Essays in Semantics and Pragmatics

ESSAYS IN SEMANTICS AND PRAGMATICS
Pragmatics & Beyond

New Series
Editors:
Jacob L. Mey
(Odense University)
Herman Parret
(Belgian National Science Foundation, Universities of Louvain and Antwerp)
Jef Verschueren
(Belgian National Science Foundation, University of Antwerp)
Editorial Address:
Linguistics (GER)
University of Antwerp (UIA)
Universiteitsplein l
B-2610 Wilrijk
Belgium
Editorial Board:
Norbert Dittmar {Free University of Berlin)
Bruce Fraser (Boston University)
John Heritage (University of California at Los Angeles)
David Holdcroft (University of Leeds)
Catherine Kerbrat-Orecchioni {University of Lyon 2)
Beatriz Lavandera (University of Buenos Aires)
Marina Sbisà (University of Trieste)
Emanuel A. Schegloff (University of California at Los Angeles)
Paul O. Takahara (Kobe City University of Foreign Studies)
Sandra Thompson (University of California at Santa Barbara)
Daniel Vanderveken (University of Quebec at Trois-Rivières)
Teun A. van Dijk {University of Amsterdam)
32
Masayoshi Shibatani and Sandra Thompson (eds)
Essays in Semantics and Pragmatics

üoo/\ï i5
IN SEMANTICS
AND PRAGMATICS
IN HONOR OF CHARLES J. FILLMORE
Edited by
MASAYOSHI SHIBATANI
Kobe University
SANDRA THOMPSON
University of California at Santa Barbara
JOHN BENJAMINS PUBLISHING COMPANY

AMSTERDAM/PHILADELPHIA
The paper used in this publication meets the minimum requirements of
American National Standard for Information Sciences — Permanence of
Paper for Printed Library Materials, ANSI Z39.48-1984.
Library of Congress Cataloging-in-Publication Data

Essays in semantics and pragmatics : in honor of Charles J. Fillmore / edited by Masayoshi
Shibatani, Sandra Thompson.
p. cm. -- (Pragmatics & beyond, ISSN 0922-842X ; new ser. 32)
Papers in honor of C. Fillmore's sixty-fifth birthday in 1994.
Includes bibliographical references.
1. Semantics. 2. Pragmatics. 3. Grammar, Comparative and general. I. Fillmore, Charles J.
II. Shibatani, Masayoshi. III. Thompson, Sandra A. IV. Series.
P325.E79 1995
401'.43--dc20 95-36112
ISBN 90 272 5044 8 (Eur.) / 1-55619-325-4 (US) (alk. paper) CIP
© Copyright 1995 - John Benjamins B.V.
No part of this book may be reproduced in any form, by print, photoprint, microfilm, or any
other means, without written permission from the publisher.
John Benjamins Publishing Co. • P.O.Box 75577 • 1070 AN Amsterdam • The Netherlands
John Benjamins North America ● P.O.Box 27519 • Philadelphia PA 19118-0519 • USA
in honor of
CHARLES J. FILLMORE
(Photograph by Paul J. Atkins)
Preface
Masayoshi Shibatani and Sandra A. Thompson
When someone is as energetic and productive as Chuck Fillmore, it is easy

to miss opportunities to celebrate those occasions that mark one's life
cycles. Asian wisdom has it that one begins a new life cycle at the age of
sixty. Another tradition suggests that sixty-five is an age to celebrate;
accordingly, we decided to celebrate Chuck's sixty-fifth birthday in 1994.
1994 also marks the 20th anniversary of the Berkeley Linguistics Society,
together with which we held a joint celebration at the 20th BLS meeting
February 18-21, 1994.
Friends of Chuck's from all over the world have joined us in this project
of expressing our appreciation of Chuck for his contributions to the field of
linguistics and for his guidance, understanding, enthusiasm, friendship,
warmth, and concern for all of us as students, colleagues, and individuals.
Responses to our invitation to contribute to a festschrift for Chuck were so
enthusiastic that we had enough papers for two volumes. Like Chuck's own
scholarship, these tributes span a wide range of areas in linguistics. One
volume, containing mainly those papers dealing with grammatical construc
tions, reflecting Chuck's more recent work on Construction Grammar, are
appearing as Grammatical Constructions: Their Form and Meaning from
Oxford University Press.
The present volume reflects the influence of Chuck's ground-breaking
work in the fields of semantics and pragmatics. The papers in this volume
pay tribute to Chuck's pioneering research into the deepest realms of the
nature of 'meaning'. An overview of this work is available elsewhere (see
especially the tribute volume of the Proceedings of the 20th Annual
Meeting of the Berkeley Linguistics Society, 1994). Suffice it to say here
that Chuck's output has reached into a remarkably wide variety of areas of
semantics and pragmatics, and that over the last 30 years he has played a
major role in showing us how to think about linguistic meaning.
The papers in this volume exuberantly celebrate this variety in the range
of approaches they represent. Taking semantics and pragmatics to define
viii PREFACE
a kind of continuum, the papers range from lexicography (Atkins) to

lexicology (Kiefer, Petruck), through the meanings of constructions
(Ackerman, Fujii, Gawron, Ohori), case (Lambrecht), metaphor (Lakoff),
and deixis (Östman), to the pragmatics of particles (Okamoto), perform-
ativity (Verschueren), kinship (Li), and face (Ervin-Tripp et al).
Putting these papers together has been a joy for us; it has been a
genuine pleasure to work with so many people who have been so enthusias
tic about sharing in an enterprise designed to honor the influence that
Chuck Fillmore has had on the field of linguistics. We thank Jef
Verschueren for his involvement in this project from its inception and for
overseeing the production of this volume. We, authors and editors alike,
offer these papers in the recognition of the excitement of doing linguistics
that Chuck himself has shared with all of us and with the field at large.
Contents
Preface
Masayoshi Shibatani and Sandra Thompson vii
A predicate construction: Inversion in Polish

Farrell Ackerman 1
The role of the example in a frame semantics dictionary

B.T.S. Atkins 25
Shifting face from Asia to Europe

Susan Ervin-Tripp, Kei Nakamura, and Jiansheng Guo 43
Mental-space builders:
Observations from Japanese and English conditionals
Seiko Yamaguchi Fujii 73
Distance in construction grammar

Jean Mark Gawron 91
Lexical information and the temporal interpretation of discourse

Ferenc Kiefer 111
Reflections on metaphor and grammar

George Lakoff 133
The pragmatics of case: On the relationship between

semantic, grammatical, and pragmatic roles in English and French
Knud Lambrecht 145
Ancestor-descendant and cultural-linguistic relativity

Charles N. Li 191
X CONTENTS
Remarks on suspended clauses:

A contribution to Japanese phraseology
Toshio Ohori 201
Pragmaticization of meaning in some sentence-final particles

in Japanese
Shigeko Okamoto 219
Recasting the deictic foundation, using physics and Finnish

Jan-Ola Östrnan 247
Frame semantics and the lexicon:

Nouns and verbs in the body frame
Miriam R.L. Petruck 279
The conceptual basis of performativity

Jef Verschueren 299
Index 323
A predicate construction:
Inversion in Polish1
Farrell Ackerman
University of California, San Diego
1. Introduction
Polish, like many languages, contains constructions in which a nominal with

certain demonstrable SUBJ(ect) properties is marked with the DATIVE case.
I will refer to these constructions as inversion constructions and to the
DATIVE argument as the inversion nominal following the terminology within
Relational Grammar. Polish, on the Relational Grammar account of
Dziwirek 1991, possesses a productive inversion pattern. That is, numerous
simple transitive and intransitive predicates have inversion analogues. This
is exemplified below:
(la) Dobrze mi się dzisiaj spalo
well 1sg-DAT rflx today sleep-3sg/neuter
I SLEPT WELL TODAY
This paper is largely based on the Relational Grammar analysis of Polish inversion
presented in K. Dziwirek (1991). Her presentation establishes a baseline for any adequate
alternative treatment, while the deftness of argumentation dependent on multi-stratal
assumptions constitutes a challenge to mono-stratal multi-level approaches and, hence, the
motivation for the present paper. All data is from Dziwirek (1991) unless otherwise
indicated. I would like to thank Katarzyna Dziwirek, Bozena Pruska, Mark Gawron,
Claudia Brugman and Phil Lesourd for useful discussion concerning various aspects of this
paper.
2 FARRELL ÀCKERMAN
(lb) (Dobrze) dzisiaj spaïam

(well) today slept-1sg
I SLEPT (WELL) TODAY
(2a) Tç ksiązke czytalo się Jankowi z przyjemnoscią

this book-ACC read-3sg/neuter rflx John-DAT with pleasure
JOHN READ THIS BOOK WITH PLEASURE
(2b) Janek czytaï tç ksiązke (z przyjemnoscią)

John-NOM read this book-ACC (with pleasure)
JOHN READ THIS BOOK (WITH PLEASURE)
The (a) sentences contain DAT marked nominals indicated by boldface,
while the (b) sentences contain either a missing NOM(INATIVE) pronominal
SUBJ, as in (lb), or an overt NOM SUBJ, as in (2b). As can be seen, (lb)
represents a simple intransitive clause, with (la) as its inversion analogue,
while (2b) represents a simple transitive clause with (2a) as its inversion
variant.
In this paper I will suggest that an analysis formulated in terms of the
principled interaction between the semantic properties of arguments, case
marking and grammatical functions provides a promising account of these
Polish facts, while relating them straightforwardly to similar phenomena
elsewhere - particularly to the languages of Southern India as analyzed in
Mohanan and Mohanan (1990). A guiding intuition of this analysis is that
the semantic properties associated with arguments play a central role in
both surface case marking and grammatical function assignment. The
interpretation of these semantic properties as developed below will follow
the seminal insights of Fillmore (1977) as well as subsequent similar
assumptions found in Gawron (1983) and especially Dowty (1991). In
particular, I suggest that the lexical semantics of the inversion predicate
determines the case marking associated with the inversion nominal and that
the grammatical function status of nominals is independent of case-marking.
On the present account, the default marking for SUBJ is NOMINATIVE, while
one of the semantically specified markings for SUBJ is DATIVE.
The Polish constructions will be analyzed in terms of predicate
formation processes operative in the Polish lexicon. In proposing this
treatment I will be focusing on a dramatic aspect of the Polish data which
arguably accords with an early view of the lexicon expressed in Jackendoff
(1975). According to this view, lexical representations need not be
INVERSION IN POLISH 3
associated with synthetic morphological expressions, i.e., wordforms

exhibiting lexical integrity. On this view, the Polish lexicon is interpretable
as containing predicates which are analytic expressions. While formally
resembling phrasal idioms, these predicates differ from idioms in that they
pattern along with morphologically derived verb forms: they participate in
a system of regularities characteristic of entities related by lexical
operations.
The organization of this paper is as follows. Section 1 presents the basic
properties of Polish inversion constructions. In section 2, I briefly
characterize the conceptual differences between two perspectives on the
role of (lexical) semantics to syntax, referred to expediently as lexical
categorial vs. constructionist. I do this in order to establish the backdrop for
the hypothesis that the inversion predicate in Polish is an analytic lexical
entity: The analysis advanced here is informed by the assumptions identified
as those which guide research within constructionist approaches to
grammar. In doing this I adopt what superficially resembles the so-called
"quirky SUBJ case marking hypothesis", but which is, better interpreted as
a "semantically motivated SUBJ case marking hypothesis": DATIVE case
marking in inversion constructions is claimed to correlate with the semantics
of the inversion PREDICATE. In section 3., I briefly illustrate how this
interpretation of inversion predicates explains several different syntactic
properties associated with these constructions-.
1. The Polish inversion construction2
Dziwirek observes that there are several characteristic properties associated

with Polish inversion constructions. In particular, they contain (1) a verbal
form in the 3rd singular (neuter) conjugation, (2) a reflexive pronoun, (3)
a (manner) adverbial or a construal of manner, and (4) a construal of
humanness and minimal agentivity for the DATIVE nominal. These
properties are exhibited by the productive inversion constructions repeated
below:
For purposes of brevity, I will focus primarily on so-called productive inversion

constructions in Polish.
4 FARRELL ACKERMAN
(3a) (Dobrze) dzisiaj spalam

(well) today slept-1sg
I SLEPT (WELL) TODAY
(3b) Dobrze mi się dzisiaj spalo

well 1sg-DAT rflx today $leep-3sg/neuter
I SLEPT WELL TODAY
(4a) Janek czytal tç ksiązkę (z przyjemnoscią)

John-N OM read this book-ACC (with pleasure)
JOHN READ THIS BOOK (WITH PLEASURE)
(4b) Te ksiązkę czytalo się Jankowi z przyjemnoscią

this book-ACC read-3sgjneuter rflx John-DAT with pleasure
JOHN READ THIS BOOK WITH PLEASURE
Whereas the verb agrees with the overt NOMINATIVE nominal in (4a), it
agrees with the "pro-dropped" NOMINATIVE 1st person singular pronominal
in (3a). In contrast, the verb exhibits 3rd singular neuter marking in the (b)
examples. Whereas there is no reflexive marker in the (a) examples, there
is a reflexive marker in the (b) examples. Whereas a manner adverbial
appears optionally in the (a) examples, it appears obligatorily in the (b)
examples. There are several other important properties associated with the
construction, but the preceding are sufficient for present purposes (see
Dziwirek (1991) and Ackerman (1994) for further details). I will assume
that the Polish inversion predicate consists of a categorial verb, a reflexive
pronoun and a manner adverbial (overt or covert).
Wierzbicka (1988) suggests that the semantic constraint on the role of
the participant encoded as a DATIVE marked nominal in (productive)
inversion constructions can be seen as the "agent viewed as experiencer".
She contextualizes this use of the DATIVE within a general characterization
of the DATIVE in all constructions in which it occurs:
"Loosely speaking, the use of the DATIVE implies a situation which is not controlled
by person Z (i.e. the individual denoted by the nominal marked in the DATIVE case
[FÀ]) but which is likely (though not certain) to have an effect on Z. Thus, whether
a person is offered something, shown something, told something, or whether things
happen in or near his body, he is likely to be affected by these events, and affected
in ways in which things - as opposed to people - cannot be affected (i.e. mentally).
Things can only be affected when something happens to THEM, but people can also
be affected when something happens to something else, because they can react
mentally to it." (1989: 427)

Wierzbicka raises an important issue with resonance for all theoretical
frameworks: is there a semantic generalization which subsumes all uses of
the DATIVE case and which, therefore, would be explanatory for the
analysis of inversion constructions? From the present perspective this can
he stated specifically as follows: are there some semantic properties
associated with the inversion predicate (i.e., the categorial verb, reflexive
pronoun and adverbial) which would explain the appearance of the DATIVE
case on inversion nominals?3
In order to address this question I will accept the fairly common
assumption that the semantics of the PREDICATE is determinative of the
semantics for the (obligatory) arguments of a clause, i.e., properties of the
PREDICATE determine its event-type and semantic entailments for its
arguments. However, correlative with this view is another concerning the
categorial expression of the PREDICATE: there is a standard identification
of the PREDICATE as a zero-level lexical head of some phrasal domain.
These two assumptions present an awkward reality for the analysis of the
Polish constructions: the entity identified as the PREDICATE here is not
expressed by a single (simplex or complex) lexical category, but rather is
expressed disjointly by several independent elements in constituent
structure. I turn now to two types of proposals regarding the mapping of
arguments to grammatical functions in order to bring this problem into
greater focus, and to propose a solution to it.
2. Mapping theories and constructions
The attempt to establish principled relations between arguments with

certain semantic properties and the grammatical functions they bear in a
clause has as estimable history in linguistics (cf. Levin and Rappaport
Hovav 1995). On the other hand, the claim that an analytic expression can
serve as the head of a clause and thereby determine the alignment of
arguments with grammatical functions requires a revision of certain
An exposition of the specific properties associated with Polish predicates is

presented and defended in Ackerman 1994. In the present paper I limit myself to
speculations concerning the general nature of an appropriate lexical analysis of these
constructions.
6 FARRELL ACKERMAN
standard assumptions concerning mapping within generative grammars.

According to this view, argument structure is projected from lexical heads
and the arguments of lexical heads (ordinarily characterized in terms of
atomic thematic roles4) are mapped into grammatical functions.
Though the mapping algorithm differs from framework to framework
one assumption tends to remain constant: argument structures are
associated with lexical heads expressed by zero level morphonologically
integrated and syntactically atomic entities. That is, the projector of
semantic information is identified with the categorial head of the clause.
In an effort to redress some of the shortcomings associated with delimiting
classes of verbs and defining a finite list of discrete semantic roles
associated with them, Gawron (1983: 289) proposes an alternative research
strategy:
"Rather than seeking an exhaustive taxonomy of verb argument relations, we are
seeking subject-inducing semantics."
Generalizing somewhat, I will claim that we are seeking grammatical
function inducing semantics and that the projector of such semantics is the
PREDICATE, rather than some lexical category such as V(erb).5 As a more
refined version of this thesis, we are also seeking surface case marking
inducing semantics, which are (often) independent of grammatical function
assignment.
Roughly speaking, two contending interpretations concerning the
relevance of lexical semantics for the explanation of certain syntactic
phenomena can be distinguished: they are distinguishable in terms of how
the projector of clausal semantics is viewed and how the semantics of
4
That there is no necessary relation between the hypothesized centrality of lexical
heads and commitments to the nature of semantic arguments as atomic is illustrated by
the fact that there have been numerous efforts over the years to decompose semantic roles
into various properties, for example Foley and Van Valin (1984) and Jackendoff (1991)
to mention just two. As observed in Van Valin (1991) and Dowty (1991), these efforts all
share the view that such decompositions yield discrete category membership. For present
purposes, I am more concerned with the presumed locus of semantics for arguments, than
with whether the arguments can be characterized as discrete or as proto-types with
overlapping features.
5
In fact, Gawron himself extends the notion head to cover the N and A of predicative
constructions, in effect, differentiating between a functional category such as PREDICATE
and the lexical exponence of this category.
associated arguments is construed. One approach, which might be referred

to as lexical categorial, derives from the work of Gruber (1965) as
interpreted through the speculations of Jackendoff (1991 and earlier) and
Fillmore (1968). It is associated with research in the MIT lexicon project
and variants of the Principles and Parameters framework, but some version
of it has found application in several different generative and functionalist
frameworks (Bresnan & Kanerva's mapping theory within Lexical
Functional Grammar (1989) and Foley and Van Valin's Role and
Reference Grammar (1984, 1991), to name a few.) The other approach,
which I will refer to as constructionist', derives from the work of Fillmore
(1977) concerning a saliency hierarchy and finds its most recent expression
in the Construction Grammar of Fillmore and Kay (ms.), Zwicky (ms.)
Goldberg (1993) and the proto-role proposal of Dowty (1991).
There are so many points of apparent convergence between the lexical
categorial and constructionist approaches that it is worthwhile keeping in
mind the following from G. Spencer-Brown's The laws of form:
"We must note that in these experiments the sign ' = ' may stand for the words, 'is
confused with'."
As suggested above, one difference between these approaches concerns
the locus of clausal semantics or what is to be best construed as the head
of the clause: in the lexical categorial approach the locus of semantics is a
lexical category (say, V), while in the constructionist approach the locus of
semantics is, arguably, the PREDICATE, i.e. an entity not necessarily
coextensive with a single lexical category and possibly expressed by several
independent entities.
A second difference concerns the interpretation of the semantic relation
between the projector of clausal semantics and its associated arguments: in
the lexical-categorial approach arguments are standardly associated with
discrete thematic or semantic roles, while in the constructionist approach
they are often associated with bundles of semantic properties determined
in some measure by the projector of clausal semantics.
The difference between commitments to the projector of clausal
semantics becomes evident with respect to Polish inversion constructions.
I would like to suggest that the difference between the ensemble of
elements in the inversion construction (e.g. a DATIVE marked nominals, a
3rd person reflexive pronoun, 3rd person neuter agreement etc.) and the
presence of similar elements in other constructions is that the former
functions as a single semantic unit. I will refer to this ensemble of
8 FARRELL ACKERMAN
properties as an INVERSION PREDICATE: this entity cannot be identified
with a single lexical head. It represents a deviation from the usual
expectations of X-bar theory whereby the categorial and semantic head of
a domain converge in a single (zero-level) expression. In some sense, the
INVERSION PREDICATE resembles what in Lexical Functional Grammar is
referred to as the f-(unctional) head of a clause, by which I mean, it serves
as the projector of the function-inducing and case-inducing semantics for its
arguments.6 I will further assume that the arguments of the INVERSION
PREDICATE are associated with entailment sets derived from this predicate,
rather than with discrete thematic role labels. In the next section I will
outline a proposal for the analysis of Polish inversion, following the work
of Dowty (1991), Fillmore (1977), and Mohanan & Mohanan (1990), which
specifies what it means for an entity composed of separate syntactic pieces
to be analyzed as a single lexical unit projecting semantic entailments to its
arguments.
2.1. An inversion predicate
The basic idea underlying the present proposal for the analysis of Polish
productive inversion is this: the Polish inversion construction contains an
INVERSION PREDICATE (expressed by elements realized disjointly in
constituent structure) whose semantics entails certain properties for its SUBJ
argument, and certain of these properties determine DATIVE case marking.
As mentioned previously, I will assume that the separate pieces constitute
a unit which I will refer to as a PREDICATE. PREDICATES as, understood
here, are information units serving as the semantic head of a clause and
whose surface expression can be larger than the standard zero-level
category, but smaller than a VP (see T. Mohanan 1991 and Ackerman &
Webelhuth for discussion of the relevant notion of PREDICATE). The
PREDICATE, by hypothesis, is an entity associated with an argument
structure: all of its arguments bear a grammatical function value.
The main difference between the inversion construction and its
analogue containing a NOMINATIVE SUBJ is, on this account, located in the
Mohanan and Mohanan (1990) present evidence that Malayalam contains complex
predicates which induce the appropriate semantics for DATIVE case marking of the
SUBJ. Since these predicates consist of syntactically separate entities in constituent
structure, they resemble the Polish constructions. On their analysis these predicates are
composed lexically by combining the participating pieces.
entailments of the PREDICATES. Different predicates may entail different

semantic properties for their arguments and this will interact with the
conventions for representing surface case marking. The present proposal
claims that DATIVE marking is semantically motivated and that the DATIVE
marked nominal is the SUBJ in these Polish constructions. On this view
there is a many-to-one relation between case markers and the SUBJ function
in Polish. In addition, for transitive variants of productive inversion I will
propose that the ACC marked nominal is an OBJ.
The operative assumption here is that an optimal account of Polish
inversion must refer to the semantic properties associated with arguments
of PREDICATES, the correspondence between such semantic properties and
surface case marking as well as to the many-to-one relation between surface
case and the grammatical function status of nominals. Finally, the exponent
of the PREDICATE can be a synthetically expressed wordform or an analytic
construction. These issues can be seen more perspicuously by reviewing
some relevant proposals by Dowty (1991), Fillmore (1977), Gawron (1983)
and Mohanan & Mohanan (1990).
2.1.1. Background for the proposal

In Gawron (1983) it is suggested that the crucial difference between
Fillmore (1968) and Fillmore (1977) concerns the determinants for which
arguments of the PREDICATE are mapped into which nuclear relations (i.e.
SUBJ and OBJ). Whereas early on Fillmore proposed that (atomic) semantic
roles were mapped into such relations, in later work he hypothesized
various features to be constitutive of a so-called saliency hierarchy and held
that these features were primarily responsible for the grammatical function
value of arguments. Gawron enumerates the following features:
(1) An active element outranks an inactive one
(2) A causal element outranks a noncausal one
(3) A human (or) animate experiencer outranks other elements
(4) A changed element outranks a nonchanged element
(5) A complete or individuated element outranks a part of an element
(6) A 'figure' outranks a 'ground'
(7) a 'definite' element outranks an 'indefinite' one.
Gawron describes the mapping procedure as follows:
"These seven statements are consulted in order and when one that fits is reached, the
ranking element becomes nuclear if a nuclear slot is still available." (1983: 28)
10 FARRELL ACKERMAN
Fillmore (1977) utilizes these features to explain the nuclear relations for
PREDICATES, e.g. knock vs. knock down. For a sentence such as:
He knocked on the door with his fist
he observes that though the activity requires more than a single participant,
i.e. a knocker and thing that is knocked, the entity that is active (and in the
present instance, also causal and human) is put into perspective and
functions as SUBJ. On the other hand, when the activity effectuates a
change of state in some participant, then that participant is also
perspectivalized and, following the mapping procedure presented above, it
functions as an OBJ. He says of such an activity, "... we express this with the
two word verb knock down, as in:
He knocked the door down". (1977: 77)
It should be observed that these features can be associated with phrasal
entities such as knock down on his account. In other words, Fillmore does
not identify the projector of features relevant for function assignment with
a single morphonologically integrated and syntactically atomic entity. In
addition, the arguments of this predicate are not explanatorily associated
with discrete thematic or semantic roles.
In similar spirit, but with an explicit disavowal of both case roles (e.g.
thematic roles) and saliency as criteria!, Dowty (1991) articulates a proto-
role view of the relation between arguments of a PREDICATE7 and their
grammatical function status. He suggests the following lists of proto-agent
and proto-patient properties.
Contributing properties for Proto-Agent Contributing properties for Proto-Patient
a. volitional involvement in event or state a. undergoes change of state

b. sentience b. incremental theme
c. causing an event or change of state in c. causally affected by another participant
another participant
d. movement (relative to position of another d. stationary relative to movement of another participant
participant
(e. exists independently of the event named (e. does not exist independently of the event, or not at all)
by the verb)
7
I am assuming that the conception of PREDICATE here is not coexistensive with
a single lexical category as on the lexical categorial approach, since, for example, Dowty
(1979) proposes lexical rules for e.g. resultative constructions such as that mentioned by
Fillmore above. In other words, there is no identification of lexical operations with an
entity that exhibits lexical integrity and is syntactically atomic.
These properties are hypothesized to be entailments following from the

semantics of PREDICATES and are associated with the arguments of a
predicate. For present purposes, the actual inventory of contributing
properties is less significant than the informing conception behind this
approach: arguments of PREDICATES are interpretable as bearing proto-
properties, while constellations of particular features (or weightings
associated with such properties) are relevant for grammatical function
assignment. The (relevant subset of) principles he suggests for effecting
such associations are the following:
Argument Selection Principle

In predicates with grammatical subject and object, the argument for which the
predicate entails the greatest number of Proto-Agent properties will be lexicalized as
the subject of the predicate; the argument having the greatest number of Proto-
Patient entailments will be lexicalized as the direct object.
Corollary 1:
If two argument of a relation have (approximately) equal numbers of entailed Proto-
Agent and Proto-Patient properties, then either or both may be lexicalized as the
subject (and similarly for objects)8
Non-Discreteness:
Proto-roles, obviously, do not classify arguments exhaustively (some arguments have
neither role) or uniquely (some arguments may share the same role) or discretely
(some arguments could qualify partially but equally for both proto-roles),
Finally, Dowty suggests that "the Proto-roles and their argument selection
principles determine hierarchies of traditional roles..." (1990: 578). The
particular hierarchy he mentions is:
AGENT > EXPERIENCER/INSTRUMENT > PATIENT > SOURCE/GOAL
In sum, Dowty articulates a view resembling the function-inducing
I interpret the reference to "both" in this passage as indicating alternity rather than
simultaneity in order to conform to uniqueness assumptions for grammatical functions.
12 FARRELL ACKERMAN
semantics also sought by Gawron and Fillmore.9 However, these linguists

do not go the further step and attempt to characterize case marking
inducing semantics independent of grammatical function.10 In contrast, this
move is made for DATIVE SUBJ constructions by Mohanan and Mohanan
(1990).
Mohanan and Mohanan (1990) propose an analysis of Malayalam
inversion constructions informed by the insight that "the selection of the
case of an argument may crucially depend upon its inherent meaning,
regardless of its grammatical function." (1990: 56). Like Wierzbicka (cited
earlier for Polish) and others, they attempt to identify the semantic
properties determinative for DATIVE marking over a range of uses which
seem palpably similar to one another. For present purposes the properties
they isolate (specifically, possessor and goal)11 are less important than the
representations they provide for encoding the insight that reference to
different dimensions of information is necessary for an explanatory account
of inversion phenomena. On the other hand, to give some (speculative)
substance to the present proposal, I hypothesize, adapting Dowty (1991),
that DATIVE marking is associated with the role he gives as EXPERIENCER
in his hierarchy: This represents a composite of the semantic properties +
sentience (in particular, (+ human)), - volitional, 4- causally affected (where
this latter is construable in terms of manner of affect, i.e.
recipient/beneficiary/maleficiary, contingent on the semantics of the
9
It would be incorrect to interpret Dowty's proposal as an effort to characterize
function-inducing semantics. Instead, as can be seen from the formulation of the Argument
Selection Principle, semantic entailments are understood as independent of the
grammatical function inventory of a predicate. The Selection Principle accounts for how
to probalistically relate entailments to functions given the co-presence of particular
entailments and particular functions. See Ackerman & Moore (1994) for discussion.
10
Just as there have been linguists who have previously argued for the independence
of case marking and grammatical function (for example, Shibatani (1977) among others,)
there have been linguists who have argued for the semantic motivation for surface case
marking (for example, Breal (1964), Wierzbicka (1980), Tsunoda (1981) among others.)
11
As often occurs in appeals to lexical semantics one enters a terminological morass:
despite the strong appearance of referring to many of the same distinctions, Wierzbicka,
Dowty and Mohanan and Mohanan, utilize different terminology for the same distinctions
and the same terms for different distinctions. I cannot attempt to reconcile or render
consistent such terminological differences here.
particular (class of) predicates which project this property, rather than
being construable as having undergone a change of state or location).
With respect to representational conventions for lexical items, T.
Mohanan (1991) distinguishes between several levels of representation for
the information associated with a PREDICATE: she argues for a semantic
level, an argument structure level, and a grammatical function level, among
others. In the present context these can be interpreted in the following way:
the argument level represents the valence of a PREDICATE, while the
semantic level possesses properties associated with the arguments (these
properties being, for example, Dowty type proto-role features indicated
schematically as P) and the grammatical function level represents the
functional value of arguments. This yields the following representation:
P P P P Semantic Level
Case Marking
ARG Argument Level
GF Grammatical Functions
CASE Case Marking
T. Mohanan suggests that, given the organization of semantic and

grammatical function information presented above, surface case marking
conventions can be keyed to either grammatical function values of
arguments (defaults as indicated by broken lines) or semantic properties of
arguments (semantically specified as indicated by unbroken lines). This
type of representation will be utilized for the Polish inversion constructions.
2.1,2. The Polish inversion predicate construction

Adopting the amalgam of views presented in the previous section, I propose
that the Polish productive inversion constructions be analyzed as containing
14 FARRELL ACKERMAN
the following representative PREDICATES serving as the semantic heads of

their clause:
Intransitives:
sleep <ARG> sleep rflx ADVERB <ARG> DAT
SUBJ SUBJ
NOM
Transitives:
read <ARG1, ARG2> read rflx ADVERB <ARG1, ARG2> DAT
SUBJ OBJ SUBJ OBJ
NOM ACC ACC

These representations can be read as follows: (1) A PREDICATE, i.e., the

information to the left of the angled brackets, is associated with a set of
arguments, i.e., placeholders within the angled brackets, (2) these
arguments are associated with semantic properties, i.e., the lines connecting
arguments to Ps, (3) the constellation of Ps (following Dowty) accounts for
function selection, i.e., the lines connecting placeholders to grammatical
function names, and finally (4) there are default and specified surface
realizations of arguments, i.e. the SUBJ is realized as NOM unless the ARG
linked to SUBJ has some specific case marking inducing properties (e.g.
DAT) and the OBJ is realized as ACC unless the ARG linked to OBJ has
some specific case marking inducing properties (e.g. INSTR(umental),
GEN(itive)).
These schematic conventions receive the following specific
interpretations: (1) the Polish PREDICATE can be expressed synthetically or
analytically, with the INVERSION PREDICATE construction representing an
analytic variant containing a rflx marker (indicating attenuated agentivity of
the SUBJ12) and a manner adverb (serving to restrict and specify the type
of experience for the SUBJ) and (2) the constellation of Ps provide the
proto-role properties determinative of function assignment, while the
properties +human, -volitional, +affected (i.e. construable in terms of the
manner in which the activity impinges on the individual, rather than on
what change of state or location is undergone by the individual) help
determine case marking.
On the present analysis, then, the difference between constructions with
inversion and without inversion does not consist of grammatical function
changes, as, for example, in the standard Relational Grammar analyis of
this phenomenon (see Dziwirek 1991): the grammatical functions within
these pairs of constructions remain constant. Evident changes in the
expression of arguments have to do with what constitutes the PREDICATE
and how the function and case inducing meanings of PREDICATES affect the
surface expression of arguments. Since, on the present account, the
inversion predicate only affects the semantics of the SUBJ there is no reason
The use of the reflexive clitic in this construction is reminiscent of its use in Polish
middle constructions which generally occur without an expressed agent: in middles there
is a NOM SUBJ and, according to my informant, if the agent is expressed, it will appear
in the DATIVE case. Finally, the claim of attentuated agentivity may be evidenced by the
inability to use zeby purposive clauses with the productive inversion constructions, while
it is possible to use them with the non-inversion analogues of these constructions.
16 FARRELL ACKERMAN
to assume that there is any effect whatsoever on the OBJ argument of the
transitive base predicate. On the present proposal, therefore, the ACC
argument is an OBJ, both for the simple predicate and its inversion
analogue» Alternatively put, the present proposal predicts the retention of
the OBJ function with productive inversion, since there is nothing in the
inversion predicate that would motivate a change in function or case
marking for this argument.
In the remainder of this paper I briefly examine several 'simple
behavioral properties of the nominals in inversion constructions and
demonstrate how they can be explained in terms of the claims that the DAT
nominal is a SUBJ and that the ACC nominal is an OBJ.13
3. Case marking and grammatical functions14
One phenomenon purportedly keyed to the SUBJ status of nominals in

Polish is verb agreement. The condition on agreement proposed in
Dziwirek is presented below:
Verb Agreement Rule: Grammatical relations
If a nominal a heads a final 1-arc in a clause b, then a determines verb agreement in
b.
On this account, the surface SUBJ (i.e., the final 1) determines verb
agreeement Since on Dziwirek's analysis there is a bi-unique relation
between surface SUBJecthood and NOMINATIVE case marking, the condition
on verb agreement is also formulable in terms of case-marking:
Verb Agreement Rule: Case marking
If a nominal a is the NOMINATIVE nominal in a clause b, then a determines verb
agreement in b.
This latter formulation suggests that agreement is triggered by
morphological marking and is consequently, in principle, independent of the
grammatical status of the determining nominal. Such a suggestion would
appear rather ad hoc if Polish were the only language for which such a
Detailed arguments for these assumptions are developed in Ackerman (1994).
The data in this section are taken from Dziwirek (1991).

condition proved to be empirically supportable. In fact, given the apparent

bi-conditional relation between NOMINATIVE case and final SUBJ in Polish
it is impossible to determine whether the agreement rule should be
formulated in terms of case, grammatical function or a conjunction of the
two. It is therefore important to consider languages which exhibit
agreement as well as a relevant many-to-many relation between case and
grammatical function. In what follows I demonstrate that the case and
agreement systems of Hindi provide an instructive pattern in this regard. In
particular, we will see that this language provides evidence for appealing to
case or a conjunction of case and function, but that the condition on
agreement is not amenable to a formulation in terms of surface
grammatical functions alone.
Following the traditional interpretation of verbal gender agreement in
Hindi, T. Mohanan (1990:136) states that the generalization accounting for
this phenomenon is that, "The verb agrees with the NOM SUBJ. If the SUBJ
is NONNOM, the verb agrees with the NOM OBJ." She adduces the following
data in support of this generalization. Hindi possesses three relevant
nominal case marking options for nominals that exhibit SUBJ properties:
NOM, ERGATIVE, DATIVE. The contrast between NOM and ERG case
marking and their effects on verbal agreement is typified in (5) and (6). In
(5) SUBJs appear in the NOM case and the verb agrees with them in gender:
(5a) reflects masculine agreement with the masculine NOM SUBJ, while (5b)
reflects feminine agreement with the feminine NOM SUBJ.
(5a) ravii baala-ko uthaaegaa
Ravi-NOM(M) boy-ACC(M) lift-FUT-M.SG.
RAVI WILL LIFT UP THE BOY
(5b) niinaa baala-ko uthaaegii

Nina-NOM(F) boy-ACC(M) lift-FUT-F.SG.
NINA WILL LIFT UP THE BOY
In contrast, the corresponding examples in (6) reflect the fact that the ERG
case appears obligatorily on the SUBJ in the past tense or with perfect
aspect. This tense and aspect induced difference in SUBJ case-marking
produces a difference in verbal agreement: in both (6a) and (6b) there is
no NOM argument and there is a correlative lack of gender agreement in
the verb. The verb in both instances appears in the default masculine form.
18 FARRELL ACKERMAN
(6a) ravii-ne baala-ko uthaayaa

Ravi-ERG(M) hoy-ÅCC(M) lift-PAST-M.SG,
RAVI LIFTED UP THE BOY
(6b) niinaa-ne baala-ko uthaayaa

Nina-ERG(F) boy-ACC(M) lift-PAST-M.SG.
NINA LIFTED UP THE BOY
In (5) and (6) the OBJ complements are in the ACC case, since this is the
convention for marking human referents. On the other hand, the NOM form
is appropriate for marking inanimate referents, as indicated in (7a) where
the 'bread' is NOM and in (7b) where the 'banana3 is NOM,
(7a) ravii-ne rotii khaayii
Ravi-ERG(M) bread-NOM(F) eat-PERF-F.SG.
RAVU ATE THE BREAD
(7b) ravii-ne kelaa khaayii

Ravi-ERG(M) banana-NOM(M) eat-PERF-M.SG.
RAVI ATE THE BANANA
The examples in (7) show that verbal gender agreement is determined by
the NOM OBJS: the verb agrees with the feminine OBJ in (7a), whereas it
agrees with the masculine OBJ in (7b).
Finally, given the conditions on case-marking, it is possible to create
clauses in which there are two NOM complements, as in (8a) and (8b)
where both the SUBJ and OBJ complements appear in this case.
(8a) ravii rotii khaaegaa
Ravi-NOM(M) bread-NOM(F) eat-PERF-M.SG.
RAVI WILL EAT THE BREAD
(8b) niina rotii khaaegii

Nina(F) bread-NOM(F) eat-PERF-F.SG.
NINA WILL EAT THE BREAD
As can be seen form these examples, gender agreement is determined by
the NOM argument which is also the SUBJ: in (8a) the verb agrees with the
masculine gender of 'Ravi', while in (8b) it agrees with the feminine gender
of 'Nina,9
What we see here, then, is that verbal agreement cannot be formulated
in terms of grammatical functions alone. Given cross-linguistic evidence that

verb agreement in some languages requires reference to surface case or a
conjunction of case and grammatical function, it is not necessary to assume
that Polish agreement must be determined exclusively by grammatical
function, rather than by a conjunction of function and case. That is, since
it is necessary to posit such a condition for other languages, there is no
principled reason to exclude this possibility for Polish, given that it is
compatible with the data. In this sense it is possible to view agreement in
Polish as determined by the NOM marked nominal in the clause, which is
sometimes also the SUBJ.
In sum, the absence of agreement between the inversion nominal and the
categorial verb in Polish inversion predicates cannot be employed as
evidence for the non-SUBJ status of the inversion nominal. This is because
the condition on verb agreement in Polish may be adequately formulated
in terms of case marking, rather than in terms of grammatical functions. On
the present account, the inversion nominal is a DATIVE marked SUBJ: the
proposed condition on verb agreement is compatible with this account.
It should be recalled that, by hypothesis, the semantics of the inversion
predicate motivates DATIVE marking and that, as mentioned above, the
inversion nominal is the surface SUBJ. The present analysis distinguishes
between two (perhaps, more) differently case marked SUBJs. If the
inversion nominal is truly a SUBJ, it is to be expected that it would exhibit
some of the behavioral properties ordinarily associated with SUBJs in Polish.
I examine below only one such property, namely, the control of reflexives.
The sentence in (9) illustrates the representative distributions for the
binding of reflexives in Polish non-inversion constructions:
(9) Janek(i) opowiedzial Ewie(j) historię
John-NOM told Eva-DAT story-ACC
swojego(i)/*(j)wycia
his-rflx life-GEN
JOHN TOLD EVE THE STORY OF HIS LIFE
The controller can only be the NOM SUBJ, as indicated by the gloss in (9).
Now, given that on the present account the inversion nominal is a DAT
SUBJ, it is predicted that inversion nominals can serve as controllers. The
relevant sentence is presented below:
20 FARRELL ACKERMAN
(10) Ewie(i) dobrze pracuje sie ze swoimi(i) kolegami

Eva-DAT well work-3sg rflx withher-rflx(i)-INSTcolleagues-INST
EVE WORKS WELL WITH HER CO-WORKERS
The inversion nominal in (10), as predicted, can serve as the controller of
the reflexive instrumental case marked possessive adjective 'her',
In this section we have seen that one representative type of syntactic
behavior in Polish is explicable in terms of case-marking (in conjunction
with grammatical functions), while another is explicable in terms of
grammatical functions. Case and grammatical function are two of the
central ingredients integral to the present proposal, so it is to be expected
that they should find independent motivation as the terms of explanation
for different sorts of syntactic behavior,
Another assumption of the present proposal is that inversion predicate
formation only semantically affects the SUBJ argument, this being reflected
in DATIVE case marking. Since the semantic effect obtains solely for the
SUBJ, it is predicted that the ACC argument in non-inversion and inversion
pairs should exhibit the same syntactic behaviors: they should both exhibit
OBJ properties.
A standard test for OBJECThood in Polish is genitive of negation. This is
a phenomenon whereby an ACC OBJ ( 11a) is expressed in the GEN(itive)
in negated clauses (l1b).15 This is exemplified by the following pair of
sentences:
(11a) Ewa czyta ksiątzkę
Eve read-3sg book-ACC
EVE IS READING A BOOK
The reader should note the following with respect to OBJs in Polish: given that
passive is a reliable diagnostic of OBJECThood, it should be observed that Polish
predicates can govern several different cases. This is why it is important when formulating
the conditions for genitive of negation to specify that it operates on ACC OBJs: OBJs
which appear in other cases do not undergo this operation. Two points are relevant in
connection with this: (1) case marking for OBJs can be specified by predicates (and
motivated by semantic considerations) just as can the case marking of SUBJs and (2) the
operation of genitive of negation must refer to the combination of function and case-
marking. Both of these phenomena are expected on the present analysis since we have
argued that semantic properties can determine case marking and our representations
contain information concerning the surface case and functional status of nominals,
(11b) Ewa nie czyta ksiązki

Eve not read-3sg book-GEN
EVE IS NOT READING A BOOK
On the assumption that this phenomenon occurs in transitive clauses (with
the proviso that it happens with some intransitives), it is to be expected that
it will occur in inversion constructions, since they are transitive on the
present account. As can be seen in example (12), genitive of negation does
obtain for the argument corresponding to the ACC argument in affirmative
productive inversion constructions:
(12) Tej ksiązki nie czytalo się Jankowi z przyjemnoscią
this book-GEN not read-rsg/neuterrfbxJohn-DAT with pleasure
JOHN DID NOT READ THIS BOOK WITH PLEASURE
As is evident in (12), the nominal 'book' appears in the GEN case. This
behavior, in fact, supports Kubinski's (1987) proposal that the ACC nominal
in productive inversion constructions is an OBJ.
In sum, we have seen here that the basic assumptions concerning the
relevance of case-marking for certain phenomena, the grammatical function
status of the inversion nominal as a SUBJ and the assumption that the ACC
argument is an OBJ are all consonant with several sorts of syntactic
phenomena in Polish.
4. Conclusions
In this paper I have demonstrated that by employing a notion of

PREDICATE distinct from its encoding as a morphophonologically integrated
and syntactically atomic unit or as a VP i.e., the difference between a
morphological word and a syntactic phrase, we can arrive at an explanatory
account of Polish inversion constructions. I have assumed that the Polish
(productive) inversion predicate is a construction consisting of certain
information content which induces both function assignments for arguments
and case marking for certain arguments. This means that function
assignments can remain invariant, while case marking associated with
arguments can vary, depending on the semantics of the predicate. We have
examined some simple evidence showing how certain assumptions
concerning the functional status of the nominals associated with the
inversion predicate relate to demonstrable syntactic behaviors.
22 FARRELL ACKERMAN
The basic point I hope to have illustrated is that a promising account of

Polish (productive) inversion constructions can be formulated by referring
to independently motivated and necessary dimensions of information such
as the entailments of predicates, grammatical functions and surface case
marking. Finally, all of these types of information are associated with an
analytically expressed inversion predicate,
References
Ackerman, F.
1994 DATIVE SUBJ predicates in Polish. Ms,
Ackerman, F. & J. Moore

1994 Valence and semantics of causatives. Ms,
Ackerman, F. & G. Webelhuth

1994 Wordhood and syntax: A theory of natural predicates. (Ms.)
Breal, M.
1964 Semantics: Studies in the science of meaning, Dover.
Bresnan, J. & J, Kanerva

1989 "Locative inversion in Chichewa." Linguistic Inquiry 20:1.147-185.
Dowty, D.
1991 "Thematic proto-roles and argument selection." Language 67:3.547-619.
Dowty, D.
1979 Word meaning and Montague grammar. Dordrecht: Reidel.
Dziwirek, K.
1991 Aspects of Polish syntax. UC San Diego doctoral dissertation.
Fillmore, C. J.
1977 "The case for case reopened." In P. Cole & J, Sadock (eds.), Grammatical
relations. New York: Academic Press.
Fillmore, C.J.
1968 "The case for case." In E. Bach & R.T. Harms (eds.), Universals in linguistic
theory, New York: Holt, Rinehart and Winston.
Fillmore, C.J. & P. Kay

1994 Outline of construction grammar. Ms.
Foley, W. & R.D. Van Valin

1984 Functional syntax and universal grammar. Cambridge: Cambridge University
Press.
Gawron, M.
1983 Lexical representation and the semantics of complentation. UC Berkeley
doctoral dissertation.
Goldberg, A.
1992 Argument structure constructions. UC Berkeley doctoral dissertation.
Gruber, J.
1965 Studies in lexical relations. MIT doctoral dissertation.
Jackendoff, R.
1991 Semantic sttuctures. Cambridge, Mass.: MIT Press.
Jackendoff, R.
1975 "Morphological and semantic regularities in the lexicon." Language 51: 639-671.
Kubinski, W.
1987 Reflexivization in English and Polish: An arc pair grammar analysis. Niemeyer.
Levin, B. & Malka Rappaport Hovav

1995 Unaccusativity at the syntax-lexical semantics interface. Cambridge, Mass.: MIT
Press.
Mohanan, T.
1990 Arguments in Hindi. Stanford doctoral dissertation.
Mohanan, K.P. & T. Mohanan

1990 Dative subjects in Malayalam: Semantic information in syntax. In M. Verma &
K.P. Mohanan (eds.), Experiencer subjects in South Asian languages.
CSLI/University of Chicago Press.
Shibatani, M.
1977 "Grammatical relations and surface cases." Language 53:789-809.
Spencer-Brown, G.
1972 Law of form. New York: Julian Press.
24 FARRELL ACKERMAN
Tsunoda, T.
1981 "Split case-marking patterns in verb-types and tense/aspect/mood." Linguistics
19:389-438.
Van Valin, R.D.

