason1974, pp. 247-270. Moortgat, M. 1988. CategorialInvestigations: Logical and Linguistic Aspectsof the LambekCalculus, Foris, Dordrecht. Moravcsik, J. M. 1975. "Aitia as GenerativeFactor in Aristotle's Philosophy," Dialogue14:622-36. Moravcsik, J. M. 1981. "How do Words get their Meanings?," Journal of Philosophy 78:5-24. Moravcsik, J. M. 1990. Thoughtand Language , Routledge, London. Morrill , G. 1992. Type-Logical Grammar, Onderzoeksinstituutvoor Taal en Spraak, Utrecht. -

Morrill

.

,

G

drecht

Morrill

.

1994

,

G

. ,

and

Theories

B

of

Mourelatos

,

losophy

- Logical

A

and - 212

.

Grammar

Carpenter

Grammar .

2 : 415

Syntax 191

Type

,

Kluwer

Academmic

Publishers

,

Dor

-

.

P

.

D

- 434 Semantics

. ,"

.

1978 .

1990

.

" Compositionality

Linguistics .

Also

" Events published

14

;

Tense

and ,

Processes in and

,

Philosophy and P

.

States

Tedeschi

Aspect

Implicational 13

and ,

Academic

:4 , " A

Logics

,

and

. Linguistics .

and

Zaenen

Press

( eds ,

New

York

. ) ,

Phi

-

1981

.

,

,

NY

.

Nebel, B. 1990. Reasoningand Revision in Hybrid RepresentationSystems , SpringerVerlag, Berlin. Nirenburg, S. 1989a. "Lexiconsfor Computer Programs and Lexicons for People," in Proceedingsof the 5th Annual Conferenceof the UW Centrefor the New Oxford English Dictionary, pp. 43-66. St. Catherine's College~ Oxford, England. Nirenburg, S. 1989b. "Knowledge -BasedMachineTranslation," Machine Translation (specialissueon knowledge - basedMT ), 4:1. Nirenburg, S. and L. Levin. 1989. "KnowledgeRepresentationand Support," Machine Translation (specialissueon knowledge -basedMT ), 4:1. Nirenburg, S. and V . Raskin. 1987. "The SubworldConceptLexiconand the Lexicon ManagementSystem," ComputationalLinguistics 13:3. Nunberg, G. 1978. The pragmatics of Reference . Indiana University Linguistics Club: Bloomington, Indiana. Nunberg, G. 1979. "The Non-uniquenessof SemanticSolutions: Polysemy," Lin guistics and Philosophy3:143-184. Nunberg, G. 1995. "Transfersof Meaning," Journal of Semantics12. Nunberg, G. and A. Zaenen. 19~2. "SystematicPolysemyin Lexicologyand Lexicography," in H. Tommola, K . Varantola, T . Tolonen, and J. Schopp(eds.), Proceedingsof Euralex II , Tampere, Finland, University of Tampere. Nunes, M. 1993. "Argument Linking in English Derived Nominals," in R. van Valin, Jr. (ed.), Advancesin Role and ReferenceGrammar, John Benjamins Publishing, Philadelphia. Quillian, M. R. 1968. "SemanticMemory," in M. Minsky (ed.), SemanticInformation Processing , MIT Press, Cambridge, MA.

283

Bibliography

Ostler,N. and B. T. Atkins.1991.PredictableMeaningShift: SomeLinguistic Propertiesof LexicalImplicationRules, in J. Pustejovskyand S. Bergler (eds.),LexicalSemanticsandKnowledge Representation, ACLSIGWorkshop Proceedings,

76-87.

Ostler,N. and B. T. Atkins.1992.PredictableMeaningShift:SomeLinguistic Propertiesof LexicalImplicationRules, in J. Pustejovskyand S. Bergler

(eds.),LexicalSemanticsand Knowledge Reperesentation, SpringerVerlag,

Berlin.

Parsons,T. 1985.UnderlyingEventsin the LogicalAnalysisof English,in E.

LePoreand B.P. McLaughlin(eds.), Actionsand Events,BasilBlackwell,Oxford, 235-267.

Parsons,T. 1989.The Progressive in English:Events,States,and Processes, Linguistics and Philosophy 12:213-241.

Parsons,T. 1990.Eventsin the Semanticsof English,MIT Press, Cambridge, MA.

Partee,B. 1992.SyntacticCategories andSemantic Type, in M.RosnerandR.