1991 Generalized semantic roles, argument selection and the syntax-semantics
interface. Ms.
Verma, M. & K.P. Mohanan

1990 Experiencer subjects in South Asian languages. CSLI/University of Chicago Press.
Wierzbicka, A.
1988 "The meaning of a case: A study of the Polish dative." In The semantics of
grammar. Amsterdam: John Benjamins.
Wierzbicka, A.
1980 The case for surface case. Karoma Press.
Zwicky, A.
1992 Syntactic constructions. Talk given at the Milwaukee Conference on Symbolic
vs. Non-symbolic Models of Language.
The role of the example
in a frame semantics dictionary
B.T.S. Atkins
Oxford University Press
The traditional dictionary description of word meaning reflects the concept

of 'senses' of words: the idea that a polysemous expression may be
adequately described by factoring out its various usages and describing
these individually, sometimes - but not always - grouping some of them
together as part of the description. In such a context, the dictionary
example exists as a distinct component of the entry, and is used not only to
exemplify facts given elsewhere, but also to introduce new information when
this is deemed necessary.
In this paper I propose to look at how the work of Charles Fillmore in
describing word meaning within the context of frame semantics offers
lexicographers a new approach to their task, at a time when the dictionary
is about to be released from the shackles of space constraints into a new
electronic universe. In particular, I shall consider how the purpose, function,
type and indeed the whole identity of dictionary examples would be
transformed in a frame semantics dictionary1.
1. The dictionary entry
Examining the dictionary example is like running a geiger counter over the
The new lexicography sketched in this paper is based on Fillmore's work on the
semantics of risk, described in Fillmore and Atkins (1992, 1994). Any idiosyncratic
divergencies from that account simply underline the need experienced by this lexicographer
for a semanticist to light the way.
26 B.T.S. ATKINS
subterranean lexicographical mass. Changes in the foundations on which the

edifice of the entry is constructed will inevitably be reflected in changes in
the type of examples it contains, in the way in which these examples are
meant to be used, and in the contribution they make to the entry as a
whole.
1.1 An entry in a traditional dictionary
The fifteen or so components (in italics, below) of the traditional dictionary

entry - although rarely all found in a single dictionary - offer the user the
following types of information about each 'sense' of the headword:
a) a general explanation of its meaning in the form of a definition (in
monolingual dictionaries) or translation(s) (in bilingual dictionaries);
b) facts about its grammatical behavior: a part-of-speech abbreviation, and
often a metalinguistic statement of its complementation pattern(s) if any;
c) examples of its use, with or without explanations of their meaning in the
form of glosses or translations;
d) facts about its participation in multiword items such as idiomatic
phrases, compounds, etc;
e) information on other aspects of language (style, register, domain,
region, currency etc.) carried by metalinguistic indicators;
f) semantic relationships indicated by inclusion of synonyms, antonyms,
etc;
g) its etymology;
h) any inflectional forms it may have;
i) its pronunciation, usually in the form of a phonetic transcription;
j) orthographical information such as its variant spelling(s);
k) its derivatives, which may be full entries in their own right, or undefined
run-ons or other subordinated entries;
1) cross-references to entries for items in some way related to the head
word.
The traditional approach to lexicographical description is thus predicated
on the existence of distinct senses of polysemous words - senses, indeed,
that are usually numbered in order to emphasize the distinction.
A FRAME SEMANTICS DICTIONARY 27
1.2 An entry in a frame semantics dictionary
Frame semantics is holistic in its approach to lexical meaning. In a frame

semantics dictionary, there is one single underlying schematization,
untrammelled by discrete dictionary senses. Individual aspects of word use
cannot be isolated from the undifferentiable whole, the frame2. The term
frame refers to a schematization consisting of three situation types (the
schemas), analyzed into and described in terms of the participant concepts
(the frame elements). The concept of 'headword' becomes obsolete, for the
whole frame is the definiendum. The components of an entry in such a
dictionary will contain information on the following aspects of the frame:
a) the schemas and their interrelationships;
b) the frame elements, including:
- what these are (possibly including an analysis of their semantic
components, which will be shared by elements in other frames);
- their syntactic and lexical modes of realization;
- the semantic neighbors of the frame;
c) the co-inhabitants of the frame (related words within it), including:
- what these are;
- their part(s) of speech;
- their complementation (semantic content and syntactic realization);
- differentiation from quasi-synonyms (other words sharing some of the
same frame elements);
- their etymology;
- their various inflectional forms;
- their various pronunciations;
- any orthographical variants of these words.
Some of the information about the words in the frame (etymological,
inflectional, phonetic and orthographical) need not concern us now; for the
present purpose, it is enough to look more closely at the broad outline of
a frame semantics dictionary entry.
It will carry three principal types of information, relating to the three
principal aspects of the frame: its schemas, its elements, and its co-inhabit
ants. These three sections - the heart of the entry - supply the type of
2
A full description of the risk frame, its schemas and frame elements, is to be found
in Fillmore and Atkins (1994).
28 B.T.S. ATKINS
information offered in a traditional dictionary by the headword definition

or translation (explanation of meaning); by the part-of-speech abbreviation,
and complementation patterns (grammar); by the statement and explana
tion of phrases and compounds in which the frame word(s) participate
(phraseology); and by the listing of synonym(s) and/or other semantically
related items. The scope of each of these sections includes all the
morphologically related words within the frame, and cross-references to
other frames will form an integral part of each description.
2. Lexically relevant facts in the 'risk' frame entry
The lexicography of frame semantics is founded on the conceptual

framework underlying the meaning and use of a word or set of related
words; its task is to explain the relationships between elements of the
conceptual frame and their syntactic and lexical realizations.
In the summary that follows of the lexically relevant information which
a frame semantics dictionary must hold for risk and its derivatives, the
names of the frame elements are capitalized when they are first introduced;
not all of them are illustrated in the figure below (the 'Protagonist', the
'Possession', and the 'Setting' are never specifically shown), but all may be
realized in some usage of risk instantiating the relevant schema. Some brief
examples of such usages are given here in order to clarify the concepts.
2.1 The schemas
The risk frame consists of three schemas, illustrated3 in Figure 1. In

Schema A, the person or thing centrally involved in the risk scenario - the
Protagonist [PR] - is in a Setting [SE] in which there are two alternative
futures; these unknown possible futures form an integral part of the
schema. In this schema, however, there is no question of the Protagonist's
making any decision that results in an action or a series of events liable to
lead to one of these alternative futures. The circle in the figure indicates
3
This diagram is taken from Fillmore and Atkins (forthcoming).
'chance' (as opposed to 'choice' represented by the square 4 in the other

schemas), and the label BA on one of the arrows stands for the Bad
outcome to which the Protagonist is exposed. This often involves the loss
of, or harm to, someone or something - the Possession [PO] - valued by the
Protagonist, and the Source [SO] of the loss or harm may be referred to.
Schema A
Schema B
Schema C
Figure 1: The RISK scenario
These figures are adapted from the kinds of directed graphs used in mathematical
decision theory (as discussed in Raiffa 1970).
30 B.T.S. ATKINS
Schema A is instantiated in the following sentences, where each frame

element is bracketed together with its realization in the sentence:
[SE There is a certain risk].
The risk to [PO their health] is not negligible.
[PR The babies] in that ward were running the risk of [BA-hypothermia] from
[SO the extreme cold],
In Schema B, the Protagonist takes a decision which may lead to the bad
outcome, although he or she is not aware of doing this. The frame element
Decision [DE] denotes both the decision and the resulting action or event.
The other frame elements in Schema B are those already introduced in
Schema A. Schema B is instantiated in:
There is little risk in [DE reflation] or [DE deciding to reflate].
[DE Such a response] would be an irrational risk.
Schema C differs from Schema B in one aspect only: the Protagonist is
aware of the two possible outcomes, knowing that one is bad, wishing to
bring about the other - the Goal [GO] (a temporary abbreviation for all of
the reasons a person might have for taking risks), and able to assess the
probabilities associated with each. The oval in figure 1 represents this
awareness. Schema C is instantiated in:
[PR She] risks [BA the hazards of the highway] [GO to catch a glimpse of
him].
The risk entry in a frame semantics dictionary must contain a comprehen
sive analysis of the schemas and their interrelationships - the central and
principal lexically relevant information about the frame that is the focus of
the entry»
2.2 The frame elements
Another piece of core information about the frame consists of a compre

hensive and detailed account of the elements in that frame. From the
foregoing it will be seen that there are seven of these:
Protagonist [PR] the central figure in the frame.
Setting [SE] the situation in which the risk exists
Possession [PO] something/someone valued by Protagonist and
endangered in the situation
Bad [BA] the possible bad outcome, or harm.

Goal [GO] the desired outcome.
Decision [DE] the decision that could trigger this.
Source [SO] the possible origin of the harm.
It is clear that an account of the frame elements cannot be separated from
the description of the relevant schemas in the frame. There is no possibility
here (unlike the 'distinct senses' approach of the traditional dictionary) of
arbitrarily sectioning off successive parts of the information and presenting
it as differentiated aspects of word behavior.
2.2,1 Their modes of realization
In the on-line entry, the discussion of the frame elements would include a
systematic description of their modes of realization; included in this would
be a comprehensive analysis of corpus citations, in which the various frame
elements would be factored out, highlighting the ways in which they are
expressed, syntactically and lexically, in the combinatory contexts of the
words in the risk frame.
It is possible here to give only a few brief glimpses into such an entry,,
In exhaustively enumerating the ways in the frame elements are realized by
sentence constituents, the dictionary would, for instance, record that three
of the frame elements (Decision, Bad and Possession) may be realized as
objects of the verb risk, as exemplified here:
He risked [DE climbing the mountain].
He risked [DE the climb].
He risked [BA falling down].
He risked [BA a fall].
He risked [PO his life].
It would note, too, that in the context of the verb risk Decision [DE] and
Bad [BA] may be expressed syntactically either as a gerund or a noun,
whereas Possession [PO] is expressed only as a noun.
Furthermore, it would record that, in the context of take the risk of
Decision [DE] may be realized as the prepositional object when it is
expressed as a gerund:
He took the risk of [DE climbing the mountain]
but not when it is expressed as a noun:
*He took the risk of [DE the climb].
32 B.T.S. ATKINS
It would also record that Bad [BA], whether in the form of a gerund or a
noun phrase, may be realized as the object of the preposition in the phrase
run the risk of:
He ran the risk of [BA falling down]
He ran the risk of [BA a fall],
And it would record that the syntactic realization of Possession [PO] would
be different with the phrase put at risk from take risks with:
He put [PO his own life) at risk.
He took risks with [PO his own life].
Thus the entry would set out in detail the relationship between the noun
risk and the verb risk, showing how both may be used to transmit an almost
identical message, and allowing an analysis of the speaker's (unconscious)
motivation in selecting one way rather than another of expressing the same
semantic content.
An account of the frame elements and their realizations would also
include an account of the relationships between the various multiword items
within the risk frame (take or run risks / a risk / the risk (of); (put or be) at
risk; at the risk of; at one's (own) risk, etc.), highlighting the way each idiom
or compound licenses different configurations of frame elements, and how
the whole offers a range of options allowing the expression of extremely
fine shades of meaning.
2.2.2 Semantic neighbors of the frame

A major task of the frame semantics dictionary would be to analyze and
record the semantic relationships linking the thousands of frames that
constitute the lexico-semantic resources of the language.
Semantic taxonomies would be quite different in such a dictionary:
these relationships traditionally link lexical units (or, broadly speaking,
'dictionary senses'), formalizing for each sense an individual network of
semantic relations; in frame semantics similar relationships may be noted
exist between frames. The relationship of frame hyponymy might be defined
as: 'if the Y frame elements constitute a subset of the X frame elements,
then Y stands in a superordinate relationship to X'. One of the facts that
the risk entry would note, for instance, is that this frame stands in a
superordinate relationship to the bet/wager frames.
A similar definition might be drawn up for frame synonymy, although
here, as in traditional semantics, partial synonymy would appear more likely

than total synonomy. Frames might be considered to be synonymous to the
extent that they share specific frame elements. Thus the dictionary entry for
the risk frame would perhaps note the danger frame as a fairly close
synonym, sharing as it does at least the Protagonist, Setting, Bad, and Source
elements, while recording, say, the value frame as a more distant relative,
sharing with risk possibly only the Protagonist and Possession elements.
As well as the frames themselves, the various words and expressions in
each frame (the 'co-inhabitants') will stand in definable semantic relation
ships to each other, and to other words from other frames. This different
level of relationship is briefly discussed in 2.3.2 below.
Another type of information to be analyzed and recorded within a
frame semantics description can only be briefly mentioned here. At an even
deeper level (the analysis of of the frame elements into their semantic
components), word meanings are once more intricately intermeshed, for the
components of the elements in the risk frame are such cognitive universals
as want, not want, uncertain, good, bad, have, cause, and act, which, in
different configurations, and in combination with other similar concepts, will
constitute other elements in other frames.
2.3 The co-inhabitants
Each frame entry would of course list and discuss all the expressions in the
frame - the most obvious access point to the information, though in an
on-line dictionary not the only one. In the case of the risk frame, these
would include:riskitself, andrisky,riskily,riskiness,risker,etc., together with
the idiomatic expressions in which these words participate.
2.3.1 Their complementation

The verb risk will provide an example of the type of information to be
recorded under this heading. The dictionary would note that this verb
requires an object, which may be in nominal or gerundial form. In the
nominal form, the object can express three frame elements, in the gerundial
only two. The frame elements expressed in this way are:
34 B.T.S. ATKINS
Possession: risk [PO your life]

Bad : risk [BA death]
risk [BA dying]
Decision : risk [DE swimming]
risk [DE a swim]
risk [DE the river]
This configuration reflects the fact that the [PO] element, naming an entity,
is always realized as a NP; the [BA] element (an event) and the [DE]
element (an act) may be represented either by a NP or by a gerund. In
addition, the entry would record the relationship between the expressions
of the [DE] element: the gerund of course designates an act, while the NP
can either be the direct designation of such an act (risk a swim), or a
metonym for it (risk the river).
There is, however, another aspect of the complementation of the verb
risk which the frame semantics dictionary must clarify We have already
seen how metonymy operates in the expression of the [DE] elements Other
features of this verb's behavior reveal another metonymic dimension, one
involving other frames, and apparent in the diverse prepositional phrases
which may accompany the verb risk, and which express one of the elements
of the risk frame, e.g.
risko.,
a. [PO more than $50,000] in [DE ranch lands]
b. [PO $5000] on [DE a roll of the dice]
[PO three months' wages] on [GO its success]
c. [PO the arrow] on [DE a shot]
d. [PO his ships] to [BA enemy attack]
e. [PO Mark] with [SO Ben]
The frame semantics dictionary would systematically record such phenome
na, noting in the risk frame entry that the verb risk "borrows" the grammar
and the meaning of other relevant frames. In the instances illustrated
above, these are (a) invest, (b) gamble, (c) lose, (d) expose, and (e) leave.
23.2 Differentiation of qua si-synonyms

A major task of the frame semantics entry would be to explicate the
relationships between the co-inhabitants of the frame and their neighbors
in semantic space; the discussion would make explicit the similarities and
differences amongst these expressions, and clarify the implications of the
way they are used. One example will suffice: why, for instance, is the verb
risk selected to express a particular concept, rather than endanger or
jeopardize, or the phrase put at risk? The answer to this and many other
questions must be furnished by the dictionary account of the frame.
In this instance, the dictionary user would learn that when risk is used
with the meaning of 'place in danger' its object is always an expression of
the [PO] element (something or someone valued by the person involved),
while there is no such constraint on endanger, jeopardize and put at risk. I
can endanger, jeopardize and put at risk people or things that I feel nothing
for; I can only risk people or things I care about. The selection of risk to
express the concept of causing someone or something to be open to
possible harm carries with it this additional implication. The dictionary entry
would also make explicit the fact that, in these parallel uses, the [PR]
element (the central person in the frame who is vulnerable to the possible
harm) can in the context of the verb risk only be realized as the subject of
the verb, whereas with endanger or put at risk [PR] may be expressed as the
verb's object (hence you endangered me ox you put me at risk, but, with this
interpretation, not*yourisked me).
2.4 Some comments on the frame semantics entry
The lexically relevant information consists both of 'frame-internal' and

'frame-external' facts. The former (such as the description of the schemas,
described in 2.1 above) would be reached normally from an access point
within the risk frame, for instance by someone wishing to have an overview
of the meaning of this word family. The latter (such as the semantic
relationships referred to in 2.2.2 and 2.3.2, or the syntactic relationships
described in 2.2.1 and 2.3.1) would be accessed both from within the risk
frame and from other entries in the dictionary: the user seeking information
about one of risk's semantic neighbors would be led to relevant facts in the
risk frame entry, possibly through a discussion of frames with elements in
common, of elements containing the same semantic components, or of
words that 'borrow' each others' syntactic complementation structures.
From this brief glimpse of some of the functions of this on-line
dictionary, it is clear that the complex network of semantic and syntactic
relationships it describes cannot be arbitrarily compartmentalized into
'senses'; nor is it possible to describe the entries in terms of the traditional
dictionary entry components referred to in section 1.1.
36 B.T.S. ATKINS
3. The role of the dictionary example
Five principal aspects of the 'example' component of the entry must be

taken into account during the design phase of a new dictionary. They form
a framework for a discussion of the dictionary example: the purpose of the
examples in the dictionary (why are they there at all?); their focus (what
fact or facts are they exemplifying?); their function within the entry (what
should the example being doing?); the source of the examples (where does
the lexicographer get them from?); and the constraints on the wording or
other features. A selection within one of these parameters may limit the
choice in the others.
3.1 Purpose of examples in a traditional dictionary
A dictionary, of whatever kind, is primarily a tool that someone uses in an

attempt to discover facts about the language. If this tool is to be effective
- if the user is to find the information without undue frustration - editors
establishing the examples policy at the design stage of a new dictionary
must be able to articulate why the examples are there. Without a clear view
of this, no decision can be made by the lexicographers about the examples
to be chosen. In traditional dictionaries, the examples may be expected to
fulfil one or more purpose(s), including the following:
¤
to help users understand the headword (passively);
¤
to help them use the headword (actively);
¤ to help them decide which sense of the headword is being defined;
¤ (in a bilingual dictionary) to give them restrictions on target-language
equivalents already proposed, or additional information regarding the
equivalence in the target language;
¤ to justify the inclusion of the word or word-sense in the dictionary;
¤ (in historical dictionaries) to give the earliest citation for a particular
usage.
3.2 Focus of examples in a traditional dictionary
Some dictionaries do not set out systematically to illustrate any particular

type of linguistic information; others have a consistent policy in this regard.
In the latter, the aspects of word use which the example is most often
called upon to illustrate are:
¤ meaning, including systematic and idiosyncratic polysemy, the differenti

ation of quasi-synonyms and 'confusibles', and diachronic meaning
change;
¤ syntax, including part-of-speech and complementation information,
transitivity alternations for verbs, countability in the noun, etc.;
¤ collocation, including participation in idiomatic expressions of various
types;
¤
morphological inflections.
3.3 Function of examples in a traditional dictionary
Intimately linked with the objective of the dictionary is the function of the
example: what does it bring to the entry in order to achieve its purpose?
Editors must be clear about this in the style guide of a dictionary, and
compilers must be aware of it every time they select an example. The
explicit / implicit dimension (see Atkins et al. 1986) is important here, for
an example may illustrate information given explicitly elsewhere in the
entry, or it may contain information that cannot be spelt out explicitly, for
lack of space. Often, it seems, information is more efficiently transmitted
to the user by means of an example than by an abstract statement of
linguistic facts,
The possible functions of a dictionary example include the following,
several of which it may carry out at the same time:
¤
to amplify, clarify or exemplify information given elsewhere in the entry,
for example by the definition, the part-of-speech abbreviation, the
grammar codes or other formal statement of complementation patterns,
or somewhere else in the dictionary, as for instance in verb lists outside
the main text;
¤ to give new information about one of the aspects (discussed in section
3.2) of the headword in use;
¤ to show the headword in use: some dictionaries do not specify the
aspects of word use that the example must illustrate, but simply supply
one or more example(s) of the headword used in the relevant sense,
chosen according to the lexicographer's intuition;
¤
to illustrate 'prototypical' usages of the word, or (at the other extreme)
to illustrate 'boundary' usages, to show how much the word may
encompass, both in meaning and grammar;
¤ to record every instance of the headword used with a particular
38 B.T.S. ATKINS
meaning in a closed corpus (for instance, in a scholarly or a terminolog

ical dictionary).
3.4 Source of examples in a traditional dictionary
The source considered acceptable for examples in a specific dictionary

depends on the type of dictionary being compiled: scholarly dictionaries on
historical principles, for instance, obviously need different kinds of examples
from dictionaries for learners of the language. The choice facing dictionary
editors includes:
¤ unedited citations: these may be attributed or unattributed; from an
electronic corpus, a reading programme, or lexicographers' personal
files; from written or transcribed spoken material; from any text or
from texts of a particular period; from any material to hand or from
specific types of writing (literary, journalistic, specialist), etc.;
¤ adapted citations (options as above);
¤ composed examples (as in this paper).
3.5 Constraints on examples in a traditional dictionary
Space places the greatest constraint on dictionary examples. With the

exception of very large scholarly dictionaries, and (future) reference works
compiled specifically for electronic access, examples have to be short, and
few in number» Of the other constraints (many of them market-driven)
liable to bear on the formulation of the dictionary example, some of the
commonest options are:
¤ on form: it must be a complete sentence, or it must be a partial
sentence;
¤ on style: it must be "natural", i.e. as close as possible to "real" language,
or it must be "canonical", i.e. with verbs in the infinitive form and the
use of metalanguage such as one's etc.;
¤ on language: it must be intelligible to the user who does not know that
sense of the headword, or it must be couched in simple syntax, or it
must contain no offensive expressions;
¤ on content: it must show the headword in an "informative" context, or
it must contain nothing personally, politically or sexually controversial.
3.6 Examples in a frame semantics dictionary
We are now in a position to contrast systematically the role of the examples

in a traditional dictionary description of the language with those in a frame
semantics description.
There are two important features that all traditional dictionaries have
in common. The first is that the examples illustrate the headword in one
'sense' only. All references to a 'headword5 or 'word' in the discussion of
traditional dictionary examples are actually to a word-sense. In a frame
semantics dictionary entry, there is certainly no place for an example
predicated on the existence of traditional 'senses' of words.
The second feature shared by all traditional works is that the example
constitutes a distinct component within the structure of the dictionary entry;
more often than not, there is an 'examples section' with its own place in the
entry relative to the other components.
This need not be the case with the frame semantics dictionary, where
the examples constitute an integral part of the description of meaning and
appear whenever they are needed to illustrate a point. (This is possible
partly, of course, because there is no space problem in an on-line work.)
There will be no 'examples section' in such a dictionary; here, the
responsibility laid on the examples is very great: they must systematically
exemplify all the lexically relevant facts of the frame, such as those
mentioned in section 2 above.
The purpose of the example in a frame semantics dictionary is similar
to that of many traditional works: it must help users to understand the
words in the frame (to put at least some of them, in some uses, into their
passive vocabulary), or to use them (put them into their active vocabulary),
or both. Like its traditional counterpart, it must also provide a navigational
aid to users, telling them by its structure and content whether they have
reached a part of the entry likely to contain the information they are
seeking.
Here it is appropriate to distinguish between the decoding user (trying
to understand a usage) and the encoding user (trying to express a concept),
who differ from each other not only in the demands they make on the
dictionary, but also in their method of access to the information in it.
Decoding users would simply key in a question, probably in the form of the
context that the user has for the unknown word. The program would parse
the example offered and use the facts extracted from it to 'open' the
40 B.T.S. ATKINS
dictionary at the most appropriate point in the entry.

Encoding users would, however, get to the information in the entry by
quite a different route. Those who already had a word in mind might simply
search for local (word-based) facts within the entry of the relevant frame.
Those who started from a - possibly vague - concept that they wished to
express would find in the thesaurus (a concept-based hierarchy of frames,
as well as of the words in the frames) a way of narrowing the search to a
handful of expressions. Those who started from two or three possible
expressions, wishing to identify the most appropriate for the task at hand,
might come to the information via the entries of related frames.
The focus of the example in a frame semantics dictionary is both very
similar to and very different from that of a traditional dictionary example
described in section 3.2. Like the latter, it must clarify and exemplify facts
relating to meaning, syntax, collocation, morphology and other major
aspects of language use. However, it takes as its focus not a word in
isolation, but a word or a group of words in the context of a particular
semantic frame.
Of the possible functions of a dictionary example mentioned in section
3.3, only the first may be ascribed to the example in a frame semantics
dictionary. Its function is to clarify and exemplify, systematically, every
statement of fact in the entry, in the same way that the examples are used
in sections 2.1 and 2.2. Every fact would be accompanied by an example -
or rather, by many examples, for the user would be able to call up more at
will; each fact stated formally would be linked to the corpus usages
exemplifying it, and these would be prioritized so as to offer the user first
the simplest and easiest to understand, and later the more complex and
difficult cases. In this cumulative approach, the examples would be there
to widen the dictionary user's horizons.
Admissible sources of examples (see section 3.4) pose no problem for
frame semantics lexicographers: everything is grist to their mill. Examples
are chosen purely for their effectiveness: priority is given to clarity and
appropriateness. When it is available and suitable, an authentic citation
(normally, one would expect, from a large electronic corpus) would be
selected; often, however, it may be necessary to start off with a composed
example in order to emphasize a particular point.
Lastly, there would be few constraints (see section 3.5) on the examples
in a frame semantics dictionary. There are no space limitations, and the
overriding consideration is clarity. The form, style, language and content of
the example are constrained only by the fact that they must communicate
their message to all types of users. In an on-line dictionary, this need not
be too restrictive; the examples may be tagged according to prototypicality,
difficulty and other features, and the dictionary customized for each user as
it is installed, so that an individual is offered only material that he or she
can cope with.
4«, Conclusion
A contrastive examination of the examples in traditional dictionaries and in

a projected on-line frame semantics dictionary has allowed us to highlight
the essential differences between the two. The dictionary of today only tells
us what the word can do, and occasionally (when the statements in the
dictionary are founded on comprehensive corpus evidence) what it actually
does. The frame semantics dictionary would go much further than that: it
would aim in addition to explain also why words behave as they do, to
make explicit the fine nuances of sense distinction, to unravel implications
of selecting one expression rather than another» Whereas the traditional
dictionary approach to word meaning and usage fragments the data, the
new dictionary would be holistic in its approach, uniting the facts in an
entry in which the example is no longer an optional extra, but a mainstream
contributor to the analysis and recording of an integrated linguistic whole5„
References
Atkins, B.T.S., J. Kegl & B. Levin
1986 "Implicit and explicit information in dictionaries." In Advances in lexicology;
Proceedings of the Second Annual Conference of the UW Centre for the New
OED, 45-63. Waterloo, Ontario: University of Waterloo.
Fillmore, Charles J. & B.T.S. Atkins

1992 "Towards a frame-based lexicon: The semantics of RISK and its neighbors."
In A Lehrer & E.F. Kittay (eds.), Frames, fields and contrasts: New essays
I should like to thank Patrick Hanks and Krista Varantola for their helpful
comments on the first version of this paper,
42 B.T.S. ATKINS
in semantic and lexical organization. Hillsdale, New Jersey: Lawrence

Erlbaum Associates, 75-102,
1994 "Starting where the dictionaries stop: The challenge of corpus lexicography."
In B.T.S. Atkins & A. Zampolli (eds.), Computational Approaches to the
Lexicon, Oxford: Oxford University Press.
Raiffa, Howard
1970 Decision analysis: Introductory lectures on choices under uncertainty. Reading,
MA: Addison Wesley.
Shifting face from Asia to Europe1
Susan Ervin-Tripp, Kei Nakamura, and Jiansheng Guo

University of California, Berkeley
1. Introduction
In 1972, anthropological linguists Brown & Levinson presented a seminal

analysis of universals of politeness centered on the concept of face, "the
kernal element in folk notions of politeness" (Brown & Levinson 1987: 57).
Their analysis began with the English vernacular sense of face as reputation
or good name, expanding to a new conceptual distinction between positive
and negative face. Subsequently it was proposed by Matsumoto (1988) that
the notions of face found in Brown and Levinson did not correspond to the
Japanese sense of the concept. Matsumoto raised questions about the
universals which they had presented, with respect to the concepts of face
and the social dynamics they implied.
In this paper, we will explore the usage of terms related to this sense
of face in English, French, Chinese, Japanese, and Korean, and discuss
how the issues raised by Brown and Levinson in their analysis are treated
by speakers of several of those languages.
This paper could not have been written without the extensive collaboration of Seiko
Yamaguchi Fujii and Jeong-Woon Park, who supplied both syntactic and semantic
analyses. In addition, Jeong-Woon Park prepared the Korean romanization and made
important corrections to the data and their interpretation. The authors wish to thank
other advisers who gave us many examples of Chinese, Japanese, and Korean speech
practices: Dan Jun, Hea-Soog Jo, Hee Won Kang, Myn-Gyun Kwon Jeong, and Hongtu
Chen. Yoshiko Matsumoto also has made helpful suggestions. Undoubtedly we have
introduced mistakes for which these collaborators are not responsible.
The Chinese in the paper employs Mainland Pinyin, the Japanese uses the Hepburn
system, and the Korean uses the Yale romanization which gives a one-to-one correspon
dence to the Korean spelling system.
44 SUSAN ERVIN-TRIPP, KEI NAKAMURA, AND JIANSHENG GUO
We have been in this work taking the first steps to an analysis of the
concept of 'face' based on collocations, as suggested in the recent study of
'risk' by Fillmore and Atkins (1992). However, we are far from reaching
more than a crude preliminary analysis, focussing on obvious cultural
contrasts, and for Chinese, Japanese, and Korean we used consultants be
cause at this point we do not have access to on-line text searches in those
languages.
English
There are over a dozen senses given for face in modern English dictionar
ies, including in one subset appearance or semblance, disguise, assurance,
effrontery, dignity, and prestige. The role of the physical human face in
expressing emotion affected how the word was extended to other uses in
English in reference to people. In Middle English face could refer to "(a)
facial appearance or expression, beren god face (b) facial or outward
appearance as contrasted to some other reality; two faces in a hod,
double-facedness, duplicity; (c) traits of personality, character; (d) a person
as having certain traits, a personage or person." (Kurath 357).
The current Oxford English Dictionary (643) includes as listed meanings
"outward show: assumed or factitious appearance; disguise, pretence; an
instance of this; a pretext." The OED gives a variety of historical examples
for this set of meanings. Wyclif in the 1382 translation of Corinthians: "Hem
that glorien in the face and not in the herte;" in 1568 Grafton wrote
"They...made good face and shewe to fight with the Englishe men."
The face was often referred to by Shakespeare, in vivid descriptions,
typically to insult: false-faced, black-faced, whey-faced, tallow-face,
brazen-faced, shame-faced and antic face.
"Thou damned tripe-visaged rascal, an the child I now go with do miscarry, thou wert
better thou hadst struck thy mother, thou paper-faced villain." (Doll Tearsheet)
2 King Henry IV
"The devil damn thee black, thou cream-faced loon! Where got'st thou that goose
look?" (Macbeth)
Macbeth
Current newpapers speak of "red-faced officials" in the same vein. Though
these phrases carry implications for character, they are fundamentally
physical descriptions, and do not have the abstract sense of reputation
alone which we find in the more recent usages reported in the OED.
SHIFTING FACE FROM ASIA TO EUROPE 45
In the OED, the first evidence of the sense of face in the figurative
sense of good name or repute is attested by an 1876 quotation from R.
Hart in These from the land of Sinim: "Arrangements by which China has
lost face," The collocations given by the OED involve only losing or saving
face, and the etymology is given as a translation from the Chinese for loss
of face diu lian. In these uses, one loses or saves one's own face, not
someone else's.
A search of social conversation databases revealed no instances even
among educated speakers. The usage for this term currently is educated,
and primarily appears to be diplomatic or psychological. The term is used
frequently in social science writing, as indicated by a sampling of journal
abstracts from library abstract collections. About a third of these uses are
in quotes, suggesting that both face and face-saving are still marginal, special
terms, with a foreign or technical-jargon flavor. In this context, the term has
closely related meanings: defending honor, displaying strength, maintaining
or restoring self-esteem, remedying embarrassment, and negotiating with a
strategic sensitivity to appearances. In the quotations below, italics are
added.
"Policemen frequently are called to restore order to situations in which one of the
parties is fighting 'for his honor' in front of a crowd. ... While the crowd looks on, a
citizen may only be able to save face by attacking the officers themselves. A citizen's
honor - his reputation for implacability & vengeance - is vital to him." (Muir 1980).
"The courtroom context sanctioned the violation of normal social norms and allowed
the public display of anger and indignation in order to restore face," (Brown 1990).
"Social face is the image of strength that a person wants to project in conflict... Face
may be so valued because it is based on the desire to feel respected as a group
member.... Research indicates that people are motivated to avoid appearing weak and
try to restore face in culturally defined ways." (Tjosvold 1983).
The notion of face in these articles is one of unblemished reputation,

in which weakness or defeat cause problems to one's esteem with self or
others. Remedies to save the reputation or sense of self are said to save
face. These include excuses, justifications, revising history, avoiding
controversy, and lying. These are all references to saving one's own face.
"Therapists seek to help patients 'change their stories' while allowing them to save
face," (Erickson 1984).
"When bisexuality appears in an otherwise heterosexual individual, it does so with

compulsive force and gradually replaces heterosexuality. Self-labeling of oneself as
bisexual is held to be a matter of face-saving, status, and denial of conflict"
(Altschuler 1984).
"A study is made of face-saving' techniques in the context of a singles dance used to
avert potentially damaging evaluations by others. Since mere attendance at a singles
dance results in some loss offace, dance management & patrons strive to offset this
aspect of disrepute" (Berk 1977).
"Experiments [....tested Goffman's [...contention that embarrassment provokes

face-restoring behavior on the part of spectators to the embarrassment" (Levin and
Arluke 1982).
The last example is a rare instance of management of the face of another

person,
A slightly different perspective occurs in negotiation, where symbolic
rewards or accounts can be used instrumentally to mitigate concessions
which might be otherwise taken as revealing weakness.
"In such cases the granting of symbolic rewards and face-saving mechanisms to the
terrorists through the ritual of negotiation can lead to their capitulation." (Miller
1978).
The collocations in which the term appears are overwhelmingly save

face and lose face, suggesting that in typical usage in English face is not
viewed as like status, a variable feature which can be improved, but as
either present or absent. Nearly all the uses found were cases in which
people lost or saved their own faces, and even diplomatic face management
was seen as providing opportunities for parties to save their own faces.
French
The French expressions are perdre la face 'lose face' in the sense of lose
prestige from a serious attack on honor, dignity, or reputation, and sauver
la face 'save face' to safeguard prestige, and dignity in defeat or a setback.
Robert (1985) gives the origin as an 1850 French translation from Chinese.
Consultants regard the expressions as always concerning the agent's own
prestige, not another's, Robert's examples emphasize putting on a good
appearance:
"Ainsi il ne perdra pas la face (itaî. sic). Depuis le dernier coolie jusqu'au premier
mandarin, il s'agit de ne pas perdre la face, leur face de bois, mais ils y tiennent et
en effet, n'y ayant pas de principes, c'est la face qui compte,2 Henri Michaux, Un
barbare en Asie, p. 165.
"Les ouvriers désiraient retrouver leur métier et les patrons leur bureau. Il ne fallait
plus que sauver la face.3 A. Maurois, Bernard Quesnay, xviii.
"Alors, l'Allemagne, compromise malgré elle, n'a plus qu'un moyen de sauver la face:
paraître intransigeante, pour effrayer l'Europe.4 Martin du Gard, Les Thibault, t. VII,
p. 18." (Robert 1935: 352).
Borrowing
A puzzle in this history is why the Chinese expression was borrowed into
French and English at that time. In both countries there was an accelerating
fad for things and ideas Chinese which peaked in the mid-eighteenth
century, with the importation of tea, porcelain, silk, chintz, and wall-paper,
the copying of Chinese and Japanese decorative elements in Chippendale
furniture and Gobelin tapestries, and the design of Chinese-style rooms and
Chinese gardens,,
The Jesuit missionaries became interested in Chinese philosophical
traditions, particularly in Confucius, and in turn the philosophes were
attracted to the idea of a secular morality unlinked to religion (Ting 1928).
Chinese literary characters were created by European writers as a voice for
making cultural criticisms of the west (Appleton 1951). As is usual in
borrowing, assimilation to the borrowers' concerns revealed a quite
superficial acquaintance with Chinese culture, though Chinese literature was
beginning to be translated as early as the seventeenth century. Ting (1928),
in a survey of French views of China in the century before 1750, tells us
that the missionaries created a stereotype, a Chinese who was imagined on
the basis of Confucius to be soft and polite, who measured as he walked
2
Thus he won't lose face. From the last coolie to the first mandarin, it's a matter of
not losing face, their wooden face. But they stick to it, and indeed having no principles all
that counts is face.
The workers wanted to go back to their work and the bosses to their offices. All
that was needed was to save face.
Thus Germany, compromised in spite of herself, had only one way to save face: to
appear intransigeant to frighten Europe.
the length of his stride, the tilt of his head, the blinking of his eyes - in
short, an absurd exaggeration of politeness. Even when the Chinese notion
of face came to be translated it was taken up in a very restricted sense into
educated usage in the west.
2. Asian face
Face terms
As described above, in English and French the term face is mainly used in
the contexts of losing face and saving face. Unlike Indo-European languages,
Asian languages have a wide range of expressions involving the term face.
In this paper, we will focus on three Asian languages, Chinese, Japanese
and Korean. We interviewed native speakers of these three languages,
asking them the following questions:
1. Are there expressions in your language which correspond to the English
expressions 'to lose face' and 'to save face'? What are they?
2. In what contexts are they used? (e.g. who loses/saves face? under what
circumstances?)
3. Are there other expressions related to the concept of face which may not
have equivalents in English? How are they used?
4. What words can replace 'face' in these verbal frames?
5. What expressions would you use in these social contexts? (listing contexts
presented as calling for face maneuvers in Brown & Levinson, p. 65-68)
In all three languages, we found one or two terms for 'face' which have
overlapping functions. They are used to describe 'physical face/complexion',
'appearance/surface' and 'reputation/dignity'. In Chinese, the relevant term
for face is lian.5 In Japanese, there are two terms: kao or tsura. Kao is the
According to Hu (1944) mian is a very ancient word which acquired a figurative

meaning by the 4th century B.C. The other word lian came from the north by the
thirteenth century, gradually supplanted mian for the physical uses, and even acquired
standard term for face, while tsura is most commonly used in colloquial
men's speech. In Korean, there are also two terms: nach and elkwul, which
are used interchangeably. An indicator of antiquity is the fact that in
Japanese and Korean indigenous terms are used in expressing the term
face, rather than terms borrowed from Chinese, of the type we will discuss
later. In addition, most of these phrases involved verbs, not adjectival
modifiers. We will discuss expressions based on these literal terms first.
No face
Certain phrases exist in all three of these languages. All have expressions
for states of extreme humiliation and shame. For example, there are expres
sions which deal with the idea of lacking face': in Mandarin, mei lian, 'lack
face', in Japanese kao ga nai, and in Korean, nach-i epsta, 'there's no face'.
In all three languages, this phrase is used in the context of T've no face' (to
see you), or 'he lacks face', because of shame. In Japanese, one might say
kimi ni awaseru kao ga nai '(there's) no face (with which) to see you'. This
might be an apology when you fail to fulfill a promise to a friend, creating
disappointment. In Chinese, the equivalent expression might be used toward
someone whose trust one has disappointed.
In Chinese, mei lian 'he has no face' said of someone else is extreme,
and is used in very serious, primarily voluntary acts, such as theft, treachery,
adultery, pregnancy prior to marriage, and failure to pass the university
entrance examination. These are losses so serious that suicide may occur.
This phrase has been in use since the 4th century B.C. in its earlier form
(Hu 1944).
Most serious in Chinese is the accusation that the person does not even
want face, which is like a curse. Unique to Chinese, the expression used is
bu yao lian 'not want face'. This is used to describe people who are
particularly cheeky or brazen, who are selfishly and shamelessly unpredict
able (Hu 1944). It can describe a man who makes inappropriate advances
to a woman. A father can use this phrase, bu yao lian 'you really don't want
your face - how shameless!' to describe a daughter who becomes pregnant
before getting married. But it is most serious when directed to superiors,
some of the figurative senses of the older term. These are not just phonological variants
but different characters.
and like mei lian, is a very severe condemnation. It is targeted at people

who selfishly defy moral standards for their own benefit (Hu 1944).
A related expression is gei lian bu yao lian which means, 'I gave you
face the first time, but you didn't take it'. This could be used to describe An
who harrasses Bai on two separate occasions. Bai might give An a chance
by not reporting the incident the first time, but could use this phrase for a
second transgression. Note that here we see a person speaking of control
ling face for another.
Lose face
In each of these languages there is a way to speak of loss of face. In

Chinese, the expression diu lian (the source of the English and French
loanwords) is used in less serious contexts than mei lian, such as when an
employee makes an error, or when a student gives a bad presentation, or
when one commits the same error one criticized in another. The difference
implies loss of some face, not absence of all face. One can diu lian due to
ignorance or inexperience, so the loss of face is less serious than having no
face at all, or not wanting face, which is considered intentional. What is
considered bad enough to involve lack of face depends on both the status
of the person and the seriousness of the offense. Though our consultants
contrasted the severity implied in the two phrases, Hu (1944) argued that
for some speakers diu lian also could refer to publicly known infractions of
the moral code, such as fraud, meanness, unfaithfulness while in office, or
cheating a customer. These lead to public shame, and thus provide a means
for inferiors to maintain their rights by exposure of superiors to criticism,
making them lose face. In addition to vulnerability in cases of failure to
maintain high standards, superiors have more dignity to maintain, and
would diu lian if their demeanor is impaired by loss of self-control "As
Chinese see it, reserve makes for depth and restraint makes for strength."
(Young 1994). Educated or superior people lose face if they quarrel with
their inferiors, and men lose face if they argue with women. Each difference
in status from gender, education, age, or marital status brings with it
respect, obligations, and risks to face. In fact, public display of extreme
anger or emotion by an adult is always likely to involve diu lian.
Since face is a collective possession, one can lose face for the family, for
a teacher, for the college, or for their country, and children are instructed
early not to lose face for their family. "Self is extended to incorporate
others" (Young 1994). A country can diu lian by failing in its obligations to
weaker nations (Hu 1944).
A common Japanese expression is kao ga tatanai, 'can't save face', (lit:
face collapses) as in the context:
kore o shinakereba kao ga tatanai 'unless I do this, my face won't stand
up',
which means that one must do something in order not to lose face. The
verb tatanai can be said of a bottle, a building, or a wave, in the sense of
their vertical dimension.
Japanese losing face' is similar to diu lian in its reference to acute
embarrassment. In all three languages face terms can be used in causative
expressions, that is to make someone lose face. In Japanese, as in Chinese,
one can lose face for someone else, that is lose their face:
kao ga tsubureru 'one's face is crushed or collapses' (intransitive like 'a
house collapses').
kao o tsubusu 'to crush someone's face' (like crushing an insect)
kao ga tsubusareru 'to have one's face crushed by someone else', (like
a house smashed by a meteorite)
kao/tsura o yogosu 'to dirty someone's face',
kao ni doro o nuru 'to paint mud on someone's face'.
These Japanese expressions are used in cases of humiliation and dis
grace. Kao ga tsubureru can be used to describe Abe's face if he recom
mends Doi for a certain job, and Doi turns out to be completely incompe
tent. A 'black sheep' of a family can be referred to as the tsura-yogoshi
'face-dirtier' of the family.
In Korean, the similar phrases are:
elkwul-ey ttongchilhata 'smear dung on someone's face'
elkwul-ey mekchilhata 'paint someone's face' (as one would paint paper
or get paint on clothes).
An expression unique to Chinese is lian fangzai nar? 'where do I put my
face?' (lit: face place-at where) said by a person who feels shame. This
same phrase can be used use for a vase or a chair, or anything movable
rather than a face. Socially, this expression is similar in contexts to wo/ni
you shenma lianmian jian ren a! 'what kind of face can I/you look at people
with!'(lit: I/you have what face to see people!), These expressions can be
used if A does something which embarrasses B (e.g. a daughter breaking

a vow of marriage or marrying down is a source of shame for her parents).
Give face
We have already shown that in these languages face can be gained or lost
either by the face-owner or by someone acting in his/her behalf. The idea
of 'giving someone face' appears in these languages, but not in English or
French. In Chinese, one can say gei lian when one of a pair of two
long-standing enemies doesn't retaliate when expected. In this case, he may
claim that he 'gave (the other party) face' by not responding. A similar
expression is gei ta ge lian ba 'give him a little face', in which Chen can
attempt to persuade An to do a favor for Bai. But both of these expressions
are extremely patronizing, since they involve a concession by someone who
has the power or the right to act against someone but abstains.
There is also a negative version. If An gives Bai a present, and Bai
refuses it, then Bai has not given face to An. Also, if Chen asks An for a
favor for Bai, or for forgiveness of Bai, and An refuses, An has not given
face to Chen. If Chen asks An for a favor to himself rather than for Bai,
face is not involved in compliance, only in refusal. If An refuses the favor,
he does not give Chen face.
In Taiwan one can still say shang lian, which is'a humble way to say
'you bestowed face on me/you granted me an honor'. This expression can
be used in contexts in which a superior or colleague gives one signs of
attention, such as bestowing privileges, awards, jewelry, honors, or even
punishments. For example, the speaker would say it to a superior if he/she
attended a party the speaker held. Mainland speakers regard this phrase as
appropriate only for joking, because they consider it too obsequious or
feudal.
All of these expressions use common verbs involving possessions one
can lack, want, not want, lose, give, or bestow money, clothes, jewelry,
documents, or people. In Chinese face is thus treated as a possession to be
handled.
Japanese expressions seem to use a more architectural view of face (fall
down, stand up, raise). In Japanese, this idea is expressed as 'raising face',
kao o tateru in which someone is given a favor which leads to their
face-gain. For example, Tanaka can organize an event, but give credit to his
superior Suzuki. In this case Tanaka 'raises the face' of Suzuki. A PhD
student can give credit for her award-winning dissertation to her professor,
'raising' the professor's face. This is different from the instances in Chinese
in that the behavior is very humble and not patronizing at all, since it is
done by the lower person for the superior.
In Japanese, unlike English, there are both transitive and intransitive
forms of this verb. The expressions are kao ga tatsu 'one's face stands up'
and kao o tateru 'raise someone's face', which is similar to the concept of
'saving someone's face'. An example would be:
kimi no kao wo tatete ippai dake nomoo
'I'll drink one to raise up your face', i.e. 'I'll drink one for you.'
This expression might be used by a person who is reluctant to drink, but is
being pressured by his friend to do so.
In Chinese, there is an expression lou lian 'look good as a result of
receiving honor or praise' (lit: reveal one's face, let one's face appear),
which would be used in the context of a student being accepted into a
university, or winning an award. If lou lian is used with the third person, it
is considered a compliment - 'he revealed his face' by his achievements -
but if it is said to a friend, with the second person 'you showed off your
face', it would be taken to be a sarcastic remark or a joke about luck or
accident. The negative connotations of this phrase stem from the view that
conformity is desirable, and sticking out in a crowd is undesirable. When
talking about light, the phrase would be said if one leaked light during a
blackout. Expressions involving this verb suggest inappropriate or unwitting
revelation, as in revealing one's lie or revealing one's body.
One can also grow face, but only in a sarcastic context. Zhen zhang lian
'really grow face' might be said to a teenager directly in the second person,
or to a friend to criticize a third person in circumstances where the third
person showed off. The verb is not appropriate for plants, only for
abstractions like power or confidence.
Although English and French texts speak most often of face-saving, in
Chinese, Japanese and Korean, there are numerous expressions involving
the concept of losing face or lacking face, but fewer phrases about saving
face.
Thick/thin face
These languages each have expressions related to the image of a 'thick-skin

ned face'. The meanings are quite similar: Korean nach-i twukkepta,
'thick-faced, shameless' or nachkacwuk-i twukkepta 'thick face-skin,

shameless', Chinese lian pi hou, and Japanese tsura-no-kawa no atsui thick
face-skin, bold. In physical contexts, these terms for thick and thin can refer
to glass, clothes, paper, blankets or walls. These expressions are used to
describe a person who is cheeky, impudent, audacious, impossible to
embarrass, and attention-attracting, like Shakespeare's brazen-faced youth.
This phrase can be used to describe a person who borrows money, but
never returns it. It can also be used to describe a person who accepts food
on the first offer, who doesn't notice when it is time to leave, or who
participates too much in the classroom - a show-off. In general it refers to
someone who imposes and is rude. The impression that western business
practices violate traditional values of trust and face is expressed in a book
title: Thick face, black heart by Chin-Ning Chu, showing how business
machinations may be seen as calling for thick face.
Chinese also has an expression for a 'thin-skinned face' (lian pi .bao),
which describes a shy and good-mannered person, "who goes to a lot of
trouble to keep his good name" (Hu 1944).
Novel face phrases
In addition to these phrases found in all three languages, each language has
unique phrases, Chinese has an expression based on face as a location:
lianshang bu hao kan 'doesn't look good on the face; unattractive face',
which is used when something is embarrassing, involving a minor loss of
face, This expression could be used to describe a student who cannot find
a job teaching, who instead resorts to washing dishes at a local restaurant.
The phrase bu hao kan 'doesn't look good' is a common descriptive term
for evaluating the physical looks of people, furniture, or clothes. The phrase
lianshang mo bu kai 'can't spread things out on one's face' (lit: face-location
spreads (intrans.) unsuccessfully/ not fully) using a verb frame employed for
paint or plaster, is found in contexts in which one cannot say something and
must hide one's feelings. The image involves a face one would like to shrink
or hide in shame. For example, this phrase can be used to describe a
situation in which An owes Bai money, but Bai cannot ask for his money
back, because Bai will feel embarrassed to ask for his money, so he
lianshang mobukai.
The expression lianshang you guang 'the face has lights' is used when
someone has something to be proud of, such as passing the university
entrance examinations, or getting a job. If a child gets good grades, his

parents will feel lianshang you guang. This frame is appropriate for any
place that can be lit, so it treats the face as a location. Likewise, there is
also the expression lianshang mei guang 'the face lacks lights', which is used
when someone has done something socially undesirable.
In Japanese, there are many other phrases using kao, such as kao ni
kakawaru 'it's a matter of face', which is used in contexts in which
something threatens one's honor or is beneath one's dignity. (One could use
'life' in this verb frame, in the phrase it's a matter of life or death') It is
also possible to say kao ga ureru 'one's face sells', which is used when one
becomes famous, just as a book or a house sells. Similarly, kao ga hiroi 'to
have a wide face' is used to describe people who are famous and well-
known, with the same adjective used for roads, gardens, or hearts. Kao ga
kiku 'one's face is strong, effective' is used to describe people who have
influence and power (an adjective also describing horseradish or medicine).
Another expression is kaomake suru 'to do face-losing', which is used to
describe situations in which, for example, Tanaka is outshone or put to
shame by Doi. Tsura-ate 'hitting the face' is used to describe spiteful and
malicious words which one person might direct at another. Tsura-no-kawa-
o-hagu 'peel someone's skin from his face' is used when a person is put to
shame or unmasked and debunked. Kao kara hi ga deru 'fire's coming out
of his face' describes a person who is extremely embarrassed or angry, and
breathing fire.
An expression unique to Korean is elkwulkaps 'the price of one's face'
which refers to the worth of a person (as in the worth of a house which is
for sale). A few Korean speakers report hearing parents say 'elkwulkaps
com hayla!', meaning 'act the price of your face' or in this case 'act your
age!' In Korean there is also nach kancilepta, which means 'one's face
itches', or feels embarrassed. This would happen when a student feels
flattered.
In Korean youth slang, especially of males, there is a phrase for acute
embarrassment: ccok-pallita, 'one's face sells', an example being the case of
a some girls who are dancing in the street thinking no one is watching them,
but who find out that they are being watched, so their face is exposed. But
it is not clear if this slang term for face is productive.
Alternative forms for face or body
In addition to these expressions which involve the usage of the literal term
face, there is also a large number of other words for face, many of which
have lost their original meaning and are now used interchangeably as
equivalents for 'reputation', 'dignity' and 'honor'. The terms for face used
in these expressions in Korean and Japanese are not the indigenous literal
terms kao, tsura, nach and elkwul, but forms based on the ancient Chinese
word mian which have taken on a figurative meaning. This ancient word for
face is used very commonly in Chinese, with a wide range of meaning inclu
ding some concrete meanings like surfaces, directions, and perspectives. It
no longer describes the front of the human head except in some fixed
combinations. Its primary use is for figurative meanings such as 'social
skills/ ba mian linglong, (lit: eight exquisite faces (as in jade)). It is
important to 'consider one's face', (and one's group or family's face) gu
mianzi, by the behavior appropriate to status (Hu 1944), or one can be
tactful about other people's face by not drawing attention to their
weaknesses, just as one can 'consider one's home'.
Many of the same kinds of dynamics described above for lian also
appear in phrases involving mianzi, but mildly. One can lack face, lose face
or give face using this form. Examples are mei you mianzi 'not have face'
meaning lacking the prestige to reach a particular goal, such as an
encounter with an important person. This word also is used to describe very
minor losses of face, A specialist may diu mianzi if he/she is unable to
answer a question in his/her field of specialization. One can say mianzi
shang bu hao kan 'the face doesn't look good' when there is disapproval.
One can give face to another in this version, too. Gei mianzi is an
action of An to increase Bai's prestige in front of other people by public
praise or by apparent deference to Bai's advice - i.e. making Bai look good
in public (Hu 1944). A slight is bu gei mianzi, 'not giving face'. Mianzi can
describe sinecures, given symbolically or by kicking upstairs to incompe
tents. But mianzi also implies collegiality or an 'old boys club', in the phrase
da jia you mianzi or 'we have shared face', (lit: everyone has face). A
related phrase is he wo you mianzi (lit: also I have face) 'have face with me,
help intercede for me', In these relationships, each would be embarrassed
to refuse favors for the other. The two don't share the same face or an
equal amount of the face, but the relationship is good enough for one to
give face to each other, and later expect a return favor. (In this same
frame, one can share a feud.)
One can also 'borrow someone's face', jie mownourende mianzi to take
advantage of another's reputation, for example in getting their reference
for a position. A reputable person in authority can ask people not to fight,
out of 'regard for my face', kan wo de mianzi (lit: look-at my face) or one's
family in the same way. But in zhaogu wo de mianzi 'be considerate of my
face9, it could be a personal situation, such as when my daughter does
something wrong but people don't spread it about since they want to
protect my face. In the third person negative, mei gu wo de mianzi 'he
lacked consideration for my face' one can refer to someone's failure to
leave some face, for instance by criticism in front of one's children. The
difference is the formal vs family relationship involved, but in both cases the
abstract mianzi is used. One can 'add face' by generosity, 'leave someone
some face' not revealing their mistakes, or 'pad someone's face' by
deference.
Unlike English and French, in these three languages, we find very few
expressions which involve the concept of saving face, the opposite of losing
when face is treated as a single whole. In Chinese wanting face for oneself,
yao mianzi, is an attempt to keep one's public image or appearance, even
if by subterfuge or excessive display. A second or third person version is lia
dian mianzi (lit: save a-bit-of face) 'keep them some face, don't make them
lose all their face,' just as one might save them space. The failure to
mention errors saves face for another person, to protect his reputation. For
the same reason, a public official might be given an empty job (Hu 1944)
to 'leave him some face'.
What is the difference between lian and mianzi since the collocations
are so similar? According to Hu (1944) the latter is more sensitive to status,
style, and connections. Hu describes the first as like integrity, and the
second as like social clout or prestige, pointing out that "Robin Hood" or
generous bandits have lian but not mianzi. To lose lian is a serious matter.
Mentsu is the Japanese equivalent of mianzi. Related expressions
include in Japanese mentsu o ushinau 'to lose face'(as one can lose time),
mentsu ni kodawaru 'to be concerned about face', mentsu no tatsu
'face-saving' (lit: face standing up), and mentsu o tamotsu 'to preserve face9
(the verb for saving money).
In addition, there are the terms mianmu in Chinese, myenmok in
Sino-Korean and menboku in Japanese, all of which are based on the same
Chinese characters for face and eye. Although mianmu is no longer used
commonly in Chinese, myenmok and menboku are both used frequently in
Korean and Japanese respectively. In Korean, there is myenmok-i epsta 'no
face'. In Japanese, there is a large variety of phrases such as menboku o
kaifukusuru 'recover face' (like recovering health), menboku ni kakeru 'com
mit one's face; pledge one's honor, put one's face on the line' (like one's
life) and menboku o tamotsu 'maintain one's honor' (like maintaining an
income), menboku o usinait 'lose face' used when one has not met expecta
tions or done something illegal or unethical.
There is also a group of phrases focusing on reputation and honor as
represented by the whole body. These are based on the Chinese characters
for body-and-face. In Chinese, these characters are read timian, and are
used in fixed phrases such as shi timian 'incur a loss of face', (using an
archaic and rare verb). Weichi timian 'maintain face, keep up appearances'
employs a verb frame for abstract nouns such as 'stability' and 'peace',
Butimian de xingwei (lit: not face REL behavior) 'dishonorable conduct' or
'undignified behavior' is a frame in which evaluative adjectives like 'elegant'
or 'conforming' can be used to replace timian de.
Similarly, in Japanese, there are expressions such as taimen o tsukurou
'put up a good face, save face; pretending or mimicking', mibunsotoo no
taimen o tamotsu 'maintain an appearance worthy of one's position' and
ikka no taimen o yogosu 'dirty/discredit the family's face'.
The related Sino-Korean word is cheymyen 'body.face'. In Korean the
usual expression for a mild loss of face uses this form. Cheymyen-ul ilhta,
'lose face' (one can also lose a job or a child) or cheymyen-i epsta 'lack
face', 'be impolite'. If a man pursues a woman over-aggressively, he will lose
face, in this sense.
Causing a face loss to another person can be expressed in this way:
cheymyen-i sonsang-toyta
body.face-NOM damage-become
'face becomes damaged'
cheymyen-ul sonsang-sikhita
body.face-ACC damage-cause
'make someone lose face'
(This is used only for abstractions like achievement or reputation.)
cheymyen-ul yucihata
body.face-ACC 'maintain face' (as one maintains health)
These are vivid expressions for losing one's own face or having one's face
lost by someone else. The most extreme loss would be cheymyen-ul
kkakkanaylita 'cut down someone's face', meaning that I am getting you to
lose your face (just as I could get you to reduce a price). This could be
directed at a woman who has a fight with her husband and leaves her
in-laws' house to return to her parents' house.
Using this 'abstract face' expression, one can also talk of saving face:
cheymyen-i seta 'one's face stands, is saved.' or the causative version:
cheymyen-ul seywuta
'get one's face saved.' (lit: make stand one's body/face) (this verb could
also be used for plastic surgery raising the nose)
And one can ask to have one's face saved:
cheymyen com salyecwe
'please save my face' (lit: make my face alive, as in saving fish or
animals)
A different set of expressions in Korean involves the use of the
character for body, as in the case of mangsin-ul tanghata 'experience a
body-loss; be disgraced, suffer loss' (using an adversative passive, as in
'experience a divorce' or 'suffer a deception') and mangsin-ul sikhita 'make
someone lose their body/self; disgrace someone' (using a causative verb as
in 'make someone do an errand' or 'make someone study'). In the case of
these expressions, the concept is similar to face, but clearly refers to loss of
the whole body and person.
Finally, another set of Korean expressions dealing with the idea of face
is represented not by words for the face of the person but by the terms for
the front of the house. For example, in Korean, there are expressions such
as kamwun-ul pichnayta 'make the family/house shine' which are used if a
family member brings face to many generations (e.g. winning an award,
entering a prestigious university). Thus the front of the house, like the front
of the head, represents honor.
Social usage
In Chinese, Japanese and Korean, we find that gain/loss of face is related

to fulfilling social expectations and conventions. Actions, such as dressing
in a distinguished manner appropriate to one's status, can make one 'gain
face'. A computer expert might 'lose face' if a friend choses to ask another
person for computer help if the expert is available and willing. Further
more, an individual's self-esteem in.interpersonal situations is based on two
kinds of complementary evaluations: those of others and those of the actor
(Kirita 1984). Therefore, self-esteem is closely integrated with public face.
In situations such as defeat in competition, prohibition of access and
betrayal of confidence, an individual may lose public face and suffer injury
to self-esteem. Kirita states that "in every interpersonal situation, there is
a risk of losing public face and experiencing failure." Here, it is obvious that
it is "public face" before the other members of the group which is the main
concern.
In addition, one Chinese speaker described "loss of face" as an
expression which is used when people are unable to use a cover, or face,
to conceal the ugly parts of their own personalities. By not being able to
hide this part of themselves, they lose face. Similarly, in Japanese, there are
two related words omote and ura, which mean 'front/facade' and 'back/rear'
respectively. In classical Japanese, omote means kao (face) and ura means
kokoro (mind, heart) (Doi 1986). Here, too, face is related to the concept
of appearance and covering (of the unsightly/true aspects of one's persona).
A large number of contexts do not seem to bring loss of face, but rather
involve simple rudeness and inappropriate behavior. Loss of face involves
relatively serious transgressions of social norms and conventions. Inability
to control one's emotions, such as laughter or crying, might be considered
not as a loss of face, but rather as rude behavior, except for very distin
guished people. It is important to distinguish which behaviors fall into the
category of 'loss of face' and which fall into the category of 'rude behavior'.
In a study on intercultural communication blocks between Japanese and
non-Japanese, Natsuoka and Sakamoto (1981) discovered Asian (e.g.
Korea, China, Taiwan, Japan) attitudes towards politeness were relatively
similar. In these cultures similar behaviors appeared in specific contexts,
such as usage of polite expressions, indirect complaints, indirect refusals,
personal questions and suppression of personal opinions. For example,
expressions of humility, such as a host saying "so-cha desu ga...", which
means "this is not very delicious, but...", upon serving tea to a guest was
considered a polite formality in China, Korea and Japan. Our consultants
emphasized that humility is an important value taught from an early age.
Parents would not report praise to children by teachers, and boasting or
showing off is strongly disapproved.
In Asian cultures, face-defensive behavior is based on avoiding behavior
that risks causing one shame (Sugiyama Lebra 1974). People must exert
self-control and rely on ritual behavior in social interactions with others.
Direct self-exposure and spontaneity are frowned upon in adults. Often
negotiations will be conducted by or mediated through a third party in
order to avoid any loss of face of the two negotiating parties. In other
situations, a person might serve as a delegate for another party, in order to
avoid direct loss of face. Writing, as opposed to speaking directly, is also a
common tactic for avoiding direct self-exposure. Another crucial strategy in
Japan is the use of anticipatory communication, in which one party
'anticipates' the other party's needs and wishes by using omoiyari, or empa
thy. In this case, the second party never has to reveal the inattentiveness of
the first party by overt speech, as their needs are constantly met by the
omoiyari of the first party. For this reason, Doi (1973) claims that a
Japanese host, unlike an American host, would never tell a guest to "help
yourself"; a Japanese host would show sensitivity in detecting what the guest
might need (e.g. drinks, food) and would provide such things without even
asking. Clancy (1986) reports that omoiyari or empathy training by mothers
of children starts in conversational interaction as early as two years of age.
Sugiyama Lebra also discusses strategies for displaying face, which she
describes as "aggressive face". The purpose of displaying "aggressive face"
is to prove one's prestige. The speaker's aggressive face can be maintained,
promoted or demoted in relation to the hearer. The most socially accept
able way of doing this is by exhibiting conspicuous generosity. This can be
done in a variety of ways, such as entertaining guests lavishly, giving gifts
and offering help. The refusal of an offer of generosity can make both the
speaker and the listener lose face. This explains why people argue over
checks in restaurants. In Chinese, this sort of competition is called zheng
mianzi 'struggling for face' (struggle to hold face up). Another example is
the escalation of gift-giving in Japanese culture. Family A will give a gift to
family B, and family B will reciprocate by giving family A an even more
expensive and elaborate gift the next time around. Consequently family A
will feel pressured to find an even more exotic and unusual gift for family
B for the next occasion, and this evolves into a vicious cycle.
Another less acceptable strategy for displaying face in Japan is
self-praise, or jiman. People sometimes boast about certain things related
to themselves, such as their children. Often this is done in an extremely
subtle and hardly detectable manner, or is preceded by the phrase "jiman
ja naindesu keredo" (I don't mean to boast, but...). Excessive boasting is
considered inappropriate. Self-deprecation is valued as one is always seen
not as alone, but as part of a supporting group. One should always speak
in self-deprecatory terms which show explicitly that one has no selfish
delusions of independent grandeur. Finally, the most aggressive and least
sociable form of displaying face is arrogance. As in the case of self-praise
and boasting, this strategy is not culturally approved and is highly discour
aged, a sign of poor upbringing.
3. Face in Brown and Levinson
Brown and Levinson (1987) have proposed that every individual has two
specific kinds of desires: (1) negative face: the basic claim to territories,
personal preserves, rights to non-distraction (i.e. to freedom of action and
freedom from imposition) and (2) positive face: the positive consistent
self-image or 'personality' claimed by interactants. They define negative face
as "the want of every 'competent adult member' that his action be unimped
ed by others" and define positive face as "the want of of every member that
his wants be desirable to at least some others" ( 1987: 61-62). However,
Matsumoto (1988) questions the universality of the proposed constituents
of face in the light of Japanese language and culture. She claims that the
notion that all individuals desire to defend their own territory from the
encroachments of others is based on Indo-European concepts of individuals
and their rights, which "cannot be considered as basic to human relations
in Japanese culture and society" (p. 405).
According to researchers such as Nakane (1970), Doi (1973) and
Sugiyama Lebra (1976), what is central to the notion of face is not one's
own territory, but one's position in relation to others in groups and accep
tance by those others. As summarized by Matsumoto, all social interaction
is governed by "acknowledgement and maintenance of the relative position
of others, rather than preservation of an individual's proper territory"
(p.405). It is assumed that all individuals depend on other members of a

family, group or society. Nakane (1970) describes the Japanese social
structure as a "vertical society", in which a highly intricate system of ranking
has evolved between persons who are related hierarchically (e.g. parent-
child, senior member-junior member, supervisor-worker). This system
explains why the first thing that Japanese businessmen do is exchange name
cards. By exchanging cards, both parties can gauge the relationship between
them in terms of relative rank, locating each other within the hierarchy of
their society, and why, in a similar society, Koreans who meet must also first
exchange information to identify their relative age and status.
Doi (1973) claims that Japanese are "group-minded", strong as a group,
but weak as individuals. Consequently they strive to be accepted by others,
basing their behavior on the concept of amae, which indicates a feeling of
dependency and a desire to be loved and accepted, such as an infant's
feeling toward his/her mother. It is rooted in "a drive to dependence" and
the child's attempt to avoid separation from his/her mother. Sugiyama
Lebra (1976) clearly describes how individuals are interdependent within
the context of society, of which they are but a mere fraction. Japanese view
society in terms of mutual dependence rather than mutual independence
(Natsuoka & Sakamoto 1981). On the other hand, Westerners tend to feel
that to confess or show dependence indicates weakness and lack of
self-reliance and independence, and they view human activities and
accomplishments largely as the product of individuals. An English term,
co-dependency, has recently been invented to describe a condition consid
ered pathological, in which one person encourages another's dependency
on persons or substances. Japanese people must interact with others with
a full "awareness of the place assigned to one in a social group, institution,
or society as a whole" and must be willing to "fulfill all obligations attached
to that place" (Sugiyama Lebra 1976: 67). These include the responsibilities
of the superior toward the subordinate.
Therefore, Matsumoto (1988) concludes that in Japanese society, "loss
of face is associated with the perception by others that one has not
comprehended and acknowledged the structure and hierarchy of the group
(p.405)." Although she agrees with universality of Brown and Levinson's
concept of positive face, she claims that the concept of "negative face",
mainly because it involves the issue of defending territory or autonomy, is
alien to the Japanese. In our interviews with native speakers of Chinese and
Korean, we also found that in general this seemed to be true for them. The
notion that independence could be violated only exists between strangers.
While interviewing our consultants, we provided them with examples of
contexts presented by Brown and Levinson(1987: 65-68) of acts which
threaten negative face and those that threaten positive face. Examples
Brown and Levinson gave of acts which primarily threaten the addressee's
negative face, by indicating that the speaker does not intend to avoid im
peding the addressee's freedom of action include orders, requests, sugges
tions, advice, remindings, threats, warnings, offers, promises, compliments
and expressions of strong emotions toward the addressee. Examples of acts
which threaten the positive face of the addressee, by indicating that the
speaker does not care about the addressee's feelings and desires include
disapproval, criticism, accusations, complaints, insults, disagreements,
challenges, mention of inappropriate topics, boasting, non-cooperation in
an activity and use of incorrect address terms.
The speakers of our three languages reported that only some of these
contexts seemed to them to suggest problems of loss of face, despite the
centrality of this concept to social behavior. Note that this is NOT the
method Brown and Levinson used, which was to examine the occurrence
of remedies as a clue to whether face was threatened, though not which
kind of face. First, let us look at those that Brown and Levinson (1987: 65-
68) say threaten negative face because they impose on the hearer's freedom
of action.
1. "Those acts that predicate some future act A of H, and in so doing put
some pressure on H to do (or refrain from doing) the act" (B & L: 65)
(a) Making requests can lead to loss of face for a requester but only
when requests are excessive relative to social position. On the contrary,
failure to ask help of an appropriate person when you need help can make
the person who could help lose face; other people might conclude he is
stingy. That is, far from always involving a threat to the face of a hearer,
making requests can be a way of honoring someone by giving them a
chance to carry out their role - a point made vividly by Matsumoto.
Correspondingly, thanking an insider, who would normally normally give
assistance, is demeaning to their face, and distancing. "Utterances of praise
or expressions like 'thank you', 'please' or 'excuse me' work to create
distance, subtly subverting the mutual support and consideration taken for
granted among family intimates" (Young 1994.)
(b) Giving advice normally seemed to our consultants to be unrelated

to face, since it is the role of specialist to give advice to a non-specialist
(e.g. supervisor to a worker, or a teacher to a student), though giving advice
can sometimes be rude. Volunteering advice could involve face if the
situation brought humiliating circumstances to light, for example, embarrass
ing status changes (e.g. a financial advisor giving advice to a bankrupt
client). Offering advice by a non-specialist to a specialist (e.g. a student
volunteering advice to a teacher) appears to our consultants to lose face for
both, though in some cases a request for the advice of a student would gain
face for the student.
2. "Those acts that predicate some positive future act of S toward H, and
in so doing put some pressure on H to accept or reject them, and possibly
to incur a debt" (B & L: 66)
Like a request, an offer is seen by Brown and Levinson as a potential
intrusion on freedom of action. An offer has to be evaluated in the context
of its symbolism of social relationship. There are very strong social customs
surrounding offers of food, for example, which may be distinct from other
situations of offering. In some situations, offers are refused until they are
repeated several times, so that the acceptance does not imply over-
eagerness by H. The excuses offered help calibrate the intentions of both
parties. An H who persistently rejects a strong offer, for instance of a gift,
makes S lose face since the cycle of favors will keep people in the same
social circle. Rejection thus implies social rejection. In this analysis we
would expect social remedies in H's excuses to mitigate the threat to S's
positive face.
3. "Those acts that predicate some desire of S toward H or H's goods,

giving H reason to think that he may have to take action to protect the
object of S's desire, or give it to S" (B & L: 66)
(a) Making compliments is usually not face-threatening to H. and is met
with modest demurrals. However, in China, a compliment to a person of
different gender means the speaker 'does not want face' (is shameless). In
China if one compliments another's possessions in some situations the
owner may offer to give them, so as not to lose face by appearing to be
materialistic; for the same reason the recipient of the offer is obliged to
refuse.
(b) An outburst of anger in some situations is just rude or inappro

priate, and not a loss of face for the addressee, unless the receiver was
negligent in his duties. But if the angry person is of high status, face is lost
because of loss of public demeanor. Any extreme emotion is face-threaten
ing for a high-status person displaying it publicly, unless it is done within
one's role, for instance anger at a subordinate for not performing services
well for a client, or anger at a servant for negligence to a guest.
Many of the acts which Brown and Levinson view as intrusive are not
so viewed, and the face they may affect is the positive face named by the
terms we have discussed. Since actors are viewed within a system of mutual
obligations, the only cases of intrusion are those of excessive demands going
beyond what is due.
There is more agreement about positive face threats, but some of these
acts are rude, and lose face for the speaker rather than the hearer.
1. "Those that show that S has a negative evaluation of H's positive face"
(B & L: 66):
Public criticism (except in politice) is likely to be avoided or and highly
mitigated. But accusations in some contexts can involve public face or
prestige, gaining face for the accuser and losing it for the target.
"Putting thoughts vaguely or obliquely helps to mitigate threats to face and person
[...] Chinese regard one's ideas as entangled with one's identity or sense of personal
worth; an attack on one's ideas is therefore an attack on one's self, or, more
specifically, one's face. In this event, Chinese try to minimize threats to face by
hedging one's claim while looking for commonality with others. Chinese avoid a
strong stance when criticizing another's argument." (Young 1994).
Admissions of guilt, or apologies involve loss of face by the speaker, but

the acceptance of an apology can save the apologist from further loss of
face. A superior may save face for a hearer by initiating an excuse for him
and thus mitigating criticism.
2. "Those that show that S doesn't care about (or is indifferent to) H's
positive face" (B & L: 66):
(a) Expressions of violent emotions lose face for S.
(b) Boasting implies a 'thick-skinned face,' and is thus a defect in the
boaster, and not in the hearer. The boaster loses face.
We also turned the situation around, and asked whether English

collocations for face made sense. From a Chinese point of view, certain
examples of the use of the term face in English seem inappropriate.
Americans talk about mediation or certain types of therapy as saving
children's face. In China children do not have face; their behavior reflects
on their families, however. An outburst of anger in a courtroom could not
restore face, as P. Brown proposed, unless it was done coolly, with "style,
manner, eloquence and reason." One can only gain face by acting appropri
ately; any sign of excess brings loss of face. When one has failed in a
responsibility, been late, made a mistake, giving excuses would not gain one
face as much as would taking responsibility with an explicit apology.
English speakers simply have very restricted and relatively unproductive
use of the notion of face, and do not speak of giving face, having face,
having no face, borrowing face, padding face or having a thick face. English
usage appears to treat face as given to the individual in childhood, like
good name or good health, and capable only of being lost, restored, or
saved. American consultants did not regard the deeds of family members
as relevant to their own face, which is individualized, and they regard
concern with face as superficial, as attention too much to image, an attitude
also reflected in the French quotation stating that Chinese are not
concerned with principles but face. In Asia, face is absent in childhood, and
accrues only in adult life, increasing with status so that a high-status person
has more to lose. Because face is shared within collective units like families
or companies, and between teacher and pupil, changes in face can be
important to social relations and to social control and are likely to be talked
about a lot.6
Such cultural differences illustrate the variability of cultural elaboration
acknowledged in Brown and Levinson's theory (e.g., what kinds of acts
threaten face, what sorts of persons have special rights to face-protection).
What is most striking about the interpretations consultants gave of the
Brown and Levinson examples is that virtually all the acts which were
described as threats to autonomy and territory were interpreted as rude and
In journalistic Chinese, mianzi, which is unambiguously abstract in meaning, occurs

relatively often, with a rank of 6650 in lexical frequency, according to Chu-Ren Huang, of
the Institute of History and Philology, Academia Sinica and Project Co-Director of the
Chinese Knowledge Information Processing (CKIP) Group. To get a sense of this rank,
Kucera and Francis (1967) found a similar frequency in English for sky, scream, mix, rug
and map.
irrelevant to face, or as affecting the positive reputation of the speaker or

the hearer.
In Korea and Japan in particular, there seems to be an extreme effort
to protect the addressee's positive face, in addition to the speaker's own
face, Sugiyama Lebra (1976) discusses how interaction in Japanese society
is characterized by mutual discretion as well as self-defense. She raises the
example of why Japanese customers in a restaurant or store don't count
their change. Such an action would hurt the feelings of the waiter or
cashier, by revealing the customer's mistrust and causing the waiter or
cashier to lose face. Similarly, a Korean customer in a restaurant may not
complain about lukewarm soup because that would make the waiter lose
face. Brown and Levinson (1987) state that the motive of speakers is to
protect the addressee's face, just as much as their own. In these examples,
tact in protecting the self-esteem of others illustrates their point, but only
with respect to positive face.
4. Conclusions
The extension of face to mean honor or prestige in English and French

occurred in the nineteenth century, but it has remained very limited both
in users and in contexts of use. The current influx of Asian immigrants
translating phrases into English and bringing their native cultural concepts
about family and about social responsibility could bring change. In contrast,
both literal and 'abstract' terms for face are shown to have a great diversity
of contexts, and often similar usages in Chinese, Korean, and Japanese,,
These uses are shown to arise from the emphasis on maintaining integrity
and honor for oneself and one's group, and on recognizing mutual obliga
tions in social relations in these countries. This was not a sociolinguistic
survey; we merely touched the surface of these usages without exploring
social differences within these speech communities, or between the three
Asian languages.
Our interviews with a small sample of consultants confirmed that the
reasoning of Korean, Chinese, and Japanese speakers about many speech
acts concerned protecting the speaker's and the hearer's positive face - their
prestige and self-esteem. Appropriate behavior usually had to be judged in
terms of the relations and mutual obligations of speaker and hearer. The
relation of these beliefs to actual linguistic behavior remains to be studied

through systemic pragmatic research.
References
Altshuler, Kenneth Z.
1984 "On the question of bisexuality." American Journal of Psychotherapy
38.484-493.
Appleton, WW
1951 A cycle of Cathay. New York: Columbia University Press.
Berk, Bernard
1977 "Face-saving at the singles dance." Social Problems 24.530-544.
Brown, Penelope
1990 "Gender, politeness, and confrontation in Tenejapa." Discourse Processes
13.123-141.
Brown, Penelope & Stephen Levinson

1987 Politeness: Some universals in language use. Cambridge: Cambridge
University Press.
Chu, Chin-Ning
1992 Thick face, black heart. Mill Valley CA: A.M.C. Publishing.
Clancy, Patricia
1986 "The acquisition of communicative style in Japanese." In Bambi Schieffelin
& Elinor Ochs (eds.), Language acquisition and socialization across cultures.
Cambridge: Cambridge University Press, 213-250.
Doi, Takeo
1973 The anatomy of dependence. Tokyo: Kodansha.
1986 The anatomy of self: The individual versus society. Tokyo: Kodansha.
Erickson, Richard C.
1984 "Tell me a story." Individual Psychology Journal of Adlerian Theory, Research
and Practice 40.277-289.
Fillmore, Charles J. & BT. Atkins

1992 "Towards a frame-based organization of the lexicon: The semantics of RISK
and its neighbors." In Adrienne Lehrer & Eva Kittay (eds.), Frames, fields,
and contrasts: New essays in semantics and lexical organization. Hillsdale:

Lawrence Erlbaum, 75-102.
Hu, Hsien Chin

1944 "The Chinese concepts of 'face'." American Anthropologist 46.45-64.
Kirita, Katsutoshi
1984 "Tasha ni yoru kyozetsu to jisonshin." [Rejection by others and self-esteem.]
Soshioroji 29.69-86.
Kucera, Henry & W. Nelson Francis

1967 Computational analysis of present-day American English. Providence: Brown
University Press.
Kurath, Hans (ed.)

1956 Middle English dictionary. Ann Arbor: University of Michigan Press.
Levin, Jack & Arnold Arluke

1982 "Embarrassment and helping behavior." Psychological Reports 51.999-1002.
Matsumoto, Yoshiko
1988 "Reexamination of the universality of face: Politeness phenomena in
Japanese." Journal of Pragmatics 12.403-426.
Miller, Abraham H.
1978 "Negotiations for hostages: implications from the police experience." Ter
rorism 1.125-146.
Muir, William Ker, Jr.

1980 "Power attracts violence," Annals of the American Academy of Political and
Social Science 452.48-52.
Nakane, Chie
1970 Japanese society. Berkeley: University of California Press.
Natsuoka, Reiko, et al.

1981 Mutual understanding of different cultures. Tokyo: Tdishukan.
Robert, Paul
1985 Dictionnaire alphabetique et analogique de la langue française. 2nd ed. Paris:
Le Robert.
Simpson, John A. & E.S.C Weiner (eds.)

1989 The Oxford English dictionary. 2nd ed. Oxford: Oxford University Press.
Sugiyama Lebra, Takie
1976 Japanese patterns of behavior. Honolulu: University of Hawaii Press.
Tjosvold, Dean
1983 "Social face in conflict: A critique." International Journal of Group Tensions
13.49-64.
Young, Linda
1994 Crosstalk and culture in Sino-American communication. Cambridge: Cam
bridge University Press.
Mental-space builders:
Observations from Japanese and English
conditionals1
Seiko Yamaguchi Fujii
University of California at Berkeley
and University of Illinois at Urbana-Champaign
1. Introduction
While English grammar easily accommodates mono-clausal conditional

expressions (to be discussed below), the Japanese counterpart often
requires conditional clause-linking morphology and syntax. This contrast in
the mechanisms for expressing conditionality is systematic, and reflects a
general divergence between the two languages. The purposes of this paper
are i) to examine English mono-clausal conditional sentences, as illustrated
in (1) through (3) below, paying particular attention to devices for building
mental spaces in the sense of Fauconnier (1985) (Section 2); ii) to examine
the corresponding expressions in Japanese, and explore the structural
divergences between English and Japanese in the preferred mental-space
building mechanisms (Section 3); and iii) to suggest several typological
characteristics which may be responsible for the observed divergences
This paper stems in part from a project on English and Japanese conditional
sentences on which I collaborated with Charles J. Fillmore in 1987-88; I have benefited
immeasurably from discussions with him; however, I am responsible for any errors or
omissions found in this paper. A slightly different version of this paper was presented at
the LACUS Forum, in Montreal, August 1992, and awarded the LACUS Presidents'
Predoctoral Prize. My thanks also go to Eve Sweetser, Gilles Fauconnier and Sandra A
Thompson for their encouragement.
74 SEIKO YAMAGUCHI FUJII
(Section 4).
(1) In France Watergate wouldn't have been a scandal.
(Fauconnier)
(2) You are so lucky. My advisor wouldn't have been so patient.
(3) A new highway would save us a lot of time.
2. English mono-clausal conditional sentences and mental-space builders
In examining conditional sentences in English, we most often focus on bi-

clausal sentences using the IF-THEN construction, such as 'If you had been
a little more careful, you would have avoided the accident,' where the IF
subordinate* clause constitutes the antecedent (protasis) for the consequent
(apodosis) presented in the main clause. But, of course, both in Japanese
and English, it is not necessary for conditional relations to be expressed by
two clauses linked by conditional clause-connectives (e.g., IF, IF ONCE etc.).
There are several types of mono-clausal conditionals. In one type, the
antecedent is presented in a certain formulaic manner with a characteristic
prosody, and the consequent is elided, as seen in (4):
(4) If only it weren't raining
Tada arne sae hutte inakereba
only rain at least fall-ASP-NEG-COND(lF)
Both the English and the Japanese utterances are to be understood here
as an expression of wish, with either positive consequence or the speaker's
positive evaluation being omitted at the end of the utterance. The choice
of the particular conditional construction in either case (e.g., the use of the
IF ONLY construction; the use of the (R)EBA conditional construction 2 in
combination with SAE) ensures this conventional interpretation even without
the consequent clause; such a construction counts as idiomatic in both
languages. 3 .
2
Devices for expressing conditionality in Japanese include particles such as to, and
nara, verb inflections such as -tara, -(r)eba, and -vmo, and the function noun toki (time)
and baai (case) optionally followed by the case-marker, ni, and/or the topic marker, wa.
3
A more detailed discussion of the idiomatic nature of conditional constructions can
be found in Fujii (1992).
MENTAL-SPACE BUILDERS 75
A second type is found where the antecedent and the consequent,

either explicit or implicit, are segmented across sentences, or across turns,
in discourse. For example, an antecedent can be interrogative, as seen in
(5) below (Jespersen 1940; De Castro Campos 1981; Bowerman 1987; and
Fillmore 1987).
(5) You want it? It's yours. (Fillmore 1987)
De Castro Campos (1981) and Bowerman (1987) note that such interac
tions in the domain of child language — where the child asks a question, the
mother confirms it, and the child then uses this agreement to draw out a
plausible inference — conform to 'the question-confirmation-implicative
assertion sequence' hypothesized by Jespersen (1940) as underling the
historical development of IF-(THEN) conditionals in English.
Another type of mono-clausal conditional occurs where adverbs, such
as OTHERWISE and THEN, are used as non-clausal antecedents to refer to
and summarize the condition given in the context (see 6a and 6b below).4
(6a) Otherwise, I won't go.
(6b) Then. I'll go.

Dattara. / Nara, watasi ga ikimasu.
Dewa, / Zvaa, watasi ga ikimasu.
The Japanese counterpart of THEN is sentence-initial DATTARA, NARA, and
DEWA. DATTARA and NARA, etymologically conditional forms of copula
like predicates, are normally used as bound clause-linking morphemes.
Here, however, they have developed into grammatically unbound (free)
morphemes, and function as sentence-initial connectives. Likewise, DEWA,
etymologically a particle cluster, is here used as a free sentence-initial
connective.
The types of English mono-clausal conditionals that we focus on here
are of yet another type, demonstrated in sentences (1), (2) and (3),
repeated below:
Fillmore (1987) discusses these circumstantial adverbs as one of two kinds of

nonclausal antecedents. The other discussed is the type where 'the consequent is
introduced by a conjunction, usually AND', and the antecedent is 'a quantified expression
indicating the extent by which some unnamed changes would have a certain result', as in
(i) below:
(i) One more stupid mistake like that and you'll be fired.
(1) In France Watergate wouldn't have been a scandal.

(Fauconnier)
(2) You are so lucky. My advisor wouldn't have been so patient.
All of these sentences express conditional relations without using the IF-
THEN construction; and, more importantly, each contains an overt
expression of the condition within the mono-clausal sentence (i.e., without
containing an explicit clausal antecedent). In each of these sentences, either
the circumstantial adverbial phrase (e.g., 1) or the subject (e.g., 2 and 3) —
here underlined — delineates the possible world in which the rest would
hold true, thus fulfilling a function similar to that served by the IF clause
in the IF-THEN construction. Although similar meanings can be expressed
using an IF-clause, as shown in paraphrases IP through 3P, English often
seems to prefer the mono-clausal version.
(1P) If it had happened in France, Watergate wouldn't have been a
scandal.
(2P) If my advisor were in your advisor's position, he wouldn't have
been so patient.
(3P) If a new highway were built, it would save us a lot of time.
The following three subsections amplify on these types of English mono-
clausal conditionals with respect to the devices used for building 'mental
spaces' in the sense of Fauconnier (1985). Briefly, a mental space is a
medium in which thoughts occur and in which conceptual entities are
located. This informal characterization will suffice for the following
discussion.5
2.1. Adverbial phrases as mental-space builders
In (1), (7), and (8) below, the adverbial phrase, here a prepositional phrase,
expresses the state of affairs in which the proposition contained in the rest
of the sentence holds true. These adverbial prepositional phrases are
'frame-setters': They set up a locational or situational frame of reference,
5
Fauconnier's (1985) discussion of mental spaces includes detailed analysis of
conceptual entities which appear in different mental spaces, and the mechanisms of how
each conceptual entity is mapped from one space to another.
taking the place of a clausal conditional antecedent. In Fauconnier's (1985)

terms, they are 'mental-space builders', for they build 'mental spaces'.
(1) In France Watergate wouldn't have been a scandal
(7) Without her help, we couldn't have found it.
(8) Prices should come down under free competition.
(8') Prices, which should come down under free competition, do not
decrease when they are bound hand and foot by controls and
restrictions.
Tensei-jingo: Vox Populi, Vox Dei, 1989, Asahi Newspaper.
In (1), the adverbial phrase in France sets up a hypothetical space

where we are asked to imagine that the Watergate incident had happened
in France rather than in Washington. Likewise in (7), the phrase without her
help sets up a hypothetical space distinct from the real space where 'her
help' was indeed provided. The original sentence from which sentence (8)
was adapted is shown in (8').6 The reality discussed in this text is that
prices (in Tokyo) 'are bound hand and foot by controls and restrictions',
and thus 'do not decrease'. The adverbial phrase, under free competition, in
(8), sets up a hypothetical space, presenting contrastive circumstances
where prices should come down.
Note that sentences (1), (7) and (8), as well as other English examples
we will look at shortly, all have an explicit conditional modal of the kind
which appears in normal consequent clauses; it is this modal which marks
the sentence as a conditional, and which ensures that the spaces set up and
described in these sentences are hypothetical (or counterfactual).
2.2. Counteridentical elements in contrast as mental-space builders
We turn next to mono-clausal conditionals where an element in contrast in

the sentence introduces a hypothetical space, by taking the place of another
entity (either person or object). For example, in sentence (2), the subject
— MY advisor — presents a hypothetical mental space, by substituting for
somebody else ('your' advisor) in the other individual's situation. In other
words, the speaker's advisor ('my advisor') is hypothetically introduced into
the hearer's advisor's situation.
This is taken from the published English version of a Japanese original; the
Japanese sentence appears as (13) below.
(2) You are so lucky. MY advisor wouldn't have been so patient.

As noted by Fillmore (1987), the hypothetically introduced individual, 'my
advisor9 in this case, has prosodic salience.
This is a variety of the counterfactual sentence type which Goodman
(1983: 6) refers to as counteridenticals (cf. Fillmore 1987). Both Goodman
and Fillmore discuss bi-clausal counteridentical sentences such as 'If I were
Larry Holmes, I would beat you up'; here the IF clause builds a hypothetical
mental space and clarifies who (which element) substitutes for whom (which
element) in the mental space. But even in a mono-clausal sentence such as
(2), the prosodically salient subject can have a similar function.
This type of mental-space builder is most commonly found with a
counteridentical subject. But, other elements of a sentence, such as the
object or oblique, can present a hypothetical mental space in the same
way,7 as shown in (9) and (10) below.
(9) (She did it to you?!) That could never have happened to ME.
(10) (I didn't take his offer, but) I would have taken YOUR offer.
2.3. Subjects in causative constructions as mental-space builders
In causative constructions, the subject, which describes the cause, constitutes

the condition for the rest of the sentence, which describes the effect. For
example, in sentence (3), the qualified subject, a new highway, presents the
condition for the rest of the sentence, once again taking the place of the
clausal conditional antecedent. 8
(3P) If a new highway were built, it would save us a lot of time.
Note in passing that when the subject has a definite article (see 3'
below) instead of an indefinite article, the hypothetical scenario that we
construe is different, as shown in its bi-clausal paraphrase (3P').
(3') The new highway would save us a lot of time.
(3P') If we took the new highway, we could save a lot of time.
This was pointed out to me by Charles J. Fillmore.

8
The causative version, of course, might not be as detailed as the bi-clausal
conditional version. Thus, (3P) is just one possible paraphrase of (3).
Another piece of data (11) comes from a T.V. commercial for a new
cereal, called Nutri-Grain Nuggets, which is compared with a similar brand,
Post Grape Nuts.
(11) <In a T.V. commercial advertising 'Kellogg's Nutri-Grain
Nuggets'>
<the initial remark>
(11a) Once you taste new Kellogg's Nutri-Grain Nuggets, you'll forget
about Post Grape Nuts.
<the final remark>
(11b) The taste will make you forget your old nuggets.
What is interesting about the discourse in this commercial is that it initially
uses a bi-clausal sentence with the connective ONCE (11a), to relate the
condition (tasting the new brand of cereal) and the result (forgetting about
the old brand of cereal); by constrast the concluding remark, shown in
(11b) uses a causative construction to repeat the same idea. This causative
sentence can be paraphrased as Tf you taste Nutri-Grain, you will forget
your old nuggets.' The two sentences exemplify the fact that English often
allows flexible alternation between these constructions.
Causality is an indispensable issue in the study of conditionals (cf.
Mackie 1965). The question of how and what kinds of causal meanings
figure in various conditional sentences has thus been one of the central
issues in the study of conditionals (see e.g., Comrie 1987; Sweetser 1990).
Causative constructions, of course, also express cause-effect relations. The
examination of these two formally different grammatical constructions,
namely bi-clausal conditionals and causatives, should prove a fruitful area
of inquiry, especially when we compare preferred event representation
patterns crosslinguistically (see Section 4,3 below).
3. Japanese devices for building mental spaces and their characteristics
We turn now to Japanese conditional sentences, to examine the Japanese

ways of building mental spaces and expressing conditionality. Japanese
differs from English in that it prefers different grammatical mechanisms to
construct sentences corresponding to the English sentences (1) through (3),
(7) through (10) and (11b). The tendency is for the Japanese counterparts
of English mono-clausal conditionals to generally involve two clauses and
to make use of explicit conditional clause-linking morphemes.
The Japanese counterparts of English sentences (1), (2), (3), (8), and
(11b), discussed earlier, are given in (12) through (16) below. (12) and
(13) are the counterparts of English sentences (1) and (8), in which
prepositional phrases set up the mental spaces. (14) is the counterpart of
the counteridentical case, (2). (15) and (16) are the counterparts of the
causative cases, (3) and (11b).
(12) Huransu de dattara wootaageeto-ziken wa
France-IN copula-coND Watergate-incident-TOPic
mondai ni naranakatta (daroo) yo.

problem-DAT become-NEG-PAST (MODAL) Particle
If it were in France, Watergate wouldn't have been a scandal.

(cf 1) In France Watergate wouldn't have been a scandal.
(13)9 Kakaku wa ziyuuna kyoosoo no moto de nara sagaru hazu da.

price-TOPIC, free, competition-GEN, under, copula-coND, come down, should
If (they are) in free competition, prices should come down.

(cf. 8) Prices should come down under free competition.
(13a) Kakaku wa ziyuuna kyoosoo no moto de wa sagaru hazu da.

price-TOPic, free, competition-GEN, under, TOPIC, come down, should
(14) Boku no adobaizaa dattara (nara) kessite sonnani

my-GEN, advisor, copula-coND, never, to that extent
gaman wa site-kurete-inakatta (daroo) yo.

'be patient for me-ASP-NEG-PAST, (MOD) Particle
If Tit had been) my advisor, he/she wouldn't have been so

patient.
(cf.2) My advisor wouldn't have been so patient.
9
The original Japanese sentence from which this sentence was adapted is given
below:
Ziyuuna kyoosoo no moto de nara sagaru hazu no kakaku mo, kisei de ganzi-garame
no zyootai dewa sagaranai.
- Tensei-jingo; Vox Populi, Vox dei, 1989, Asahi Newspaper.
(15) Atarasii haiuee ga dekireba zuibun zikan no

new, highway-NOM, be built/be completed, a lot of time-GEN
setuyaku ni naru (daroo) (ne).

saving-DAT, become, (MOD) (Particle)
(If a new highway were built, it would be a lot of time-saving.)

(cf.3 A new highway would save us a lot of time.
(16) Kono axi o sireba, mukasi no nagetto nado

this taste ACC know-coND old nuggets
warurete-simau (daroo).
forget (MOD)
If (you get to) know this taste, you will forget your old nuggets.
(cf.11b) The taste will make you forget your old nuggets.
Note that all of these Japanese sentences contain conditional markers,
which are used as clause-linking morphemes (i.e., TARA in (12) and (14),
NARA in (13), and (R)EBA in (15) and (16)). General statements such as
sentence (13a) could be expressed with the topic marker WA, instead of
NARA. But it would be difficult for the other sentences to express
'conditionally' without marking the conditional adverbial adjuncts with
these conditional linkages.
I must add here that it is not simple to determine the syntactic status
of the adverbial adjuncts in these Japanese conditional sentences. For
example, do we analyze 'boku no adobaizaa dattara9 in (14) as a genuine
subordinate clause, or a constituent of the main clause? To a degree, the
answer depends on the syntactic theory which we use to describe the
various linkage types.10
But aside from the precise syntactic status of the different types of
linkage (which we need to explore in another context), the point to be
10
For example, to capture the difference in syntactic status between 'boku no
adobaizaa dattara' in (14) and 'Atarasii haiuee ga dekireba' in (15), Role and Reference
Grammar (Foley & Van Valin 1984; Van Valin 1984) might treat the former linkage at
the 'core' level while treating the latter at the 'clause' level. The hierarchical theory of
linkage in Mihami's Grammar (Minami 1964,1974) might categorize the former as a 'Type
A linkage', and the latter as a 'Type B linkage'.
made here pertains to Japanese and English devices for building mental
spaces: Japanese, by using conditional linkage, marks the mental-space
builder (antecedent), whereas English does not have to mark the mental-
space builder with a conditional expression. This contrast in the mechanisms
for building mental spaces and expressing conditionality is systematic, and
reflects general divergences between the two languages, in particular,
regarding i) the mechanisms for expressing modality; ii) the clause-linking
mechanisms; and iii) the argument structures.
4. Typological motivation for divergences
4.1. Different mechanisms for expressing conditional modality
First, the mechanisms for expressing conditional modality differ.

In English, when the main clause has a conditional modal, the
'conditionality' of the sentence as a whole is made apparent11, and it thus
becomes possible to omit an explicit antecedent clause and still have the
sentence be understood as having a conditional meaning.12 By contrast,
Japanese does not have any clear modal which unambiguously makes
'conditionality' explicit in the main clause alone (DAROO also has noncon-
ditional uses). Instead, 'conditionality' is made explicit by marking the
antecedent adjunct which delineates the condition. In a sense, conditional
clause-linking markers in Japanese take the place of English modal
expressions semantically. Thus, in Japanese sentences (12) through (16), the
modal expression DAROO is optional; the primary mark of conditionality is
that the antecedents should be marked with appropriate conditional
linkages.
Actually, 'would' has other uses (e.g., past habituais). But they are quite marked.
12
This idea was first presented in a plenary speech entitled 'Mysteries of Conditional
Sentences in English and Japanese', by Charles J. Fillmore, in collaboration with me, at
the Annual Meeting of the ATJ (Association of Teachers of Japanese) in San Francisco,
March 1988.
4.2. Different clause-linking mechanisms and ellipsis
Second, clause-linking mechanisms are different between the two languages.

Japanese conditional markers are clause-linking morphemes; in fact,
most of them are verb endings, unlike the free-standing English IF. In
addition, Japanese clause-linking morphemes appear at the end of the
subordinate clause, immediately preceding the main clause — a feature
common among verb-final and head-final languages. These facts, taken
together with the very free use of ellipsis in Japanese, enable not only full
clauses but also various kinds of truncated clauses (appearing as phrases on
the surface) to be marked by these linking morphemes and thereby
delineate mental spaces.
To understand this better, we will here discuss some of the subtleties
involved in counteridentical cases, and examine Japanese 'truncated
antecedents'.
Fillmore (1987) discusses two kinds of situations in which a sentence
like (17) can be used:
(17) If I were your father, I would spank you.
It can be uttered 'first, by somebody who would like to be in the father
role, so that he could effect a punishment which the real father is
apparently unwilling to do; and second, by a temporary caretaker who
wants the child to be grateful for his lenient tendencies, since the real
father in this situation would be more punitive' (p. 181). As noted by
Fillmore, prosodic salience here would go with either 'I' or 'YOUR FATHER',
respectively, thereby selecting the individual to be hypothetically introduced
into the other individual's situation.
The Japanese sentences appropriate to the two situations possible in
(17) are shown in (18) through (20).
(18) Watasi ga kimi no otoosan dattara hippataite osioki sum yo.
I NOM you GEN father copula-Tara(coND), spank punish PART
(If I were your father. I would spank you.)

Either reading of (17).
(19) Watasi dattara, hippataite osioki suru yo.

I copula-TARA(coND), spank-punish PART
(I would spank you.)
(If it were I, (I) would spank you.)
First reading of (17): < Your father doesn't spank you, but if I
were taking your father's position, I would spank you.>
cf. (19') ... isioki sitai yo. (# nasaru yo. / # sitagaru yo.)
(20) Kimi no otoosan dattara. hippataite osioki suru yo,

your father copuIa-TARA(coND), spank-punish PART
(Your father would spank you.)

(If it were your father, (he) would spank you.)
Second reading of (17): <I won't spank you, but if your father
were here, he would spank you.>
cf. (20') ... nasaru yo. /... sitagaru yo. (# osioki sitai yo.)
Sentence (18), which is the structural equivalent of the English sentence
(17), containing a full clause antecedent, can be uttered in either situation,
as in (17).13 But, as shown in (19), with the truncated antecedent Watasi
dattara (watasi, T, followed by DATTARA), only the first interpretation is
possible. The sentence (20) with the different truncated antecedent Kimi
no otoosan dattara (kimi no otoosan 'your father' followed by DATTARA)
is uttered only in the second situation.14 In either case, the mental-space
It seems that sentence (18) more typically receives the first reading while sentence
(ii) below more typically receives the second reading.
(ii) Kimi no otoosan ga watasi dattara, hippataite isioki suru yo.
you GEN father NOM I copuIa-TARA(COND), spank punish PART
(If your father were I, he would spank you.)
14
Another case in which the sentence gets only one reading is when a certain
property of the main-clause verb identifies the subject - whether it is a first person (I) or
a third person (your father) (Thanks to Yoko Hasegawa for drawing my attention to this).
For example, sentence (18) with 'sitai' only gets the first reading, whereas, with 'sitagaru'
or 'nasaru' (honorific), it only gets the second reading. Thus, notice that 'sitai' can appear
in sentence (19) but not in (20); 'sitagaru' or 'nasaru' can appear in (20) but not in (19),
as shown in (19') and (20').
builder is marked by the conditional marker, DATTARA15. The unambig

uous readings of these truncated antecedents (in 19 and 20) support the
claim that one of the functions of the conditonal marker DATTARA in
(19) and (20) is precisely to profile the hypothetically introduced individual
(mental-space builder).
Notice also that here again, by virtue of this conditional marking, the
hearer knows that the speaker has set up and is talking within the
hypothetical space, even though the main clause does not contain any
specific conditional modality.
4.3. Different argument structures and transitivity
Finally, the argument structures of verbs differ in English and Japanese.