Johnson(eds.),Computational Linguistics andFormalSemantics,Cambridge

University

Press

Partee,B. andM.Rooth.1983.Generalized Conjunction andTypeAmbiguity, in Meaning,Use,andInterpretation ofLanguage, Buerle,Schwarze, andvon Stechow eds., Walter de Gruyter.

Passonneau,R. 1988.A Computational Modelof the Semanticsof Tenseand Aspect, Computational Linguistics, 14.2.

Paul, H. 1880.Prinzipiender Sprachgeschichte, 9th Edition,1975,Niemeyer, Tbingen.

Pelletier,F. J. andL. K. Schubert.1988.Problemsin the Representation ofthe Logical FormofGenerics, Plurals,andMassNouns,in B. LePore(ed), New Directions in Semantics, Academic Press, London.

Pelletier,F. J. andL.K.Schubert. 1989.MassExpressions,inD.GabbayandF.

Guenthner(eds.),Handbook ofPhilosophical Logic,Vol.IV,Reidel,Dordrecht. Perlmutter,D. M. 1970.On the TwoVerbsBegin, in R. Jacobsand P. Rosenbaum (eds.), 107-119.

Perlmutter,D. M. 1978.ImpersonalPassivesandthe Unaccusative Hypothesis, BLS 4, 157-189.

Pinker,5. 1984.Language Learnability andLanguage Development, HarvardUniversity Press, Cambridge, MA.

Pinker,5. 1989.Learnability andCognition: TheAcquisition ofArgument Structure, MIT Press, Cambridge, MA.

Pinker,S.andJ. Mehler.1988.Connections andSymbols, MITPress,Cambridge, MA.

Pollack,M. 1986.A ModelofPlanInference that distinguishes betweenthe Beliefsof Actorsand Observers, in Proceedings24th Annual Conferenceof the Association of Computational Linguistics,New York, 207-214.

Pollard,C. and I. Sag.1987.Information-Based Syntaxand Semantics,CSLI Lecture Notes Number 13, Stanford,

CA.

Pollard,C. andI. Sag.1994.Head-Driven PhraseStructureGrammar,University of Chicago Press and Stanford CSLI, Chicago.

Postal,P. 1970.On the SurfaceVerbRemind, LinguisticInquiry1:37-120.

Postal,P. 1971.Cross-over Phenomena, Holt,Rinehart, andWinston, NewYork. Prince,A. andP. Smolensky 1994.Optimality Theory, ms.JohnsHopkins University and Rutgers University. Pustejovsky, J. 1988.The Geometry of Events,in Studiesin Generative Approachesto Aspect,C. Tenny,ed.,Lexicon ProjectWorking Papers24,MIT, Cambridge,

MA.

284

Bibliography

1991a . "The

Generative Lexicon ," Computational Linguistics ,

Bibliography

285

Roeper , T. 1987 . "ImplicitArguments andtheHead Complement Relation ," Linguistic Inquiry18. Roeper , T. 1993 . "ExplicitSyntax in theLexicon : theRepresentation of Nomi nalizations ," in J. Pustejovsky (ed.), Semantics andtheLexicon , Kluwer Aca demic Publishers , Dordrecht . Rooth , M. 1985 . Association withFocus , Ph.D. Dissertation , University ofMas sachusetts , Amherst . Rudanko , J. 1989 . Complementation andCase Grammar : A Syntactic andSemanticStudyof Selected Patterns of Complementation in Present -DayEnglish , SUNY Press , Albany , NY. Ruhl,C. 1989 . OnMonosemy : A Study in Linguistic Semantics , State University ofNewYorkPress , Albany , NY. Rumelhart , D.E. andJ.L. McClellan . 1986 . Parallel distributed processing : explorations in themicrostructure ofcognition , MITPress , Cambridge , MA. Russell , G., J. Carroll , andS. Warwick . 1990 . "Multiple Default Inheritance in a Unification -Based Lexicon ," in Proceedings of Workshop onInheritance in Natural Language , Tilburg , August . Russell , G., A. Ballim , J. Carroll , andS. Warwick -Armstrong . 1993 . "A Practical Approach to Multiple Default Inheritance forUnification -Based Lexicons ," in T. Briscoe , V. dePaiva , andA. Copestake (eds .), Inheritance , Defaults , and theLexicon , Cambridge University Press , Cambridge . Ryle , G. 1966 . TheConcept oftheMind , Hutchinson , London . Sag , I. A. andC. Pollard . 1991 . "AnIntegrated Theory ofComplement Control ,') Language 67:63-113 . Sag , I. A. andA. Szabolcsi , eds . 1992 . Lexical Matters , CSLILecture Notes 24, Center fortheStudy ofLanguage andInformation , Stanford University , Stan ford,CA. Saint-Dizier , P. 1995 . "Constraint Propagation Techniques forLexical Semantic Descriptions ," in P. Saint -DizierandE. Viegas , eds ., Computational Lexical Semantics , Cambridge University Press . Saint -Dizier , P. andE. Viegas , eds . 1995 . Computational Lexical Semantics , Cam bridge University Press , Cambridge . Saksena , A. 1980 . "TheAffected Agent ," Language 56:812 -826 . Saksena , A. 1982 . "Contact in Causation ," Language 58:820 -831 . Sanfilippo , A. 1990 . Grammatical Relations , Thematic Roles , andVerb Semantics . Ph.D. Dissertation , Centre forCognitive Science , University ofEdinburgh . Sanfilippo , A. 1991a . "Thematic andAspectual Information in VerbSemantics ,)' unpublished ms., University ofCambridge , Cambridge , UK. Sanfilippo , A. 1991b . "Argument Selection andSelection Change : AnIntegrated Approach ," CLS27. Sanfilippo , A. 1993 . "LKBEncoding of Lexical Knowledge ," in T. Briscoe , V. dePaiva , andA. Copestake (eds .), Inheritance , Defaults , andtheLexicon , Cambridge University Press , Cambridge . Sanfilippo , A. K. Benkerimi , andD. Dwehus . 1994 . "VirtualPolysemy ," in Proceedings of COLING -1994 , Nantes , France . Scha , R. 1983 . "Logical Foundations forQuestion Answering ," MS12.331Philips Research Laboratories , Eindhoven , TheNetherlands . Schank , R. C. 1972 . "Conceptual Dependency : A Theory of Natural Language Understanding ," Cognitive Psychology 3:552 -631 . Schank , R.C. 1973 . "Identification ofConceptualizations Underlying Natural Language " in R.C. Schank andK.M. Colby(eds .), Computer Models of Thought andLanguage , W.R. Freeman , SanFrancisco , CA, 187 -247 . Schank , R. 1975 . Conceptual Information Processing , Amsterdam , North -Holland . Searle , J. 1979 . Expression andMeaning , Cambridge , Cambridge University Press .

286

Bibliography

Seuren,P. 1985.DiscourseSemantics, Blackwell Publishers,Oxford.

Shibatani, M.,ed.1976.Japanese Generative Grammar, Academic Press,New York.

Shieber, S.1992.Gonstraint-based grammar formalisms: parsing andtypeinferencefor naturalandcomputer languages, MITPress,Cambridge, MA. Small, S. I., G. W. Cottrell, and M. K. Tanenhaus, eds. 1988. Lexical Ambiguity Resolution,MorganKaufmann,San Mateo, CA. SparckJones,K. 1986.Synonymy andSemanticClassification.. Ph.D.Disserta-

tionwithnewForeword. Edinburgh Information Technology Series(EDITS). Smith, B.andK.Mulligan. 1983.Framework forFormal Ontology,Topoi 2:7385. Edinburgh: Edinburgh University Press.

Smolka,G. 1988.A FeatureLogicwithSubsorts,Wissenschaftliches Zentrum der IBM Deutschland,LILOG-Report33.

Soames, 5. 1989.Presupposition, inD.Gabbay adF. Guenthner (eds.),Handbook of Philosophical LogicVol. IV, 553-616. Sowa,J. 1984.ConceptualStructures,Addison-Wesley, Boston.

Sowa, J. 1993.Lexical Structures andConceptual Structures, inJ. Pustejovsky (ed.), Semanticsand the Lexicon,Kluwer,Dordrecht. Steinberg, D. andL. Jakobovits, eds.1971.Semantics, Cambridge University Press, Cambridge. Strachey, C. 1967.Fundamental Concepts in Programming Languages, Lecture Notesfor International Summer School in Computer Programming, Copenhagen. Talmy, L.1975.Semantics andSyntax ofMotion,inJ. P.Kimball (ed),Syntax and Semantics 4, AcademicPress, New York, 181-238. Talmy,L. 1976.SemanticCausative Types, in M.Shibatani(ed),43-116. Talmy, L.1985.Lexicalization Patterns:Semantic Structure inLexical Forms,in T. Shopen(ed), LanguageTypology andSyntacticDescription 3, Grammatical

Categories andtheLexicon, Cambridge University Press,Cambridge, UK,57149. Tedeschi, P. andA. Zaenen,eds.1981.TenseandAspect,SyntaxandSemantics 14, Academic Press, New York, NY.