To look at this difference, we will examine the Japanese and English
ways of expressing causation, by comparing a number of Japanese
conditional sentences (taken from a junior high school science textbook)
and their English translations (21 - 23).16
In each pair of English translations, translation A is a first translation
produced by a native Japanese speaker fluent in English; translation B is
a later version produced by a team — including native English speakers —
which edited the first translation to make it sound more natural and
stylistically appropriate in English discourse.
(21) Tikara no heekoo sihenkee no kankee o
power GEN parallelogrammic relations ACC
tukau TO, hitotu no tikara o,
use WHEN/IF 'one power' ACC
I am using only examples with TARA here, but other conditional markers (e.g.,
NARA) can also be used similarly.
The on-line text of this science textbook (written in Japanese, and used by junior
high school students in Japan) was made available by the Communications and
Information Processing Laboratory, Nippon Telegraph and Telephone Corporation (NTT),
Yokosuka, Japan. Sponsored by NTT, this text was used in a larger project on Japanese
clause-linkage, directed by Charles J. Fillmore. As part of the project, the original
Japanese text was translated into English by a team of bilingual speakers of Japanese and
English; first by the native speakers of Japanese, then by the native speakers of English.
Translation and editing took place through active verbal interaction, with the team
(normally four to five participants) recording various versions of the translations.
kore to onazi hataraki o sum

this same function ACC do
hutatu no tikara ni wakeru koto ga dekiru.
'two power' DAT divide can (potential)
(21A) When we use forces' parallelogrammic relations, we can divide

one single force into two forces which function the same way as
the original force does when they get together.
(2IB) Exploiting the relationships in a parallelogram of forces enables

us to decompose a single force into two forces which produce
the same effect.
(22) Sankagin o kanetu sum TO, gin to sanso ga

silver dioxide ACC heat WHEN/IF silver and oxygen NOM
dekiru.
be produced
(22A) When one heats silver dioxide, silver and oxygen are produced.
(22B) Heating silver dioxide produces silver and oxygen.
(23) Pasukam no genri o ooyoo mm TO,

Pascal GEN principle ACC apply WHEN/IF
tiisana tikara o ookina tikara ni kaem koto ga dekiru.
small force ACC great force DAT change can
(23A) By applying Pascal's principle, we can transform a small force

into a large force.
(23B) An application of Pascal's principle allows us to transform a

small force into a large force.
Notice that the first version (A) preserves the bi-clausal structure of the
original Japanese sentences. In each case, the adverbial clause constitutes
the antecedent, expressing the condition (or cause). In version (B), on the
other hand, it is the element taking the grammatical subject position —
whether a sentential or a non-sentential subject — that delineates the
condition. These pairs of English sentences show not only that alternation
is possible, but also that the causative version is sometimes considered more
appropriate in scientific discourse in English.

Both versions (A) and (B), like the Japanese original text, express
relations of causation and enablement between two events. But, whereas
English can. do so by using such verbs as enable, produce, and allow, the
meanings of which directly show the relations of enablement, Japanese
relies on the clause-linking construction marked by the TO particle.
In Japanese, the causative version is often not possible because
Japanese tends not to allow inanimate subjects for transitive verbs,
especially for causative verbs. Thus, certain argument structures — as seen
in the English sentences (3), (11b), (21B) - (23B) — are unnatural in
Japanese. Even when verbs which can take such argument structures are
available in Japanese, causative version is obviously not a preferred event
representation pattern — a result of the general tendency in Japanese
toward lower transitivity in Japanese.17
5. Conclusions
We have shown that whereas English grammar easily accommodates mono-

clausal conditional expressions, Japanese often requires that the counter
parts of these expressions use conditional clause-linking morphology. This
contrast in the mechanisms for building mental-spaces and expressing
conditionality is systematic, and reflects general typological divergences
between the two languages regarding i) mechanisms for expressing
modality; ii) clause-linking mechanisms; and iii) argument structure and
transitivity. Unlike English, Japanese does not have any clear modal which
makes conditionality explicit in the main clause alone. Instead, conditionali
ty is expressed by marking the space-building antecedent with a clause-
linking morpheme, so that clause-linking mechanisms in effect take the
place of modal expressions semantically. In addition, unlike English,
Japanese clause-linking mechanisms, taken together with the free use of
It has been noted that Japanese tends toward lower transitivity, compared with
English. This tendency has been discussed in various approaches, including analyses of
morpho-syntactic properties and case markings of intransitive and transitive verbs
(Jacobsen 1982,1985; Sugamoto 1982; Teramura 1978), and analyses of fine-graned lexical
semantics of verbs (Ikegami 1985), Ikegami (1981, 1982, 1991), among others, has also
investigated the issue of agentivity and transitivity from the point of view of event
realization patterns, which he attributes to cultural rhetorical patterns.
ellipsis in Japanese, allow various kinds of truncated antecedents which are

marked with conditional clause-linking morphemes to delineate mental
spaces. Furthermore, unlike English, the Japanese mode of expressing
causation tends to appeal to conditional clause-linking constructions instead
of causative constructions. This style reflects the tendency in Japanese
toward lower transitivity, and the difference in its preferred event represen
tation pattern and in the argument structure of verbs.
Although we focused on conditional sentences and the mental-space
building functions in the current case study, the phenomena discussed in
this paper are in fact not limited to conditionals. Given the above-
mentioned typological divergences between the two languages, the findings
are generalizable to a large extent to other clause-linking mechanisms in
Japanese and their comparisons with their counterparts in English.
References
Bowerman, Melissa
1986 "First steps in acquiring conditionals." In E.G. Traugott, A. Meulen, J.S.
Reilly & C.A. Ferguson (eds.), On conditionals. Cambridge: Cambridge
University Press, 285-308.
Comrie, Bernard
1986 "Conditionals: A typology." In E.C. Traugott, A. Meulen, J.S. Reilly & C.A.
Ferguson (eds.), On Conditionals. Cambridge: Cambridge University Press,
77-99.
de Castro Campos, Maria Fausta P.

1981 "On conditionals as dialogue constructs." Paper for International Encounter
in the Philosophy of Language. State University of Campinas, Brazil.
Fauconnier, Gilles
1985 Mental spaces: Aspects of meaning construction in natural language.
Cambridge, MA: M.I.T. Press.
Fillmore, Charles J.
1987 "Varieties of conditional sentences." Proceedings of the Third Eastern States
Conference on Linguistics, 163-182.
Foley, William & Robert Van Valin

Press.
Fujii, Seiko Yamaguchi
1992 "On the idiomaticity of conditional constructions in Japanese." In J.
Boulanger & C. Ouellon (eds.), Proceedings of the Fifteenth International
Congress of Linguists. Québec, Canada.
Goodman, Nelson
1983 Fact, fiction, and forecast. Cambridge: Harvard University Press.
Ikegami, Yoshihiko
1981 SURU to NARU no Gengogaku [Linguistics of DOING and BECOMING].
Tokyo: Taishuukan.
Ikegami, Yoshihiko
1982 Hyoogen Koozoo no Hikaku: SURU-teki na gengo to NARU-teki na gengo
[Contrast of linguistic representation types: 'DO-language' and 'BECOME-
language']. In Tetsuya Kunihiro (ed.), Nichieigo-hikaku-kooza: Hassoo to
Hyoogen [Japanese-English comparative grammar]. Tokyo: Taishuukan, 286-
326.
Ikegami, Yoshihiko
1985 "'Activity'-'Accmplishment'-'Achievement': A language that can't say 'I
burned it, but it didn't burn' and one that can." In Adam Makkai, & Alan
K. Melby (eds.), Linguistics and philosophy: Essays in honor of rules S. Wells.
Amsterdam/Philadelphia: John Benjamins Publishing Company, 265-304.
1991 "'DO-language' and 'BECOME-language': Two contrasting types of linguistic

representation." In Yoshihiko Ikegami (ed.), The empire of signs: Semiotic
essays on Japanese culture. Amsterdam/Philadelphia: John Benjamins
Publishing Company, 286-326.
Jacobsen, Wesley
1982 Transitivity in the Japanese verbal system. Indiana: Indiana University
Linguistics Club.
Jacobsen, Wesley
1985 "Morphosyntactic transitivity and semantic markedness." In CLS 21: Papers
from the Parasession on Causatives and Agentivity. Chicago: Chicago
Linguistics Society.
Jespersen, Otto
1940 A modern English grammar on historical principle, Vol. v.: Syntax. London:
George Allen and Unwin.
Mackie, J.L.
1965 "Cause and Conditions." American Philosophical Quarterly 2:4.245-264.
Minami, Fujio
1964 Fukubun [Complex sentences]. Tokyo: Meijishoin.
1974 Gendai Nihongo no Kozo [The structure of modern Japanese]. Tokyo:

Taishukan.
Sugamoto, Nobuko
1982 "Transitivity and objecthood in Japanese." In Paul J. Hopper & Sandra A.
Thompson (eds.), Syntax and Semantics, Vol 15: Studies in transitivity. New
York: Academic Press.
Sweetser, Eve
1990 From etymology to pragmatics: Metaphorical and cultural aspects of semantic
structure. Cambridge: Cambridge University Press.
Teramura, Hideo
1978 'NARU' hyoogen to 'SURU hyoogen: Nichiei 'tai' hyoogen no hikaku." In
Nihongo To Nihongo-kyooiku [Japanese and Japanese Language Education].
Tokyo: Kokuritsu Kokugo Kenkyuusho.
Van Valin, Robert

1984 "A typology of syntactic relations in clause-linkage." BLS 10.
Distance in construction grammar
Jean Mark Gawron

SRI International
1. General considerations
The last twenty years in linguistics has seen a great enrichment of the
conception of the lexicon. This movement, usually called Lexicalism, can be
viewed as a shaping force in all of the currently active theories. What
became evident in the years following 1970 was that a great number of the
properties of a grammar could be explained as projections of lexical
properties. Much of the variation in the grammar that was previously
handled by mappings between trees could be more perspicuously handled
as variation of lexical properties, which were then projected into the
grammar by very general principles. As a consequence, an account of the
projection of lexical properties lies at the heart of all current syntactic
theories.
The grand success of this enterprise and the great fruitfulness of this
paradigm might lead one to suppose that all of grammar can be accounted
for as the projection of lexical properties. This is clearly not the prevailing
view in contemporary syntactic theory. In the Governement Binding Theory
of Chomsky 1981, for example, one might say that the Government half
(and everything connected with the Projection Principle) deals with lexical
projection, but the Binding Theory clearly does not. Recent overhauls of
the theory of phrase-structure have put a great deal more weight on
projections of material from inflectional morphology and closed class items
such as complementizers, but the basic separation between lexically
projected and non-lexical still remains.
It is thus one of the central tasks for a linguistic theory to characterize
which properties of a grammar are projections of lexical properties and
which are not. Construction Grammar as developed in Fillmore 1987 and
92 JEAN MARK GAWRON
Fillmore, Kay, and O'Connor 1988 is a framework which makes a fairly

reactionary proposal about the non-lexical part of grammar. It says that
there are a number of syntactic and semantic properties of particular
syntactic configurations that are not projections of lexical properties.
Connecting the syntax and semantics of these configurations defines what
is called a construction. Now there are two responses that one can make to
this at the level of the linguistic formalism: One can generalize the notion
of lexical item so that it embraces a great deal more than it does in
standard accounts, including discontinuous words and complex tree
structures; or one can leave the concept of word as it was and extend the
vocabulary of the grammar to make a greater variety of non-lexically
projected information available. On the basis of the work that has appeared
so far, I'm not sure that Construction Grammar has shown much of a
strong preference for either direction,
It is natural that the central problem for such an approach is to make
clear the relationship between the productive and non-productive parts of
the grammar, to show how constructions with their lexically unpredictable
properties interact with the lexically projected part of grammar. This paper
is not an attempt to solve that very difficult problem, but an exploration of
a small domain of grammar which contains a small group of LEXICALLY
MOTIVATED CONSTRUCTIONS.
Lexically motivated constructions are constructions which can be
defined by a lexical class. But there is also a larger class of items that enter
into the construction, these lexical items will exhibit novel semantic features.
Goldberg 1991b argues that the English ditransitive construction is a
lexically motivated construction in this sense. Thus, the particular sense of
give observable in:
(1) Mary's behavior gave John an idea
is not available in the non-double object valence. Rather, it is the construc
tion which confers the new meaning.
Clearly, lexically motivated constructions present a crucial test case for
any account of the interaction of the lexically projected and the construct
ional What Goldberg is in effect saying is that the boundary between these
domains may be fluid. What starts out as the paradigm lexically projected
property, a subcategorization frame, may end up as a construction imposing
new meanings on the verbs which enter into it.
Below I will propose a different instance of a lexically motivated
construction, which I will call the distance-noun adjunct construction.
DISTANCE IN CONSTRUCTION GRAMMAR 93
Crucial to any development of construction grammar will be the ability

to make grammatical statements pairing particular syntactic configurations
with a particular semantics. We begin with some simple examples:
(2) From Boston to Chicago is a long way
Here we have what appears to be a sequence of two prepositional phrases
acting as a subject. One view of grammar holds that there should be a
productive (and autonomous) component of phrase-structure rules that
licenses all possible constituents. Thus there would have to be a rule
building a constituent from a sequence of PPs. CG, on the other hand,
could say that there is a special construction here, licensing an ad hoc
constituent because a path is denoted, tying this special syntax to that
special semantics.
As another, more celebrated example:
(3) Is John coming?
Here, the grammar needs to make a statement linking syntactic inversion
with question semantics.
There is no linguistic level called the construction level. Instead of
having a level, constructions are arbitrary associations of material from
various linguistic levels. Much like words (or better, lexical entries). Just as
the sound of a word and its meaning are brought together by extra-
grammatical forces, so are the pieces of a construction.
Thus, as noted in Fillmore, Kay, and O'Connor 1988 (FKO), construc
tions have a word-like nature. How lexical are constructions? Do they
undergo lexical rules? Well, if the notion construction is formulated so as
to include lexical entries as in FKO, the answer is non-trivially yes. But are
there non-trivial examples? I shall argue in section 2.2 that there are cases
of constructions being related by rulelike processes.
We would also expect constructions to have something like a blocking
effect:
(4) a. How much time does it take to get to Rome?
b. How long does it take to get to Rome?
c. How much (food) did John eat?
d. *How much does it take to get to Rome? ('How much
time' reading)
There is going to be a productive rule which allows how much to occur as
a headless NP (4a) shows that how much is semantically compatible with
94 JEAN MARK GAWRON
time-expressions, yet it is not compatible with headless time-expressions. A

natural account of this would be to posit a headless NP long as a special
construction with temporal semantics. This explains the appearance of long
in environments that seem to call for NPs:
(5) a. He didn't stay long
b. He didn't wait for long
c. It didn't take long
If constructions can block semantically and structurally close alternatives,
then the unacceptability of (4d) would also be explained.
The crucial feature for any formulation of CG is that arbitrary
connections between different linguitic levels be statable. Thus, a particular
syntactic structure such as the path structure in (2) must be able to be
associated a particular semantics; similarly it must be possible to associate
the inversion structure of (3) with question semantics. Because of its
particular approach to the syntax/semantics interface, The theory of Head-
Driven Phrase Structure (HPSG see Pollard and Sag 1992) already has a
body of work in which such statements are made. The case for LFG is less
clear. There is in some sense an autonomous phrase-structure component
in LFG, but the fact that the phrase-structure rules are annotated with
functional information makes if possible in effect to semantically constrain
them. Moreover, there are a variety of architectures for implementing
semantics in LFG, which might enable the semantics to directly constrain
the phrase-structure without going by way of the Functional Structure.
In sum, the basic idea of a construction does not immediately force a
particular variety of linguistic formalism. However, some of the concepts of
HPSG, particularly the HPSG analysis of adjuncts, have proven useful in
the analysis of distance expressions that follows.
2. Distance expressions
2.1 A construction
We start by looking into the syntactic and semantic structure of expressions

like:
(6) 300 miles from Boston
This this is a constituent is suggested by:
(7) How many miles from Boston is New York?

One possibility is that the distance expression 300 miles is a specifier on the
preposition or the preposition phrase, the most natural analysis for
modifiers like 20 minutes after the earthquake in:
(8) a. an aftershock 20 minutes after the earthquake
b. *an aftershock 20 minutes
c. an aftershock after the earthquake
But there is a crucial difference with the distance expressions. Expressions
like 300 miles from Boston are essentially functioning as locative modifiers.
In order for the from-phrase to function as a locative modifier, the measure
modifier and the from -phrase must both be there:
(9) a. a city 300 miles from Boston
b. *a city 300 miles
c. *a city from Boston
This suggests that something non-compositional is going on. To put it
semantically, expressions like after the earthquake can be thought of as
denoting the property (of events) of occurring after the earthquake; 20
minutes after the earthquake restricts that to a property of occurring 20
minutes after the quake. We want to think of six inches from the pin as
designating the property of events or objects of being six inches from the
pin. But from the pin does not denote the property of events or objects of
being some distance from the pin. This is what explains the followin
contrast:
(10) The ball rolled to a stop *from the pin
6 inches from the pin
There are a variety of modifiers that have similar semantics:
(11) a city a long journey from Boston
pretty far
quite a distance
a long way
300 miles
Note that some of these post-nominal modifiers appear to be NPs. In
general, post-nominal modifiers cannot be NPs:
(12) *a man the mayor of Boston
96 JEAN MARK GAWRON
There are however a few very special cases, involving NPs that describe a
property:
(13) a. A dog the size of a small pickup truck was procured.
b. A birthmark the color of an eggplant floated on his fore
head.
What is going on in (12) appears different from what is going on with the
distance nouns: There seems to be no reason to think 200 miles from Boston
is a description true of the property of being 200 miles from Boston, while
there is good reason to think the color of an eggplant is a description true
of a color property.
One more point of importance: The very same distance NPs can also
function adverbially:
(14) a. The meteorite landed a short distance from the city.
b. Mary lived twenty miles from the sea.
c. Jasha lived two day's journey from Lublin.
Again, there are other NPs that act as adverbial modifiers, but their
semantics is special:
(15) a. John shaved the way he ate aspic. Reluctantly.
b. Henryka lived on Opaczewska Street her entire stay in
Warsaw.
c. Irena left the day Miszka arrived.
The question is: What about the distance nouns links them to these other
adverbial heads?
Summing up the problem: There is a class of NPs denoting distances
which can be adnominal or adverbial modifiers. Thus, the preliminary
answer to the syntactic half of our initial question about (6) seems to be
that it is an NP whose head is miles. This answer in turn raises other
questions: Why are these NPs allowed in contexts where others are not?
What we are essentially, trying to explain here is the possibility of
certain noun-phrases acting as adjuncts. In order to take a step toward such
an account, I want to borrow an idea from HPSG. This is the idea that
adjuncts select for their heads. In HPSG, that reduces to explaining why
these noun phrases can select nominal and verbal heads to modify. I will
adopt a simplified version of this approach, trying to restrict myself to the
apparatus necessary for the problem at hand. Let us assume that lexical
entries have attributes of various sorts associated with them; among these
are three that will concern us in particular, MOD, DENOTATION and MOD-
CONTENT. MOD gives a syntactic and semantic specification of what
projections of this head may modify. DENOTATION tells us what it dnots.
MOD-CONTENT tells us the semantic result of applying this modifier to this
head. The role of these three different attributes can be illustrated with an
adverbial noun like way. Let us suppose that way is a property of manners,
and way that John eats aspic a property of manners that John eats aspic.
The NP the way that John eats aspic should denote a particular manner
John eats aspic in. This is what I propose to call the denotation of the NP.
This denotation will be used the way the denotation of any NP is used in
sentences like:
(16) I like the way John eats aspic.
Now the MOD feature of the NP will telll us that it can modify verb
phrases, perhaps imposing semantic restrictions that determine which verb
phrases. The MOD-CONTENT of the NP will tell us the resulting semantics
of the verb phrase once the NP modifies it. In effect adjuncts are allowed
to carry their own semantic rules for combining with their heads.
Let us assume a Davidsonian semantics, and that the desired semantics
for the meteor landed some distance from Boston is the following:
(17) [land(e) actor(e, the-meteor) distance (e, Boston,d)]
Here e is an event variable, and d a distance variable. The relations
DISTANCE holds between two things and a distance when the things are
that distance apart. Then what the NP some distance from Boston should
specify for its three relevant features is:
This describes a sign of category N that modifies something of category N

or V (other syntactic details such as bar level restrictions, have been left
out). The semantics of the modified element is IT, and the semantics of the
sign is d, which is restricted to be a distance which holds between Boston
and some object. The semantics that results when this sign modifies
something is just the conjunction shown in MODS-CONTENT. Existential
98 JEAN MARK GAWRON
quantifications have been left as the work of semantic interpretation

machinery, although there are some sticky issues we will turn to below.
Essentially, when the NP modifies something as an adjunct, it adds the
restriction that whatever object it modifies be located that distance from
Boston.
Note that the story for adnominal modification would be no different;
instead of an event being located some distance from Boston, it would be
a different kind of object, such as a city.
What we have done is to very informally gather together some semantic
and syntactic information. There is an NP which can function both as a
denotational NP denoting a distance, and as an adjuncts modifier. Now in
effect, with the two attributes DENOTATION and MOD-CONTENT, we have
allowed a single constituent to make two different kinds of contributions to
semantic content, one kind when it is an argument (an ordinary NP),
another when it is an adjunct. I will call this sort of lexical entry
bifunctional This is a curious move for which HPSG cannot be blamed.
HPSG makes due with a single CONTENT attribute. With only a single
CONTENT attribute, the most natural account is that NPs like some distance
from Boston are ambiguous between adjunct and denoting NP readings,
with two readings related by something like a lexical rule. Since the lexical-
rule account is available without splitting semantic contributions in two, it
has something going for it from the outset. In fact, the data in the section
2.2 argue strongly for a lexical rule account, but I believe that a bifunctional
account may well be motivated for other adverbial nouns like way and day.
One might also want a bifunctional semantics to account for the uses of
locative PPs to denote places, in sentences like:
(19) Under the bed is a good place to hide.
Whether we posit a lexical rule or generalize the information in (18) to
cover a class of nouns, we would still have something we might well call a
construction, but note that it is one we could state at the lexical level. So
far we seem to be talking about an unusual class of nouns. Nothing yet has
suggested that non-lexically projected properties are involved. For the time
being, nevertheless, I will continue to speak of a construction, deferring
justification until later.
The syntactic and semantics details of what such a construction would
look like would require a more explicit theory of adjuncts modification than
there is room for in this paper. What I want to focus on is some valence-
changing properties of the construction which will help draw out the
differences between a lexical rule account and a bifunctional account.
The correct account needs to capture a basic fact: The complements
that can attach to distance nouns when they are ordinary NPs are different
from those in the adjunct function:
(20) a. the distance from New York to Boston
b. a city some distance from New York
c. *a city some distance from New York to Boston
d. *a city some distance to Boston
To see how this can make sense on the bifunctional account, let's propose
a noun entry that includes some details left out above:
[CAT N I
CAT V OR N 7
MOD DENOTATION π
[ DENOTATION-VAR g J
(21) CAT P CAT P
SUBCAT PFORM from, PFORM to
DENOTATION s
DENOTATION g\ J
MOD-CONTENT π A distance (g, s, d)
[DENOTATION d such that distance (g, s, d) J
I will refer to s,g, and d as the source, goal, and distance, respectively. The
idea here is that under the attribute SUBCAT, the noun contain a list of
subcaterorized-for complements, as is standard in HPSG. The crucial thing
to note is that the to-phrase fills the same argument position of the
distance-relation as the heads the NP adjunct can modify. The DENOTA
TION-VAR of the MOD and the DENOTATION of to-phrase fill the goal
argument-position of the distance relation; and both fill the source
argument-role of the distance-relation.
Given (21), the unacceptability of (20c) and (20d) will follow now if we
have some principle that precludes filling the same argument position of the
same relation twice in a single syntactic configuration. Now the competing
lexical rule account will accomplish the same by not permitting the adjunct
NP to subcategorize for a to-phrase. This is quite natural if we think of the
rule as a distant cousin of passivization: What starts out in an important
syntactic-semantic slot (the denotation role of the NP) becomes a non-
player; something else that began as oblique (the to-phrase) gets 'promoted'
100 JEAN MARK GAWRON
to be the argument of the adjunct-NP. In effect, the lexical rule just splits
the lexical entry in (21) in two. We start with:
And we end with:
It is worth pointing out here that what I am calling a lexical rule account
does not need to be thought of as a rule at all. If we simply think of (22)
and (23) as two distinct types licensing classes of nouns, with the relation
distance itself a type admitting further specification, then (22) and (23)
together say that there will be nouns with distance semantics with the
indicated properties. Developing this view would probably give us a more
constructional view of what is going on. In the interests of employing
familiar notions where possible, I will, however, continue to refer to a
lexical rule account. When I refer to the lexical rule account as defining a
construction, I will mean the syntax and semantics in (23).
The striking feature of the lexical rule account is that one needs to posit
non-denotational NPs (or rather, NPs that can only denote properties, like
adjectives or adverbials). The striking feature of the bifunctional account is
double-barreled semantics. One might think that tests of the denotational
status of NPs like:
(24) a. the distance from Boston
b. some distance from Boston
will help one decide between the two accounts. This turns out not to be the
case. For exammple, NPs with the definite article appear to preclude the
adjunct use, whereas those with some appear to preclude uses referring to
distances:
(25) a. the distance from Boston is 300 miles.

b. ?some distance from Boston is 300 miles.
On the other hand, some NPs are switch hitters:
(26) a. 300 miles from Boston is a long distance.
b. ?A long journey from Boston is any distance over 300 miles.
c. A long journey from Boston would cheer me up.
Sentence (26a) shows a distance noun referring to a distance; sentence
(26b) shows that travel nouns referring to distances are awkward, while
(26c) shows that reference to a traveling event is quite simple.
Another parameter of variation shows up in demonstrative or anaphoric
uses of distances:
(27) a. Two cities the same distance from Boston competed for the
contract.
b. No city that distance from Boston will be a contender.
One might on the basis of the surface forms want to call these both
definites, but they pattern with the indefinite cases in allowing the adjunct
use.
What is showing up here is the corner of a complex array of facts. But
either of the two approaches shows some promise of extending to cover the
data. Consider the bifunctional account first: The contrasts in (25), (26),
and (27) have to dowithspecifiers of the head noun, either particles or the
adjective same; if we assume that these have the possibility of binding the
DENOTATION attribute of the NP, then they can force denoting uses.1
On the lexical rule account, however, we can rule out definite articles
with non-denoting NPs for an analogous reason. Suppose definites expect
an individual-level denotation from the nouns they combine with. Then they
will be the wrong semantic type for adject nouns, which have proposition
level entities as their denotation.2
It is a little more difficult to force adjuncts uses, but perhaps this can be done with
a special some.
There is an interesting difference between (25a) and (26a). (25a) requires a

discourse bound reference point, the point to which we are measuring the distance from
Boston. On the other hand, (26a) appears to have quantified away the same argument
position: Any point 300 miles from Boston will do. No doubt this contrast has something
to do with the definiteness and indefiniteness of the NPs in question. Beyond making that
Now I want to turn to characterizing the class of nouns that allow this
strange construction. In effect, we want to characterize a class of nouns that
are going to be allowed to have an adjunct semantics as well as a denota-
tional semantics. I will view this as the licensing of another meaning, even
though the account I propose does not posit a distinct lexical entry to carry
that meaning.
I propose a rule of semantic transfer or a semantic license (in the sense
of Nunberg and Zaenen 1992) of something like the following form:
Semantic License A
A noun which denotes a basic three-place relation among a SOURCE,
a GOAL and the distance EXTENT between them, and which functions
as a property of distances when saturated, can also function as a
property of GOAL»
I use the term saturated here not inn usual sense (having proposition or
term semantics), but to mean something which has gathered together all its
syntactically realizable arguments. Thus a relational noun like picture is
saturated when it has picked up its of-phrase complement. When the noun
distance has picked up its two complements as in distance from Boston to
New York, it is saturated. Since it denotes a property of distances when
saturated, it is a noun of the right sort to undergo license A. On the
bifunctional account, License A is in effect a semantic statement of the
conditions that must be fulfilled for nouns to have the sort of DENOTATION
and MOD- attributes posited in (21). On the lexical rule account, it is an
input condition on the rule. On both accounts one might well fold the
semantic conditions into the rule itself, so that a separate statement of
License A would be redundant, but it will be useful for our purposes to
have a characterization of the semantics of the nouns independently from
the construction.
But A has been stated in a very restricted fashion, so that it includes
only properties of distances. What about the following cases:
(28) a. That city is a 200 mile drive from Boston.
b. New York is a 200 mile journey from Boston.
observation, I have no detailed proposal to make. This, however, is the sticking point in
the general existential quantification invoked above in the semantic machinery:
Distinctions need to be made between those things that need to be anchored in the
discourse and those that do not.
One might try to bring these nouns in under the umbrella of License A on
the basis of a metonymy account. Thus, note:
(29) The drive to Boston is 200 miles.
Here a drive too Boston seems to be being equated with a distance. A
metonymic map from travel events to the distances they cover would
account for such uses. Moreover, if travel nouns can metonymically be
mapped to distance nouns, then they will come under License A, and
adjunct uses will be licensed. The trouble with this is that travel nouns do
not fit in well everywhere that distance nouns do:
(30) a. ?John ran a 2 hour drive.
b. ?The airliner flew a two-day journey over the jungle.
Now these judgements are perfectly explicable if nouns like journey and
drive denote travel events in most contexts. But then how do we handle the
cases in (28)?
I assume that heads like drive and journey can be handled by the
following license:
Semantic License B
A noun which denotes a basic three-place relation among a SOURCE,
a GOAL and the event of moving between them and which functions as
a property of movement events when saturated, can also function as
property GOAL.
This is simply license A with movement events replacing distances. Now
under License B travel nouns still denote travel events in their adjunct uses,
and verbs that select for distances may still select against them. This is
presumably what is happening in (30).
It is License B, I suggest, that truly establishes the constructional nature
of (21) or of (23). We have now proposed that the same unusual syntactic
property extends to distinct but semantically related classes. The crucial
analogue is the extension of the ditransitive construction from events of
intended or successful transfer of objects (give, win) to verbs enabling a
transfer (permit, allow) and to verbs describing actions that will cause a
transfer at a future time (reserve). This is what Goldberg 1991b calls
constructional polysemy. What it illustrates is a construction exhibiting the
wordlike property of acquiring distinct but related meanings. Moreover the
polysemic mapping between distance and travel event is independently
motivated in other constructions, as shown in (29).
On a view of grammar which recognizes only lexically projected

properties in this domain, such wholesale transfer of a syntactic property
from one semantic class to another is mysterious. There is no account of
why the polysemy is limited to certain syntactic contexts. On a view which
treats constructions on all fours with words, as basic entities of the grammar
with independent semantic properties, such polysemy is to be expected.
One point I have not addressed thus far is the role of the notions
SOURCE and GOAL in the account. These notions, originally introduced as
deep case roles in Fillmore 1968, have often been invoked since in
discussions of the semantics of movement. It is interesting here to see the
two English prepositions that feature most prominently in those discussions
showing up in a number of examples which involve no movement at all. I
have the space here only to point out that there are a number of interesting
questions that await further inquiry: Why should expressions of distance
invoke notions borrowed from the domain of movement rather than notions
borrowed from the domain of static location? Why does the adjunct in the
adjunct uses fill the role identified with the goal argument and not the
source argument? Why do we say far from but close to?3
We have now supplied some syntactic and semantic answers to our
questions about noun-phrases like some distance from New York, It is a
Noun Phrase whose head subcategorizes for a single complement marked
with from. Presumably, because of the unacceptability of a city some
distance, that complement is obligatory.4 The example we started with, (6),
is only a little different from distance: Mile denotes a function from
numbers to a three-place relation among distance source, and goal. The
umbrella of License A thus needs to be slight widened to include functions
to three-place distance relations.
3
Thanks to G. Nunberg for pointing out this question.
4
In the interests of fair play, I point out that this fact presents a problem for the
account being presented here which is not shared by the lexical rule account. Capturing
the fact that the from -phrase is obligatory in the adjunct function is not going to be easy
in a single lexical entry. Matters may not be as bad as they first appear, however. NOte
that postnominal modifiers with nothing following the head are in general, marginal: *a
man happy versus a man happy with his lot. Note also that predicative uses without the
from -phrase do appear possible:
(i) Boston is a long journey.
(ii) Denver is a two-hour walk.
Now actually, if it is something close to right, Semantic License A is

quite remarkable, because it has the semantics of a category-changing rule
without doing any category-changing. But where does the construction come
from? Why does semantic license A exist? Why aren't transfers like this
more common? I suggest the transfer works here at least in part because
we are dealing with abstract objects. Reference to things like measures is
not particularly fixed because measures are rather abstract objects, defined
relationally, so the step from denoting a measure to denoting the relation
between objects that defines it is not a great one.
2.2 Syntactic variations
We begin by noting the occurrence of a special dummy-it in certain distance

constructions:
(31) *Boston is a long journey (from Boston) to New York.
It pretty far
quite a distance
a long way
300 miles
These contrast with cases where the distance head is not followed by a to-
phrase:
(32) Boston is a long journey from New York.
It pretty far
quite a distance
a long way
300 miles
Further evidence of dummy-hood of ft in (31) can be found in post-nominal
constructions. As noted in section 2.1, the distance heads cannot function
as postnominal complements when the to-phrase is present:
(33) a. a city 300 miles from New York
b. *a city 300 miles to New York
It is in accounting for these facts that I think one can drive a wedge
between the bifunctional and lexical rule account in 2.1. Recall that on the
lexical rule account, the output in (23) did not allow the noun a goal
complement in its SUBCAT list. The rationale there was that since that
structure was for the noun in the adjunct function and the adjunct-head
always filled the goal's argument position, no goal complement could ever
occur. But suppose we did allow a goal complement on the subcat list. We
could still appeal to a principle requiring that argument roles be filled only
once per structure to block:
(34) *a city some distance to New York
The question is: What would then be the status of the adjunct NP? With all
its argument positions saturated it would denote a proposition and would
be unable to function as an ordinary NP (recall that the basic purpose of
(23) was to provide a lexical entry that could only function as an adjunct.
With this semantic nature, it would not be at all surprising to find such an
NP showing up in a predicative construction with a dummy subject.
Consider:
(35) a. That New York is far from Boston is obvious to everyone.
b. It is obvious to everyone that New York is far from Boston.
Sentence(35) shows a somewhat different case of a dummy subject. If we
think of obvious as denoting a relation between an EXPERIENCER and a
PROPOSITION, then in (35b), both arguments have been supplied within the
verb phrase, and a dummy subject is the natural way to fill the requirement
of English grammar that each sentence have a subject. Now the option of
taking a dummy subject and supplying all arguments within the verb phrase
is a marked one in English grammar. Except for our path nouns, the other
cases all seem be heads taking propositional arguments. 5 It thus appears
that some stipulation would be needed to license the option for distance
nouns. But that stipulation would be a good deal less remarkable if the
dummy subjects were licensed for non-denotational nouns with all their
arguments in the VP
Consider on the other hand, the bifunctional account of (21). The point
there was to deal with adjunct and denoting uses in a single lexical entry.
But on such an account some distance from Boston to New York is just an
ordinary denoting NP Although we could still posit a lexical rule licensing
a dummy subject construction for this NP, there is no particular reason to
expect it.
5
Possibly property arguments are just as good, as in:
It costs twenty dollars to get to Scranton.
It is worth comparing the distance nouns to other nouns associated with

dummy subjects, to suggest that there is more to this story than has yet
been told:
(36) a. It is a fact that New York is far from Boston.
b. it was the day after Christmas.
In (36a) we see a proposition-complement noun fact taking a dummy-it
subject.6 Sentence (36b) is more closely analogous to the case of the
distance nouns, in that another adjunct noun is implicated:
(37) John arrived the day after Christmas
What is interesting is that it is harder to argue for (36b) that the dummy
subject is licensed because all availabale argument positions have been
filled up. The adjunct noun still ought to have its event argument position
left open. I can only suggest that in (36b) the event argument position is
filled by the reference time of the discourse.
One loose end still left is the acceptability of
(38) It is a long distance to Boston.
Here the SOURCE is not present, but the dummy subject is still licensed.
This would be analogous to the cases where the argument of a proposition-
al adjective like obvious is missing, but a dummy is still possible:
(39) It's obvious!
I think it is plausible to argue that the source in (38) is discourse-bound.
That there is a semantic position available for discourse and other kinds of
binding is argued by sentences like:
(40) Wherever you live in Connecticut, it is still a long distance to
Boston
What I still have no explanation for is why this kind of omission is not
possible for the GOAL:
(41) *It is a long distance from New York, (dummy-reading)
6
It is striking here that the same kind of definteness restriction we saw on the
distance nouns seems to apply to the propositional nouns with dummy subjects:
?It was the fact that Bill coming.
As noted in 2.1, there are several asymmetries between SOURCE and GOAL
in these constructions. The missing piece here, and perhaps the most
interesting piece of all, is the role of these notions in the semantics of
distance.
Here is what looks like another example of such an asymmetry. If the
sentences in (31) are analogous to ordinary cases of extraposition, one
would not expect either of the two prepositional phrases to be special.
Thus, the following contrast may still need to be accounted for:
(42) a. *How far from New York is it to Boston?
b. ?How far to Boston is it from New York?
3. Conclusion
We find distance heads in the following kinds of uses. In each case, the
head in question has been italicized:
(43) a. The distance from New York to Boston is 300 miles.
b. Boston is a short distance from New York. (License A)
c. Boston is a short drive from New York. (License B)
d. It is a short distance from New York to Boston. (Dummy-
Subject)
e. It is a short drive from New York to Boston. (Dummy-
Subject)
I would like to conclude by focusing on the significance of the entry of
dummy subjects into this construction, particularly in cases of travel nouns,
as in (43e). I suggested in 2.1 that occurrences of travel nouns in the
adjunct function were a case of constructional polysemy. Sentence (43e)
shows that constructions feed lexical rules. More precisely, it shows that
extending the basic meaning of the distance-noun adjunct construction to
travel nouns can license lexical entries to feed other lexical processes, here
the rule that licenses dummy subjects.
References
Chomsky, Noam
1981 Lectures on Government and Binding. Dordrecht: Foris.
Fillmore, Charles
1968 "The Case for Case." In E. Bach, and R. Harms (eds.), Universals of
Linguistic Theory. New York: Holt, Rinehart, and Winston.
1987 "The mechanisms of construction grammar." BLS 14.
Fillmore, Charles, Paul Kay & M.C. O'Connor

1988 "Regularity and idiomaticity in grammatical constructions: The case of Let
Alone" Language 64.501-538.
Gawron, Jean Mark

1988 Lexical representations and the semantics of complementation. New York:
Garland.
Goldberg, Adele
1991a "On problems with lexical rule accounts of argument structure." Cognitive
Science Conference Proceedings.
1991b "The inferent semantics of argument structure: The case of the English
ditransitive construction." Cognitive Linguistics
Morgan, Jerry
1978 "Two types of convention in indirect speech acts." In P. Cole (ed.), Syntax
and Semantics 9: Pragmatics. New York: Academic Press.
Nunberg, Geoffrey and Ann Zaenen

1992 "Systemic polysemy in lexicology and lexicography." Proceedings of Euralex
1992, Tampere, Finland.
Pollard, Carl and Ivan A. Sag

1992 Head-driven phrase structure grammar. University of Chicago and CSLI
Publications.
Lexical information and the temporal
interpretation of discourse
Ferenc Kiefer
Research Institute for Linguistics,
Hungarian Academy of Sciences
The question of how to account for the temporal structure of narrative

discourse has often been addressed in linguistic literature. The principles
which have been proposed were mostly based on aspect and/or tense (Dow-
ty 1986; Hopper 1979; Kamp 1979 and 1981). However, accounts relying
solely on these categories failed to explain the ways in which temporal
interpretation of texts works. Lexical information was not taken into
account except - in certain cases at least - for the information concerning
aspectual class membership. A number of observations seem to suggest that
temporal structure can best be accounted for in terms of practical
inferences based on defeasable knowledge. By adopting this view, however,
we give up the idea that significant clues for temporal interpretation are
encoded in grammar. In the present paper it will be argued that lexical
information is of prime importance for temporal interpretation and that
practical inferencing is only needed in cases where lexical information falls
short.
Both lexical information and everyday knowledge are integrated into
what Fillmore calls 'scenes': "The scenes we construct for texts are partly
justified by the lexical and grammatical material in the text and partly by
the interpreter's own contributions, the latter being based on what he
knows about the current context, what he knows about the world in general,
and what he assumes the speaker's intentions might be." (Fillmore 1976: 92)
If it is true that temporal structure is largely based on lexical information
then this information must be enriched by information about event
structure. Though Fillmore has never spelled out this claim, it is implicit in
112 FERENC KIEFER
his account of lexical items in terms of frames. (Fillmore 1975, 1976 and
1982)
The paper begins with a brief overview of some of the earlier work on
temporal interpretation. Particular attention will be paid to the exceptions
(or counterexamples) which were noted with respect to each of these
approaches. It will also be pointed out that a number of these exceptions
have been known for some time from research in text linguistics.
The second part of the paper will be devoted to the discussion of the
types of lexical information which may be relevant for temporal interpreta
tion. It will be shown that a significant number of problematic cases can
easily be handled if this information is taken into account.
1. Perfectivity
As far as I know, the first mention of the role of perfectivity in the

temporal interpretation of discourse can be found in Wacha 1978. Using
Hungarian examples Wacha points out that only perfective verbs can be
used to express temporal progression, imperfective verbs do not have this
function. In Hungarian most perfective verbs are formed by prefixation.
Thus, in example (1) the verb form összeszedte 'picked up' contains the
prefix össze, and the verb form elment lit. Vent away, i.e. left' the prefix el.
The verb forms in (2), on the other hand, do not contain any prefixes, they
are imperfective.
(1) Összeszedte a holmiját és elment
'He picked up his things and left'
(2) Írt és olvasott egész nap
'He was writing and reading all day'
In (1) the second event follows the first event, hence the two events
express temporal advancement. In (2), on the other hand, the two events
are overlapping, there is no temporal advancement. It should be made clear
that Wacha did not claim that perfective verbs always express temporal
advancement; he did claim, however, that only perfective verbs are capable
of expressing temporal advancement without any other lexical material or
contextual support. In fact, he used this property to define perfectivity: a
verb is perfective if and only if it can express temporal advancement.
Since aspect is compositional (Verkuyl 1972) Wacha's claim can be
reformulated in the following fashion.
THE TEMPORAL INTERPRETATION OF DISCOURSE 113
(3) Perfective sentences may be used to express temporal advance

ment, imperfective sentences cannot.
According to (3) there is room for cases where perfectivity does not
express temporal progression. In fact, some such cases were noted in the
literature (Pete 1983). Consider
(4) Peter megkopaszodott es elhizott
'Peter got bald and fat'
(5) Átkutattak es összeturkáltak mindent
They rummaged and ransacked everything'
(6) Amikor elmentem hazulról, megpillantottam a tolvajt
'When I was leaving home, I caught sight of the burglar'
'When I had left home, I caught sight of the burglar'
In (4) we have to do with resultative change of state verbs. The verb
forms contain the prefixes meg- and el-, respectively. Resultativity goes hand
in hand with perfectivity in these cases. The two events in (4) are overlap
ping, which would remain true even if the two events were expressed by
separate sentences. Apparently, resultative changes of state suggest
simultaneity rather than progression.
As to (5) notice that the two verbs are almost synonymous; they both
mean something like 'to search a place thoroughly' and the verb ransack,
more so than rummage, may also mean that 'the place was left in a mess'.
On the other hand, rummage, but perhaps not ransack, means that the
search was carried out 'by moving things around in a hurried or careless
way'.1 The English verbs ransack and rummage, as well as their Hungarian
counterparts, may be considered to be quasi-synonymous. Example (5) thus
suggests that quasi-synonymous verbs -even if they are perfective (notice
that the Hungarian verbs contain the prefixes össze and at, respectively)
-express overlapping events.
Sentence (6), once again, contains two perfective verb forms elmentem
lit. 'I went away' and megpillantottam 'I caught sight' which contain the
perfectivizing prefixes el- and meg-, respectively. Though (6) may express
temporal progression, i.e. I first left home and then I caught sight of the
burglar, it may also indicate simultaneity, hence it is ambiguous. This
ambiguity is made explicit by the English translations of (6).
The Hungarian verb összeturkal 'ransack' has clearly the meaning that 'the place was
left in a mess', while the verb átkutat doesn't have such an implication.
114 FERENC KIEFER
2. Tense
Kamp and Rohrer have proposed in a number of works (Kamp 1979, 1981,
Kamp & Rohrer 1983) that in French temporal structure is determined by
tense. The passé simple (replaced by the passé composé in spoken language)
is used to express temporal advancement whereas the imperfait expresses
the background or the discourse situation. The contrast between the two
tenses can clearly be demonstrated by sentence pairs such as (7)(a) and
(7)(b).
(7) (a) Quand Pierre entra, Marie téléphonait
'When Peter entered, Mary was making a phone call9
(b) Quand Pierre entra, Marie téléphona
'When Peter entered, Mary made a phone call'
In (7)(a) the first verb form is in the passé simple, the second one in the
imperfait. In (7)(b), on the other hand, both verb forms are in the passé
simple. In (7)(a) the second event expresses the discourse situation, i.e. the
background for the first event and there is no temporal advancement. In
(7)(b), on the other hand, the second event follows the first event. This
observation has led to the formulation of the following principle.
(8) If the sentence is in the passe simplé, its event follows and does
not overlap with the event of the previous sentence. But if this
sentence is in the imparfait, its event overlaps temporally with
that of the previous sentence.
This principle is obviously too strong. In the literature the following
exceptions were noted (Kamp - Rohrer 1983; Vet - Molendijk 1986).
(9) L'été de cette année là vit plusieurs changements dans la vie de
nos héros. François épousa Adèle, Jean-Louis partit pour le Brésil
et Paul s'acheta une maison à la campagne
'Last summer there were several changes in the lifes of our
heroes. Frank married Adela, John left for Brazil and Paul
bought a house in the country-side'
(10) Marie chanta et Pierre l'accompagna au piano
'Mary sang and Peter accompanied her on the piano'
(11) Nadine s'assit et lut la lettre. Elle fut étonnée de constater que la
lettre la touchait beaucoup
'Nadine sat down and read the letter. She was surprised to find
out that the letter affected her a lot'
(12) Jean monta dans sa chambre. Pierre le suivit
'John went up to his room. Peter followed him'
The first sentence in (9) expresses the main event and the three clauses
of the second sentence represent subevents of this main event. Notice that
the subevents are unordered: They may have occurred in any order within
the time interval defined by the main event. Notice, furthermore, that the
first sentence of (9) contains the quantifier plusieurs 'several'. It would seem
that, in general, subevents of a main event do not bring about temporal
advancement.
In (10) the second clause contains the verb accompagner 'accompany'.
When you accompany a singer or a musician, you play one part of a piece
of music while they sing or play the main tune. The accompanying event is
always simultaneous with another performing event. This is equally true
when accompany means 'escort': If you accompany someone to the church,
the accompanying event is simultaneous with the going-to-the-church event.
Consequently, it is not at all surprising that (10) expresses two simultaneous
events.
The texts in (11) and (12) exemplify yet another case. Notice that the
second event in the first sentence in (11) may slightly overlap with the first
event. Nadine could have started reading the letter while still standing. The
first event of the second sentence, too, contains the passé simple, one would
thus expect that its event follows and does not overlap with the previous
event. This is, of course, not the case: Nadine might already have felt the
surprise while reading the letter. The situation is quite similar in the case
of (12): A considerable overlap between the two events is quite conceivable.
Evidently, principle (8) cannot be true. Thus Kamp and Rohrer 1983
propose to replace it by principle (13).
(13) The event introduced by the next passé simple may not be
encoded in the DRS (= Discourse Representation Structure) as
entirely preceding the event or time which at that stage of the
representation construction functions as reference point.
Though this principle is couched into the terminology of discourse

representation theory, its impact should be completely clear. The only case
which this principle excludes is the case when an earlier event is expressed
later. To be a little bit more precise, let us assume that the text at hand
116 FERENC KIEFER
contains thé following sentences: S1 S2,...,Sn- Each sentence expresses an

event, that is, the sequence of sentences can be mapped into the sequence
of events e1 e2, ..., en. What is then excluded by principle (13) is the case
whene1precedes entirely ei-1. All other possibilities are admitted: Complete
overlap, partial overlap (in both directions) and 'unmarked' order without
overlap (i.e. when ei follows ei-1). Consequently, this is a rather week
principle indeed and even this week principle is untenable, as we will see
later on.
3. Tense and aspectual class membership
Dowty (1986) has argued that the Past Tense of certain verbs expresses
temporal advancement. One of his examples is repeated under (14).
(14) John arrived at 8 o'clock. He went into his room and started
reading a book. He felt warm and opened the window
The verbs which can be used in this fashion are either achievement
verbs or accomplishment verbs. The relevant principle can be found under
(15).2
(15) The Past Tense of an achievement or accomplishment verb
moves the narrative forward.
Achievement and accomplishment verbs are aspectually perfective, but
there are perfective verbs which are neither achievement nor accomplish
ment verbs. We will return to this question presently.
Dowty mentions some counterexamples to his principle (15) (Dowty
1986: 58), which are repeated under (16)-(18).3
(16) Pedro dined at Madame Gilbert's. First there was an
hors-d'oeuvre. Then the fish. After that the butler brought a glazed
chicken. The repast ended with a flaming dessert
2
It should he made clear that Dowty's formulation is different from principle (15).
His temporal discourse interpretation principle (TDIP) doesn't make any reference to
aspectual class membership. Temporal structure is inferred on the basis of the interplay
between TDIP and aspectual class membership. (Dowty 1986: 45).
3
Dowty attributes (16) to Kamp and (18) to Dry. (Dowty 1986: 58).
(17) At the signal, everyone went to work at once. Mary searched the
room for any of the items on the list that might be there. John
went next door to do the same in Bill's apartment Susan orga
nized the rest of the players to canvass the block
(18) John knelt at the edge of the stream and washed his face and
hands. He washed slowly, feeling the welcome sensation of the icy
water on his parched skin
Example (16) is in a way parallel to (9). The first sentence represents
the main event followed by a sequence of sentences each describing a
subevent of the event mentioned in the first sentence. The subevents are
ordered: The order in which they occurred is indicated by lexical means:
First, then, after, end. Notice, however, that if we omitted from (16) the
lexical indication of temporal order, we would still interpret the subevents
in (16) in the same order. The interpretation would then be based on
everyday knowledge rather than on lexical information. We simply know
that the hors-d'oeuvre precedes the main dish and that a meal is normally
finished by a dessert.
Example (17) is not really different from (16). The first sentence
describes the main event and the subsequent events are interpreted as
subevents of the main event. The subevents are unordered: Neither lexical
nor everyday knowledge indicates any temporal order. Notice, incidentally,
that the sentence describing the main event contains the quantifier everyone.
As to (18), at first glance it is not quite clear whether it constitutes a
genuine counterexample to principle (15): Neither kneel nor wash seems to
be an accomplishment/achievement verb. However, (18) is ambiguous: It
has an imperfective and a perfective interpretation. In order to show this
the only thing one has to do is to translate (18) into a language which
makes a systematic distinction between imperfective and perfective verb
forms. Hungarian is such a language. And, indeed, kneel can be rendered
either as térdel (imperfective) or as letérdel (lit. kneel down, perfective) and
wash can be either mos (imperfective) or megmos (perfective). On the
imperfective interpretation we don't expect any temporal progression. If, on
the other hand, (18) is interpreted perfectively, the sequence of events is
expected to express temporal advancement. Contrary to this expectation,
however, the first and second sentences in (18) describe overlapping events
on both interpretations. A characteristic feature of this example is the
repetition in the second sentence of one of the predicates of the first
sentence.
118 FERENC KIEFER
Dowty goes much further in his conclusion than Kamp and Rohrer
1983. He suggests that the temporal discourse interpretation principle "must
be modified to allow that if the discourse itself conveys some implication as
to how events are to be specifically ordered, this should take priority over
(the principle) that orders reference times successively. In other word, (this
latter principle) is the 'default case' to be followed when neither time
adverbials nor entailments and implicatures of the discourse itself give clues
to the ordering of events." (Dowty 1986: 58)
Dowty's conclusion seems further be corroborated by an example due
to Ritchie 1979. Compare
(19) When they built the 39th Street bridge...
(a) a local architect drew up the plans
(b) they used the best materials
(c) they solved most of their traffic problems
Depending on the context the event described by the when-clause can
be prior to (cf. (19) - (c)), simultaneous with (cf. (19) - (b)), or posterior to
(cf. (19) - (a)) the event described in the main clause. Notice, however, that
this temporal interpretation can also be considered to be a consequence of
lexical information. Let's first have a closer look at (19)-(a). When you plan
something that you are going to make, build, or create, you decide what the
main parts of it will be and how they will be arranged. In other words, the
planning event must precede the building or creating event. As to (19)-(b),
the verb use implies simultaneity. If you use something such as a tool,
service, material, or a course of action, you do something with it in order
to do a job, achieve something, or solve a problem. On this reading the
verb use means 'to use something for doing something'. In other words, the
using event is simultaneous with the building event. Finally, to solve a
problem means to find or provide a solution to it. The solution is a
consequence of certain actions: You do something which eliminates a
problem. This means that in (19)-(c) the building event must precede the
solving event, the latter being in a way the culmination of the former.
4. Text linguistics
Research in text linguistics has long since claimed that there is a finite
number of text constituting principles which ensure the coherence of texts.
(van Dijk 1977; de Beaugrande and Dressier 1981; Longacre 1983; Sperber
and Wilson 1986) Some of these principles have direct bearing on temporal
interpretation. They are exemplified in (20) - (26) (Isenberg 1971).
(20) Common topic
Mein Bruder hat sich einen Anzug gekauft. Peter ist die Treppe
heruntergefallen. Meine Tante hat sich den Arm gebrochen. Dies
alles erfuhr ich gestern morgen
'My brother bought a suit for himself. Peter fell from the
staircase. My aunt broke her arm. I was told all this yesterday'
(21) Specification
Gestern ist ein Unglück geschehen. Peter hat sich den Arm
gebrochen
'An accident happened yesterday. Peter broke his arm.'
(22) Causal link
Anna hat sich den Arm gebrochen. Sie ist gestern vor ihrem Haus
ausgerutscht und gefallen
Ann broke her arm. Yesterday she slipped in front of her
house and fell.'
(23) Diagnostic interpretation
Peter ist angekommen. Anna hat sein Fahrrad gesehen
Teter arrived. Ann saw his bicycle.'
(24) Pragmatic presuppositions
Der Junge ist ins Kino gegangen. Irgendjemand hat ihm Geld
gegeben
The boy went to the movies. Somebody gave him money.'
(25) Specification of the reason ('Motiv-Anknüpfung')
Hans ist hinausgegangen. Jemand hat das Fenster eingeschlagen
'Hans went out. Somebody broke the window.'
(26) Adversative contrast
Peter hat sich ein neues Auto gekauft. Sein Bruder hat dagegen
seine Wohnung neu eingerichtet
Teter bought a new car. His brother, on the other hand,
refurnished his apartment.'
Notice that in (22), (23), (24) and (25) the second event must precede
the first event, which means that these examples constitute violations of any
temporal interpretation principle which claims to be 'universal' and excludes
cases in which from among two events the earlier event is described in the
second sentence. Examples (20) and (26) express simultaneous or
overlapping events and example (21) contains one single event.
120 FERENC KIEFER
The 'common topic' case is essentially the same as the examples in

French and English illustrating the 'main event - subevents' case. Notice
also that the last sentence in (20) contains the quantifier alles 'everything'.
Examples (22) through (25) are different from Ritchie's bridge-building
example since the succession of events is not predictable on the basis of
lexical information. It can only be deduced from text constituting principles
and everyday knowledge. The temporal structure of (21), on the other
hand, can to some extent be predicted. Notice that the first sentence does
not provide sufficient information. We also may wish to know what the
accident was. The information deficit exhibited by the first sentence in (21)
is similar to the lack of informativeness in question-answer pairs such as
(27).
(27) Has anything happened?
-Yes
Finally, the adversative contrast in (26) is in part indicated by the
presence of the adversative conjunction dagegen 'on the other hand' and by
the parallel structure of the two sentences.
If causality is conceived very broadly, the examples (22) through (25)
can all be brought under the heading of causal relations.
By way of conclusion it should perhaps be noted that much work could
have been saved if scholars working on the temporal structure of discourse
had been paid due attention to earlier work in text linguistics.
5, The internal structure of events
It has become a generally accepted view that events are not primitive
entities but are internally structured. (Bach 1986; Moens-Steedman 1988)
Events contrast with states, they may be momentary or durative, or may,
but need not, contain a culmination point. Both achievement and accom
plishment verbs contain a culmination point.4 More precisely, they both
exhibit the following structure (Moens-Steedman 1988: 18).
4
The taxonomy of events was first introduced in Vendler 1967. For more recent
accounts cf. Croft 1991 and Smith 1992.
preparatory phase culmination consequent state
For example, the event described by They reached the top contains a
preparatory phase which consists of all the activities which precede the
achievement, the achievement itself is the culmination, and 'being on the
top' is the consequent state.
The distinction between events and states, duration and culmination gives
rise to the following taxonomy of events.
The taxonomy of events
events states
momentary durative
with culmination realize build a house love

reach the top eat a sandwich hate
win the race write a letter know
no culmination hiccup write perceive

sneeze read see
wink eat
run
Notice that all momentary verbs are by definition perfective. Conse

quently, momentary verbs with no culmination ('semelfactives') are
perfective verbs which are neither achievement nor accomplishment verbs.
Depending on context the very same sentence may exhibit different
event structures. This can be illustrated by the various interpretations of the
bridge-building example. Consider
(28) (a) they prepare they have built

to build the bridge
they build the bridge

122 FERENC KIEFER
(b) they build they have completed

the bridge
they complete the bridge
Moens and Steedman argue (1988: 22-23) that in the case of (19)(a) the
entire culminated process of building the bridge tends to become a
culmination, which is associated with preparations for, and consequences of,
the entire business, as in (28)(a), The drawing up of the plans is then, for
reasons which have to do with knowledge of the world, situated in the
preparatory phase. In (19)(b), "in contrast, people tend to see the building
of the bridge as decomposed into a quite different preparatory process of
building, a quite different culmination of completing the bridge and some
consequences". The resulting decomposition is given in (28)(b). The use of
the best materials is, once again, situated in the preparatory phase, which
is, however, different from the preparatory phase of the previous example.
Finally, (19)(c) has the same decomposition as (19)(a), but on the basis of
everyday knowledge the event described in the main clause is situated in
the consequent state.
On Moens' and Steedman's account, then, knowledge of the world
enters into play (i) when we decompose an event into a preparatory phase,
a culmination and a consequent state and (ii) when we situate another
event with respect to this decomposition. Lexical knowledge is not called
for.
The above account suggests the following interpretation principle:
(29) From among two events ei and ej
(i) ej precedes ej if it is placed into the preparatory phase
of ej;
(ii) ei follows ej if it is placed into the consequent state of
e j;
(iii) ei is simultaneous with ej if it can neither be placed into
the preparatory phase, nor into the subsequent state of
e j.
If one allows for more than one decomposition, as Moens and
Steedman do, then simultaneity can also be arrived at in different ways, i.e.
both ei and ej can fall either into the preparatory phase or into the
consequent state.
Though Moens' and Steedman's assumptions about the internal

structure of achievements and accomplishments seem to be well-founded,
there are several problems with their proposal, (i) First, the 'algorithm'
does not work if ei and ej are not directly interrelated. It wouldn't make
much sense to order the events It started raining and I finished my paper for
Chuck with respect to each other, (ii) Second, though knowledge of the
world plays an important role in the interpretation of temporal structure (as
it does in discourse interpretation, in general), it would be misguided to rely
solely on everyday knowledge if linguistic clues are available. Linguistic
clues should take precedence over extralinguistic ones, (iii) Third, it would
follow from the internal structure of events and their possible ordering
relations that momentary events with no internal structure cannot be
ordered with respect to each other. Consider, however, (30).
(30) Bill began to feel cold. He sneezed
World knowledge helps us to temporally order events even if they don't
have any internal structure and hence cannot be decomposed into a
preparatory phase, a culmination point and a consequent state.
6. Lexical information and temporal structure
In this section we are going to inquire into the role of lexical information
in the temporal interpretation of discourse in more detail. It would seem
that the relevant lexical information may basically be of two kinds. One type
of lexical information is directly related to temporal structure, i.e. temporal
ity is encoded in the lexicon. In this case the predicate contains an event
variable as one of its arguments. Since events presuppose temporality, there
must be a temporal relationship between the event expressed by the
predicate and the one expressed by its argument. This relationship, too,
must be part of lexical representation.5 In the second case lexical informa
tion is only indirectly related to temporal interpretation. In some cases we
9
This is also implicit in Fillmore's description of scenes and frames. For example, the
elements in a prototypic commercial event scene are the buyer, the seller, the money that
changes hands, and the goods that change hands. (Fillmore 1976: 102-103) Consequently,
x bought y from z, z sold y to x, and x gave z money for y express essentially the same event
and cannot therefore be used to advance the narrative.
124 FERENC KIEFER
will have to stipulate interpretation principles on the basis of lexical

information, in the case of others, temporal interpretation is a corollary of
lexical information and pragmatic principles. For ease of exposition we will
use the examples cited earlier in this paper. Reference to these examples
will be given in parentheses after each heading.
A. Lexical information directly related to temporal structure

(a) The verbs use and accompany (Examples (19)(b) and (10))
The meaning of the verb use which is relevant in the present context
can be paraphrased as 'to use something for something' or 'to use
something while doing something'. The verb use can thus be assumed to
have three arguments, the third argument being an event argument. In the
case of the bridge-building example this event argument is the building
event. That is, we get the following representation. 6
(31) (a) USE(x,y,z) and z = BUILD(x,u)

The using event and the building event must be simultaneous:
(31) (b) SIMULTANEOUS(ei, ej)
ei = using event, ej = building event
The verb accompany must contain the meaning element 'to do
something while someone else is doing something' in its lexical characteriza
tion, as we already saw. For a musical event we get the representation
(32)(a).
(32) (a) ACCOMPANY(u,v) where both arguments are events:
u = PLAY-PIANO(x) and v = SING(y)
The verb accompany temporally relates the two events with each other.
That is, we have to add (32)(b) to the lexical characterization of accompa
ny:
(32) (b) SIMULTANEOUS(ei,ej)
ej = piano-playing event, ej = singing event
6
The lexical representations to be proposed here are utterly oversimplified. They are
process semantics-like representations and thus similar to the ones proposed in Miller &
Johnson-Laird 1976.
(b) The predicate 'to draw up a plan' plan (Ex. (19)(a))
When you draw up a plan for something that you are going to make,
build, or create, you decide what the main parts of it will be and how they
will be arranged. The planning event presupposes two participants: An
agent and the object to be planned. Let us denote the predicate 'to draw
up a plan' by PLAN, which thus contains two arguments of which the second
argument is an event argument.
(33) (a) PLAN(x,y) and y = BUILD(u,v)
And, once again, the temporal relationship between the planning event
ei and the building event ej must be made part of the lexical representation.
(33) (b) PRIOR-TO(ei,ej)
(c) The verb solve (Ex. (19)(c))

If you solve something you do something which eliminates a problem.
The actions you take in order to solve a problem must always be prior to
the solution. Thus, the verb solve contains two arguments, the first
argument is an event argument and the temporal relationship between the
solving event ei and the building event ej must be part of the lexical
representation.
(34) (a) SOLVE(x,y) and x=BUILD(u,v)
(b) POSTERIOR-TO(ei,ej)
ei = solving event, ej = building event
There is a generalization which seems to emerge from the above lexical
characterizations: If an argument of a predicate is itself an event, then the
temporal relation of the two events (the event expressed by the predicate
and the event expressed by one of its arguments) must be part of its lexical
representation.
If my conclusion is correct, the bridge-building example can be
interpreted without reliance on everyday knowledge, the temporal
interpretation being a consequence of lexical representation. This observa
tion is further corroborated by the following example.
(d) The verb follow (Ex. (12)
If you follow someone or something that is moving or travelling, you
move along behind them or you go to the place that they have recently
126 FERENC KIEFER
gone to, and where they are now staying. Compare They followed after me
and He followed them to Venice. In principle, with verbs of movement both
interpretations are possible. Notice, however, that if the first event is not
yet completed, only the first interpretation is available. E.g. Bill got up and
made for the stairs. Mary followed. On one of its readings the verb follow
thus means 'to move behind in the same direction'. As to temporal
structure note that one of the arguments of follow is an event variable. That
is,
(35) FOLLOW(x,y) and y = MOVE(u,v)
which means that 'x follows the event y which consists in u's moving toward
v'. Since the lexical characterization of follow contains an event variable,
according to the above generalization the temporal relationship of the two
events (the moving event and the following event) must be made part of
the lexical representation.
(36) (a) OVERLAP(ei,ej)
e j = following event, ej = moving event
Condition- ej is not as yet completed
(b) POSTERIOR-TO^ej) otherwise
B. Lexical information indirectly related to temporal structure

(a) Resultative change-of-state verbs (Ex.(4))
Resultative change-of-state verbs denote, as we saw above, overlapping
events. This state of affairs is independent of whether the two change-of-
-state verbs occur coordinated in one sentence or in separate sentences.
Overlap does not mean simultaneity, that is, in addition to simultaneity, we
have to postulate the temporal relationship OVERLAP (x,y) between two
events. Furthermore, this temporal relationship cannot be encoded in the
lexicon, and apparently it cannot be deduced from other properties of
change-of-state verbs either. Consequently, it must be stipulated. The
temporal relationship OVERLAP(x,y) can, of course, be overriden by temporal
adverbials, (First he got fat, then he got bald) but it cannot be overriden by
any other lexical material. The situation is different in the case of the
temporal relationships directly encoded in the lexicon which cannot be
overriden by temporal adverbials either.
Resultative change-of-state verbs all contain the predicate BECOME-
(x,Adj). We may thus stipulate the following interpretation principle.
(37) If there is no indication to the contrary interpret two consecu
tive resultative change-of-state verbs as expressing overlapping
events. That is, if ei=BECOME(x,Adji) and ej=BECOME(x,Adjj),
then OVERLAP(ei,ej).
It is perhaps not necessary to restrict (37) to identical subjects: Two
different subjects do not seem to affect temporal interpretation: He got old
and she got fat,
(b) The verbs rummage and ransack (Ex. (5))
The verbs in question have a common core meaning which can be
rendered by means of the predicate SEARCH(x,y). Though it may be unclear
where to draw the line between core meaning and peripheral meaning on
a principled basis, for the two verbs at hand the solution is quite straightfor
ward. We may refer to verbs whose meaning differs only in the periphery
as being quasi-synonymous. The corresponding events are simultaneous
because they denote, in fact, different aspects of the very same event. That
is, the following generalization seems to hold.
(38) Quasi-synonymous verbs denote different aspects of the same
event.
The temporal interpretation is a corollary of (38): The temporal
relationship between the two events is that of simultaneity.
Similar things hold for (18), where we have the core meaning WASH(x,y)
in the relevant part of the text, which occurs twice: In the second case it is
modified by SLOWLY:
SLOWLY(WASH)(x,y))
(c) Emotives (Ex. (11))
Emotive reactions may accompany an event or they may be a conse
quence of an event. That is, emotions create ambiguity. One may be
surprised while reading a letter or after having read it. Since, however, the
letter is the cause of the emotional event, the latter cannot precede the
event of reading the letter. Let ei be the causing event and ej the emotional
event. The temporal relationship between ei and ej can be either (39)(a) or
(b).
(39) (a) OVERLAP(ei ,ej)
(b) PRlOR-TO(ei,ej)
128 FERENC KIEFER
(39) (b) corresponds to the default case, therefore only (39)(a) has to
be stipulated
(d) Quantifiers (examples (9), (17), (20))

In sentences representing the 'common topic' we often find quantifiers
or quantifier-like expressions. In example (9) we find plusieurs changements
'several changes', in (17) everyone and in (20) alles 'everything'. In all these
cases we have to do with quantification over events. In a sequence of events
the quantified event is either the first one or the last one. The following
principle seems to hold.
(40) If in a sequence of events e1,e2,...,en either e1 or en is a quanti
fied event, interpret the other events in the sequence as being
simultaneous or overlapping with the quantified event.
(e) Nouns containing an event variable (Ex. (21))

Nouns may contain an event variable which may, but need not, be
specified in the context. Such nouns include fact, the noun event itself,
incident, occurrence (on one of its readings), truth, deed, accident, etc. For
the noun accident, for example, we need a representation such as (41).
(41) HAPPENING(e) and UNPLEASANT(e)
If a noun contains an event variable, the Gricean maxim of quantity
requires the specification of this variable. ('Give as much information as
necessary'.) Consequently, if ei is the event containing the unspecified event
variable, the subsequent event ej will normally be interpreted as the
specification of this variable. But if ej is the specification of ei, ei and ej will
denote the very same event. The temporal relationship between them will
be that of identity.
The examples could easily be multiplied but this may suffice to show
how lexical representation may incorporate temporal information and how
temporal interpretation may make use of lexical representation.
7. Conclusion
In the present paper I have argued that lexical information can be used to
explain a great number of exceptions to the principles proposed in the
pertinent literature in order to account for the temporal interpretation of

texts. From among the exceptions I have offered explanations for (4), (5),
(9), (10), (11), (12), (17), (18), (19)(a)-(c), (20), (21) and (26). In the case
of examples (10), (12) and (19)(a)-(c) the temporal relationship is explicitly
encoded in the lexical representation. In the remaining cases temporal
interpretation is only indirectly related to lexical representation. The
consequences of lexical representation for temporal structure had to be
stipulated in some cases, while in others temporal interpretation was
considered to be due to the interaction of lexical representation and
pragmatics.
From among the exceptions cited in this paper (6) and (22)-(25) were
left unexplained. In (6) the temporal relationship between the event
expressed by the main clause and by that of the subordinate clause can only
be inferred on the basis of world knowledge. The same holds true for the
causal relations expressed by examples (22) through (25). Lascarides and
his collaborators (Lascarides 1990; Lascarides - Asher 1991; Lascarides -
Oberlander 1991) have argued that temporal structure of texts can best be
explained in terms of defeasible rules characterizing causal laws and
Gricean-style pragmatic maxims. For example, the text Max fell. John
pushed him. can be explained by assuming that the text is coherent and by
postulating the following defeasible rule: I believe that if Max fell and John
pushed him, then the pushing caused the falling, unless I already believe
something to the contrary. Except for aspectual class membership the theory
does not make use of lexical information. I have attempted to show that
lexical information plays a much more crucial role in temporal interpreta
tion of texts than assumed so far. To be sure, world knowledge cannot be
dispensed with totally, but one should perhaps minimize its role and restrict
its use to cases where linguistic clues are not available.
References
Bach, Emmon
1986 "The algebra of events". Linguistics and Philosophy 9.5-16.
de Beaugrande, Robert-Alain & Wolfgang Dressier

1981 Introduction to Text Linguistics. London: Longman.
Croft, William
130 FERENC KIEFER
1991 Syntactic Categories and Grammatical Relations: The Cognitive Organization

of Information. Chicago: The Chicago University Press.
van Dijk, Teun A.

1977 Text and Context. London: Longman.
Dowty, David R.
1986 "The effects of aspectual class on the temporal structure of discourse:
Semantics or pragmatics?" Linguistics and Philosophy 9.37-61.
Fillmore, Charles
1975 "An alternative to checklist theories of meaning". Proceedings of the First
Annual Meeting of the Berkeley Linguistic Society, 123-131.
1976 "Topics in lexical semantics". In P. Cole (ed.), Current Issues in Linguistic

Theory. Bloomington: Indiana University Press, 76-138.
1982 Frame semantics. In The Linguistic Society of Korea (ed.), Linguistics in the
Morning Calm. Seoul: Hanshin, 111-138.
Hopper, Paul
1979 "Aspect and foregrounding in discourse." In T. Givón (ed), Syntax and
Semantics Vol 12: Discourse and Syntax. New York: Academic Press, 213-241.
Isenberg, Horst
1971 "Überlegungen zur Texttheorie." In Jens Ihwe (ed). Literaturwissenschaft
und Linguistik. Frankfurt: Athenäum, 150-172.
Kamp, Hans
1979 "Events, instants and temporal reference." In Rainer Bäuerle, Urs Egli,
Arnim von Stechow (eds), Semantics from Different Points of View. Berlin:
de Gruyter, 376-417.
1981 "Evénements, représentations discursives et référence temporelle." Langages

64.39-64.
Kamp, Hans Sc Christian Rohrer

1983 "Tense in texts." In Rainer Bäuerle, Christoph Schwarze, Arnim von
Stechow (eds.), Meaning Use, and Interpretation of Language. Berlin: de
Gruyter, 250-269.
Lascarides, Alex
1990 "Knowledge, causality and temporal representation." Edinburgh: The
University of Edinburgh, Human Communication Research Centre.
Lascarides, Alex & Nicholas Asher

1991 "Discourse relations and defeasible knowledge." Human Communication
Research Centre. Edinburgh: The University of Edinburgh.
Lascarides, Alex & Jon Oberlander

1991 "Temporal coherence and defeasible knowledge." Edinburgh: The University
of Edinburgh, Human Communication Research Centre.
Longacre, Robert
1983 The Grammar of Discourse. New York and London: Plenum.
Miller, George A. & Philip Johnson-Laird

1976 Language and Perception. Cambridge: Cambridge University Press.
Moens, Marc & Mark Steedman

1988 "Temporal ontology and temporal reference." Computational Linguistics
14.15-28.
Pete, István
1983 "Az igeszemlélet, a cselekvés megvalosulásának foka, a cselekvés módja és
minősége a magyar nyelvben" (Aspect, resultativity and aktionsart in
Hungarian'). Magyar Nyelv 79.137-149.
Ritchie, George D.
1979 "Temporal clauses in English." Theoretical Linguistics 6.87-115.
Smith, Carlota S.
1992 The Parameter of Aspect. Dordrecht: Kluwer Academic Publishers.
Sperber, Dan & Deirdre Wilson

1986 Relevance: Communication and Cognition. Oxford: Blackwell.
Vendler, Zeno
1967 Linguistics in Philosophy. Ithaca: Cornell University Press.
Verkuyl, Henk J.
1972 On the Compositional Nature of the Aspects. Dordrecht: D.Reidel.
Vet, Co and Arie Molendijk

1985 "The discourse functions of the past tenses in French." In Vincenzo Lo
Cascio & Co Vet (eds.), Temporal Structure in Sentence and Discourse.
Dordrecht: Foris, 133-159.
132 FERENC KIEFER
Wacha, Balâzs
1978 "Aspect in predicate constructions (in Hungarian)." Doctoral Dissertation,
The University of Budapest.
Reflections on metaphor and grammar
George Lakoff
The University of California at Berkeley
Chuck Fillmore walked into my office one day and asked coyly if I had ever
thought about the sentence,
- From my office, I can see the bay
The sentence, he observed, raised the issue of how it is possible to
semantically unify the adverbial phrase "from my office", which designates
a source, with "I can see the bay" which is about perception. The answer,
he observed, had to be a metaphorical one, in which seeing involves the
metaphorical motion of the gaze from my eyes to the bay, as in the
sentence,
- From my office, I can see all the way to the bay.
Chuck was aware, of course, that Johnson and I, in Metaphors We Live
By, had pointed out the general conceptual metaphor SEEING IS TOUCHING;
EYES ARE LIMBS, which as as instances cases like:
My eyes picked out every detail of the pattern. I ran my eyes over the
walls. I couldn't take my eyes off her. His eyes are glued to the tv. My
gaze is out over the bay. Their eyes met.
In this conceptual metaphor, there is a mapping from the source domain
of TOUCHING to the target domain of SEEING. The word "gaze" designates the
metaphorical Visual limb' that reaches out, just was the word "eyes" can
designate such visual limbs. The metaphor maps the following source
domain knowledge onto target domain knowledge:
Source: Limbs can be directed.
Target: Vision can be directed.
134 GEORGE LAKOFF
Source: A limb can go in only one direction at a time.

Target: Vision can go in only one direction at a time.
Source: Limbs can extend from the body to other objects.

Target: Vision can move from the body to other objects.
Source: Tactile perception occurs when a limb touches an object.

Target: Visual perception occurs when the eye-gaze touches an object.
Source: Limbs can pick out up objects.

Target: Vision can pick out objects.
The metaphor, of course, is a conceptual mapping, permitting one to
comprehend vision in terms of touch. There are linguistic correlates to the
conceptual mapping, but it is primarily conceptual The evidence for the
mapping that Johnson and I had cited was primarily lexical. In each of the
sentences given above, vision is lexically portrayed in terms of touch.
However, in
- From my office, I can see the bay.
the main clause, "I can see the bay" has no lexical material indicating the
use of this metaphor. Instead, it is the syntax, the occurrence of a preposi
tional phrase indicating a source, that shows that the main clause is being
conceptualized metaphorically. And the semantic intergration of the from-
phrase and the main clause applies to the metaphoric interpretation of that
clause. This is anything but a surprise, given the theory of conceptual
metaphor, but it is eloquent confirmation of that theory, as well as
disconfirmation of classical semantics, in which there is no such thing as
conceptual metaphor.
Metaphorical contradictions - An aside
Of course, there is no reason to suppose that the simple sentence

- I can see the bay
is interpreted metaphorically. Thus, it appears that the metaphoric
interpretation of the main clause in the above sentence is triggered by the
from-phrase. Given this, an obvious question arises. The from-phrase "from
REFLECTIONS ON METAPHOR AND GRAMMAR 135
my office" requires a source interpretation for the perceiver. Would that

source interpretation be inconsistent with a main clause where the perceiver
is understood to be a receiver rather than a source. Consider, for example,
sentences that are understood via the metaphor PERCEPTION IS RECEPTION.
- The view blew me away
- The view knocks me over
- The mountain peeked at me through the fog.
In each of these sentences, vision is understood in terms of visual sense
impressions reaching the eyes. In the first two examples, the sense
impressions metaphorically strike the perceiver with force. In the third
sentence, the mountain is personified as as looking at the perceived. The
sight of the mountain is conceptualized as coming through the fog to the
perceiver. Suppose we take each of these sentences, where the perceiver
is metaphorically conceptualized as a receiver of sense impressions, and ask
whether they can be paired with a from-phrase, which requires that the
perceiver be conceptualized - via metaphorical contradiction - as a source.
The answer is no.
- *From my office, the view blew me away.
- *From my office, the view knocks me over.
- *From my office, the mountain peeked at me through the fog.
Given the theory of metaphor, this can be accounted for in a straightfor
ward way: The PERCEPTION IS RECEPTION metaphor conceptualized the
perceiver as a perceptual receiver, while the from-phrase conceptualizes the
perceiver as a perceptual source.
It should be noted, incidentally, that the theory of conceptual metaphor
is needed for this explanation to go through. However, the theory of
conceptual metaphor, by itself, does not predict these facts. What is
necessary is an auxiliary mechanism that puts together the metaphorical
perceiver interpretation with the source interpretation. It is simply a fact
that the metaphorical interpretation of the main clause does happen to be
combined semantically with locational adverbial phrases in such cases.
It is important to remember this because of the existence of the
following sentences:
- The view from my office blew me away.
- The view from my office knocks me over.
136 GEORGE LAKOFF
Here, the from-phrase is a modifier inside the noun phrase instead of an

adverbial modifier of the main clause. Again, the main clause metaphorical
ly presents the perceiver as a perceptual receiver. Again, the from-phrase
indicates that the perceiver is a perceptual source. BUT THERE IS NO
METAPHORICAL CONTRADICTION! It seems that the internal metaphorical
semantics of the noun phrase is segregated from the metaphorical semantics
of the main clause. It is as though "from my office" has the job of specifying
which view is being referred to, and then, its job done, ceases to be
relevant. The view thus specified then enters into the PERCEPTION IS
RECEPTION metaphor without contradiction. However, this is true only of
metaphorical contradictions, not literal contradictions.
-* T h eview that I couldn't see knocked me over.
- *The view that I couldn't see blew me away.
To my knowledge, none of the present theories of how modifiers are
interpreted can account for these phenomena. I have no non-ad-hoc
suggestion to make as to how to account for this phenomenon, and I invite
all semanticists to try to explain why this should happen.
The general case
Let us return to the main thread of the narrative. Chuck Fillmore's

purpose in walking into my office went beyond giving me the present of a
lovely confirmation of a decade-old metaphorical analysis. The very next
thing he asked was whether I had thought about sentences like
- From my office, I can hear the trains.
and
- From my office, I can smell the bay.
Though Johnson and I had observed the tactile metaphor for vision, we had
not observed any such metaphor for hearing or for smell. Yet here was
syntactic evidence for a more general metaphor of perception (visual,
auditory, and olfactory) in terms of touch. Fillmore's implicit question (he
loves to ask implicit questions!) was whether there is any lexical evidence
for such a more general metaphorical mapping.
Once the question is posed, an answer is possible: Yes! Take the
following sentences, which contain the expressions "pick up" and "pick out"
from the tactile domain, and apply them to the auditory and olfactory
domains:
- From the top of the mountain, I can pick up broadcasts from Moscow
on my radio.
- From my deck, I can pick out the sounds of construction on campus.
- Through the walls, I can pick out the aroma of the saffron in the
bouillabaisse being cooked in the next apartment.
- Through the piles of leaves that had fallen since the storm last night,
the dogs picked up the scent of the escaped prisoner.
So far, so good. But there were further implicit questions raised by the
sentences Fillmore gave me. The SEEING IS TOUCHING metaphor maps the
directability of limbs onto the directability of vision. Hearing and smell are
not directed in the way vision is. Sounds and smells are perceivable from
all directions at once. Could a single, more general metaphorical mapping
map directability for vision but not map directability for smell?
As one might expect with Chuck, his implicit question presupposed even
more implicit information. At the time, he and I were both teaching
seminars on inheritance in construction grammar. We differed in our
theories of inheritance. He claimed that inheritance could not have
overrides, and I, on the basis of my there-construction research (Lakoff
1987, case study 3), claimed that overrides were necessary. I had also
claimed that the Invariance Principle (Lakoff 1990) was a form of
inheritance (the source domain inferences are inherited by the target
domain) to which target domain overrides could apply. The directional
inferences about touch are mapped in the vision case, but not in the cases
of hearing and smell. Can this be due to target domain overrides of the
Invariance Principle?
Again the answer is yes. Suppose we assume that the more general
metaphor is of the following form:
PERCEPTION IS TOUCH
PERCEPTUAL ORGANS ARE LIMBS
and the knowledge mapped is:
Source: Limbs can be directed.
Target: Perceptual organs can be directed.
Source: A limb can go in only one direction at a time.

138 GEORGE LAKOFF
Target: A perceptual organ can go in only one direction at a time.
Source: Limbs can extend from the body to other objects.

Target: Perceptual organs can move from the body to other objects.
Source: Tactile perception occurs when a limb touches an object.

Target: Perception occurs when the perceptual organ touches an object.
Source: Limbs can pick out up objects.

Target: Perceptual organs can pick out objects.
There will, of course,, be special cases of this mapping, and as we saw
above, the special case of vision works. What about hearing and smell? The
Invariance Principle, as we know, requires target domain overrides; that is,
knowledge of the target domain cancels any knowledge mapped from the
source domain that is inconsistent with it. Since hearing and smell are
known not to be directed, the first two of the knowledge mappings given
above will be overridden. Thus, not only will these inferences be overridden,
but many of the lexical items that would be sanctioned by such mappings
are disallowed. Thus, we do not say:
- *He ran his ears over the crowd.
- *He ran his nose over the shoreline.
- *His nose was glued to the mountaintop.
- *He couldn't take his ears off the violinist.
Indeed, just about the only lexical items sanctioned by the remaining
inferences are "pick out" and "pick up".
But this more general metaphor itself raises some questions. Why
should the general version of the metaphor exist at all if it does so little
inferential and lexical work? One possible answer is that it does do some
work, and a more general principle that does some additional inferential
and lexical work is better than a more specific principle that accomplishes
less.
That is a sufficient explanation, but I think that there is an additional
reason why the general version of this metaphor should exist. That reason
is the phenomenon of duality: The dualities that have been found so far all
have the following character:
- There are two mappings from domain A to domain B.
- Both mappings involve relative motion between two entities, E and F.
- In one of the mappings E is fixed and F moves to E; in the other F

is fixed and E moves to F.
- The two mappings are really two special cases of a single more
general mapping in which only relative motion between E and F is
specified. In one special case, E moves to F, while in the other, F
moves to E.
The simplest example of such a case is the very widespread metaphor
for time, in which the passage of time is understood as the relative motion
between an entity understood as a time and an observer. In the case where
time is a moving object, the observer is facing toward the future, and future
times move toward him. In the case, where times are locations on a
landscape, the observer moves toward future times. Thus, we get the
minimal pair:
- Christmas is coming up on us. (moving time)
- We're coming up on Christmas, (moving observer)
For a full discussion, see (Lakoff 1993).
Duality also occurs for perception metaphors. In the general metaphor
we have seen so far, perception occurs when the perceiver moves his organs
of perception to the thing perceived and touches it. But, as we noted above,
there is also a metaphor of perception in which the thing perceived moves
to the perceiver's organs of perception. This has been referred to as
PERCEPTION IS RECEPTION. Examples include:
- A comet came into my sight.
- The noise came through the walls.
- The smell of the bay came through the fog.
- The aroma of garlic hit me as I walked into the restaurant.
- He was struck dumb by the sight of her.
We can now analyze PERCEPTION IS RECEPTION and PERCEIVING IS TOUCHING as
metaphorical duals. That is, they are special cases of a more general
metaphor.
PERCEPTION IS CONTACT BETWEEN PERCEIVER AND PERCEIVED.
Thus, we can attribute the existence of a general PERCEPTION IS TOUCHING
metaphor to the tendency for metaphorical duals to form, and for
generalizations to be stated at a level that is neutral between figure and
ground.
140 GEORGE LAKOFF
Let us now return to the example first brought to me by Fillmore:

- From my office, I can see the bay.
Fillmore's point was that the semantic unification of the source phrase
"from my office" and "I can see the bay" can come about only if "I can see
the bay" is understood metaphorically, so that seeing is construed as a form
of reaching out and touching. The office then becomes the location from
with the reaching out originates.
This means that that traditional semantic theories, which do not take
conceptual metaphor into account, cannot handle such examples. It also
means that such examples are handled naturally and straightforwardly by
the contemporary theory of metaphor, which sees conceptual metaphors as
fixed, conventional correspondences across domains that can impose
structure from the source domain onto the target domain. Here the
structure imposed is one in which the visual organs can reach out from the
location of the body to what is perceived.
Metaphorical experiences
But the implications do not stop there. There are implications as well for
the classical theory of semantic roles first suggested by Fillmore himself.
Consider Fillmore's Experiencer role. In analyses dating back to the 196Q's,
Fillmore has insisted on distinguishing the Experiencer from semantic roles
like Agent, Patient, Source, and Goal. An Experiencer is someone who
experiences perception, emotion, or cognition, no matter what grammatical
role that person plays in a given sentence. Examples of Experiencers
include the italicized first-person pronouns in the following sentences:
- I can see the bay.
- That idea just came to me.
- John looks sick to me.
- The idea struck me as silly.
- I was letting off steam.
- I'm picking up Radio Moscow.
What I would like to suggest is that the Experiencer role is a very general
role, and that it does not strictly contrast with Agent, Patient, Source, and
Goal. Rather, conceptual metaphors can impose all of those roles on the
experiencer role, with the result that the metaphors, for the most part,
determine the grammatical relations borne by the experiencers.
Let us begin by returning to the sentence,
- I can see the bay from my office.
As Fillmore observed, "I can see the bay" must be metaphorically under
stood in order for semantic unification to take place with the source phrase
"from my office". But what are the details of that unification? What, for
example, is the role played by "the bay"? In the metaphor, my gaze is
extending to the bay. That is, the bay is the metaphorical goal. And, indeed,
it is marked as such in sentences like:
- From my office I can see all the way to the bay.
Here, "the bay" which is the Content of the perception in "I can see the
bay" is also the metaphorical goal, and is grammatically marked as such in
the Path+Goal phrase "all the way to the bay".
Now consider:
- That idea just came to me out of the blue.
"Me" would be merely an Experiencer on Fillmore's semantic role analysis,
since it indicates the person experiencing cognition. But "me" is marked
grammatically as a goal, which is just what it is metaphorically. The general
conceptual metaphor involved here is the CONDUIT metaphor, according to
which ideas are objects that can be sent and received. "Out of the blue" is
a metaphorical source phrase, and "That idea" is not just the Content of the
cognitive experience, but is also the metaphorical Theme that moves to
"me." The grammar of the sentence is a reflection of the metaphor. That
is, it has the grammar of a literal Theme-Goal-Source sentence, like the
literal "The dog came to me out of the kennel." To put it another way, the
sentence has source domain syntax. This contrasts with "I can see the bay,"
which has target domain syntax.
Now consider
- John looks sick to me from the pictures.
This sentence is about the acquisition of information via perception; that
is, "me" is an Experiencer of perception and cognition. Again we have a
case of metaphor, in which "from the pictures" is a source phrase; the
pictures are the metaphorical source of both perception and information.
"Me" is the metaphorical goal, and is expressed grammatically as a goal.
142 GEORGE LAKOFF
But, whereas "came" in the previous example is taken from the source
domain, in this sentence "look" is taken from the target domain. That is, the
grammar (TO X FROM Y) is taken from the source domain of the metaphor,
while the lexical item is taken from the target domain.
In the above examples, there are Experiencers which are metaphorical
Goals, and have the syntax of Goals, Now let us turn to a case where an
Experiencer is a metaphorical Patient and has the syntax of a Patient.
- The idea struck me out of the blue,
Again, we have the CONDUIT metaphor, with an idea that is conceptualized
as an object that comes from a source "out of the blue" to me, not just
reaching me as a goal but striking me. Thus, "me" is not merely a Goal, but
moreover, a Patient that is affected by being struck. The verb "struck" is
from the source domain, as is the syntax, in which "me" is direct object,
which is the natural grammatical relation for a Patient to have.
We have just seen that Experiencers can be metaphorical Goals and
Patients, with the corresponding syntax. They can also be metaphorical
Agents. Consider:
- I was letting off steam very carefully.
- I'm picking up Radio Moscow with some difficulty.
In both these sentences "I" is an Experiencer, an Experiencer of emotion
in the first sentence and an Experiencer of perception in the second. But
in both cases "I" is a metaphorical Agent, taking two of the grammatical
correlates that go with Agents but not with literal Experiencers: A
progressive auxiliary and a manner adverbial » "very carefully" and "with
some difficulty" select for agentive verbs. "Let off steam" and "pick up" are
both agentive predicates in the source domain.
In sum, Experiencers can play at least the metaphorical roles of Goal,
Patient, and Agent, and can have the corresponding source domain syntax
associated with those roles. The syntactic results could also be achieved by
arbitrary linking rules or arbitrary lexical constraints. But the conceptual
metaphors are needed independently, and since they are there to be used,
the most parsimonious solution is to make reference to them and avoid
linking rules that would only do the same work as the independently needed
metaphors.
The moral: Metaphor enters a grammar in at least three ways. (1)
Metaphor may be needed for semantic unification. (2) The clause grammar
may be a reflection of metaphorical semantic roles. And (3), a metaphorical
lexical item may be used. Thus, linking rules that specify how semantic roles
are paired with grammatical relations can often be dispensed with when we
take into account metaphorical semantic roles characterized by indepen
dently existing metaphors.
References
Lakoff, George & Mark Johnson

1980 Metaphors We Live By. Chicago: University of Chicago Press.
Lakoff, George
1987 Women, Fire, and Dangerous Things. Chicago: University of Chicago Press.
1990 "The invariance hypothesis." Cognitive Linguistics 1:1.39-74.
1993 "The contemporary theory of metaphor." In A. Ortony (ed.), Metaphor and

Thought (Second edition), Cambridge, U.K.: Cambridge University Press.
144 KNUD LAMBRECHT
D E N N I S T H E M E N A C E / H a n k Ketcham
Do those shoes h u r t ? N o , b u t m y feet are

KILLING me!
The pragmatics of case:
On the relationship between semantic,
grammatical, and pragmatic roles
in English and French
Knud Lambrecht
University of Texas, Austin1
1. Introduction
This paper is an investigation of the three-termed relation between the

semantic structure of propositions, the grammatical structure of sentences,
and the pragmatic structure of utterances. More specifically, it deals with
the relationship between semantic roles (agent and patient), grammatical
roles (subject and object), and pragmatic roles (topic and focus). It is an
attempt at integrating into the Fillmorean model of case grammar a
linguistic dimension without which I think the relationship between semantic
and grammatical roles cannot be fully captured: The dimension of the
informational structuring of propositions in discourse. My investigation is
centered on a small set of data (the English predicate hurt and its French
counterparts faire mal and avoir mal) and on a restricted set of construction
types. Given this limitation, the conclusions which I will draw will be
necessarily tentative. I nevertheless believe that the kind of analysis
1
Earlier versions of this paper were presented at the Second Colloquium on
Romance Linguistics at UT Austin, at the Department of Modern Languages and
Linguistics at Cornell University, and at the UC Berkeley Linguistics Colloquium. I am
indebted to John Aske, Suzanne Fleischman, Mirjam Fried, Adele Goldberg, Laura
Michaelis, Sue Schmerling, Eve Sweetser, and Robert Van Valin for valuable comments
on various versions of the manuscripts.
146 KNUD LAMBRECHT
attempted here, in which syntactic, semantic, and pragmatic facts are

analyzed in relation to each other, is of sufficient interest to linguistic theory
to be presented even in such limited form.
My analysis is based on the assumption that the relationship between
the structure of sentences and the information function of utterances is
regulated by rules and conventions of sentence grammar, at a grammatical
level which (following Halliday 1967) I call INFORMATION STRUCTURE, and
which I define as follows:
INFORMATION STRUCTURE: That component of grammar in which
propositions as conceptual representations of states of affairs are
paired with lexicogrammatical structures in accordance with the mental
states of interlocutors who use and interpret these structures as units of
information in given discourse contexts.
The information-structure component mediates between sentence meaning
and sentence form by creating PRAGMATICALLY STRUCTURED PROPOSI
TIONS. Such pragmatically structured propositions are then expressed as
sentences with lexical, morphosyntactic, and prosodic structure.
The theoretical framework used here is that developed in Lambrecht
1994 (see Section 2 below). There are relatively few references in this
paper to previous treatments of the interrelations between semantic,
syntactic, and pragmatic roles. This is not because I consider the work of
my predecessors unimportant, but because the present analysis, based on
a theory of FOCUS TYPES, is different enough from existing analyses to
make the comparison difficult. For summaries of previous and alternative
approaches to the issues explored here see Van Oosten 1985, Chapter 1 (in
particular pp. 15-17) and Lambrecht 1994, Chapter 1.
The paper is organized as follows. In Section 2, I will briefly define 5
the information-structure categories needed for my analysis. In Section 3,
I will summarize some well-known English facts pertaining to the semantic
subject-selection hierarchy established by Fillmore. I will then add to the
data the dimension of information structure and I will argue that parallel
to the semantic role hierarchy there is a PRAGMATIC ROLE HIERARCHY,
which determines the ways in which the pragmatic relations topic and focus
are matched with the grammatical roles subject and object. This pragmatic
role hierarchy interacts with the semantic hierarchy, resulting in a complex
system of interrelations between pragmatics, semantics, and syntax. I will
show, in particular, that the construal of the semantic role of an argument
in a sentence may be influenced by the information function of the sentence
THE PRAGMATICS OF CASE 147
in discourse. In Section 4, the situation in English will be contrasted with

that in French, a language in which the pragmatic role hierarchy has more
obvious consequences for the syntactic structure of a sentence than in
English, where differences in information structure are often coded by
prosody alone. When the requirements of discourse threaten to lead to
impermissible mappings of pragmatic and grammatical roles, French resorts
to certain relation-changing devices in order to prevent violations of the
hierarchy. These devices are of two kinds: Valence-changing bi-clausal
constructions and alternative lexical predicates permitting alternative
mappings of semantic and syntactic roles. In Section 5,I will summarize the
findings of this paper and I will argue that the phenomena analyzed here
support a view of the grammatical system in which pragmatic information-
structuring principles intervene directly at all meaning-bearing levels of the
linguistic system: Prosody, morphosyntax, and the lexicon.
2. Topic and focus
To describe the phenomena I have in mind I need to make reference to a

number of pragmatic concepts, which I have discussed in detail elsewhere
(Lambrecht 1994). For the present analysis the most important are
PRAGMATIC PRESUPPOSITION, PRAGMATIC ASSERTION, FOCUS, and
TOPIC. These concepts are defined in item (1) below. As they stand, the
definitions, especially that of 'topic', may seem insufficiently explicit to some
readers; considerations of brevity prevent me from further elaborating them
here. I trust, however, that these definitions will do for the purposes of the
present analysis:
(1) PRAGMATIC PRESUPPOSITION: The set of propositions lexico-
grammatically evoked in a sentence which the speaker assumes
the hearer already knows or believes or is ready to take for
granted at the time the sentence is uttered.
PRAGMATIC ASSERTION: The proposition expressed by a

sentence which the hearer is expected to know or believe or
take for granted as a result of hearing the utterance.
FOCUS: The component of a proposition whereby the assertion

differs from the presupposition.
148 KNUD LAMBRECHT
TOPIC: The referent of a linguistic expression which, in a given
utterance context, a sentence is construed to be about.
The definitions of 'focus' and 'pragmatic assertion' imply that if a sentence
evokes no presupposition (as e.g. It's raining), focus and assertion coincide.
Both 'focus' and 'topic' involve PRAGMATIC RELATIONS, i.e. they have
to do with pragmatically construed relationships between propositions and
their components. When talking about the grammatical manifestions of
these relations in given sentences I will use terms like 'topic expression',
'focus constituent', 'focus accent', etc. The definition of 'topic' crucially
involves the concept of ABOUTNESS, which I unfortunately cannot explicate
here (for details, the reader is referred to Reinhart 1982 and to Lambrecht,
1994: Chapter 4). For a proposition to be construable as being 'about' a
referent, this referent must be assumed to be presently under discussion
or pragmatically accommodatable as such. The topic of a proposition
therefore necessarily belongs in the realm of the pragmatic presupposition.
Topic and focus are not necessarily complements of each other in a given
assertion. An assertion may lack a topic, but not a focus.
My definition of 'focus' is compatible with the definitions found in much
current work on focus and focus structure, but it differs from those of most
authors in one important respect: Not all sentence accents count as focus
accents. In particular I exclude accents which fall on topic constituents (see
footnote 2 below and Lambrecht 1992). I will not deal in any detail in this
paper with the question of the 'projection' of the focus (Höhle 1982), i.e.
of the mechanisms whereby a focus accent may flag a semantic and
syntactic domain that properly includes the accented constituent. A theory
of focus projection is presented in Chapter 5 of Lambrecht 1994. In the
present paper I will be concerned only with the focus status of ARGUMENT
expressions, in particular subjects and objects.
In making a pragmatic assertion, a speaker expresses a PRAGMATICAL
LY STRUCTURED PROPOSITION, i.e. a proposition which reflects not only
a state of affairs but also the interlocutors' state of mind with respect to this
state of affairs at the time of an utterance. The information structure of a
proposition is partly defined in terms of differences in the respective
domains of the presupposition and the focus. I will refer to such domain
differences as different FOCUS ARTICULATIONS of a proposition. Focus
articulations may be expressed prosodically and/or morphosyntactically,
depending on the language. I distinguish three major TYPES of focus
articulation, according to the domain of the focus in the proposition:
PREDICATE FOCUS, SENTENCE FOCUS, and ARGUMENT FOCUS. T h e

distinction of these three focus types is empirically justified by the fact that
they are expressed in distinct formal categories across languages.
The three types of focus-articulation are illustrated in items (2), (3), and
(4) with examples from English, Italian, and French. The examples illustrate
crosslinguistically widely attested grammatical strategies for expressing these
focus-articulations. They do not, however, exhaust the grammatical
possibilities found across languages nor even within a given language; in
particular I have not included examples of morphologically marked focus,
as found e.g. in various African languages (see e.g. Givón 1975; Watters
1979; Sasse 1987). The questions in parentheses before each example
suggest minimal discourse contexts in which the given focus articulations are
likely to occur. These questions do not by themselves determine the focus
type of the replies (i.e. a given question may be followed by an answer with
a different focus articulation). Small caps indicate the focus accent, i.e. the
pitch accent which - either by itself or in combination with morphosyntactic
features - marks the focus articulation of the sentence:2
(2) PREDICATE FOCUS ARTICULATION
(Have you recovered from your accident? How's your foot and
your knee?)
a. My foot still HURTS (but the rest is OK).
b. Il piede mi fa ancora MALE (ma il resto va bene).
the foot to-me makes still pain
c. Mon pied me fait encore MAL (mais le reste ça va).
my foot to-me makes still pain
(3) SENTENCE FOCUS ARTICULATION:

(Why are you walking so slowly?)
a. My FOOT hurts.
b. Mi fa MALE un PIEDE.
to-me makes pain a foot
c. J'ai mon PIED qui me fait MAL.
I have my foot that to-me makes pain
2
I will ignore in this paper intonational differences coding semantic or pragmatic
distinctions other than the focus structure of a sentence, such as the marking of speech-act
differences or of speakers' attitudes towards propositions. See Chapter 5 of Lambrecht
forthcoming for justification of this procedure.
150 KNUD LAMBRECHT
(4) ARGUMENT FOCUS ARTICULATION:

(Is your knee hurting?)
a. No, my FOOT hurts.
b. No, è il PIEDE che mi fa male. / No, mi fa male il PIEDE.
no it-is the foot that to-me makes pain / no to-me makes
pain the foot.
c. Non, c'est mon PIED qui me fait mal
no it is my foot that to-me makes pain
In the PREDICATE-FOCUS sentences in (2), information is requested among
other things about the addressee's foot. In the answer, the speaker asserts
ABOUT her foot that it hurts. The subject NP my foot (il piede, mon pied)
is a TOPIC expression, i.e. the referent of the NP stands in a TOPIC
RELATION (or ABOUTNESS RELATION) to the proposition expressed by the
sentence. The accent on the predicate hurts (fa male, fait mal) indicates
that the focus ('the new information') is expressed in the predicate phrase
alone and excludes the subject NP. The designatum of the predicate phrase
has a FOCUS RELATION to the proposition as a whole. 3
In the SENTENCE-FOCUS structures in (3), in contrast to (2), the object
of inquiry in the question is not primarily the addressee or her foot but a
state of affairs (i.e. the fact that the addressee is walking slowly). The focus
of the answer includes both the predicate and the subject, hence the entire
proposition (with the exception of the referent of the determiner my, an
exception to which I will return). The status of the SUBJECT (or primary
argument) as a NON-TOPIC, i.e. a referent which, together with the
predicate, stands in a FOCUS RELATION to the proposition, crucially
distinguishes the articulation in (3) from that in (2). It is this non-topic
status of the subject argument that is grammatically marked in the
sentence-focus structures in (3). In English, it is marked by prosodic
prominence on the subject noun and lack of prosodic prominence on the
verb phrase. 4 In Italian it is marked both by prosodie prominence on the
3
In the Gontext of (2) the subject noun foot is likely to receive what is often called
a 'contrastive accent'. This accent, called 'topic-establishment accent' in Lambrecht (1994,
Chapter 5.4) is functionally different from the focus accent and is irrelevant for the
argument at hand.
4
For an explanation of the projection mechanism whereby a single subject accent
signals a focus which extends over an entire proposition see Lambrecht 1992.
subject noun and by postverbal position of the subject NP.5 In French it is

marked by means of a special bi-clausal construction (called a 'presenta
tional cleft construction' in Lambrecht 1988a) in which the lexical (i.e. non-
pronominal) constituent which codes the semantic subject of the proposition
(the speaker's foot) is marked as a non-topic by being coded as an object
in its clause. Further details concerning this French construction will be
presented in Section 4.
As I have argued elsewhere (Lambrecht 1987b), the pragmatic
distinction between 'predicate focus' and 'sentence focus' is related to the
logical distinction made by the 19th century philosophers Brentano and
Marty between a CATEGORICAL (or 'double') and a THETIC (or 'simple')
judgment type, a distinction whose relevance for generative grammar is
discussed in Kuroda (1972) and whose importance for language typology
has been emphasized by Sasse (1987). The difference between the two
focus types crucially involves the relationship between the subject (or
primary argument) and the predicate. In the predicate-focus articulation
(expressing a categorical judgment), the predicate conveys information
about the referent of the subject expression. This referent has a salient
status in the discourse independently of its association with the predicate
of the given sentence. In the sentence-focus articulation (expressing a
thetic judgment), the subject and the predicate form an informational unit,
in which the subject referent has no discourse salience prior to its
association with the predicate in the given utterance. Constructions with
sentence-focus articulation have been discussed in the literature under
various labels (without necessarily being recognized as distinct categories),
such as 'neutral descriptions' (Kuno 1972), 'news sentences' (Schmerling
1976), 'all-new utterances' (Allerton & Cruttendon 1979; Fuchs 1980),
'eventive sentences' (Faber 1987), and 'event-reporting sentences'
(Lambrecht 1987b, 1988a).
It is important for the following discussion to understand that the
utterances in (2) and (3) (as well as in (4)), in addition to providing
information about the speaker's foot (as in (2)) or about the situation (as
in (3)), also provide information about the speaker herself, i.e. the
individual who feels pain in her foot and who is designated in the questions
5
I am using the term 'subject' here in the traditional sense, in which it encompasses
postverbal NPs like un piede in (3). Whether such NPs have certain object properties or
may be considered 'objects' at some more abstract level of analysis is not at issue here.
152 KNUD LAMBRECHT
with the pronoun you. Therefore the linguistic expressions which refer to
the speaker in the answer sentences, i.e. the possessive determiner my in
English, the dative pronoun mi in Italian, and the determiner mon as well
as the subject and object pronouns je and me in French, are TOPIC
expressions; However, since these expressions do not code the primary
argument of the proposition, their topic status does not affect the focus-
articulation type of the sentence. The fundamental criterion for distinguish
ing the predicate-focus and the sentence-focus articulations is that in
sentences of type (2) the SUBJECT ARGUMENT of the underlying proposi
tion has a TOPIC relation to the proposition while in sentences of type (3)
the subject has a NON-TOPIC (i.e. FOCUS) relation to the proposition.
In the ARGUMENT-FOCUS sentences in (4), the focus domain is
'narrow", i.e. restricted to the designatum of the accented NP argument.
This pragmatic feature distinguishes the argument-focus articulation both
from the predicate-focus and the sentence-focus articulations. The context
question in (4) evokes a discourse situation in which the fact that the
speaker feels pain is assumed to be known to the addressee. The informa
tion conveyed by the answer consists merely in identifying the body part in
which the pain is felt.6 The argument-focus articulation corresponds to the
well-known 'focus-presupposition' type discussed e.g. in Chomsky (1972)
and Jackendoff (1972: Ch.6), in which the focus constituent corresponds to
a variable in a pragmatically presupposed open proposition. The argument-
focus articulation is only indirectly relevant to the issues presented in this
paper, which concern mainly the contrast between the predicate-focus and
the sentence-focus articulations. I will therefore not discuss it here at any
length.
Notice that the English sentence in (4a) is formally identical to the
sentence-focus structure in (3a). Given that the difference between sentence
focus and argument focus is a universal information-structural distinction
which is grammatically reflected in other languages (e.g. French or Italian),
this identity of form in English is to be seen as a case of focus-structure
homophony. This is confirmed by the fact, to be illustrated later on, that the
argument-focus type enjoys greater distributional freedom than the
sentence-focus type in English. All sentence-focus structures involving
6
In Lambrecht (1994, Chapter 5) I refer to sentences with argument-focus
articulation as 'identificational sentences', contrasting with 'topic-comment sentences'
(predicate-focus) and 'event-reporting' sentences (sentence focus).
subject accentuation also have argument-focus readings, but the reverse is

not true: Not all argument-focus structures involving subject accentuation
also have sentence-focus readings. The same asymmetry holds for Italian
VS sentences (see Wandruzka 1982). Unless indicated otherwise, subject-
accented sentences are to be understood throughout this paper as having
sentence-focus rather than argument-focus articulation.
3. The interdependence of role types in English
3.1. Semantic and syntactic roles
It is a generally accepted fact that predicates are associated in the lexicon

with one or more arguments and that these arguments are in turn
associated with certain semantic roles or participant roles. In the Fillmore-
an tradition, the set of possible participant roles associated with a given
predicate has been referred to as that predicate's CASE FRAME. In recent
work, Fillmore (1988b) and Fillmore & Kay (1993) distinguish between two
kinds of participant roles, FRAME-SPECIFIC ROLES and THEMATIC ROLES
(or THETA ROLES). The former are the roles as they fit the meaning of the
semantic frames associated with individual verbs or sets of verbs, such as
the lover' and the 'lovee' associated with the predicate love. The latter are
general semantic notions, such as 'agent', 'patient', 'experiencer', 'goal',
'source' etc, "which represent the linguistic schematizing of acts, events,
and states-of-affairs into very general patterns" (Fillmore & Kay, Chapter
The set of theta roles associated with a predicate is called the

predicate's THETA FRAME. A theta frame is "a structured set of theta roles
which corresponds to a perspective for conceptualizing or notionally
parsing an event (or state of affairs)" (ibidem, Chapter 8). It is theta roles,
not frame-specific roles, that get linked with grammatical functions in
sentences. For example the verbs buy and sell have the same array of
frame-specific roles (buyer, seller, goods, and money), but with buy the
7
I will avoid the term 'thematic role' because in information-structure research the
term 'theme' is traditionally understood as referring to a pragmatic rather than semantic
relation. Somewhat reluctantly, I will go with the trend and sometimes use the term 'theta
role', in the sense in which it is used by Fillmore & Kay.
154 KNUD LAMBRECHT
buyer is the agent and the seller the source, while with sell the buyer is the
goal and the seller the agent. But with both verbs, it is the agent role that
gets matched with the grammatical role of subject in a sentence.
Let us look at the participant roles associated with the predicator hurt,
The semantic frame of this predicator involves at least the roles listed in
(5), which I will first designate with frame-specific names:
(5) (a) a sentient being feeling the pain, or SUFFERER
(b) a body part where the pain is felt, or LOCUS OF PAIN
(c) a (non-volitional) stimulus causing the pain, or CAUSE OF
PAIN.
(d) a (volitional) agent inflicting the pain, or INFLICTOR.
(e) an instrument with which an agent inflicts pain, or INSTRU
MENT.
One or more (but rarely all) of the roles in a semantic frame will be
expressed in syntactic argument positions in a sentence when the verb is
called upon to describe a certain state of affairs, or SCENE, as Fillmore calls
it. In (5), roles (c) and (d) are mutually exclusive in a given scene. Pain is
caused either by a volitional agent or a non-sentient stimulus, but not by
both at the same time. Moreover (e) conceptually presupposes (d). I will
be concerned mostly with roles (a), (b), and (c) in this paper, i.e. most
examples to be discussed are to be understood as involving non-volitional
stimuli.
The frame-specific role names in (5) translate into the following theta
roles: The sufferer is an EXPERIENCER (EXP), the locus of pain a
LOCATION (LOC), the cause of pain a STIMULUS (STI), the inflictor an
AGENT (AGT), and the pain-causing instrument an INSTRUMENT (INS).
Notice that the mapping from frame-specific to theta roles is not an
automatic translation procedure but depends on what Fillmore & Kay call
the "notional parsing of the event" in a given scene. Since I am dealing in
this paper with a limited set of predicators, requiring few theta-frame
generalizations, the distinction between frame-specific roles and theta roles
is relatively unimportant for my argument. I will therefore allow myself
relative terminological freedom with respect to this otherwise important
distinction.
There is notorious and vast disagreement among scholars about the
nature and number of semantic roles in linguistic theory.8 Fortunately,
there is much less disagreement concerning the relative ranking of semantic
roles with respect to each other. In the following discussion I will be mainly
concerned with questions of relative ranking, therefore the controversy over
the nature and number of semantic roles has no major consequence for the
present analysis.
Since Gruber (1965) and especially Fillmore (1968) it has been
generally recognized that the syntactic arguments associated with a
predicator in a sentence may be assigned different semantic roles (or,
alternatively, that the semantic roles associated with a predicator may be
mapped with different syntactic arguments), depending on the scene
depicted by the predicator in a given sentence. Consider, for example, the
difference in the semantic role of the single argument of the predicate hurt
in (6) and in (7) (examples from Fillmore 1988b). The theta role of the
nominal argument is indicated in angled brackets after the sentence:
(6) My foot hurts. <LOC>
(7) This shoe hurts. <STI>
In (6) the subject NP indicates the body part where the pain is felt, or the
location of the pain, whereas in (7) it indicates the entity which causes the
pain, or the stimulus.
As Fillmore has shown (1968, 1988b), in the case of predicates with
more than one argument the various possible mappings from semantic roles
to syntactic positions are at least in part regulated by general principles,
which can be expressed in the form of SEMANTIC ROLE HIERARCHIES.
These role hierarchies determine which of the two or more semantic roles
associated with a predicate will appear in which syntactic position, in
particular which one will be selected for the role of subject. For example,
if in a given scene the predicate hurt is associated with two arguments, as
in (8)
(8) This shoe hurts my foot <STI - LOC>, <*LOC - STI>
the subject argument can only be the pain-causing stimulus, and the locus
of the pain must be the object. If the order is reversed, i.e. if the NP my
foot is selected as the subject and the NP this shoe as the object, the subject
will still be interpreted as the cause of the pain and the object as the entity
For a useful recent overview over the debate see Dowty (1991).
156 KNUD LAMBRECHT
affected by the pain, in pragmatic incongruity with the respective lexical

meanings of the arguments. Thus sentence (8) cannot have the same
meaning as sentence (9). The symbol @ indicates pragmatic anomaly or
incongruity:
(9) @ My foot hurts this shoe. <STI - LOC>, <*LOC - STI>
The stimulus argument ranks higher than the location argument for subject
selection. As a result, the only possible interpretation of (9) is pragmatical
ly anomalous.
The contrast between (8) and (9) may be seen as the manifestation of
a general semantic principle according to which, all else being equal, a
more 'agentive' argument outranks a more 'patient-like' argument for
subject selection. Indeed, even though in our example the patient proper
is not the foot but the individual experiencing the pain, it is clear that the
foot is affected by the shoe and not the other way around. The general
subject-selection hierarchy originally suggested by Fillmore (1968: 33) ranks
the roles of agent, instrument, and object with respect to each other. A
terminologically slightly different hierarchy is formulated in Fillmore
(1988b):
(10) Agent > Instrument > Patient
The hierarchy in (10) accounts e.g. for the possible and impossible
sequences in (11) (examples (a), (b), and (c) are from Fillmore, 1988b):
(11) a. The child broke the balloon. <AGT-PAT>
b. The pin broke the balloon. <INS-PAT>
c. The child broke the balloon with the pin.
<AGT-PAT-INS>
d. *The balloon broke the child. <PAT-AGT>
e. *The balloon broke the pin. <PAT-INS>
The ungrammaticality of (11) (d) and (e) in the interpretation suggested
by the semantic roles in angled brackets is due to the fact that the subject
is represented by an argument which is lower on the hierarchy in (10) than
the non-subject.
The idea of the hierarchical ordering of semantic roles is generalized
in Foley & Van Valin (1984: 59), who postulate the linear hierarchy in (12):
(12) Agent > Effector > Experiencer > Locative > Theme >
Patient
(A somewhat similar generalized hierarchy is postulated by Bresnan &

Kannerva 1989.) The hierarchy in (12) accounts for tne above-mentioned
contrast between the stimulus and the location arguments in (8) and (9).
Taking our stimulus argument to be an 'effector' in Foley & Van Valin's
terms, and our locus of pain to be either a locative' or a 'patient', this
hierarchy predicts the difference between (8) and (9).
The notion of a generalized linear hierarchy applying to all predicates,
such as that proposed by Foley & Van Valin or Bresnan & Kannerva, is
rejected by Fillmore & Kay (1993), on the grounds that it is the theta
frames, each schematizing an event type, that are ontologically primary and
that generalized hierarchies allow for a great number of theta arrays which
never occur in fact. Fillmore & Kay assume instead a set of individual
theta frames (such as agent-patient, experiencer-content, stimulus-experien-
cer, etc), each of which has a distinguished argument, which is realized as
the subject.
Without adopting any specific proposal, I will simply assume the
existence of a relative ranking constraint such that the more agent-like of
two arguments will be selected as the subject. Following Foley & Van
Valin, we may refer to the two relatively ranked roles in a clause as two
MACRO-ROLES, one ACTOR and one UNDERGOER, the 'actor' being the
more agent-like and the 'undergoer' the more patient-like of the two. In our
examples, the stimulus would be the actor and the location the undergoer.
It is only the respective ranking of such macro-roles that I am concerned
with in this paper.
For the following discussion it is relevant to observe that the English
predicate break can also be used intransitively, as in (13):
(13) The balloon broke. <PAT>
In this case, the subject argument does play the role of patient, i.e. of the
entity to which something happens or which undergoes a change, contrast
ing with the transitive (11d), where the association of patient and subject
caused the sentence to be ungrammatical. On the basis of facts such as
this, and following the insights from Case Grammar, typologists like
Hawkins (1981) and Comrie (1981) have observed that English differs from
many other languages in that it allows a greater variety of semantic roles
to be mapped onto the syntactic argument position of subject, without
requiring a morphologically or syntactically marked valence change (such
as reflexivization or passivization). Compare e.g. the English pair I opened
the door - The door opened with the French pair J'ai ouvert la porte - La
158 KNUD LAMBRECHT
porte s'est ouverte, where the verb of the intransitive sentence must have
reflexive morphology (cf. *La porte a ouvert).9
Returning to the English predicate hurt, we notice that while in (6),
(7), and (8) the semantic roles of the locus and/or the cause of pain are
overtly expressed as arguments of the predicate, the role of the EXPERIEN-
CER of the pain, which is a necessary participant role at the conceptual
level (it is the person, not the foot, that feels the pain), does not appear
as an overt argument of hurt. Rather the experiencer role is either
expressed indirectly, via the possessive NP determiner my (example (6) and
(8)), or it is omitted altogether, as in (7). As we will see, this is a feature
whereby English markedly differs from French (and other Romance
languages), where such a separate experiencer argument plays an important
syntactic as well as pragmatic role,10
While certain restrictions are placed upon the coding of the experiencer
as an object argument with predicates like hurt, it is possible, at least in
certain dialects or with a certain semantic specialization, to use the
experiencer as the unique argument, as in (14):
(14) 1 hurt (ail over). / I'm hurting. <EXP>
(14) is comparable to (13), where a patient argument could appear as a
subject, provided that no other argument was competing with it for this
position. Again, English is different in this respect from other languages,
As we will see in Section 4, in French it is necessary, in order to express
the meaning of (14), to choose a lexical predicate different from the one
9
This restriction is not absolute in French. For example under certain semantic
conditions the transitive casser 'to break' may be used intransitively, as in Le verre a cassé
'the glass broke'.
10
Among the English speakers I have consulted on the matter, several rejected
sentences like (i) (a) and (b) below, in which the experiencer is expressed as a separate
pronominal object argument, although some accepted corresponding versions involving
progressive aspect such as (i) (c) and (d):
(i) a. ?My foot hurts me. <LOC - EXP>
b. ?This shoe hurts me. <STI - EXP>
c. My foot is hurting me more every day. <LOC - EXP>
d. My shoes are hurting me. <STI - EXP>
Be that as it may, the general preference in English is not to express the experiencer in
the form of a separate argument, i.e. speakers tend to say (6) or (7) rather than (i) (a) or
(b).
used to express the transitive sentence (8).
3.2. Semantic, syntactic, and pragmatic roles
Thus far we have looked at the relationship between semantic roles and
syntactic functions in English. In accordance with the goal of this paper, I
will now add a third dimension, by asking how the syntactic and semantic
levels relate to the level of INFORMATION STRUCTURE, in particular to the
pragmatic roles of TOPIC and FOCUS.
Looking first at the relationship between SEMANTIC and PRAGMATIC
roles, without reference to syntax, we notice that the two are in principle
independent of each other. The independence of semantic and pragmatic
roles is an obvious consequence of the fact that information structure has
to do with the USE of SENTENCES, rather than with the MEANING of
PROPOSITIONS. Information-structure distinctions are necessarily manifested
at the level of the UTTERANCE, i.e. at the level where propositional
contents are communicated via sentences with prosodic and morphosyn-
tactic structure. Since, as we have seen above, the same semantic role can
appear in different syntactic functions, and since it is syntactic positions that
express pragmatic roles, semantic roles per se cannot be pragmatically
constrained. Any pragmatically determined mapping constraints intervene
necessarily at the point where semantic roles are manifested as syntactic
constituents in a sentence.
Let us then consider the relationship between PRAGMATIC and
SYNTACTIC roles. Here we observe a revealing parallel with the relation
ship between syntax and semantics. As noted, English allows both agent
like and patient-like roles, including that of patient or highly affected
experiencer. (Foley & Van Valin's undergoer), to appear as subjects,
without concomitant morphosyntactic or lexical changes (see e.g. (13) and
(14)). In the same fashion, English allows in principle both the topic
relation and the focus relation to be mapped with the syntactic role of
subject, without concomitant changes in syntax or morphology. In this
respect, English differs from many other languages, including Italian and
French (see the contrasts illustrated in (2), (3), and (4) above). In English
the same unmarked SV(O) sequence may be used to express different
focus articulations, although, as we will see, the sentence-focus articulation
is semantically more constrained than the predicate-focus and the
argument-focus articulations. Differences in focus articulations are typically
160 KNUD LAMBRECHT
(though by no means exclusively) expressed in English via prosodic rather

than syntactic contrasts. 11
Consider again the English sentences in (2) and (3), repeated here for
convenience as (15) and (16). The angled brackets after each example
indicate the semantic, syntactic, and pragmatic role of the NP argument:
(15) My foot still HURTS. <LOC/SUBJ/TOP>
(16) My FOOT hurts. <LOC/SUBJ/FOC>
As noted earlier, in the predicate-focus structure in (15) the subject NP my
foot is a topic expression, i.e. it has a topic relation to the proposition,
whereas in the sentence focus structure in (16) it has a focus relation to the
proposition. (Recall that in both examples the possessive determiner my
is a topic expression.) (16) shows that a LOCATION may be the focal subject
argument in a sentence-focus construction.
The same information-structure contrast as that illustrated in (15) and
(16) is possible in English if the subject has the semantic role of STIMULUS,
as it does in (7). Given the proposition expressed in (7) (with a change
from singular to plural to make for greater pragmatic naturalness), (17)
presents the predicate focus version and (18) the sentence focus version:
(17) These shoes HURT. <STI/SUBJ/TOP>
(18) These SHOES hurt. <STI/SUBJ/FOC>
Example (17), a predicate-focus structure or topic-comment sentence,
could be used in a context in which the shoes are the topic under
discussion and in which the proposition expressed by the sentence conveys
further information ABOUT these shoes. Example (18), a sentence focus
structure, could be uttered out of the blue, e.g. as a sudden complaint by
the person wearing the shoes.
Even though in the case of the predicate hurt both the location (ex.
(16)) and the stimulus (ex. (18)) may appear as focal subject arguments in
sentence-focus structures, it must be emphasized that intransitive sentence-
focus structures such as (16) and (18) are a grammatically much more
constrained expression type than the corresponding predicate-focus
structures. While all of the possible semantic roles listed in item (5) for the
predicate hurt may be used freely for TOPICAL SUBJECTS in PREDICATE-
11
For discussion of a particular type of constructionally coded focus articulation in
English see Lambrecht 1988c.
FOCUS STRUCTURES, as shown in (19):12

(19) a. John hurts all OVER. <EXP/TOP>
b. That slap HURT. <INS?/TOP>
c. My foot still HURTS. <LOC/TOP>
d. These shoes HURT. <STI/TOP>
e. These shoes hurt my FEET <STI/TOP>
f. John hurt his SISTER. <AGT/TOP>
g. The shoe she threw at him HURT him. <INS/TOP>
only the location and the stimulus may appear as FOCAL SUBJECTS in
SENTENCE-FOCUS STRUCTURES, as shown in (20) (the # sign indicates
unacceptability of the example in the sentence-focus reading):
(20) a. # JOHN hurts all over. <EXP/FOC>
b. #That SLAP hurt. <INS?/FOC>
c. My FOOT hurts. <LOC/FOC>
d. These SHOES hurt. <STI/FOC>
e. #These SHOES hurt my feet. <STI/FOC>
f. # JOHN hurt his sister. <AGT/FOC>
g. # The SHOE she threw at him hurt him. <INS/FOC>
The intransitive (20) (a) and (b) show that it is not sufficient for a sentence
to contain a single argument for a sentence-focus reading to be available.
It is the semantic nature of the syntactic argument that seems to count. In
(20a) the experiencer role of the subject precludes the sentence-focus
interpretation. (20b), which implies an agent, is less acceptable than the
semantically rather similar (20d), where no agent is understood, suggesting
the presence or absence of volition as a determining factor. If the sentence
is transitive (cf. (20) (e), (f), (g)), no sentence-focus structure is possible
(with one exception involving pronominal arguments to be discussed below,
ex. (23)). In this respect, the contrast between (20) (d) and (e) is particular
ly revealing. It shows that the constraint against the presence of an object
NP holds even for sentences whose subjects have semantic roles which are
(19a) is rather unnatural, compared to (14), in which the subject is a first person
pronoun subject. The lexical NP is needed for the present argument because a correspond
ing sentence-focus structure containing an accented pronominal subject would be
unacceptaable for independent reasons (see below). In (19b), the question mark indicates
a certain difficulty in assigning a semantic role to the subject argument. According to
DeLancey 1991, the subject that slap takes on the properties of the agent in that sentence.
162 KNUD LAMBRECHT
in principle compatible with the sentence-focus articulation.

The reader may have noticed that I have been finessing the question of
whether transitive sentences with two intonation peaks, like (19e')
(19e') These SHOES hurt my FEET.
count as having sentence-focus articulation in contexts where the referent
of the subject NP is not naturally construable as a topic. It is easy to
imagine situations in which the subjects of such two-accent sentences are
non-topical in the discourse, allowing the sentences to serve an event-
reporting function. Utterances like Oh my God, the STORM has destroyed our
HOUSE! or Guess what, the POLICE have arrested my GIRL FRIEND! are
natural examples. I therefore see no reason not to count sentences like
(19e') as having sentence-focus articulation, depending on the context in
which they are uttered. Notice, however, that such SVO sentences with two
intonation peaks are systematically ambiguous between a sentence-focus
and a predicate-focus reading, since the accented subject NP may in
principle always receive a (contrastive) topic interpretation (see footnote
2 and the reference therein). In other words, such sentences are not
marked as belonging to a formal category 'sentence-focus construction'. I
will leave this difficult matter open here. 13 What counts for the purposes
of the present argument is the fact that with a single accent on the subject
noun transitive sentences cannot be construed as having sentence-focus
articulation, while with a single accent on the object noun they are naturally
interpreted as having predicate-focus articulation.
I should mention that correlations between the transitivity of a clause
and the focus articulation of the proposition expressed by it do exist and
have been observed in various languages. For example Italian, which
requires VS for sentence focus, disallows sentence-focus readings for VSO
or VOS structures, which only have argument-focus readings (see
Wandruzka 1982). A correlation is observable also in English, in existential
and deictic there-sentences, which count as having sentence-focus structure
in my framework since their subjects are non-topics and their predicates
non-presupposed. It is well-known that there-sentences are restricted to a
subclass of intransitive predicates (compare There came a storm with *There
In Lambrecht forthcoming (Chapter 5) I argue that the availability of two focus

construals for such sentences is explained by the nature of the predicate-focus articulation
as the unmarked focus type.
destroyed a storm a house). However, the existence of such construction-

specific syntactic constraints on the expression of sentence focus has no
bearing on the question of whether sentences like (19e') can be contextually
construed as having sentence-focus articulations.
A contrast similar to that between (19) and (20) is observable with
many predicates, such as the above-mentioned break. With a single focus
accent on the subject, none of the transitive sentences involving break in
(11) could be used in a context in which a sentence-focus structure is called
for. For example a one-accent sentence like (11a')
(11a') The CHILD broke the balloon.
in which the subject is an agent, would be as unacceptable as (20e) in a
sentence-focus reading. Only the intransitive sentence in (13), in which the
subject is a patient (or undergoer), has a sentence focus reading, as shown
in (13'):
(13') The BALLOON broke.
The same is true for the predicate open. While the intransitive The DOOR
opened, in which the subject has a patient or theme role, has a sentence-
focus reading, the equally intransitive JOHN opened, in which the subject is
an agent, can only be construed with argument focus.
The data show that predicate-focus readings, involving topical subjects,
may be freely assigned to any sentence, while sentence-focus readings,
involving focal subjects, may be assigned only to a restricted set of
sentences. We may therefore characterize the position of the preverbal
subject in English as the UNMARKED TOPIC POSITION, taking 'unmarked'
in its distributional sense (see Lyons 1977: 305ff). On the other hand, while
the mapping from FOCUS to SUBJECT is restricted (ignoring argument-focus
readings, which are totally unconstrained in English), the mapping from
FOCUS to OBJECT is free, as any transitive sentence with a lexical object
NP will show (cf. e.g. (19) (e) and (f)). In other words, while many
intransitive predicates disallow sentence focus, there are no transitive
predicates which disallow predicate focus. The position of the (postverbal)
object may therefore be characterized as the UNMARKED FOCUS POSITION.
By 'focus position' I mean the position of the argument having a focus
relation to the proposition and bearing the focus accent. It should be borne
in mind here that both in the predicate-focus and the sentence-focus
articulation an accent on an NP argument represents a larger focus
domain, which extends either over the predicate phrase or the entire
164 KNUD LAMBRECHT
sentence.14
Given the restrictions against the co-mapping of focus and subject and
the absence of such restrictions in the case of topics, we may posit as a
general information-structure principle, for English at least, that the topic
argument OUTRANKS the focus argument for subject selection. This general
ranking difference between topic and focus arguments may be expressed
in the form of a PRAGMATIC ROLE HIERARCHY:
(21) THE PRAGMATIC ROLE HIERARCHY: topic > focus
The pragmatic hierarchy in (21) parallels the semantic subject-selection

hierarchies discussed above, in particular Foley & Van Valin's two-termed
actor-undergoer hierarchy. It seems that topic arguments are strongly
associated with agentlike roles (actors) and focus arguments with
patientlike roles (undergoers).
The parallel between the pragmatic hierarchy and the semantic
hierarchy may be pushed one step further. Just as a patientlike argument
may often become a subject only if no other semantic role competes with
it for this position (cf. (13), (14) and discussion), the focus relation may be
mapped onto the subject role only if no (lexical) object argument is present
which competes with the subject for the focus role (see (20) (d) vs. (e) and
below).
At this point it would be tempting to draw the conclusion that a one-to-
one relationship holds between semantic and pragmatic roles in English,
such that only a subclass of patient-like arguments, which would form a
natural class, may function as focal subjects in sentence-focus structures.
Such a conclusion would be of high theoretical interest if it could be
shown, for example, that the set of predicates permitting sentence-focus
readings is coextensive with the subset of intransitive predicates now often
referred to .as 'unaccusative verbs' (Perlmutter 1978) or 'ergative verbs'
(Burzio 1981). The generalization would seem reasonable since these
predicates appear regularly (though by no means obligatorily) as POST-
Not having analyzed the matter in sufficient depth, I leave open the question of
whether the mapping of TOPIC and OBJECT is cross-linguistically constrained in the
same way as that of focus and subject. For strong statistical evidence that topics are rare
as lexical object arguments in French see Lambrecht (1986b: Chapter 6).
VERBAL subjects in languages like Italian.15

That this tempting conclusion cannot quite be correct is shown in the
fact that a given predicate may permit or prohibit a sentence-focus reading
depending on the semantic role of its subject argument, as the various
contrasts in (20) have revealed. Since the possibility of sentence-focus
construal is influenced by the semantic role of the subject in a given scene,
the deciding criterion cannot be the lexical predicate itself. To the extent
that unaccusative status is seen as a lexical property of a predicate, there
can be no one-to-one relationship between unaccusativity and sentence-
focus articulation. Moreover there are predicates, such as Italian far male
'hurt', which permit sentence-focus articulation (see example (3b)) but
which do not satisfy the criteria for unaccusative status (such as auxiliary
choice and so-called ne-cliticization). But most important, certain
'unergative' verbs may appear in sentence-focus constructions even when
their subjects are actors rather than undergoers:
(22) A: What happened?
B: JOHN called. / A telefonato GIOVANNI.
(23) A: Why is Mary so upset?

B: Her HUSBAND left her. / L'ha lasciata il MARITO.
The subject arguments of the predicates 'to call s.o. (on the telephone)' or
'to leave s.o.' are clearly agents, in the sense that sentences containing
these predicates express actions intentionally performed by the subject
referents. Nevertheless these sentences have possible sentence-focus
readings. As (23) shows, it is even possible for a sentence-focus structure
to be syntactically transitive, i.e. to contain a constituent expressing the
entity affected by the action of the subject. However such a patient
constituent must be PRONOMINAL. This constraint is shown also in the
contrast between the acceptable My SHOES are hurting me (see footnote 9)
and the unacceptable (20e) #These SHOES hurt my feet.16
Thus the possibility or impossibility of sentence-focus construal does
For a recent critique of syntactic approaches to the analysis of unaccusative verbs

see Van Valin (1990).
See Lambrecht 1988b for further discussion and for a proposed explanation of the
asymmetry between lexical and pronominal arguments with respect to the expression of
focus structure.
166 KNUD LAMBRECHT
not depend on the type of predicate alone. But neither does it entirely
depend on the semantic role of the subject argument. The latter conclusion
is confirmed by the fact that the availability of a sentence-focus reading for
a sentence with a given semantic structure often depends on the scene
which it is used to evoke. For example, the slightly varied version of the
unacceptable (20b) in (20b')
(20b') The SLAP hurt him!
has a possible sentence-focus reading in a context where a nurse is testing
the pain sensitivity of a patient emerging from anesthesia. 17 We must
conclude, then, that the pragmatic subject-selection hierarchy does not
directly match any natural semantic hierarchy such as e.g. Foley & Van
Valin's hierarchy in (12). It does not seem possible to define a cut-off point
on a semantic role hierarchy beyond which semantic roles can no longer
serve as the subject arguments of sentence-focus structures.
In spite of examples such as (22) and (23), it is nevertheless possible,
I think, to establish an inherent connection (i.e. one that goes beyond a
mere discourse tendency) between sentence-focus subject and non-agentive
semantic role. In order to see this connection it is necessary to take into
account an important fact concerning the interpretation of argument roles,
which in my opinion has not received sufficient attention: The construal of
the role of an argument may be influenced by discourse factors, i.e. by
factors other than the inherent lexical meaning of the predicate.
The discourse factor I have in mind is well-illustrated in examples (22)
and (23). Although in (22) the subject John designates the person who
makes the phone call, and although the activity of making a phone call
undoubtedly involves a willful agent, it is intuitively clear that the individual
making the phone call is not primarily CONCEPTUALIZED in (22) as an
agent on the level of the discourse. This intuition correlates with the fact
that (22) could not serve as an answer to the question What did John do?,
i.e. that the subject NP John has the pragmatic status of a NON-TOPIC in the
discourse. On the level of information structure, the event described by (22)
is not PREDICATED of the referent of the subject expression John in the
given context, i.e. the communicative purpose of the utterance is not to
convey information ABOUT this referent. The function of (22) is not to
17
This fact was pointed out to me by Laura Michaelis. For analogous examples and
discussion see Fuchs (1980) and Faber (1987).
inform the addressee that the caller has the property of having made a
phone call but rather to announce an event, in which the caller is involved
as a necessary participant. This fact necessarily leads to non-agentive
construal of the argument in the given context. Similarly, the reply in (23)
is not primarily to be understood as describing an action done by the
husband, but as announcing an event which happened to the topic referent
'Mary'. Pragmatically, the action of leaving Mary is not predicated of the
individual designated by the subject NP her husband. Rather (23) is a
predication about the topic 'Mary', which is expressed in the unaccented
object pronoun her.18
The above notion of a relationship between semantic roles and
information function is by no means obvious. (It has been received
skeptically by colleagues to whom I have presented it). How can the
semantic interpretation of the subject role of a predicate, say to call (on the
phone), be influenced by the degree of topicality or salience of the caller in
the discourse? As a first step towards making this notion less objectionable
I would like to remind the reader that theta roles themselves - as opposed
to the frame-specific roles of individual predicators - represent schematiza-
tions of events which apply to large classes of predicates in relative
independence of their specific lexical meanings. And pragmatic roles may
be seen to correlate with such abstract theta roles rather than with frame-
specific roles.
But the real explanation for the interpretive phenomenon at hand is
found, I believe, in the notion of CONSTRUCTIONAL MEANING, as
developed in the framework of CONSTRUCTION GRAMMAR (Fillmore
1988a, 1988b; Fillmore, Kay, & O'Connor 1988; Fillmore & Kay 1993), in
particular in recent work by Goldberg (1992):19 Goldberg argues,
convincingly I think, that the meaning of grammatical constructions cannot
be seen as a mere projection of the meanings of the lexical items they
contain, as postulated e.g. in the 'Projection Principle' of the Government
18
The idea that agentive arguments are not interpreted as fully agentive when used
in certain "detopicalizing" constructions is already discussed in Kirsner (1973) in a
comparison of active and corresponding 'pseudo-passive* versions of the same sentence in
Dutch.
19
See also Michaelis 1993, especially Chapter 2. Preliminary remarks concerning the
importance of constructional meaning for the information-structure interpretation of
sentences are found at the end of Lambrecht 1987b.
168 KNUD LAMBRECHT
and Binding framework. To use one of Goldberg's examples, it is difficult
to see how the meaning and form of a transitive sentence like Sam sneezed
the napkin off the table can be computed on the basis of the lexical meaning
of the verb sneeze alone, since sneeze is an intransitive verb (cf. *John
sneezed the napkin). Rather the interpretation of such a sentence is
determined by the combined meanings of the lexical items it contains and
of the grammatical construction instantiated by the sentence, in this case
the verb phrase construction [V NP PP], where V is a verb of motion and
PP a directional phrase.
I believe that the notion of constructional meaning can be applied
fruitfully to the issue of the pragmatic construal of the semantic role of
sentence-focus subjects. Such an application requires that we first define the
basic constructional meaning of sentence-focus structures.20 As I have
argued elsewhere (Lambrecht 1987b and 1988b), while certain unaccusative
predicates (like arrive, disappear, die etc.) crosslinguistically favor sentence-
focus construal by their lexical meaning, the criterion determining which
predicates are permissible in sentence-focus constructions is not primarily
lexical but PRAGMATIC, having to do with the ability of the predicate to be
used in a PRESENTATIONAL function. I term a sentence 'presentational' if
its communicative purpose is not primarily to depict a referent as playing
a role in some state, event, or action, but to introduce the referent into a
discourse or else to report an event which involves the referent not as a
topic but merely as a necessary participant (in the latter case it is the event
involving the referent which is 'presented'). The presentational discourse
function corresponds on the semantic level to the earlier-mentioned
THETIC judgment type.21
One of the strongest pieces of evidence in favor of the presentational
analysis of sentence-focus constructions lies in the fact, first observed by
Fuchs (1980), that the subjects of such sentences can never be PRONOUNS
i.e. cannot have referents which are assumed by a speaker to be already at
20
The present argument presupposes, of course, that subject-accented sentences of
the kind analyzed here are indeed grammatical constructions, i.e. that their semantico-
pragmatic structure is not entirely predictable from their form, and vice-versa. That this
is indeed the case is demonstrated in detail in Lambrecht 1992 and in Chapter 5 of
Lambrecht (1994).
21
Sasse (1987) refers to the two kinds of presentational sentences as 'entity-central'
and 'event-central thetic sentences.
the forefront of the addressee's mind (cf. e.g. the contrast between My
FEET hurt and #THEY hurt.) The behavior of pronominal subjects in
sentence-focus constructions is thus the exact opposite of that of pronomi
nal objects. The discourse status of a subject referent which can appropri
ately be coded with a pronoun, i.e. which is taken to be presently under
discussion, clashes with the inherent presentational function of the sentence-
focus structure, resulting in pragmatic unacceptability of the sentence.22
Apparent exceptions to the rule that sentence focus subjects must have
'new', i.e. not yet topical, referents can be accounted for in terms of the
notion of 'constructional accommodation' (Lambrecht 1994), whereby a
pragmatic feature which is conventionally attached to a grammatical
construction - such as the 'new' requirement for the referent of the subject
in a sentence-focus construction - is inherited by an expression appearing
in the construction even if the actual pragmatic status of the expression in
the discourse is not that normally required by the construction. In a
presentational construction the pragmatic feature of 'newness' or 'unexpect
edness' which is conventionally attached to the construction can be
inherited by an expression appearing in it even when the referent of the
expression is 'objectively' not new in the particular discourse context in
which the construction is used (for further discussion see Lambrecht 1987b).
I would like to claim, then, that the non-agentive construal of certain
agentive sentence-focus subjects, such as those in (22) or (23), is a result of
the basic presentational meaning inherent in the sentence-focus CON
STRUCTION. Verbs with agentive subjects take on the function of
presentational predicates when they head such constructions. It is of course
necessary, for the agentive subject of an intransitive sentence to be
pragmatically construed as non-agentive, that the basic meaning of the
predicate be amenable to the presentational function of the sentence-focus
construction. For example, the reason the verb call can be used as a
presentational predicate has no doubt to do with our knowledge of the
phone-call scenario: Announcing an unsuccessful phone call is a conven
tional way of introducing the caller into the consciousness of the addressee.
The above argument concerning non-agentive construal of agentive
22
An apparent exception to the constraint against pronominal subjects in sentence-
focus structures is the use of a heavily stressed pronoun as a 'euphemism' for a proper
name, as in When I drove by Dad's house, I could tell that SHE was there, where 'she' is
Dad's scandalous mistress (example provided by Sue Schmerling).
170 KNUD LAMBRECHT
subject arguments on the basis of constructional meaning finds further

support in the following observation, which I have developed in detail in
Lambrecht 1988b. It is a crosslinguistically widely attested fact that the
subject participants in sentence-focus constructions are grammatically
expressed more like objects than like subjects. (Like object NPs, they carry
the main sentence accent in English, they occupy postverbal object position
in Italian, they appear as objects of a transitive verb in French, etc.) It is
therefore not surprising that the semantic construal of the focal subject
argument is at least, in part similar to that ordinarily given to objects of
transitive verbs. Since the semantic roles of transitive objects are inherently
less agentive than those of transitive subjects, non-agentive construal of
sentence-focus subjects follows naturally.
On the basis of these facts concerning the semantic construal of
sentence-focus subjects, I would like to postulate the following general
principle of interpretation: The construal of the semantic role of an
argument in a sentence is not only determined by the lexical structure of
the predicate and by the conceptual schematization of the given event via
a theta-frame but at least in part also by the INFORMATION STRUCTURE
of the sentence, i.e. by the communicative function which is conventionally
associated with the sentence construction. This suggests that the study of
the relationship between semantic and grammatical roles must take into
account not only the logical relations between predicates and arguments, or-
-as more aptly described by Fillmore - the relationship between semantic
frames and the participants in these frames, but also the relationship
between a sentence expressing these frame relations and the mental states
of the interlocutors who use and interpret the sentence as a piece of
propositional information.
3.3. Summary
The situation in English concerning the various mapping relations between

semantic, syntactic, and pragmatic roles may be summarized as follows for
our model predicate hurt. Concerning the relationship between SEMANTIC
and GRAMMATICAL roles, we observe that, along with many other semantic
roles, the location and the stimulus argument may each be associated with
the position of subject (exx. (6) and (7)). However, the more agentive of
these two roles (stimulus) outranks the less agentive (location) for subject
selection, as predicted by the subject selection hierarchy, i.e. if both are
expressed in the same clause, the cause of pain wins out (exx. (8) and (9)).
The locus of pain appears as a direct object (even though semantically it
designates a place). If the cause of pain and/or the locus of pain are
expressed in a clause, the role of the experiencer tends to remain unex
pressed or else to be combined with that of the possessor coded by the
possessive determiner within the NP expressing the locus-of-pain or the
cause-of-pain argument. If the predicate has no other roles associated with
it in the sentence, the experiencer role may appear as subject, without the
concomitant lexical, morphological, or syntactic changes in the verb which
are required in other languages (ex. (14)).23
As for the relationship between PRAGMATIC and GRAMMATICAL roles,
we observe that in principle both the topic relation and the focus relation
may be mapped onto the role of subject, without concomitant changes in
constituent order. However, just as with semantic roles, the pragmatic
mapping possibilities are constrained by a hierarchy: The co-mapping of
subject and focus is much more restricted than the co-mapping of subject
and topic. Subjects with strongly agentive semantic roles are normally not
compatible with the focus relation. Moreover, focus subjects cannot cooccur
with lexical objects (cf. (20)). There is thus a relationship between the
semantic and the pragmatic role hierarchies, even though the two cannot
be conflated. The existence of such a relationship is confirmed by the fact
that the construal of the semantic role of an argument may be influenced
by the communicative function of the sentence in context. In sentence-focus
constructions, the semantic role of any agentive subjectNP is neutralized or
backgrounded and the proposition is given a presentational or event-
reporting interpretation.
In the English examples discussed, the vehicle of the pragmatic
information is not the syntactic but the prosodic structure of the sentence.
The general typological picture emerging from these facts (and from many
others not discussed in this paper) is that English is a language in which
SVO constituent order can be used for a variety of semantic and pragmatic
purposes and in which the canonical sentence-initial subject NP position is
23
The statement that the mapping of experiencer and subject is contingent upon the
absence of any other argument is not meant as a general statement about English, cf. e.g.
the existence of such sentences as I feel pain in my foot, in which an experiencer/subject
cooccurs with a location/object. Moreover, as pointed out to me by Sue Schmerling, in
sentence (14), in which the subject is an experiencer, the phrase all over may be seen as
expressing a location argument.
172 KNUD LAMBRECHT
a position of great functional prominence. English is a highly subject-

prominent language (in the sense of Li & Thompson 1976), in which word
order is dictated to a great extent (though by no means exclusively) by
grammatical rather than pragmatic considerations (cf. also Mathesius 1928
and Thompson 1978).
4. The interdependence of role types in French
4.1. Semantic and syntactic roles
As I mentioned earlier, French has two lexical expressions corresponding

to the English verb hurt: Faire mal, literally 'to make (i.e. cause) pain', and
avoir mal, literally 'to have pain'. The two expressions are illustrated in
(24): 24
(24) a. Le/Mon pied me fait mal. <LOC - EXP> (see (2c)
above)
the/my foot to-me makes pain 'My foot hurts'
b. J'ai mal au pied. <EXP - L O O