Tenny, C.,ed.1988. Studies in Generative Approaches toAspect, Lexicon Project Working Papers24,CenterforCognitive Science, MIT,Cambridge, MA.

Tenny, C. 1992.The Aspectual Interface Hypothesis,in I. A. SagandA. Szabolcsi (eds.), 1-27.

Thomason, R., ed. 1974.FormalPhilosophy: TheCollected Papersof Richard Montague,New Haven,Yale UniversityPress.

Touretzky, D.1986.TheMathematics ofInheritance Systems, Morgan Kaufmann, Los Altos, CA. Travis,L. and E. Williams 1983.Externalization of Arguments in MalayoPolynesianLanguages The LinguisticReview2.

Trier, J. 1931. Der deutsche Wortschatzim Sinnbezirkdes Verstandes:Die

Geschichte einessprachlichen Feldes.BandI. Heidelberg. VanValin,ft. D. Jr. 1987.The Unaccusative Hypothesisvs. LexicalSemantics:

Syntacticvs.Semantic Approaches to VerbClassification, NELS17,641-661.

Van Valin,ft.D.Jr. 1990.Semantic Parameters ofSplitIntransitivity, Language 66:221-260. VanValin,RD. Jr. 1993.A Synopsis ofRoleandReference Grammar,in ft. van Valin, Jr. (ed) Advances in Role and Reference Grammar, JohnBenjamins Publishing, Philadelphia. VanValin,RD. Jr. andD.Wilkins. 1993.Predicting SyntaxfromSemantics, in ft. van Valin,Jr. (ed.)Advances in Roleand Reference Grammar,John

287

Bibliography

Benjamins Publishing , Philadelphia . Van Voorst , J . 1986 . Event Structure . Ph .D . Dissertation Ottawa

, Ontario

, University

of Ottawa ,

.

Van Voorst , J . 1992 . "The Aspectual Semantics of Psychological Verbs ," Linguis tics and Philosophy 15:65- 92 . Vendler , Z . 1967 . Linguistics in Philosophy , Cornell University Press , Ithaca . Vendler , Z . 1984 . " Agency and Causation ," Midwest Studies in Philosophy 9:371 ~ .Q .4 -

Verkuyl , H . 1972 . On the Compositional Nature of the Aspects , Reidel , Dordrecht . Verkuyl , H . . 1973 . " Temporal Prepositions as Quantifiers ," in F . Kiefer and N .

Ruwet (eds.) , 582-615. Verkuyl , H . 1989 . "Aspectual Classes and Aspectual Composition ," Linguistics and Philosophy 12 :39-94 . Verkuyl , H . 1993 . A Theory of Aspectuality , Cambridge University Press , Camr bridge . Verkuyl , H . and J . Zwarts . 1988 . "Time and Space in Conceptual and Logical Semantics : the notion of Path ," Linguistics 30 :483 -511 . Verma ) M . and K . P . Mohanan 1991 . Experiencer Subjects in South Asian Lan guages , CSLI ) University of Chicago Press . Viegas , E . 1995 . "The Semantics - Pragmatics of Lexicalization ," in P . Saint - Dizier and E . Viegas (eds .) , Computational Lexical Semantics , Cambridge University Press

.

Vikner , C . and P . Hansen . 1994 . " Knowledge Bases and Generative Lexicons ," paper presented at Workshop on Lexical Semantics , November 11, University of Copenhagen . Vlach , R . 1981 . "The Semantics of the Progressive ," Syntax and Semantics : Tense and Aspect 14, 271 - 92 . Vossen P . 1991 . "The End of the Chain : Where Does Decomposition of Lexical Knowledge Lead Us Eventually ?," unpublished ms ., University of Amsterdam , Amsterdam Walker

.