I have pain at-the foot 'My foot hurts'
(24a) and (24b) have the same truth conditions, but they differ with
respect to the mapping possibilities from semantic roles to syntactic
positions. This difference follows in part at least from the different valence
requirements or semantic frames of the verbs faire 'do, make' and avoir
'have'. Given the theoretical orientation of this paper, I will pay special
attention to the fact that this difference has PRAGMATIC implications: The
two sentences cannot be used in the same discourse contexts, i.e. the choice
of one or the other will depend on the informational structuring of the
proposition in the discourse.
Let us start with faire mai The semantic frame of this predicate
resembles that of English hurt in that it contains all five roles listed in (5),
24
That in (24a) both mon pied and le pied are in principle possible has to do with the
grammar of inalienable possession in French. This fact is not directly relevant for the
argument that follows. It is mentioned for completeness and in order to avoid misunder
standings due to the use of one rather than the other construction.
i.e. experiencer, location, stimulus, agent, and instrument. And as in English,

both the location and the stimulus may appear as subjects, as (25a) and
(25b) show:
(25) a. Le/Mon pied me fait mal. (=24a) <LOC - EXP>
b. Ces chaussures me font mal. <STI - EXP>
these shoes to-me make pain 'these shoes hurt'
However there is an important difference between the two languages
concerning the role of the EXPERIENCER: With faire mal, the experiencer
cannot appear as a subject. Rather it must appear as a dative argument (in
our examples me). Moreover, while in transitive sentences involving
English hurt the experiencer role typically remains unexpressed or is
combined with the possessor role within the body part NP (see exx. (6)
through (8)), with French faire mal the experiencer must be overtly
expressed. Thus the corresponding versions without the pronoun me in (25')
are ungrammatical:
(25') a. *Le/Mon pied fait mal. <LOC>
b. *Ces chaussures font mal. <STI>
Given the contrast between (25) and (25'), we may say that there is one
more obligatory syntactic argument position in French than in English.25
If both the location and the stimulus argument are present in the
sentence, the situation with faire mal is similar to that with English hurt in
that the stimulus outranks the location for subject selection. This is shown
in the contrast between (26) and (27), which parallels the contrast between
(8) and (9) in English. As in (9), the symbol @ indicates pragmatic
anomaly:
(26) Ces chaussures me font mal aux pieds. <STI - EXP - LOC>,
<*LOC-EXP-STI>
these shoes to-me make pain at the feet, 'these shoes hurt my
feet'
(27) @ Les pieds (*me) font mal à ces ch. <STI - (EXP) - LOC>,
<*LOC-(EXP)- STI>
25
The experiencer argument may remain unexpressed in French in cases of
'Indefinite Null Complementation' (Fillmore 1986), i.e. if the referent is not specific, as
e.g. in Ca fait mal 'that hurts',
174 KNUD LAMBRECHT
the feet (to-me) make pain at these shoes 'my feet hurt these
shoes'
(If the experiencer argument is present in (27), the sentence is not only
pragmatically anomalous but uninterpretable; hence the star in front of me.)
The similarity between French and English with respect to the ranking of
the two semantic roles is predicted by the subject selection hierarchies in
(10) and (12), which are assumed to be universal (see Van Valin 1993).
Alongside this similarity in the respective ranking of the stimulus and
the location there are again important differences between French faire mal
and English hurt. With faire mal, the locus of pain does not appear as a
direct object argument, as my foot does in the English (8). Rather, it must
be expressed in the form of the locative prepositional phrase au pied.
meaning 'at the foot'. This makes semantic sense, since the locus of the
pain is a body part and since locations tend to be expressed by adpositions
across languages (ci English She hit me on the head). And as in (25), we
observe in (26) that the experiencer of the pain must be coded as a
separate pronominal argument. The alternative version of (26) in (26'),
which lacks this argument, is as unacceptable as (25') (in a non-metaphori
cal reading, as indicated by the semantic roles in angled brackets):
(26') *Ces chaussures font mal à mes pieds. <STI - LOC>
these shoes make pain at my feet 'these shoes hurt my feet'
Finally, unlike with English hurt (ex. (14) and discussion), the experiencer
can never appear as the subject of faire mal.
This leads us naturally to the alternative lexical expression avoir mal.
The semantic frame of this predicate involves only two roles: Experiencer
and location. The experiencer is necessarily expressed as a subject, and the
location can only appear as a locative prepositional phrase. The predicate
avoir mal thus replaces faire mal if the experiencer argument is to appear
as subject. The locative argument of avoir mal is optional, as (28b) shows:
(28) a. J'ai mal au pied, 'my foot hurts' (=25b) <EXP - L O O
b. J'ai mal. 'I feel pain/I hurt' <EXP>
c. *Mon pied a mal 'my foot hurts' <LOC>
Sentence (28c) is ungrammatical (in the intended non-metaphorical
reading) because only an experiencer can be the subject of avoir mal
Moreover, since the non-subject argument of avoir mal can only be the
locus of pain, a sentence such as (29) can only have a facetious reading:
(29) T a i mal aux chaussures, 'my shoes hurt/I feel pain at my

shoes' <EXP-STI>
Given the semantic frame of avoir mal, the prepositional phrase in (29) is
necessarily interpreted as a body part, hence the comical effect.26
4.2. Syntactic, semantic, and pragmatic roles
I will now turn to the three-termed relation between semantic, syntactic,

and pragmatic roles in French. Here the emerging picture differs radically
from that in English. As I have shown in previous work (Lambrecht 1984,
1987a, 1988a), there is a quasi-absolute correlation in French between topic,
preverbal position, and subject on the one hand, and focus, postverbal
position, and non-subject on the other, especially in the spoken language.
The information-structure constraint expressed in the pragmatic role
hierarchy in (21) appears in spoken French in the form of strict one-to-one
correspondence between syntactic and pragmatic roles.27
Let us first assume a discourse context in which the propositional
content 'My foot hurts' is to be expressed in the form of a PREDICATE-
FOCUS STRUCTURE, as in our model example (2c).28 The French
predicate-focus sentence corresponding to the English and Italian predicate-
focus sentences in (2) is shown in (30):
26
The same effect is possible (but not necessary) in English with the gloss My shoes
hurt, in particular with the possessive my instead of the demonstrative these, and with
sentence-focus articulation. The effect is exploited in the Dennis-the-Menace cartoon
reproduced at the beginning of this paper.
27
One (apparent) exception to the equation between focus and non-subject in French
is the presentational [il V-unacc NP] construction, illustrated in (i)
(i) Il est venu trois femmes. 'There came three women'.
it is come three women
in which a focal lexical subject FOLLOWS the verb. In such constructions the subject
properties are distributed over the non-preverbal pronoun il (which can be neither a topic
nor a focus expression) and the postverbal lexical subject constituent.
28
In spoken French, predicate-focus structures involving lexical topic NPs typically
require a 'detachment' (or 'dislocation') construction, in which a bound preverbal pronoun
cooccurs with the lexical NP (see Barnes 1985; Lambrecht 1987a), while in the written
standard the canonical SV(O) structure is preferred. For reasons of simplicity, I will use
the canonical versions in the example sentences below.
176 KNUD LAMBRECHT
(30) Le pied me fait encore MAL <LOC/SUBJ/TOP - EXP/OBJ/

TOP>
the foot to-me makes still pain 'My foot still HURTS'
In (30), the primary topic (the locus of pain) is the subject NP le pied. The
secondary topic (the experiencer) referring to the speaker, is syntactically
expressed with an indirect object pronoun, as in the Italian sentence (2b).
The focus accent falls on the last word of the predicate phrase, i.e. on mal.
Since the lexical structure of faire mal allows the body part to become a
subject, hence a topic, no clash between syntax and pragmatics arises in
(30). The structure of this sentence conforms to the information-structure
requirements of French.
Let us now assume that the proposition underlying (30) is to be uttered
in the discourse context indicated in example (3). Neither the structure in
(31a) nor that in (31b) is acceptable (the symbol # indicates an informa-
tion-structure violation and the asterisk, as usual, ungrammaticality):
(31) a. #Le PIED me fait mal.
b. *Me fait mal le PIED.
The structure in (31a) with its accented preverbal subject noun, which
represents the English strategy for the expression of sentence focus, is not
viable in French because of the constraint against the co-mapping of
subject and focus in preverbal position (a constraint which is no doubt
explainable at least in part by the oxytonic rhythmic pattern of French,
which prevents the sentence accent from moving leftward in the sen
tence).29 The subject-verb inversion structure in (31b), which represents
the Italian (and general Romance) strategy, is impossible because it
violates the constituent order constraints of modern French.
As I mentioned at the beginning, the grammar of French provides two
solutions to this conflict between the requirements of information structure
and the structural constraints of the sentence, one constructional, the other
lexical. The lexical solution involves selection of the alternative lexical
structure avoirmal.Since in this structure the locus of pain (the body part)
The exact nature of the violation in (31a) is unclear to me. Given the oxytonic
rhythm of French, one might consider (31a) an instance of phonological (prosodic) ill-
formedness rather than a violation of an information structure constraint. However (31a)
is not, strictly speaking, ill-formed. Leftward 'accent movement' does occur occasionally
in French, especially in metalinguistic corrections of previous statements.
necessarily appears as a non-subject argument, a sentence containing this
structure allows the focus accent to fall on a postverbal NP, i.e. in its
unmarked position. The resulting sentence is given in (32):
(32) J'ai mal au PIED (=24b) <EXP/SUBJ/TOP - LOC/OBJ/
FOO
I have pain at-the foot 'My FOOT hurts'
Moreover, since the experiencer argument of avoir mal is necessarily the
subject and since, as I have emphasized, the experiencer (i.e. the owner of
the body part) is pragmatically a topic in both contexts, it becomes possible
for the subject and the topic to be combined in preverbal position.
Notice that formally the resulting sentence (32) does not have sentence-
focus but predicate-focus articulation, given that its subject is a topic.
French thus expresses here via a predicate-focus construction a pragmatical
ly structured proposition which English and Italian express via sentence-
focus constructions. This is a noteworthy fact. It shows that it is not always
possible to predict from a given discourse context (such as the one assumed
in the model example (3)) which focus articulation will be used in a given
language to express a given propositional content. What IS predictable from
the utterance context, and what remains constant across languages, is that
a certain NP will be marked as a focus rather than a topic constituent (in
our case the NP designating the speaker's foot). But whether this focus
constituent will appear as a subject, resulting in a sentence-focus structure,
or as an object, resulting in the unmarked predicate-focus structure,
depends on the valence of the predicate provided by the lexicon of the
language. It is important to understand that the term 'sentence-focus
structure' designates a GRAMMATICAL CATEGORY, whose occurrence,
though entirely motivated by pragmatic considerations, is contingent upon
the structural and semantic properties of the sentence in which it is
instantiated. Information structure MOTIVATES grammatical form, it does
not DETERMINE it.30
The lexical alternation involving the predicates faire mal and avoir mal
and its relevance for the informational structuring of the French sentence
30
The observation that a given pragmatically structured PROPOSITION may be
formally expressed either by a sentence-focus or by a predicate-focus STRUCTURE makes
it necessary to abandon the notion, which I indirectly defended in Lambrecht 1987b, that
there is a necessary correlation between the thetic judgment type and the sentence-focus
structure.
178 KNUD LAMBRECHT
is not an isolated phenomenon. The two expressions are formally and

semantically related to a set of pairs of psychological predicates, all of
which involve the verbs faire and avoir. Some members of this set are listed
in (33):
(33) X fait envie à Y - Y a envie de X
'X looks appealing to Y' 'Y feels like (having) X'
X fait peur à Y Y a peur de X
'X scares Y' 'Y is scared of X9
X fait pitié à Y Y a pitié de X
'X inspires pity in Y' 'Y pities X'
X fait honte à Y - Y a honte de X
'X makes Y feel ashamed' 'Y is ashamed of X'
One difference between faire/avoir mal and the predicates in (33) is the
semantic role of the X-argument, which is that of psychological stimulus in
(33) rather than of physical cause, as with 'hurt'. This difference is formally
manifested in the use of the preposition de rather than à in the right hand
column (à cannot introduce a stimulus role). Moreover, unlike the pair faire
mal/avoir mal the members of the pairs in (33) have the same theta frames
(they each involve a stimulus and an experiencer role). Finally, the
predicate pairs in (33) do not necessarily correspond to single predicates in
English, as in the case of hurt. Such lack of strict correspondence is to be
expected from lexical rather than syntactic alternations. But the pairs in
(33) have in common with faire/avoir mal that they allow speakers to
express identical states of affairs under differing pragmatic circumstances
in such a way that the same semantic role is allowed to appear as a topic-
subject in one sentence and as a focus-object in the other.
Another, well-known, pair of predicates allowing for alternative
mappings of pragmatic and syntactic roles is the pair avoir 'have' vs. être
'be', used in such contrasting synonymous sentences as illustrated in (34)
and (35):
(34) a. J'ai les mains FROIDES 'My HANDS are cold'
b. Mes mains sont FROIDES 'My hands are COLD'
(35) a. Il a les cheveux LONGS 'His HAIR is long / He has long

HAIR'
b. Ses cheveux sont LONGS 'His hair is LONG'
The lexical alternations in (34) and (35), like the ones in (33), may be said
to be PRAGMATICALLY MOTIVATED in ways similar to the alternation

involving faire mal and avoir mal?31
The second solution to the conflict between information-structure
constraints and the rules of sentence structure involves the special bi-clausal
sequence which I have called the avoir-cleft construction' (Lambrecht 1986a,
1988a). Since I have presented detailed descriptions of the syntax,
semantics, and pragmatics of this construction in the above-mentioned
papers, I will content myself here with a brief summary.
Example (36) illustrates the avoir-cleft version of the proposition
underlying (32):
(36) J'ai mon PIED qui me fait MAL. 'My FOOT hurts.'
I have my foot which to-me makes pain
<LOC/SUBJ/TOP - THEME/OBJ/FOC> <LOC/SUBJ/TOP -
EXP/OBJ/TOP>32
In (36), the proposition 'My foot hurts' is syntactically expressed via two
clauses, each of which shows the required mappings of semantic, syntactic,
and pragmatic roles. The avoir-clause has the purely pragmatic function of
introducing the referent 'my foot' in lexical NP form, in the postverbal
position French grammar reserves for that function. The subject pronoun
je, serves to deictically anchor the new referent in the discourse. In the
avoir-cleft construction the sequence j'ai stands in more or less free
variation with the sequence (il) y a NP 'there is NP', where il is a dummy
placeholder (which is optional in spoken French) and y an adverbial
pronoun (as e.g. in (Il) y a le téléphone qui sonne 'The PHONE is ringing';
see also example (41a) below). The second, relative, clause is predicational
in nature, i.e. by its form it suggests an assertion ABOUT the (by now
pragmatically activated) referent 'foot' which is coded in the relative
pronoun qui.
To understand the nature of the avoir-cleft it is important to see that
the association of the avoir-clause and the relative clause forms a GRAM-
31
As the English glosses 'He has long hair' and 'His hair is long' show, the contrast
between the predicates have and be in English allows in principle for the same pragmatic
contrasts as that between French avoir and être. For the analysis of the pragmatic function
of have in English cf. Lambrecht 1988c.
32
Following Foley and Van Valin 1948: 48, I assume that the semantic role of the
subject of 'have' or 'avoir' is locative.
180 KNUD LAMBRECHT
MATICAL CONSTRUCTION in the sense of the framework of Construction
Grammar, i.e. a complex grammatical unit with semantic, pragmatic, and
syntactic properties which are not predictable from other properties of the
grammar. In this construction, the first clause, e.g. J'ai mon pied 'I have my
foot' in (36), is semantically non-compositional in that the proposition it
expresses becomes meaningful only in combination with the proposition
expressed in the following relative clause. Both clauses present themselves
formally as topic-comment structures, with the pronominal subjects as the
topics; yet the combination of the two has the specific pragmatic function
of expressing a single proposition with sentence-focus articulation.33
Let us now consider the proposition These shoes hurt9, in which the
subject argument is the cause of pain. If the sentence expressing this
proposition is to be construed pragmatically as conveying information about
the topic 'these shoes', it will have the form of the predicate-focus structure
in (37):
(37) Ces chaussures me font MAL. <STI/SUBJ/TOP - EXP/OBJ/
TOP>
these shoes to-me make pain These shoes HURT'
(37) is pragmatically and syntactically well-formed. The cause-of-pain
argument (the shoes), which is a topic, appears as the subject NP ces
chaussures; the experiencer, which, as in (31), is a secondary topic, appears
again as the atonic pronominal object me.
But how can the proposition underlying (37) be articulated in a
discourse context in which the NP representing the cause of pain is to be
in focus, as in the English sentence-focus structure in (18) {These SHOES
hurt)? Both (38a) and (38b) are unacceptable, for the same reasons the
parallel structures in (31) were unacceptable:
(38) a. #Ces CHAUSSURES me font mal.
b. *Me font mal ces CHAUSSURES.
In this case, the lexicon provides no alternative structure (as it did in the
case of the ill-formed sentences in (31)) which would allow the experiencer
to become a preverbal subject-topic and the cause of pain a postverbal
33
As far as I can tell, (36) is not only semantically but also pragmatically equivalent
to (32). Such cases of free variation are not the norm, since the occurrence of lexical
alternatives such as (32) hinges on the existence of particular predicates in the lexicon.
object-focus. As we saw earlier (example (29)), a sentence like (39), in

which the cause-of-pain argument appears in the form of a locative
prepositional phrase, is ungrammatical (on a non-facetious reading):
(39) *J'ai mal à ces CHAUSSURES. <EXP/SUBJ/TOP - STI/OBJ/
FOC>
These SHOES hurt'
In (39) the information-structure constraints are respected, in the sense that
the topic and the focus argument appear in their required positions, but the
mapping from semantics to syntax violates the semantic frame of the
predicate avoir mal. The way out of the dilemma is offered by the avoir-
cleft construction, as shown in (40):
(40) J'ai ces CHAUSSURES qui me font MAL. These SHOES hurt'
I have these shoes which to-me make pain
<LOC/SUBJ/TOP - THEME/OBJ/FOC> <STI/SUBJ/TOP -
EXP/OBJ/TOP>
As in (36), the avoir-cleft permits the focus NP (here ces chaussures) to
appear in object position, while simultaneously allowing its referent to
appear as a subject in the relative clause.
The avoir-cleft construction is highly productive, in the sense that it
permits the expression of sentence-focus articulations with a great variety
of semantic frames. This is due to the fact that the construction is made up
of two clauses, each of which formally has its own focus structure. As a
result, the avoir-cleft is not subject to the same semantic and syntactic
constraints as the English (and Italian) sentence focus constructions (and
as the earlier-mentioned French [il V-unacc NP] construction). Thus all of
the unacceptable English sentence-focus readings in (20) can in principle
be expressed via avoir-clefting in French. For example the proposition in
(20a) ('John hurts all over') can be expressed as (41a), and the transitive
(20e) (These shoes hurt my feet') as (41b):
(41) a. (Il) y a JEAN qui a mal PARTOUT. (cf. (20a))
(it) there has Jean who has pain everywhere
b. J'ai ces CHAUSSURES qui me font mal aux PIEDS. (cf.
(20e))
I have these shoes which to-me make pain at-the feet
The sentences in (41) syntactically force sentence-focus construal by virtue
of being avoir-cleft constructions. In the corresponding English sentences,
182 KNUD LAMBRECHT
sentence-focus marking via subject accentuation was not possible because

of the semantic role of the subject (as in (20a)) or because of a cooccurring
lexical object argument (as in (20e)).
4.3. Summary
Let us summarize the situation in French concerning the relationship

between the various role types, beginning with the mapping of SEMANTIC
and GRAMMATICAL roles. French has two lexical predicates meaning 'hurt'
which differ from each other in their semantic frames, hence in the
mapping of semantic and grammatical roles. Faire mal has the same frame
as English hurt, and like the latter it allows both the locus-of-pain and the
cause-of-pain arguments to appear as subjects; and as in English, the
stimulus argument outranks the location argument when both are present
in a sentence. Unlike English however, French expresses the experiencer
role in the. form of a separate dative argument and the locus-of-pain
argument appears not as a direct object but as a prepositional phrase.
Moreover, faire mal does not permit the experiencer role to appear as a
subject» If the experiencer is to be the subject, avoir mal must be used, a
predicate whose frame involves only two semantic roles, which have fixed
grammatical positions: An experiencer/subject and a locus-of-pain/object
Confirming the tendency alluded to at the end of Section 3.1, the mapping
of semantic and grammatical roles is more constrained in French than in
English: Fewer semantic roles may be mapped onto the primary grammati
cal relations subject and direct object.
The mapping from PRAGMATIC to GRAMMATICAL roles is also more
constrained in French. Only the topic relation may be mapped onto the
grammatical role of (preverbal) subject, at least in the spoken language, the
focus relation being restricted to postverbal non-subject arguments. This is
true not only for the predicates discussed in this paper but for all predi
cates. Given this constraint, and given the syntactic constraints on constitu
ent order, French must resort to grammatical devices which allow
alternative mappings of pragmatic roles and syntactic positions while leaving
the semantic structure of the proposition intact. These devices appear
either in the form of special syntactic constructions, such as the avoir-cleft,
or in the form of alternative lexical structures permitting different
mappings of semantic and syntactic (hence pragmatic) roles.
In French, the main vehicle of pragmatic information is not prosody,
but syntax. -French however does not easily fit into the binary typology
proposed by Thompson (1978), according to which languages may be
divided into two major types: Those with 'grammatically controlled word
order' (say, English) and those with 'pragmatically controlled word order'
(say, Latin or Russian). In French, constituent order is grammatically
controlled as strongly as in English. Nevertheless French does not express
focus-structure alternations prosodically but via syntax, by using a variety of
pragmatically specialized grammatical constructions.
5. Conclusion
Let me recapitulate my observations about the various interdependence

relations between semantic, pragmatic, and grammatical roles in the
languages examined. Looking first at the relationship between SEMANTIC
and GRAMMATICAL roles, we notice that it is subject to constraints,
expressed in the semantic role hierarchies of Case Grammar, according to
which more agent-like arguments outrank more patient-like arguments for
subject selection. While the respective ranking of semantic roles seems to
be invariable across languages, there is crosslinguistic variation with respect
to the degree to which more patient-like roles may be mapped onto subject
position without requiring morphological or syntactic adjustments. English
allows relative freedom in this respect. This is tantamount to saying that in
English a relatively large number of propositional contents may be
expressed with the same simple morpho-syntactic structure using the
semantic frame of the same predicate. In a language like French, in which
fewer of the semantic roles of a given frame can be mapped onto subject
position than in English, valence-changing devices such as reflexivization,
or different lexical predicates, must be used to permit an undergoer
argument to appear as subject. Crosslinguistically, the mapping of actor and
subject (and of undergoer and object) is unmarked, while the mapping of
undergoer and subject (and of actor and object) is marked.34
As for the relationship between PRAGMATIC and GRAMMATICAL roles,
34
As the reader will have noticed, I have not addressed in this paper the question of
the criteria according to which a syntactic constituent may be said to be the subject of its
clause, in particular I have not addressed the issue of ergative languages. The validity of
my crosslinguistic generalization remains to be investigated for such languages.
184 KNUD LAMBRECHT
it too is subject to grammatical constraints, expressed in what I have called

a 'pragmatic role hierarchy9 (item (21)), according to which a topic
argument outranks a focus argument for subject selection. The pragmatic
role hierarchy necessarily correlates with the semantic role hierarchy
because of the fact that both interact with syntax. The pragmatic role
hierarchy has no grammatical repercussions as long as the discourse context
permits the use of the unmarked predicate-focus structure, in which topic
and subject, on the one hand, and focus and object, on the other, coincide.
However if for pragmatic reasons a referent must be expressed as a focus
constituent (because it is not what the proposition is about) and if the
semantic role hierarchy requires this constituent to appear as a subject, a
clash arises between formal structure and information structure. This clash
requires a change in the prosodic, morphosyntactic, or lexical make-up of
the sentence.
The grammatical manifestation of the pragmatic role hierarchy is
subject to crosslinguistic variation. In a language like English, in which the
co-mapping of focus and subject is in principle permitted, it is sometimes
possible to change a predicate-focus structure (object focus) into a
sentence-focus structure (subject focus) by merely changing the prosodic
structure of the sentence. But this possibility is subject to various semantic
and syntactic conditions. The subject must have a certain type of non-
agentive role, the predicate must be amenable to a presentational or
eventive interpretation of the sentence, and no lexical object NP may
cooccur with the subject. In a language like Italian, the co-mapping of focus
and subject is permitted under similar conditions, but in addition to
receiving the sentence accent, the subject is placed in postverbal, i.e. object
position, preventing any object arguments from appearing after the verb
(see Wandruszka 1982). The constraint against cooccurring postverbal
object arguments in Italian sentence focus structures is clearly reminiscent
of, and no doubt governed by the same principle as, the constraint against
cooccurring lexical objects in English sentence-focus structures (see
Lambrecht 1988b). In French, where the pragmatic role hierarchy is strictly
adhered to, i.e. where co-mapping of subject and focus is in principle
prohibited, recourse is necessary either to special grammatical constructions,
resulting in radical alteration of the constituent structure of the sentence,
or to alternative synonymous lexical structures in which the unmarked
mapping of pragmatic and syntactic roles is preserved.
The prosodic, syntactic, and lexical facts described in this paper have
implications for the view of the overall organization of grammar and of the
role played by information structure in the grammatical system. The fact
that information structure may motivate PROSODIC STRUCTURE is well
known. Sentence-focus structures involving accented subjects, such as those
found in English, have been discussed in the literature at least since
Bolinger 1954. A particularly abundant discussion has arisen from the
stimulating observations in Schmerling 1976 (in particular her Chapter 5).
The approach to focus structure sketched in the present paper (and
discussed in more detail in Lambrecht 1988b, 1992, and 1994) is one in
which subject accentuation is seen as an information-structure device for
indicating the marked mapping of a pragmatic and a syntactic role.
The influence of information structure on sentence form is not limited
to prosody. Like other pragmatically motivated grammatical constructions,
such as e.g. the often-discussed English it-cleft and wh-cleft constructions
(and their equivalents in other languages), the French avoir-cleft construc
tion demonstrates the far-reaching consequences information structure may
have for the SYNTACTIC organization of sentences. Since such constructions
have the specific effect of changing the mapping relation between syntactic
and pragmatic roles found in the corresponding canonical sentences, they
may be referred to as PRAGMATIC-RELATION-CHANGING constructions.
These are in many ways comparable to such constructions as passive,
dative shift, tough-movement, etc., which are often referred to in generative
grammar as 'grammatical-relation-changing' constructions, a term which
primarily evokes their semantic and syntactic properties. However such
constructions may equally well be seen as permitting alternative mappings
of syntactic and pragmatic roles.
Finally, the data analyzed in this paper show that information structure
may manifest itself in certain aspects of the LEXICAL SELECTION process.
Crucial in this respect is the contrast between the structures in (30) and
(32) {Le pied me fait mal - J'ai mal au pied). If the mental process of
sentence formation only involved procedures for relating syntactic structure,
semantic structure (or logical form) and phonetic form, it would be difficult
to understand why alternative lexical structures like faire mal and avoir mal
should EXIST in a language and why a speaker would ever use one rather
than the other. The contrast between (31) and (32) shows that the
selection of a predicating expression from the lexicon is sometimes
determined not only by the contribution of the expression to the meaning
of the sentence but also by information-structure constraints on the
186 KNUD LAMBRECHT
syntactic form of the sentence in which the predicate is used to express this
meaning. There is a point where lexical structure, grammatical structure,
and information structure meet.
In selecting a predicate to express a proposition, a speaker must take
into account rules determining which of the roles in the semantic frame of
the predicate is to be mapped onto which syntactic position in the sentence.
But the speaker must also take into account pragmatic rules of information
structure determining which pragmatic role is to be associated with which
syntactic position, in accordance with her assumptions about the mental
state of the addressee. These pragmatic rules may in turn have conse
quences for the formal organization of the sentence (focus accent on the
subject vs. on the predicate, preverbal vs. postverbal position of the subject,
use of a special grammatical construction, choice of a different predicate,
etc.).
Facts such as those described in this paper suggest that a complete
explanation of the grammatical structure of a sentence requires not only
a theory of the relationship between semantics and syntax on the one hand,
and between syntax and prosody on the other, but a theory of the four-
termed relation between semantics, syntax, prosody, and information
structure. This paper has suggested one way in which this relation may be
described.
References
Allerton, DJ. & A. Cruttendon

1979 "Three reasons for accenting a definite subject." Journal of Linguistics
15.49-53.
Barnes, Betsy
1985 The pragmatics of left detachment in spoken standard French. Amsterdam:
John Benjamins.
Bolinger, Dwight
1954 "English prosodic stress and Spanish sentence order." Hispania 37.152-156.
Bresnan, Joan & J.M. Kannerva

1989 "Locative inversion in Chichewa: A case study of factorization in grammar."
Linguistic Inquiry 20.1-50.
Burzio, Luigi
1981 Intransitive verbs and Italian auxiliaries. MIT dissertation.
Chomsky, Noam
1972 "Deep structure, surface structure, and semantic interpretation." In Studies
on semantics in generative grammar. The Hague: Mouton, 62-119.
Comrie, Bernard
1981 Language universals and linguistic typology. Chicago: The University of
Chicago Press.
DeLancey, Scott
1991 "Event construal and case role assignment." BLS 17.338-353.
Dowty, David
1991 "Thematic proto-roles and argument selection. Language 67. 547-619.
Faber, David
1987 ''The accentuation of intransitive sentences in English. " Journal of Linguistics
23.341-358.
1968 "The case for case." In E. Bach & R. Harms (eds), Universals in linguistic
theory. New York: Holt, Rinehart & Winston, 1-88.
1986 "Pragmatically controlled zero anaphora." BLS 12.95-107.
1988a. "The mechanisms of 'Construction Grammar'." BLS 14.35-55.
1988b. On grammatical constructions. University of California, Berkeley.
Fillmore, Charles J. & Paul Kay

1993 Construction Grammar coursebook. University of California, Berkeley.
Fillmore, Charles J., Paul Kay & Mary Catherine O'Connor

1988 "Regularity and idiomaticity in grammatical constructions: The case of 'Let
alone'." Language 64.501-538.
Foley, William & Robert D. Van Valin, Jr.

Press.
Fuchs, Anna
1980 "Accented subjects in 'all-new' utterances." In G. Brettschneider & Chr.
Lehmann (eds.), Wege zur Universalienforschung. (Festschrift für Hansjakob
Seiler). Tübingen: Gunter Narr Verlag.
188 KNUD LAMBRECHT
Givón, Talmy '

1975 "Focus and the scope of assertion: Some Bantu evidence." Studies in African
Linguistics 6.185-205.
Goldberg, Adele
1992 Argument structure constructions. PhD dissertation, University of
California, Berkeley.
Gruber, Jeffrey
1965 Studies in lexical relations. PhD dissertation, MIT.
Halliday, Michael AK.

1967 "Notes on transitivity and theme in English, part II." Journal of Linguistics
3.199-244.
Hawkins, John
1981 "The semantic diversity of basic grammatical relations in English and
German." Linguistische Berichte 75.1-25.
Höhle, Tilman
1982 "Explikationen für "normale Betonung" und "normale Wortstellung"." In W.
Abraham (ed.), Satzglieder in Deutschen. Tübingen: Gunter Narr Verlag,
75454.
Jackendoff, Ray
1972 Semantic interpretation in generative grammar. Cambridge: MIT Press.
Kirsner, Robert S.
1973 "Natural focus and agentive interpretation: On the semantics of Dutch
expletive er." Stanford Occasional Papers in Linguistics 3.101-13.
Kuno, Susumo
1972 "Functional sentence perspective: A case study from Japanese and English."
Linguistic Inquiry 3.299-320.
Kuroda, S.-Y.
1972 "The categorical and the thetic judgment. Evidence from Japanese syntax."
Foundations of Language 9.153-185.
Lambrecht, Knud
1984 "A pragmatic constraint on lexical subjects in spoken French." CLS
12.239-256.
1986a. "Pragmatically motivated syntax: Presentational cleft constructions in spoken

French." CLS 22, Papers from the Parasession on Pragmatics and Grammat-
ical Theory, 115-126.
1986b. Topic, focus, and the grammar of spoken French. Unpublished PhD
dissertation, University of California, Berkeley.
1987a. "On the status of SVO sentences in French discourse." In R. Tomlin (ed.),
Coherence and grounding in discourse. Amsterdam: John Benjamins, 217-262.
1987b. "Sentence focus, information structure, and the thetic-categorical distinc

tion." BLS 13366-382.
1988a. "Presentational cleft constructions in spoken French." In J. Haiman &

Thompson (eds.), Clause combining in grammar and discourse. Amsterdam:
John Benjamins, 135-179.
1988b. "When subjects behave like objects: A markedness analysis of sentence focus
constructions across languages." Expanded version of a paper read at the
1987 LSA Meeting.
1988c. "There was a farmer had a dog. Syntactic amalgams revisited." BLS
14.319-339
1992 "Sentential-focus structures as grammatical constructions." Paper read at

the Annual Meeting of the Linguistic Society of America in Philadelphia.
1994 Information structure and sentence form: A theory of topic, focus, and the
mental representations of discourse referents. Cambridge: Cambridge
University Press.
Li, Charles & Sandra Thompson

1976 "Subject and topic: A new typology of language." In Charles Li (ed), 1976.
Subject and Topic. New York: Academic Press, 457-489.
Lyons, John
1977 Semantics, 2 vols. Cambridge: Cambridge University Press.
Mathesius, Vilém
1928 (1964) "On linguistic characterology with illustrations from modern English."
Actes du Premier Congrès international de linguistes à la Haye. Reprinted
in J. Vachek (ed.), 1964. A Prague School Reader in Linguistics. Blooming
ton: Indiana University Press.
Michaelis, Laura A.
1993 Toward a grammar of aspect: The case of the English perfect construction.
PhD dissertation, University of California, Berkeley.
190 KNUD LAMBRECHT
Perlmutter, David
1978 "Impersonal passives and the unaccusative hypothesis." BLS 4.157-189.
Reinhart, Tanya
1982 Pragmatics and linguistics: An analysis of sentence topics. Distributed by the
Indiana University Linguistics Club.
Sasse, Hans-Jürgen
1987 "The thetic/categorical distinction revisited." Linguistics 25.511-580.
Schmerling, Susan F.
1976 Aspects of English sentence stress. Austin: University of Texas Press.
Thompson, Sandra A.
1978 "Modern English from a typological point of view: Some implications of the
function of word order." Linguistische Berichte 54.19-35.
Van Oosten, Jeanne

1985 The nature of subjects, topics, and agents: A cognitive explanation.
Distributed by the Indiana University Linguistics Club.
Van Valin, Robert D.

1990 "Semantic parameters of split intransitivity." Language 66.221-260.
1993 "A synopsis of Role and Reference Grammar." In Van Valin (ed.), Advances
in Role and Reference Grammar. Amsterdam: Benjamins Publishing
Company.
Walters, John R.
1979 "Focus in Aghem: A study of its formal correlates and typology." Southern
California Occasional Papers in Linguistics 7.137-197.
Wandruszka, Ulrich
1982 Studien zur italienischen Wortstellung. Tübingen: Gunter Narr Verlag.
Ancestor-descendant and cultural-linguistic
relativity
Charles N. Li
University of California, Santa Barbara
1. Introduction
Fillmore's work on lexical semantics, especially in the area of deixis, e.g.

Fillmore (1971), (1972), (1982) ranks among the most seminal and
influential contributions in modern linguistics. Following this inspiring
Fillmorian tradition, I will examine the use of spatial-temporal terms in
Chinese expressions of ancestor-descendant relations and its implication for
cultural-linguistic relativity.
2. Ancestor-descendant relations in Chinese language and culture
It is universally known that Chinese civilization is steeped in the philosophi

cal tradition of Confucius, and the essence of this tradition revolves around
human relations of which the most important is the relation between
ancestor and descendant. In this relation as prescribed by the Confucian
ethic, the ancestor is accorded all of the power and privileges, and the
descendant all of the burden and duties. Thus, filial piety serves as the
primary Confucian virtue. All other virtues presuppose filial piety. It is a
millennia-old practice in China that all youngsters, educated or uneducated,
must memorize the "Twenty-four Stories of Filial Piety" which teach them
how they should obey, serve, and care for their ancestors. Severe condem
nation and social ostracization of a descendant who does not live up to the
expected level of filial piety is usually swift and automatic. But if a parent
chooses to neglect his/her descendant, there is no social convention that
192 CHARLES N. LI
condemns such a parent. The importance of filial piety in the culture has
given rise to ancestor worship, which has the characteristics of a religious
ritual. Each traditional Chinese family would typically display an altar at a
prominent spot in the house to commemorate the ancestors. On certain
days of the year, incense and candles would be burned and food would be
placed on the altar. Hence, an ancestor even after death commands a great
deal of power over his/her living descendants. In short, we can describe the
ancestor-descendant relation in China as an asymmetrical dimension of
ethical value with a positive direction toward the ancestor. The ancestor,
dead or alive, is always worthier than the descendant. The descendant is
forever beholden to the ancestor because it is the ancestor who gave life to
the descendant, and it is within the realm of social legitimacy for an
ancestor to take life away from a descendant.
The expressions of the ancestor-descendant relation in the Chinese
language reflect the asymmetry of cultural values. First, the temporal aspect
of the relation is by and large stated in terms of spatial metaphors.1
Among the pairs of spatial terms denoting height and positions: gao 'high' -
di low', qian 'front' - hou 'back', shang 'up' - xia 'down', the positive
member of each pair is used to express 'ancestor', and the negative member
of each pair is used to express 'descendant'. Consider the following data:
(1) (a.) gao yi bèi
high one generation
'One generation older'
(b.) qián yi bèi

front one generation
(c.) shàng yi bèi

up one generation
(2) (a.) di yi bèi

low one generation
1
There are also a few terms which employ temporal expressions, e.g. xian ren 'early
person = ancestor', wan ren 'late person = descendant'.
LINGUISTIC RELATIVITY 193
'One generation younger'
(b.) hôu yî bèi

back one generation
(c.) xià yi bèi

down one generation
(3) (a.) qiân - bèi

front generation
'ancestor or person of any ancestral generation'
(b.) hôu - bèi

back generation
'descendant or person of any descendant generation'
(c.) qián - rén

front person
'ancestor'
(d.) hôu - rén

back person
'descendant'
(e.) hòu - dài

back generation
'descendant generation'
Three pairs of spatial terms occur in (1) - (3): high-low, front-back, up-down.
The positive member of each pair, high,front,up, is used to designate the
ancestral direction, whereas the negative member, low, back, down,
designates the descendant direction. The features rendering high, front, up
'positive' and low, back, down 'negative' in Chinese are the same as those
in English discussed in Clark (1973) and Lakoff & Johnson (1980).
Thus in denoting the ancestor-descendant relation, the Chinese
language exhibits the same prejudice in favor of 'ancestor' as Chinese
culture does.
194 CHARLES N. LI
Correlating with the cultural-linguistic bias in favor of 'ancestor' is the

general Chinese perception attributing greater wisdom, ability and
knowledge to the 'ancestor'. This perception holds even when it runs
counter to obvious biological facts. For example, in traditional martial arts,
the master-disciple relation takes on the characteristics of the father-son
relation. Thus, the terminology for relations in the martial arts incorporates
most of the paternal kinship terminology, e.g.,
(4) (a.) shï fù
teacher - father
'master'
(b.) shi » xiǒng

teacher » elder brother
'a senior colleague trained by the same master'
(c.) shi - dì
teacher - younger brother
'a junior colleague trained by the same master'
(d.) shi - zŭ
teacher - ancestor
'founding master'
(e.) tu - dî
disciple - youth
'disciple'
(f.) tu - sŭn
disciple - paternal grandson
'disciple of a disciple'
In spite of the obvious biological correlation between old age and the
loss of physical prowess, the opposite correspondence is upheld in Chinese
folklore and novels on martial arts, i.e. the older a master of martial arts,
the greater the physical prowess he possesses.
To the Chinese mind, then, age is not only a repository of authority,
knowledge and wisdom, but also a source of physical prowess and agility.
The correlation of this mind set with the cultural practices and linguistic
expressions in China serves as an example of a Whorfian hypothesis - a
connection between language, culture, and mind. What is interesting about
this example, however, are the questions it poses: Does the Confucian ethic
still dominate the cultures of the various Chinese populations scattered
across the world? If the Confucian ethic is no longer upheld and the
Chinese language retains the positive-negative polarity in its expressions of
ancestor-descendant relation, what can one say about the Whorfian
hypothesis? Answers to those questions will shed light on the fundamental
nature of language.
3. The Whorfian hypothesis and language
The first and most obvious caveat for any Whorfian hypothesis is that there
isn't any necessary and sufficient condition for establishing correlations
between language, culture and mind. Even if one substantiates an
association between a certain cultural practice, X, with a certain language
pattern, Y, it does not mean that the existence of X necessarily implies the
existence of Y, or vice versa. In case of the ancestor-descendant relation,
the correlation which I have described does not imply that if the Confucian
ethic has been adopted by a society, the language of that society will
necessarily reflect its people's perception of age, ancestor and descendant.
Conversely the correlation which holds in the Chinese civilization does not
imply that if a language exhibits the Chinese pattern of expressions
concerning ancestor and descendant, the speakers of that language will
necessarily subscribe to the Confucian ethic.
To illustrate the former case, I will cite a Mongolian language, Baonan,
spoken in the eastern edge of Inner Asia in China. For centuries, the
Baonan people have completely adopted the Confucian ethic,2 and
borrowed a number of kinship terms from Chinese through contact with the
2
In a field trip to the Baonan villages, I was struck by the rigidity of the unequal
relation between parent and offspring. All power rests in the hands of parents, especially
the father, regardless of the age of the offspring. There is no question that nowadays, the
Baonan adheres much more strictly to such Confucian ethical norms as filial piety and
parental authority than the Chinese does. At one point I asked my Baonan language
consultant, whose son as an adult had already taken over the primary responsibility of
cultivating the family farm, what his son would do if he had ordered his son to follow a
particular farming practice that was clearly inappropriate. He responded with
laughter,"Well, that's unlikely. But if it did happen, my son would always do as I say.
Whether I am right or wrong would be irrelevant." His son confirmed this assertion of
parental authority.
196 CHARLES N. LI
Chinese-speaking Hui ethnic group in Inner Asia. But their language does
not show a positive direction toward 'ancestor' along the ancestor-descen
dant axis, and it does not have terminology of the type cited in (l)-(3). In
fact, Baonan does not have expressions for 'ancestor' and 'descendant'. The
Baonan word for parents is borrowed from a Chinese expression, xian-ren,
which literally means 'early people' in Chinese, and the Baonan term for
offspring is an indigenous word, rarla.
As an illustration of the latter case where the language exhibits a
consistent pattern of bias in favor of ancestor in the ancestor-descendant
dimension, but the people do not adhere to the Confucian ethic, I will cite
the transformation in contemporary China.
Over the past four decades the various political movements promoted
by Mao and his cohorts in China severely discredited the validity of the
Confucian ethic. Ancestor worship was prohibited. During the Cultural
Revolution and other political campaigns, rebelling against authority to the
point of denouncing one's own parents as counter-revolutionaries was
encouraged. The dismantling of ancestors as an exalted caste was also
enhanced by the policy of one child per family which has been strictly
enforced in urban areas. One of the consequences of having only one child
is that many Chinese parents have become such doting and indulging
parents that Chinese children have been described as "lording over their
parents". In other words there appears to be a shift from filial piety toward
parents to parental devotion to offspring. In addition, contact with western
civilization has resulted in significant changes of attitudes regarding the
ancestor-descendant relation, especially among Chinese populations in
Hong Kong, Singapore and other places.3 I would not suggest that the
Confucian ethic with respect to the ancestor-descendant relation is no
longer valid in Chinese civilization, but there is no doubt that it has moved
towards Western framework of values. Yet the Chinese language has not
changed or abandoned the terminological skewing shown in (l)-(3).
We may conclude that correlations between cultural principles,
linguistic patterns, and human perception do exist, but they are not
predictable. In other words, a language may or may not reflect a particular
pattern of social cognition and culture of its speakers. The reason for this
unpredictability lies in the nature of language and language change which
3
One example is that ancestor worship is no longer practiced in most families in
urban China, Hong Kong and Singapore.
I will try to capture in the following.

Beginning with Saussure (1916), the demarcation between synchrony -
the structure of language at a fixed point in time, and diachrony - the
change of language in time, became nearly absolute. The driving force in
twentieth century linguistics from Saussure to Chomsky has been structur
alism in synchronic linguistics characterized by the assumption that linguistic
principles and generalizations share the same properties as rules, laws and
theorems in physical or mathematical sciences.4 By now it is widely
recognized that linguistic generalizations and principles are far from being
water-tight and absolute. Almost all of them are context-sensitive tendencies
which depend heavily on Rosch's notion of prototype (Rosch 1977). One
of the primary reasons for the fuzzy nature of linguistic generalizations is
that linguistic data always represent extractions from long, intricate and
perpetual historical processes. This makes the 'science' of language
significantly different from the physical sciences. It is true that physical
objects such as stars, rocks and even sub-atomic particles are also subject
to change and may be seen as the end results of long historical processes.
But the change affecting physical objects does not involve unpredictable
behavioral, attitudinal, cultural, and socio-political changes. The unpredict
ability of language change is underscored by contact between languages, a
universal phenomenon. The evidence amassed over the past few decades
clearly show that socio-political and cultural factors, not linguistic factors,
constitute the determining forces shaping contact-induced language change
(Thomason and Kaufman 1988).
There is, however, one discipline in the natural sciences which does
resemble language in terms of perpetual change and the fuzzy nature of its
principles and generalizations. This discipline is evolutionary biology. Like
linguistic change, evolutionary change builds on pre-existing states.5 Like
linguistic phenomena, biological phenomena are layers and layers of
historical processes that may or may not be valid across the board. But
unlike biologists, who recognize the importance of evolution as a powerful
tool for explaining biological phenomena, most linguists, especially those
who subscribe to the various versions of the generative paradigm, have
See Givon (1979) for a critique of structuralism in 20th century linguistics.
The pre-existing states in evolution may be described as phylogenetic inertia and

preadaptation.
198 CHARLES N. LI
relegated diachronic linguistics to a position ancillary to synchronic

linguistics. In fact, theoretical linguistics in the twentieth century has been
played out almost exclusively in the synchronic arena dominated by such
concerns as simplicity, elegance, and so-called formalism.6 On the issue of
simplicity and elegance, I will quote Francis Crick (1988), who was a
physicist before becoming a biologist,
"Elegance and a deep simplicity, often expressed in a very abstract mathematical
form, are useful guides in physics, but in biology such intellectual tools can be very
misleading. For this reason, a theorist in biology has to receive much more guidance
from the experimental evidence, however cloudy and confused, than is usually
necessary in physics. [...] The basic laws of physics can usually be expressed in exact
mathematical form and they are probably the same throughout the universe. The
"laws" of biology, by contrast, are often only broad generalizations with significant
exceptions to them." (All italics are mine)
The relevance of Crick's insightful observation to the study of language
should be obvious.
4. Conclusion
I began this paper by citing the importance of Fillmore's work on lexical

semantics and went on to present some Chinese deictic terms relating to
the ancestor-descendant relationship, which led to the revelation that
because of the unpredictablility of language change and the fuzzy nature of
language, the ancestor-descendant terminology may or may not reflect the
social, cultural condition of a contemporary Chinese community. I
concluded by stating that language was not a coherent, logical, mathematics
like system as some linguists would like it to be. The empirical data
amassed over the last half century clearly demonstrate that the grammatical
system of a language is characterized by fuzzy edges. These fuzzy edges are
part of the prerequisite for a system which changes continuously in time
while maintaining its function as the primary communicative vehicle in
human interaction. While the fuzzy nature of language does not entail that
6
The so-called formalism in modern linguistics is no more than a set of notations
with a few simple conventions. Notations and conventions can be useful for the scholar
to present his/her analyses more succinctly and rigorously. As a former student of
mathematics and physics, I have always found the use of the term 'formalism' in linguistics
somewhat pretentious.
we should abandon our effort to seek out linguistic generalization whenever

we can, it does highlight the importance of lexical semantics, including
Fillmore's work in the study of deixis. Someday, linguists might discover that
for many languages, there are few morpho-syntactic principles governing
them, and the most appropriate analysis of such languages may be a set of
lexical and discourse principles rather than morpho-syntactic rules.
References
Clark, Herbert
1973 "Space, time semantics, and the child." In T.E. Moore (ed.), Cognitive
Development and the Acquisition of Language. New York: Academic Press,
28-63.
Crick, Francis
1988 What mad pursuit: A personal view of scientific discovery. New York: Basic
Books, Inc.
1971 Lectures on deixis. University of California, Santa Cruz, Summer Program
in Linguistics. Manuscript.
1972 "How to know whether you are coming or going." In K. Hylgaard-Jensen
(ed.), Linguistik 1971. Frankfurt: Athenäum, 369-379.
1982 "Toward a descriptive framework for spatial deixis." In Robert G. Garvella
& Wofgang Klein (eds.), Speech, place and action. Chichester: John Wiley
and Sons Ltd, 31-59.
Givón, Talmy
1979 On understanding grammar. New York: Academic Press.
Lakoff, George & M. Johnson

1980 Metaphors we live by. Chicago: The University of Chicago Press.
Rosch, Eleanor
1977 "Human categorization." In N. Warren (ed.). Advances in cross-cultural
psychology, Vol. 1. New York: Academic Press.
Saussure, Ferdinand de
1916 Cours de linguistique generale. Paris: Payot. Translated by Wade Baskin as:
Course in general linguistics. New York: McGraw-Hill, 1959.
200 CHARLES N. Lï
Thomason, Sarah Grey & T. Kaufman

1988 Language contact, creolization, and genetic linguistics. Berkeley: University
of California Press.
Remarks on suspended clauses:
A contribution to Japanese phraseology1
Toshio Ohori
University of Tokyo at Komaba
0. In natural discourse, it is a widely recognized fact that not all utterances

are grammatically complete. Furthermore, incomplete patterns often have
their own functions in context and indeed can be regarded as independent
constructions (in the sense of Fillmore, Kay, and O'Connor 1988). In this
paper, I take up what I would call suspended clauses in Japanese, and
examine their pragmatic properties. By suspended clauses I mean clauses
that bear markers for subordination (or more broadly interclausal
dependency) and yet are not accompanied by main clauses. Thus the
problem I am going to deal with is formulated in the following way:
(1) Under what conditions can a clause 'marked for subordination'
1
The present paper originates in part from Ohori (1990), and the question
formulated in (1) was originally presented by Charles J. Fillmore. As such, I consider it
is fitting to dedicate this study to him, with deep gratitude for his guidance in our
linguistic research.
Romanization of examples is phonemic, basically following Kunrei-system. Long
vowels are represented by doubling, e.g. sensee 'teacher'. Interlinears are given according
to the following conventions: (i) function words are glossed in capitals (linkage markers
are treated as primitives, so they are also in capitals); (ii) when a single word in one
language corresponds to more than one word in the other, dots are used instead of spaces,
e.g. ryuugaku-'study.abroad'; (iii) elements that are not grammatically realized in the
original sentence, e.g. subject NPs, are put in the parentheses in the English translation.
Abbreviations for function words are as follows: ACC(usative), BEN(efactive),
COMP(lementizer), DAT(ive), EMPH(atic), NEG(ative), NOM(inative),
NZ(=nominalizer), PERF(ective), POL(ite), PRE(fix), PRED(ication), PRT( = Particle),
STAT(ive), TOP(ic), VOL(itional). When case markers serve non-canonical functions, they
are glossed PRT.
202 TOSHIO OHORI
not be accompanied by a following main clause (assuming that

the intended utterance is completed)?
Besides genuine interest in the descriptive details of Japanese grammar, the
general significance of this study is twofold. First, it sheds light on the
formulaic nature of linguistic structure and is expected to offer a case for
the analysis of language based on the notion of constructions. Second, given
recent interest in the pragmatics of connective expressions (cf. Blakemore
1987; Blass 1990), the analysis of what functions a linkage marker may have
in the absence of the following clause should pose an interesting question
for pragmatic theory.
Let us start with the following example:2
(2) A: Zyugyoo yasumu no?
class skip PRT
'Are you going to skip the class?'
B: Tukare-tyat-te ne.
exhausted-PERF-TE PRT
'(I)'m exhausted.'
[The form tyat- (citation form tyau), glossed PERF(ective), is a
colloquial form of te-simau.]
In this mini-dialogue, B's utterance is incomplete in the sense that the
clause tukare-tyat-te tired-PERF-TE, '(I)'m exhausted' is marked as depen
dent by the desinence V-te, which is usually glossed 'participial'. Hence its
more exact translation would be '(I)'m exhausted, and ...', or 'Being
exhausted, ...'. Despite this grammatical incompleteness, the TE~marked
clause constitutes a complete utterance, and is marked by the pragmatic
particle ne, which frequently occurs utterance-finally (cf. Cook 1990 for an
accessible description). In the following, I will take up a variety of such
suspended constructions and discuss how they arise.
This paper is organized in the following way. In Section 1, I will present
some more illustrative examples of structurally incomplete sentences, which
are indeed heterogeneous. Then, in Section 2, I will sort them out according
to the paths of grammaticization, and focus on a few typical cases. In
Section 3, after discussing residual cases, I will offer some general remarks.
2
The data for this paper come from both introspection and casual observation of
everyday talk, reflecting the preliminary nature of the study.
REMARKS ON SUSPENDED CLAUSES 203
1. The following examples may give a more specific idea of suspended

clauses in Japanese. Morphological details are given in the brackets.
(3) A: Doo.si.te.mo ryuugaku-si-tai-n.desu.
by.all.means study.abroad-do-VOL-PRED
1 want to study abroad by all means.'
B: Zya sure-ba.
then do-BA
Then, please.'
[The desinence V-ba is normally glossed as 'conditional' or
'provisional'.]
(4) A: Sensee-no kai-ta-no kai-masi-ta yo.