, D . E . , and R . A . Amsler

. 1986 . " The

Use of Machine

- Readable

Dictionaries

in Sublanguage Analysis ," in R . Grishman and R . Kittredge , eds ., Analyzing Language in Restricted Domains , Lawrence Erlbaum : Hillsdale , N J . Walker , D ., A . Zampolli , and N . Calzolari , eds . 1994 . Automating the Lexicon , Oxford University Press . Weinreich , U . 1959 . "Travels through Semantic Space ," Word 14:346 - 366 . Weinreich , U . 1963 . " On the Semantic Structure of Language ," in J . Greenberg

(ed.) , Universal of Language, MIT Press, Cambridge , MA . Weinreich , U . 1964 . " Webster 's Third : A Critique of its Semantics ," International Journal of American Linguistics 30 :405 -409 . Weinreich , U . 1972 . Explorations in Semantic Theory , Mouton , The Hague . Wheeler , R . S. 1990 . " Sense and Subsense : The Meanings of the English Verb Understand ," InS., Utah State University . Whorf , B . L . 1956 . Language , Thought , and Reality , MIT Press , Cambridge , MA . Wierzbicka

, A . 1972 . Semantic

Wierzbicka

, A . 1980 . Lingua

demic

Press , New

York

Primitives

, Athenaum

Mentalis : The Semantics

, Frankfurt

of Natural

.

Language , Aca -

, NY .

Wierzbicka , A . 1982 . "Why Can You Have a Drink When You Can 't Have an Eat ?," Language 58 :753- 799 . Wierzbicka , A . 1988 . The Semantics of Grammar , Amsterdam , John Benjamins . Wilensky , R . 1986 . " Some Problems and Proposals for Knowledge Representa tion ," Cognitive Science Report 40 , University of California , Berkeley . Wilensky , R . 1991 . " Extending the Lexicon by Exploiting Subregularities ," Report

UCB / CSD 91/ 618, Computer Science Department , University of California ,

288

Bibliography

Berkeley . Wilks , . 1975a . " A Preferential Pattern Seeking Semantics for Natural Inference ," Artificial Intelligence , 6 :53- 74 . Wilks , . 1975b . " An Intelligent Analyser munications of the ACM , 18 :264- 274 . Wilks , Y . 1978 . "Making

, Y . , D . Fass , C . M . Guo , J . McDonald

Tractable

Machine

for English ," Com -

Preferences More Active ," Artificial

97 .

Wilks

and Understander

Dictionary

, T . Plate

Language

, and

Intelligence B . Slator

a8 a Resource for Computational

10:75-

. 1988 . " A

Semantics ," in

B . Boguraev and T . Briscoe (eds.), Computational Lexicography for Natural Language Wilks

Processing , Harlow , Essex , Longman .

, Y . , D . Fas , C . M . Guo , J . McDonald

, T . Plate

, and B . Slator

. 1993 . " Provid

-

ing Machine Tractable Dictionary Tools," in J. Pustejovsky (ed.) , Semantics and Williams

the Lexicon , Kluwer Academic Publishers , Dordrecht . , E . 1980 . " Predication ," Linguistic Inquiry 11 :203 - 238 .

Williams

, E . 1981 . " Argument

1 : 81 - 114 .

Structure

Williams , E . 1982 . " Another Argument guistic Inquiry 13: 160- 163 .

and Morphology that

," The Linguistic

Passive is Transformational

Review ," Lin -

Williams , E . 1983 . " Against Small Clauses ," Linguistic Inquiry 14:287-308 . Williams , E . 1983 . " Semantic vs . Syntactic Categories ," Linguistics and Philoso phy 6 :423 -446 . Williams , E . 1987 . " Implicit Arguments , the Binding Theory , and Control ," Nat ural Language and Linguistic Theory 5 :151- 180 . Williams , E . 1994 . Thematic Structure in Syntax , MIT Press , Cambridge . Woods

, W . 1975 . " What

' s in a link : foundations

of semantic

networks

," in D . G .

Bobrow and A . Collins (eds.) , Representation and Understanding : Studies in Cognitive

Science , Academic

von Wright , G . 1963 . Norm Paul , London

Press , New York .

and Action : A Logical Inquiry , Routledge

and Kegan

.

Wunderlich , D . 1987 . " An Investigation of Lexical German be-Verbs ," Linguistics 25 , 283 -331 .

Composition

: The

Case of

Wunderlich , D . 1991 . "How Do Prepositional Phrases Fit into Compositional Syntax and Semantics ," Linguistics 29 :591 -621 . Wunderlich , D . 1994 . " Models of Lexical Decomposition ," in E . Weigand and F .

Hundsnurscher (eds.) , Proceedings of the International

Conference on 'Lexi -

cology and Lexical Semantics ', Munster , Tubingen , Niemeyer . Zaenen , A . 1993 . " Unaccusativity in Dutch : The Interface of Syntax

and Lexi -

cal Semantics," in J. Pustejovsky (ed.) , Semantics and the Lexicon, Kluwer , Dordrecht

Zubizarreta

.

, M . L . 1987 . Levels of Representation

Faris , Dordrecht

Zucchi , A . 1993 . The Language Publishers

in the Lexicon and in the Syntax ,

.

, Dordrecht

.

of Propositions

and Events , Kluwer

Academic

Author

Index

A Abraham, W ., 188 A busch, D., 259 Ackerman, F., 259 Allen, J., 69, 183, 246 Alsina, A., 249, 259 Anderson, M., 263 Anick

,

P

. ,

Apresjan

55

,

Aristotle

,

244

,

Asher

,

,

Atkins

.

. ,

,

B

.

,

247

,

234

,

. ,

91

,

20

,

T

,

242

30

233

62

,

17

246

N

231

,

J

,

76

,

248

,

241

121

5

,

,

,

199

,

77

,

247

244

47

167

,

,

,

,

48

85

176

,

247

,

,

,

122

97

,

243

,

177

,

184

,

215

,

216

,

D

248

,

241

B Bach

Baker

,

E

,

. ,

C

15

. L

,

. ,

16

,

184

,

250

,

264

50

. Ballim, A., 255 Beierle, C., 114 Belletti, A., 213 Benthem, J. van, 57, 58, 69, 92, 184, 245, 246, 255 Bergler, S., 241 Bes, G., 247 Bierwisch, M., 18, 152, 242, 244, 266 Boguraev, B., 5, 44, 51, 90, 91, 116, 141, 248, 255, 264 Bookman, L., 243 Bouillon, P., 116, 199, 200, 201, 207, 248, 258, 260 Brachman, R. J., 24, 58 Bresnan, J., 10, 63, 249, 254 Briscoe, T ., 5, 78, 90, 225, 233, 247, 255, 264 Brugman, C., 268 B uitelaar) P., 248 Burge, T ., 75 Busa) F., 68, 74, 187, 188, 189, 195, 196, 197, 198, 266 Baker

,

M

. ,

245

,

Chomsky, N., 8, 36, 40, 55, 63, 103, 108, 117, 118, 132, 135, 137, 181, 184, 194, 207, 213, 242, 245, 249, 260 Cohen, P.R., 24, 241 Collins, A., 58 Comrie, B., 159, 246, 259 Copestake , A., 78, 81, 90, 94, 114, 141, 189, 224, 233, 247, 248, 264 Cottrell, G. W , 36 Croft, W ., 209 Crouch, R., 246 Cruse) D. A., 23, 24

259

c Cardelli, L., 58, 111, 245 Carlson, G., 15, 225 Carnap, R., 58 Carpenter, B., 81, 94, 145, 244 Carroll, J., 255 Cattell, R., 258 Charniak, E., 7 Chierchia, G., 106, 189, 195, 198, 249

.. Dahl, 0 ., 246 Davidson, D., 67, 165 Descles , J.-P., 246, 258 Di Scuillo, A.-M., 253, 260 Diesing, M., 226 Dixon, R. M. W., 19, 22, 23, 135, 241 Dolling, J., 264 Dor, D., 50 Dowty, D. R., 13, 14, 16, 25, 54, 56, 71, 107, 108, 109, 110, 174, 201, 207, 228, 242, 245, 249, 251, 259, 260

E Eberle,K., 183 Elliott, D., 52, 244 En

The Generative Lexicon

The Generative Lexicon

Lexicon

Generative Art

Lexicon Athanasianum

The Leadership Lexicon

The Demon's Lexicon

The acquisition of the lexicon

The Ayn Rand Lexicon

The Multilingual Lexicon

The Philosophical Lexicon

Ayn Rand Lexicon, The

The Leadership Lexicon

Lexicon Plotinianum

Lexicon botanicum

The Acquisition of the Lexicon

The New Analytical Greek Lexicon

The Mental Lexicon: Core Perspectives

The Mental Lexicon: Core Perspectives

Formulaic Language and the Lexicon

Einfach generieren - Generative Programmierung

Arabic-english lexicon

Arabic-english lexicon

A Lexicon to Herodotus

Das Kakteen Lexicon

Formulaic Language and the Lexicon

The Demon's Covenant (Demon's Lexicon)

Arabic-english lexicon

Arabic-english lexicon

Loudspeaker handbook and lexicon

Weaving a Lexicon

The Generative Lexicon