teacher-PRT write-PAST-NZ buy-POL-PAST PRT
'(I)'ve bought (the book) that you (=teacher) wrote.'
B: It-te.kurere-ba age-ta-noni.
say-BEN-BA give-PAST-NONI
'(I)'d have given (you) (a copy) if (you) had requested.'
[NONI is a concessive connective, normally glossed 'although'.]
(5) Daremo tasuke ya si-nai-toiu.

anybody help PRT do-NEG-TOIU
'(The story is) nobody helped (me).'
[TOIU is a kind of complementizer, analyzable as to (=COMP)
and iu (='say', noun-modifying ending). As such, it is used for
noun complementation with the structure S-TOUIN, 'such N that
S'.]
Utterances such as these are commonly found in natural discourse. They
are structurally incomplete, but in discourse they are very often interpreted
as complete, sometimes urging turn-taking. In (3)-(5), what seems to be
going on is a functional shift from clause-linking particles to utterance-final
pragmatic particles. There is little feeling that the consequent clause is
missing in these examples. Here it should be made clear that there are two
sets of examples that must be put aside, namely echo questions and
204 TOSHIO OHORI
elliptical answers.3 For both of these uses, virtually any linking device
allows suspension. The criterion relevant to the present discussion is
whether a given linking device can be used in an utterance that is
self-initiated and yet structurally incomplete.
2. Of the several conditions that characterize suspended clauses, the most

obvious one is that they occur more frequently in the spoken register. This
is natural since ideas are expressed in 'spurts' in conversation (to borrow
Chafe's term, e.g. 1980), and do not have to be syntactically complete. Also,
considering the great negotiability of conversational interaction, the hearer's
active collaboration with the speaker may be one important factor for the
occurrence of suspended clauses. Further observations reveal, however,
much more interesting properties of suspended clauses. Importantly, the
suspended clauses have different paths of grammaticization. In (3), V-ba is
a part of an idiom (for example, V-ba ii, 'to do V would be fine'), so the
BA-marked clause stands alone as a result of reduction of a fixed expres.
sion. In (4), the suspended clause is not a part of any idiom, but the
reduction is largely due to the conventionalization of pragmatic inferences
associated with NONI-linkage. In (5), the linkage is complementation, unlike
in the other examples. The reduction takes place with the head noun
modified by the clause linked by TOIU, rather than with clauses. Below I
will focus on the first two types, and turn to the third type in Section 3.
3
Examples (i) and (ii) represent echo question and elliptical answer, respectively:
(i) A: Moo sukosi nomu-to, ...
more little drink-TO
'If (I) drink a little more, ...'
B: Moo sukosi nomu-to?
more little drink-TO
'If (you) drink a little more, then ... [what will happen]?'
(ii) A: Doo suru-to fairu-ga utuse-ru no?
how do-TO file-NOM transfer-can PRT
'How can (I) transfer a file?'
B: Koosite kurikku-sita mama aikon-o ugokasu-to, ...
thus click-do whle icon-A€C move-TO
'Move the icon while (you) click on it like this,
then...[you can transfer the file]'
Note that the sequential marker TO, which is not used in the suspended construction, is
possible in the above sentences.
2.1. The first class of examples to be identified is the idiom-fragment, which

is exemplified by B's utterance in (3), zya sure-ba. In this example, the
linkage marker BA is in fact a part of an idiom. In its full form, (3) could
be (6), with ii 'good; fine' following BA (with rising intonation).
(6) Zya, sure-ba ii-n.zya-nai?
then do-BA fine-PRED-NEG
Then, please,' lit. Then, if (you) do so, (that)'ll be fine, won't
it?'
Here, V-ba ii is an idiom that expresses suggestion for or permission of the
hearer's activity. Example (3) is a reduced form of (6), and in the latter the
desinence V-ba stands for the whole idiom in the absence of the rest. Since
V-ba-ii is highly fixed as an idiom, BA alone can be seen as a marker of
suggestion. Another conditional marker, namely TARA, also occurs in the
same context, though NARA happens not to, as the following examples
illustrate:
(7) Zya, si-tara/??suru-nara?
then do-TARA/do-NARA
Then, please.'
I am inclined to call this type of reduction constructional synecdoche because
a single conjunctive particle stands for the whole idiomatic construction
which it is a part of. Interestingly, because of the elliptical nature of the
expression, the suspended clause marked by BA or TARA tends to sound
rude and imply indifference (hence example (3) can be glossed, Then
please; nobody stops you').
Another example of suspended clause that is analyzed as an idiom
fragment involves TE, as in the following example:
(8) Tyotto kotti ki-te.
a.little(?) here come-TE
'Come here now.'
Here, TE can be a part of the idiom V-te-kudasai, V-te-kure, or V-te-tyoodai
(all V-TE-givë; or simply V-BEN, 'Will you please do the favor of V-ing?').
AS such, TE may be described as a clause-final particle of request.
Besides idiom fragments whose missing parts are recoverable in the
form of morphemes, there are examples of the suspended clause that are
still idiomatic but less fixed:
206 TOSHIO OHORI
(9) Hontoo.ni, o-sewa-ni nat-te.simat-te.

truly PRE-favor-DAT become(?)-PERF-TE
'(Thank you) so much for doing (me) such a favor/
In this example too, what comes after TE is crystal-clear: The expression of
thanks. However, what is predictable is its speech act type, not morphologi
cal form. Thus the 'complete' version of (9) could be (10), with arigatoo»
gozaimasu 'thank you', or (11), with sumi.mase.n 'I feel sorry', which in
Japanese cultural context can mean T m obliged3 and hence the feeling of
thanks.
(10) Hontoo.ni, o-sewa-ni nat-te.simat-te arigatoo.gozai.masu.
truly PRE-favor-DAT become(?)-PERF-TE thank.you
Thank you so much for doing (me) such a favor.'
(11) Hontoo.ni, o-sewa-ni nat-te.simat-te sumi.mase.n.

truly PRE-favor-DAT become(?)-PERF-TE sorry( = obliged)
Thank you so much for doing (me) such a favor/ lit. '(I)'m
really obliged to have received such a favor/
In a way, constructions like (9) can be viewed as idiom fragments, but
idiom' in this case is a highly abstract notion: It is a combination of a
specific morpheme (namely v-TE) and the speech act of expressing
gratitude, as exemplified by (10) and (11). This point raises an interesting
issue for the discussion of idiomaticity, as idioms here are defined with
reference to speech act type.
In some cases, (9) can be followed by another clause, as exemplified by
the following sentence:
(12) Hontoo.ni, o-sewa-ni nat-te.simat-te, kore-kara-mo yorosiku
o-negai si-masu.
truly PRE-favor-DAT become(?)-PERF-TE this-from-PRT consid
eration PRE-beg do-POL
"(Thank you) so much for doing (me) such a favor, and (I)
would like to ask your continuous support/
Here, the linkage between the two clauses is not direct, and the TE-marked
clause is indeed an instance of suspended construction. Literally, this
sentence can be paraphrased as '(I) owe so much to (you)-TE[-thank you],
and (I) would like to ask your continuous support'. The acknowledgement
of obligation and the request for continuous support is mediated by an
unexpressed and conventionalized message of thanks as shown in (10)-(11).
2.2. The second class of suspended clauses typically bears a (pseudo-)logical

connective, e.g. reason (because-type) or concession (though-type). In
Japanese, reason is most frequently marked by NODE and KARA, and
concession by KEDO (= reduced form of KEREDOMO) and NONI. Let us
start with suspended clauses that bear markers that express concession or
counterexpectation. Example (4), reproduced below as (13), is of this type.
(13) A: Sensee-no kondo kai-ta-no kai-masi-ta yo.
teacher-PRT lately write-PAST-NZ buy-POL-PAST PRT
'(I)'ve bought (the book) that you (=teacher) wrote lately.'
B: It-te.kurere-ba age-ta-noni.
say-BEN-BA give-PAST-NONI
'(I)'d have given (you) (a copy) if (you) had requested.'
Here, suppose A is a student, and B is a teacher. The student tells the
teacher that she bought a book that he wrote.4 The teacher is happy about
it, but he would be even happier to give a copy of his book to his student
personally. What is remarkable about B's utterance is that, although the
utterance itself can be glossed Though (I) would have given (you) (a copy)
if (you) had requested', or '(I)'d have given (you) (a copy) if (you) had
requested, but...', the missing portion is not recoverable in terms of either
morphemes or speech act types. It could be, for example, like (14), but in
this case the utterance is completely tautological.
(14) B: It-te.kurere-ba age-ta-noni, iwa-nakat-ta.
say-BEN-BA give-PAST-NONI say-NEG-PAST
'(I)'d have given (you) (a copy) if (you) had requested, but
(you) didn't.'
NONI in (13) implies that the clause to which it is attached is not true, just
like its counterfactuality is marked by subjunctive (had requested) in English.
As such, the additional clause iwa-nakat-ta '(you) didn't (request)' in (14)
provides no new information. This type of suspended clauses can be
characterized by the irrelevance of the (assumed) main clause. The
4
Here I refrain from using cumbersome non-sexist pronominal reference such as s/he,
his/her, and the like, in favor of clarity of prose. Besides, this example is a real quote
anyway, the teacher being myself.
208 TOSHIO OHORI
reduction of the main clause thus takes place due to its pragmatic
redundancy, without involving any sort of idioms as in the constructions
discussed in 2.1. In NONI-rnarked suspended clauses, because only the
semantic relation of counterexpectation is realized without the consequent
statement, their pragmatic effect involves regret or very mild blaming tone.
By now, it seems reasonable to consider a clause-linking particle as having
its own discourse function in the suspended construction.
The mechanism of inference that gives rise to this kind of suspension
is by no means simple. Here, NONI no longer codes any logical relation
between the linked clauses, but marks some aspect of the participant's
involvement in the conversational context. The pragmatic effects of such
suspended clauses are not easy to describe, but, generally speaking, the
effect of NÖNI may be rendered approximately, 'Alright, but the due
consequence has not followed in reality, you know'. In (13), the 'due
consequence' is the teacher's giving a copy of his book to his student. In
this example, there is only a feeling of weak regret in the teacher ( = B)'s
utterance. In some cases, however, concessivity or counterexpectation
encoded by NONI can give rise to a strong blaming tone. See also the
following example:
(15) Konnani hituyoo.na-noni.
this.much necessary-NONI
'(You) never know how much (I) need (you)', lit. Though (I)
need (you) this much, ...' or '(I) need (you) this much, but ...'
Here, what is to be inferred is not actually the consequent clause, but the
state of affairs that give rise to the unhappy situation, so a more exact
translation would be 'Although I need you so much, the due consequence
does not follow, which I regret'. In this case, an appropriate rendering of
NONI would be 'Alright, but I'm/you're not getting what I/you ought to'. As
a result, one of the main pragmatic effects of NONI is to urge the hearer to
think of the cause of misfortune.
In addition to NONI, there is another concessive marker, namely KEDO,
which is also frequently used in the suspended construction:
(16) Undookai-tte.ie.ba watasi are-o omoidasi-masu-kedo ne.
athletic.meet-TOP I that-ACC recall-POL-KEDO PRT
'Athletic meet? I recall that (incident).'
KEDO is a shortened form of KEREDOMO (the former is limited to the
spoken register). The pragmatic effect of KEDO in this context may be
That's right, but that's not the whole story' with a warning tone. Here the
challenged presupposition can be any aspect of what is being talked about,
for example, its positive/negative evaluation. Example (16) is taken from a
DJ's talk about athletic meets. To take part in them is usually supposed to
be fun. But the DJ comes to think of her own experience and says 'I recall
that incident, but...', the assumption being 'right, many people have fun in
athletic meets, but I don't totally agree on that point'. After this, she starts
to talk about her miserable experience, which however is funny to others.5
By now the paths for the functional extension from linkage markers to
pragmatic particles are schematized as follows:
(17) X-NONI, Y (Though X, Y') = > X-NONI, ø (Though X, the due
consequence hasn't followed,' or 'X, but I regret that the
adversative condition holds.')
(18) X-KEDO, Y (Though X, Y') = > X-KEDO, ø (Though X, that's

hot the whole story,' or 'X, but prepare for more.')
The difference between NONI and KEDO seems to arise, at least partially,
from the strength of counterexpectation (in some cases adversity) they
embody. In certain styles of conversation, the utterance-final KEDO is
almost bleached, and in some cases it is used merely to imply that the
speaker has certain reservation to what s/he is saying (the 'polite' tone
associated with this form comes from this function, cf. fn. 5). Example (19)
is from a broadcast of a baseball game:
(19) Kono.hen-de hasiru-to omosiroi-desu-kedo ne.
here-at run-TO interesting-PRED-KEDO PRT
'Here (it) will be interesting if (the runner) tries to steal a base.'
5
Itani (1992) recently discusses the utterance-final use of KEDO (together with
NODE), focusing on its 'softening' and hence 'polite' effect. One example is (romanization
and the gloss are slightly regularized):
(iii) Otya-ga hairi-masi-ta-kedo
tea-NOM ready-POL-PAST-KEDO
Tea is ready.'
By using KEDO, 'but', the speaker gives the hearer an option of either having or not
having the tea (in which case the omitted clause would be something like '...but would you
like to have it?'). I do not consider that the suspended version is always 'polite', but the
feeling of reservation as I described above in 2.2 may give rise to indirectness, which is an
important element of politeness.
210 TOSHIO OHORI
Here the speaker (=the broadcaster) is not saying that the runner really
ought to steal the next base. He is just wondering,with certain detachment,
how it would be if that happens and KEDO is used to express the 'prepare
for more' part of (18). In addition to NONI and KEDO, GA also participates
in this construction, though less frequently (when suspended, its function is
closer to KEDO).
Next, markers for reason, especially KARA, are frequently used in the
suspended construction. See the following examples:
(20) A: Kyoo hima?
today free
'Are (you) free today [i.e. this evening]?'
B: Ee, demo, tukare-teru kara.
yes but tired-STAT KARA
'Yes, but (I)'m tired, so...[I can't make it]'
[teru is a reduced and colloquial form of te.iru which is a stative
marker]
(21) Nani it-te.mo iu koto kika-nai-n.da-kara.
what say-TEMO say thing listen-NEG-PRED-KARA
'Whatever (I) say (s/he) won't listen to (me), so...'
In both of these examples, reduction takes place because the message of
the consequent clause is inferable. But they do differ with respect to the
significance of the suspended clause. In (20), since A wants to know
whether B can go out with him (for dinner maybe), B's giving an excuse
means a rejection of A's invitation. To this extent, the unexpressed message
of rejection is central to the communicative goal. In (21), on the other
hand, the reason clause is suspended even though the clause that follows
it is not identifiable. Here, the speaker is not giving any excuse, but is
complaining about somebody's stubbornness. Since KARA is a marker of
reason, a possible non-suspended version of (21) might be:
(22) Nani-o it-temo iu koto kika-nai-n.da-kara komaru yo.
what-ACC say-TEMO say thing listen-NEG-PRED-KARA troubled
PRT
'Whatever (I) say (s/he) won't listen to (me), so (I)'m troubled.'
The speaker's complaining tone comes from the meaning of KARA and the
inference it induces. In this example, the consequent part, that the speaker
is troubled, is inferred only indirectly. The speaker utters a statement, which
seems to provide a reason for some condition, but this consequent condition
is not expressed. Then, assuming that the speaker is talking about
something relevant to his/her present condition, the hearer makes an
inference that the KARA-marked clause expresses the speaker's current
concern. In this sense, the main clause in (22), '(I)'m troubled', is only one
possibility. A maximally broad interpretation of the KARA-marked
suspended clause is that the expressed statement has a strong concern on
the speaker and possibly the hearer. In an extreme case, KARA 'so' solely
marks the speaker's urge for the hearer's involvement, as in the following
example:
(23) Watasi nante motto taihen-dat-ta-n.da kara ne.
1sg EMPH more troubled-PRED-PAST-PRED KARA PRT
'I was even more troubled,' or 'I had an even tougher time, so...'
The pragmatic effect of suspension in this case may be paraphrased as 'I
had an even tougher time, and so you know what'. An expected response
would be, for example, 'Oh yeah, that's really bad'. Thus the hearer does
not simply recover the omitted part, but actively responds to the hearer by
showing sympathy.
Pragmatically, the suspension of the reason clause has the following
mechanism:
(24) X-KARA, Y ('X, so Y') = > X-KARA, ø ('X, so you know what,'
or 'X, so it concerns me/you')
NODE is also used with the same function, though less frequently. With
NODE, the speaker-hearer's concern seems a little weaker, partly because
it is associated with the formal register and is less forceful.
An important point to be noted in this connection is that linkage
markers that are not specialized for reason, such as SI and TE, can only
have a reason reading when they are suspended. Thus the Si-marked clause,
which by itself codes a very loose linkage as in (25), can be suspended when
it is put into an explanation- or excuse-frame, as in (26):
(25) Kinoo-wa eega-o mi-ta-si, kyoo-wa iti.niti hon-o yon-de.i-ta.
yesterday-TOP movie-ACC see-PAST-SI today-TOP all.day
book-ACC read-STAT-PAST
'(I) went to see a movie yesterday and was reading a book all
day today.'
212 TOSHIO OHORI
(26) A: Kyoo hima?

today free
'Are (you) free today [i.e. this evening]?'
B: Ee, demo, tukare-teru-si.
yes but tired-STAT-SI
'Yes, but (I)'m tired, so...[I can't make it]'
In the latter example, the direct answer to A's question is omitted in B's
utterance, and its explanation alone is given. This is the same with (20), and
the point here is that when suspended, Si-marked clause is used to solicit
inference (which is fairly straightforward in this case). If SI in B's utterance
in (26) were interpreted as having merely an 'additive' function (to use
Halliday and Hasan's term, cf. 1976), it would not convey the intended
message.
In the same vein, the reason interpretation of TE-linkage is predominant
when it is suspended, and it is perhaps the only interpretation of the
suspended use. This is in fact the case in (2), reproduced as (27) below:
(27) A: Zyugyoo yameru no?
class skip PRT
'Are you going to skip the class?'
B: Tukare-tyat-te ne.
exhausted-PERF-TE PRT
'(I)'m exhausted.'
While TE-linkage can be used to realize many semantic relations, including
sequential and juxtapositional ones, the function of the suspended
TE-marked clause is almost limited to causal relations. Of the various uses
of TE-linkage, concession as illustrated in (28) is perhaps the only possible
interpretation, besides reason, that may be assigned to a suspended
constrution, as in (29).
(28) Are.dake hatarai-te kono teido-no kyuuryoo ka yo.
that.much work-TE this amount salary PRT PRT
'Even though (I) worked that much, how should (I) get such a
low salary?'
(29) ?Are.dake hatarai-te.

that.much work-TE
'Even though (I) worked that much...'
The acceptability of (29) seems to vary according to speakers. In my dialect,

it is OK when the subject is third person and the conveyed message is
something like 'He worked that much, but oh, poor guy', with some
pejorative tone. In this case, the clause-final TE is lengthened (-TEE), and
bears an intonational peak, indicating its pragmatically loaded quality.
23. By now the relevant conditions for suspended clauses that have been
discussed so far are summarized as follows:
(30) A suspended clause occurs most frequently when:
(a) it is in the spoken register, especially conversation, and
(b) it is part of an idiomatic expression and can code the
meaning of the whole without the rest, or
(c) it is marked by a (pseudo-)logical connective, and the
intended message is either contextually inferable or convention
alized.
Condition (a) is self explanatory. Some linkage types, for example the
linkage by the 'gerundive' ending (ren'yoo-kee) of verbals, do not occur in
the suspended construction because they are mostly limited to the written
or formal register. Condition (b) refers to such idioms as V-BA-ii (suggestion
or permission), and V-TE-'thanks'. As discussed in 2.1, a clause is suspended
when the linkage marker can stand for the whole idiom in the communica
tive context. Condition (c) refers to cases in which the main clause is
omitted because it is inferable, and the pragmatic effect, for example regret
or blaming, is conventionalized, as we have seen in 2.2. This is the case with
NONI. Another marker of counterexpectation, KEDO, adds a reservational
nuance, implying that there is room for further consideration. The reason
marker KARA indicates the speaker's concern and consequently expresses
an urge for the hearer's involvement, usually sympathy. What is common
to these constructions is that they occur when the (pseudo-)logical conse
quence is contextually inferable. Markers for concession and reason are
typically inference-intensive (i.e. they put various sorts of constraints on
interpretation), and with these markers, the hearer tries to infer what is
coming after the dependent clause more actively than otherwise. This
means that, when faced with suspended clauses, the hearer must infer what
may come after them. In doing so, relevant aspects of the context are
picked up and generalized. When they are sufficiently generalized, they
become conventionalized part of the construction.
214 TOSHIO OHORI
3. In this section, I will first turn to a few other miscellaneous cases of

suspended clauses and then speculate on the cross-linguistic aspects of the
grammaticization of clause linkage.
3.1. The remaining class of suspended clauses involves the complementizer

TO or its variant. Example (5), reproduced as (31) below, is of this type.
(31) Daremo tasuke ya si-nai-toiu.
anybody help PRT do-NEG-TOIU
'(The story is) nobody helped (me).'
As remarked earlier, the structure S-TOIUN 'such N that S' is used for noun
complementation, and the head noun typically designates some kind of
event, such as wake and hanasi (both translatable as 'story'). Thus a
non-suspended version of (31) would be:
(32) Daremo tasuke ya si-nai-toiu wake yo.
anybody help PRT do-NEG-TOIU story PRT
'The story is nobody helped (me)/
It is understandable that, because the noun wake has only a general
meaning and hence is not very informative, the complement clause can
stand by itself in discourse. However, there is an interesting pragmatic
effect of this suspended construction. With TOIU attached clause-finally, the
speaker may imply detachment or discommitment to the conveyed message,
Example (31) could be like (33), without TOIU, but there is a pragmatic
difference:
(33) Daremo tasuke ya si-nakat-ta.
anybody help PRT do-NEG-PAST
'Nobody helped (me)/
In (31), the speaker reports the event as if she was an observer when it
happened. Even though one who suffered is the speaker herself, the use of
TOIU renders the utterance as if it were about somebody else. But in (33),
this detachment is not felt, and the speaker simply talks about her
experience from her own point of view. Were I to coin a term, I would
name the function of TOIU in the suspended construction as anti-evidential
because it obscures the credibility of information by impersonalizing it.
Here too, the linkage marker undergoes a functional shift and becomes
grammaticized as a clause-final pragmatic particle, though the mechanism
is different from those of the markers examined above.

Two other suspended constructions that involve complementizers are
given below, though only brief descriptions are provided in want of space.
(34) Katto-nat-te yat-ta-toiu.ka.
mad-become-TE do-PAST-TOIU.KA
'(I) became mad and did (that), sort of.'
[TOIU is a complementizer as in (31). KA is homophonous to a
question particle, and also to a disjunction marker, as in A ka
B (ka), 'A or B'. Thus a highly literal translation of (34) would
be, 'Rather than to say (I) became mad and did that,...' TOIUKA
is here used in order to avoid the speaker's responsibility for
the expressed state of affairs.]
(35) Sonna koto nai-tte.

that thing NEG-TTE
That can't be true.'
[TTE is a reduced form of (T)TE-BA, which is a kind of emphatic
marker by itself. The original and full construction is X TO-ie-BA
X, X COMP-say-BA X 'if (I) say X, then (it)'s X'. The comple
mentizer TO is realized as (T)TE because of the following vowel.
Considering the pragmatic function of TTE, the translation may
better be, That can't be true, I bet'. I owe this analysis to
Derek Herforth (p.c.).]
In (34), the speaker admits that he became mad and did something wrong,
but the use of the particle KA provides room for an alternative account.
However, he is not certain about what to say in fact, so the indeterminacy
expressed by KA serves to convey the speaker's abandoning of responsibili
ty. In this sense, the function of TOIU has considerably shifted from its
original use as a marker of noun complementation to a pragmatic particle.
This use of TOIU, which is often a sign of insincerity, is spreading among
young people quite rapidly.6 The use of TTE in (35) is a kind of reduction
Sociologically, it would be intriguing to think about the rise of these constructions

in conjunction with the claim that younger Japanese people are more assertive and
straightforward, perhaps being influenced by the Western culture. I believe that the matter
is not that simple, and the existence of constructions whose primary purpose is to convey
detachment from responsibility can be interpreted as reflecting the continuing mentality
to obscure the ultimate source of responsibility.
216 TOSHIO OHORI
from a fixed expression, but it is doubly reduced, first from X TO-ie~BA X to
X-TTE-BA, and then from X-TTE-BA to X-TTE. To this extent, this example
is comparable to other suspended constructions that are based on idiom
fragments, as discussed in 2.1.
3.2. Here, I would like to discuss the nature of suspended clauses from a
broader perspective.
First, suspended clauses can be aptly considered as constructions in
their own right, in the sense of Fillmore (e.g. 1988; also cf. Fillmore, Kay,
and O'Connor 1988). That is, suspended clauses have their own discourse
functions that are not manifest in a non-suspended version, and their
well-formedness conditions must be specified as a set of features that
pertain to particular lexico-grammatical configurations. Notice that the
pragmatic effects of suspended clauses we have examined so far are only
partially reducible to the properties of clause-linking devices themselves.
For example, the blaming or regretting tone of NONI, or the urge for the
hearer's involvement with KARA should best be described as a property of
the suspended construction, for such effects are found only when suspended
(also cf. the case of SI and TE in (26)-(27)). At the same time, it does not
seem to be reasonable to reduce everything to pragmatic inferences and
make no reference to structural properties. To do so would fail to recognize
the highly conventionalized nature of the pragmatic effects associated with
suspended constructions. The foregoing analysis thus seems to support the
view that the notion of construction is important and indeed indispensable
for understanding the structure of language.
Second, one important issue to be put on the agenda is the explication
of the pragmatic mechanism involved in the suspension of dependent
clauses. Since reinforcement of pragmatic inferences and their conventional
ization are crucial aspects of grammaticization, a closer look at the
inferential processes associated with suspended clauses will be necessary.
In this connection, suspended clauses seem to raise an interesting issue for
the pragmatic theory of connectives. In Blakemore (1987), it is argued that
the function of connectives is to put semantic constraints on the way an
utterance is interpreted. On the one hand, suspended clauses can be seen
as demonstrating the usefulness of the theory, because it says that
connectives guide the hearer's inference and this function may not change
whether the dependent clause is suspended or not. On the other hand, the
'semantic constraint' theory runs into a trouble when faced with the fact
that pontentially polysemous markers such as the participial TE can only

have reason for its normal reading in the suspended construction. There is
no way to predict that this particular semantic relation is picked up when
the TE-marked clause is suspended. This difficulty can be avoided if we
resort to a construction-based theory and say that the suspended clause, as
a lexico-grammatical frame, best fits an inference-intensive semantic
relation.
Finally, suspended clauses offer an interesting case of grammaticization.
Very often grammaticization is taken to be a process in which a lexical item
is converted to a grammatical morpheme. Yet it may not be misleading to
conceive of the rise of polyfunctionality as another mode of grammaticiza
tion in the sense that a morpheme which is already a part of grammar
acquires a new grammatical function. In the present case, the emergence
of suspended clauses can be seen as conventionalization of discourse
functions with clause-linking constructions. Cross-linguistically, the
examination of comparable constructions in other languages will be
interesting. For example, Haiman (1988)'s discussion of 'inconsequentials'
in Hua, a Papuan language, appears to have something to do with the
present discussion of suspended clauses.
In this paper, I hope to have demonstrated some wrinkles about the

pragmatics of clause linkage constructions in Japanese, together with their
significance for general linguistic research. A more in-depth analysis of
suspended clauses, based on larger amount of data, will be our future task.
References
Blakemore, Diane
1987 Semantic constraints on relevance. Oxford: Blackwell.
Blass, Regina
1990 Relevance relations in discourse: A study with special reference to Sissala.
Cambridge: Cambridge University Press.
Chafe, Wallace L.
1980 "The deployment of consciousness in the production of a narrative". In
Wallace L. Chafe (ed.), The pear stories: Cognitive, cultural, and linguistic
aspects of narrative production. Norwood, NJ: Ablex, 9-50.
218 TOSHIO OHORI
Cook, Haruko Minegishi

1990 "The sentence-final particle ne as a tool for cooperation in Japanese
conversation". In Hajime Hoji (ed.), Japanese/Korean linguistics. Stanford:
CSLI, 29-44.
1988 "The mechanisms of 'Construction Grammar'". Berkeley Linguistic Society
14.35-55.

1988 "Regularity and idiomaticity in grammatical constructions: The case of let
alone" Language 64:3.501-538.
Haiman, John
1988 "Inconsequentials in Hua and the typology of clauses". In John Haiman &
Sandra A. Thompson (eds.), Clause combining in grammar and discourse.
Amsterdam: John Benjamins, 49-69.
Halliday, Michael A.K. & Ruqaiya Hasan

1976 Cohesion in English. London: Longman.
Itani, Reiko
1992 "Japanese conjunction kedo ('but') in utterance-final use: A relevance-based
analysis". English Linguistics 9.265-283.
Ohori, Toshio
1990 Clause linkage in Japanese: A synopsis. Special Field Examination, The
University of California, Berkeley. To be distributed by the Indiana
University Linguistics Club.
Pragmaticization of meaning in some
sentence-final particles in Japanese
Shigeko Okamoto
California State University, Fresno
1. Introduction1
The complementizers (COMPs) no, koto, and to/tte (tte is a colloquial form
of to) in Japanese appear at the end of a complement clause, as illustrated
in (l)-(3): 2
(1) Boku wa Yoshio ga Masao o butu no o (*koto o/*to)mita.

I TM SM OM hit COMP OM see PST
T saw Yoshio hitting Masao.'
(2) John wa nihongo ga muzukasii koto o (no ol*to) mananda.

TM Japanese SM difficult COMP OM learn PST
'John learned that Japanese is difficult.' (Kuno 1973)
1
An earlier version of this paper was presented at the 66th Annual Meeting of the
Linguistic Society of America (1992). I would like to thank Charles Fillmore, Orin
Gensler, Kiyoko Hirose Ohara, Yoshiko Matsumoto, Masayoshi Shibatani for their
valuable comments and discussions.
The following abbreviations are used in this paper: AUX (auxiliary verb), COMP
(complementizer), COP (copula), DMT (diminutive suffix), GEN (genitive case marker),
IMP (imperative mood marker), NEG (negative auxiliary), OM (object marker), PASS
(passive voice), PPX (polite prefix), PRG (progressive aspect marker), PRT (particle), PST
(past tense marker), SEP (sentence-final particle), SM (subject marker), TM (topic
marker), and TQ (tag question marker).
220 SHIGEKO OKAMOTO
(3) Yoshio wa Masao ga sinda to/tte (*no o/*koto o) itta.

TM SM die PST COMP say PST
'Yoshio said that Masao died.'
The distribution of these COMPs is largely controlled by epistemic factors -
how the speaker has acquired the knowledge in question. Although there
are many Irregularities', no is typically used for a concrete or directly
perceived event (e.g. (1)), and koto for an abstract or indirectly perceived
event (e.g. (2)); to/tte is a quotative marker (e.g. (3)) (Kuno 1973; Josephs
1976; McCawley 1976; Maynard 1984).3
These uses of no, koto, and to/tte as COMPs are well known. There are,
however, other less well-recognized or well-defined uses of these mor-
phemes in a sentence-final position, as illustrated in (4) and (5):
(4) Saku no naka ni hairanai koto.
fence GEN inside in enter NEG
'You must not go inside the fence.'
(5) Nan da to. Tumaranai da to.

what COP boring COP
'What?! Boring?!'
In (4), koto is used as a directive; in (5), to expresses defiance. The present
paper investigates the functional divergence of such sentence-final usages
of no, koto, to, and tte.
As shown in (4a), one might analyze sentences such as (4) and (5) as
elliptical, consisting of a complement, a COMP, and an implicit main clause:
(4a) Saku no naka ni hairanai koto (o meeziru).
COMP OM order
T order you not to go inside the fence.'
Such an analysis, however, is too simplistic. In this study, I will closely
examine the properties of these sentence-final morphemes and demonstrate
that these morphemes are best regarded as sentence-final particles (SFPs)
rather than COMPs. (Martin's reference grammar of Japanese (1975) lists
most of the usages of these morphemes discussed here, treating them as
3
Although it is beyond the scope of the present paper, the actual distribution of the
complementizers no, koto, and to/tte is much more complicated than the generalizations
given here and requires further study.
PRAGMATICIZATION OF MEANING 221
sentence-final particles.) Employing the notion of grammatical construction
advanced by Fillmore and others (Fillmore 1985, 1988; Fillmore, Kay, and
O'Connor 1987),4 I will argue that the sentence-types S no/koto/to/tte (e.g.
(4) and (5)) are not context-dependent elliptical expressions, but indepen
dent grammatical constructions used for specific purposes, and that they are
not complements, but main clauses, in which these SFPs serve as modality
indicators. These main clause constructions are, however, closely related to
complement clause constructions. It is then argued that the usages of these
SFPs are instances of reanalysis of morphological functions through
pragmaticization of meaning (Traugott 1988, 1989; Traugott and Konig
1991).5
The paper is organized as follows: Section 2 outlines the semantic-
pragmatic properties of no, koto, to, and tte as SFPs. Section 3 compares
these final morphemes and the corresponding COMPs with regard to
morphosyntactic and semantic-pragmatic properties, and argues for the
status of the fomer as SFPs. Section 4 discusses the implications of the
analysis given in section 3.
2. Semantic-pragmatic properties of no, koto, to, and tte as SFPs
2.1. No
No as an SFP is typically used when the speaker gives, or asks for, an

explanatory or clarifying comment with regard to a certain situation in the
discourse context (cf. Kuno 1973; Martin 1975; Teramura 1984; Cook 1990;
Maynard 1992).6
4
Other studies that advocate the view of grammar based on the notion of
grammatical construction and other releant notions (e.g. formulas, preferred clauses,
preferred argument structures) include Du Bois (1985), Lakoff (1987), Lambrecht (1984,
1988), and Hopper (1987). Okamoto (e.g. 1993, 1994) analyzes a number of instances of
grammatical constructions in Japanese.
5
See also Heine et al. (1991) and Matsumoto (1988).
Kuno and Teramura discuss only no da, an expression related to no. I discuss the
difference between no da and no in section 3.1.
222 SHIGEKO OKAMOTO
(6) <from natural conversation>

Akira-tyan, Tie-tyan, onaka suita no. Onaka suita mitai ne.
DMT DMT hungry hungry look SFP
'Akira, Chie, are you hungry? You look hungry, am I right?'
For example, in (6) two children, Akira and Chie, are near the speaker,
who is preparing lunch. The children's behavior suggests that they are
hungry, and the speaker wants them to clarify their behavior. So the
speaker asks a question marked by no. If the children had not been showing
any signs of being hungry, the question would not include no:
(6a) Akira-tyan, tie-tyan, onaka suita.
Similarly, in (7), A uses no to ask for clarification of her observation of how
B prepared the zucchini. Further, in (7) and (8) B uses no to provide an
explanation in response to A's question:
A: Nee, kono zukini wa tada tiizu o mabusita dake na no
hey this zucchini TM only cheese OM coat PST only COP
"Hey, did you just coat this zucchini only with cheese?'
B: Aa, are, panko to hanhan na no.
oh that bread crumbs with half half COP
'Oh, that is half bread crumbs and half cheese.'

A: Go-syuzin, kyoo wa Ohiru.
PPX husband today TM lunch
'Is your husband going to be here for lunch today?
B: Ohiru wa ne, kyoo are ni Uta no, Montoree ni itta no.
lunch TM PRT today that to go PST. Monterey to go PST
'Lunch, you know, today, he went there. He went to
Monterey.'
Here, the use of no indicates that B is not making the remark abruptly, but
rather clarifying a situation evoked in the discourse context.
2.2. Koto
There are two usages of koto as an SFP: (a) to express exclamation (Martin
1975: 945) and (b) to give an order or direction. The former usage is
illustrated in (9) and (10):

(9) Ara kiree da koto.
oh pretty COP
'Oh, how pretty !'
(10) Maa yoku o-benkyoo-suru koto.

my! hard PPX study
'My! How hard you study!'
This use of koto is considered feminine (Martin ibid.). Sentences with koto
in this usage are thus often accompanied by a feminine interjection, such
as ara 'oh' or maa 'my/wow', as shown in (9) and (10). It is awkward to use
a nonfeminine interjection, such as o 'oh' or iyaa 'boy', as in (9a):
(9a) *O kiree da koto.
In the second usage, koto is used to give an order or direction, as
shown in the following examples:
(11) Oya o daizi ni sunt koto.

parent OM take good care of
'You must take good care of your parents.'
(12) <from Asahi Shinbun >

Toku ni ta no habatu ni tuite wa issai hurenai koto.
especially other factions about TM at all discuss NEG
Miyazawa-ha no zimusyo ni konna ohuregaki ga haridasareta.
faction GEN office in such order notice SM post PASS PST
'In particular, you must not discuss other factions at any time.
Such a notice of orders was posted in the office of the
Miyazawa faction.'
Compared to regular imperative sentences (e.g. (4b) and (11a)), sentences
with koto (e.g. (4) and (11)) are less direct as orders, in that the proposition
is presented as one's duty or a rule:
224 SHIGEKO OKAMOTO
(4b) Saku no naka ni hairu na.

IMP
'Do not go inside the fence.'
(11a) Oya o daizi ni si nasal

IMP
'Take good care of your parents.'
When stating one's duty or a rule is not relevant, as in (13), where an
immediate command is required, the use of koto is inappropriate, as shown
in (13a):
(13) Nee, oki nasai Moo hati-zi yo.
hey wake up IMP already eight o'clock SFP
'Hey, wake up. It's already eight o'clock.'
(13a) *Nee, okiru koto. Moo hati-zi yo.

'Hey, you must wake up. It's already eight o'clock.'
It is probably because of its function of rule-presentation that koto in the
second usage is often employed in writing.
2.3. To
To as an SFP has three usages, all of which are restricted to informal

speech: (a) 'to express defiance, (b) to make a self-affirmation, and (c) to
make a declaration in a casual manner (cf. Martin 1975: 1014). The first
usage is illustrated in (5):
(5) Nan da to. Tumaranai da to.
what COP boring COP
'What?! Boring?!'
Expressions such as those in (5) are considered blunt or vulgar; they are
used mainly by men in response to a remark made by the addressee. The
speaker challenges the addressee by repeating the addressee's remark back
to him. The particle to reflects this quotative nature of the utterance.
In the second usage, to is used to make a self-affirmation, as if the
speaker were talking to himself or herself:
(14) Are wa moo yatta to.

that TM already do PST
'Let me see, that, I've already finished.'
(15) <from Croissant, a women's magazine> (The speaker is at a

department store.)
Itiban takai mono kara zyun ni attate ikkai
most expensive things from one by one check first floor
akusesarii wa saigo-ni sum to.
accessories TM lastly do
'Let me see, I will start with the most expensive things, and
look at the accessories on the first floor last.'
Here, the use of to indicates that the speaker's soliloquy in making a self-
affirmation is quoted. But to cannot be used for a simple quotation of any
soliloquy; for example, the utterance *Aa, tukaretato'Oh, (I)'m tired' is not
acceptable because it is a direct expression of the speaker's emotion, and
not a self-affirmation.
In the third usage, to is used for making a declaration in a casual
manner:
(16) Moo neyoo tto,
now go to bed will
'I'll go to bed now.'
(17) Boku siranai tto.

I know NEG
'I don't know.'
Here, the form of the particle is tto, not to. The use of tto in (16) and (17)
suggests that the speaker's thought or soliloquy is quoted as a declaration.
The nuance conveyed by tto becomes clearer if we compare it with that of
the particle yo. In (17), for example, the speaker simply declares a
proposition and is indifferent to what the addressee thinks about it, whereas
in (17a) the speaker is actively trying to make the addressee believe the
proposition:
(17a) Boku siranai yo.
'I don't know.'
Thus, a vocative particle such as nee 'hey' and an emphatic phrase such as
226 SHIGEKO OKAMOTO
hontoo ni 'really' can be used in (17a), but not in (17):

(17b) *Nee, boku siranai tto, Hontoo ni.
hey really
'Hey, I dont' know. Really.'
(17c) Nee, boku siranai yo. Hontoo ni

'Hey, I don't know. Really.'
2.4. Tte
Tte as an SFP has two usages: (a) to report a proposition expressed by

someone else (Martin 1975: 939), and (b) to express the speaker's
insistence on a proposition. The first usage is very common in speech:
(18) <from a TV drama>
Yama tte7 abunai n desu tte.
mountain dangerous AUX
'Mountains are dangerous, I hear.'

Kanozyo, kekkon enki sitai n da tte.
she marriage postpone want AUX
'She wants to postpone the marriage, I hear.'
(20) <from natural conversation> (A is talking about a friend of

her friend who is coming to America to study.)
A: Sugoi no yo. Okusan ga daigaku ni hairu n da tte.
amazing SFP wife SM university OM enter AUX
'It's amazing. The wife is going to the university, I hear.'
B: Go-syuzin wa.
husband TM
'What about the husband?'
A: Nihon ni oite kuru n da tte.
Japan in leave come AUX
7
Tte here is a contracted form of to yuu no wa - i.e. COMP (to) + say (yuu) +
pronoun (no) + TM (wa)), which translates into 'what is called (mountain).'
Kodomo hutari turete kuru n da tte, (laughter)

children two bring AUX
'She is going to leave him in Japan, I hear. She is going to bring
her two children, I hear.'
In these examples, tte indicates that the information in question is hearsay
and hence something for which the speaker is not directly responsible. This
usage of tte is very similar to the use of the evidential auxiliary soo da 'they
say/I hear':
(18a) Yama tte abunai soo desu.
AUX
In fact, all instances of tte in (18)-(20) could be replaced by soo da?
This usage of tte may also appear in interrogative sentences:
Daigakusee da tte.
college student COP
'You are a college student, I hear; is that true?'
(22) <from natural conversation> (A and B are talking about

hemorrhoids.)
A: Ti ga, ti ga ikanai.
blood SM blood SM circulate NEG
'The blood, the blood doesn't circulate.'
B: Nan da tte. Kawai soo.
what COP pitiful
'What? How pitiful!'
In the case of interrogative sentences, the information source may be the
addressee rather than a third party (e.g. (22)).
The second usage of tte as an SFP expresses the speaker's insistence:
Hontoo da tte. Ore puropoozu nanka sitenai tte.
true COP I propose such a thing do NEG
Though interesting, it is beyond the scope of the present study to investigate the
difference between tte and soo da.
228 SHIGEKO OKAMOTO
'It's true, really. I didn't propose marriage to her, really.'

A: le yo.
say/tell SFP
Tell me about it.'
B: Nan demo nai tte. Nan demo nai yo.

anything NEG anything NEG SFP
It's nothing, really. It's nothing.'
The nuance conveyed is that the speaker is repeating something he or she
has said earlier and still holds to be true. This usage thus often implies the
speaker's desperation or frustration in trying to persuade the addressee.
3. The status of no, koto, to, and tte as SFPs
In section 2, I have discussed the morphemes in question as SFPs. However,

as mentioned in section 1, one might argue that sentences ending with these
morphemes are elliptical expressions or the result of the main clause
deletion, and hence that these morphemes are not truly SFPs at all, but
rather are COMPs (e.g. (4), (4a)). Such analogies, however, ignore the
significant functional differences between these SFPs and the corresponding
COMPs. In this section, I compare these sentence-final morphemes and the
corresponding COMPs with regard to their morphosyntactic and semantic-
pragmatic properties, and argue for the status of the former as SFPs, serving
as modality markers.
3.1. No
As outlined in section 2.1., the SFP no is typically used to indicate that the
utterance is an explanatory comment about a certain situation in the
discourse context. One might argue that this use of no, as in (8a), is an
abbreviation of no da, as in (8b):
(8a) Montoree ni Uta no.
to go PST
'He went to Monterey.'
(8b) Montoree ni itta no da.

'He went to Monterey.'
If we examine the function of no da, however, we discover that the final
morpheme no is not a functional equivalent and therefore not an abbrevia
tion of no da.9
The expression no da is often analyzed as consisting of a COMP or
nominalizer {no) and a copula {da). However, as discussed by Teramura
(1984: 306-311), its syntactic and semantic-pragmatic properties suggest that
no da as a whole functions as a modal auxiliary verb, which indicates that
the utterance is an explanatory comment about a certain situation evoked
in the discourse context (cf. Maynard (1992), in which the no da construc
tion is characterized as a commentary predicate, functioning as a 'Discourse
Modality strategy'). The analysis of no da as a COMP and a copula assumes
that a sentence with no da has the syntactic structure X wa Y no da 'X is
that Y', which is analogous to the 'basic' topic-comment sentence structure
X wa Y da, as shown in (25):
(25) Boku wa gakusee da.
I TM student COP
T am a student.'
For example, (8b) above might be given an underlying structure such as
(8c):
(8c) Kare ga inai to yuu koto wa Montoree ni itta no da.

he SM be NEG COMP fact TM to go PST
'The fact that he is not there is (means) that he went to
Monterey.'
However, (8c) is not exactly parallel to (25), in that in (8c) the predicate
phrase does not provide an attribute or identity of the subject, as in (25);
rather it provides an interpretation or explanation of the situation expressed
by the topic NP. In other words, in (8c) no is not a mere nominalizer and
da does not have the same copulative function as in (25). Rather, no da as
a whole is taken as indicating the explanatory nature of the sentence.
o
The no da construction has been studied extensively. Maynard (1992) provides a
good review of previous studies on no da.
230 SHIGEKO OKAMOTO
Moreover, it is often the case that sentences with no da cannot be

assigned a topic NP representing a situation to be explained; for example,
Doo sita no da. 'What happened?' (*Kimi ga naite iru to yuu koto wa doo
sita no da. The fact that you are crying is (means) what happened?').
Further, in sentences like (8d), the topic NP does not represent the
situation to be explained by the predicate:
(8d) Kare wo Montoree ni Uta no da.
he TM
'He went to Monterey.5
That is, (8d) does not mean 'He is (means) that he went to Monterey'.
Consider also (25a):
(25a) Boku wa gakusee na no da.
I TM student COP
'I am a student.'
Here, the topic NP (boku wa) is linked to the predicate nominal (gakusee)
by the copula na, not by da in no da. The use of no da in (25a) indicates
that the proposition T am a student' is an explanation about some situation
evoked in the context (e.g. the fact that the speaker cannot work full time).
Thus, while it is true that the use of no da assumes a topic in the sense
that the utterance is a response to a situation in the context, such a
situation need not be coded syntactically as a topic NP (see also Teramura
1984: 309; Maynard 1992). The form no da suggests that the no da
construction is based on the structure X wa Y no da, in which two
situations X and Y are simply linked to each other. However, as discussed
above, the role of Y is construed in a specific way, suggesting that no da as
a whole is used, or has been renalyzed, as an explanatory modal auxiliary;
hence the relevant situation to be explained need not be coded as a topic
NP.
Note, however, that although no da is best regarded as a modal
auxiliary, its function seems to reflect the function of no as a nominalizer
and that of da as a copula. In explaining the cognitive basis of no da as a
commentary predicate, Maynard (1992) characterizes the function of no in
no da as objectification óf the event or state being described through
nominalization, and that of da as personalization of utterances through
speaker-foregrounding. That is, no in no da seems to reflect the function of
no as a nominalizer to the extent that the situation being described
becomes objectified so that it can be related, as one unit, to some other
unit or situation in the discourse context.10

It appears then that the function of no as an SFP is the same as that of
no da in that both function as an explanatory or commentary modality
marker. In this repect, no is definitely related to no da. However, no and
no da cannot be considered paraphrases. As mentioned above, Maynard
(1992: 587-591) explains that da in no da personalizes the utterance by
foregrounding the speakerhood. Accordingly, the use of no alone, or the
lack of da, makes the utterance "less blatant," "less emphatic" (ibid.: 598),
gentler, and less assertive, and it is probably because of this nature of da
that women and children use no by itself rather than no da more frequently
than men (ibid.: 598). Furthermore, da in no da, but not no by itself, may
inflect according to modal features such as tense (e.g. datta), negation (e.g
dewanai), and formality (e.g. desu). In addition, no may follow desu - a
formal variant of da - in a feminine speech (e.g., Are musuko desu no.
T h a t is my son.'); no here cannot be replaced by no desu (i.e. *desu no
desu). (Note also that desu is not a prenominal form, which suggests that
no in desu no is not a COMP.)
Thus, no and no da are not functionally equivalent; hence, the former
cannot be regarded as a simple abbreviation of the latter. In other words,
the patterns S no and S no da should be considered instances of related,
but different, grammatical constructions, in which no and no da serve the
same basic function, but bring about different pragmatic effects.11 It is
thus most appropriate to categorize the morpheme no in S no as an SFP
used to mark an explanatory or commentary modality with particular
stylistic nuances such as gentleness and feminimity.
3.2. Koto
3.2.1. Koto for exclamation

The morpheme koto for exclamation (e.g. (26)) might be analyzed as a
COMP, that is, as an abbreviation of an expression such as koto ni odoroita
'(I) was amazed that' (e.g. (26a)) or koto to ittara nai 'it is indescribable':
10
I assume that the role of no for objectification is relevant at the cognitive level. As
discussed, syntactically no da as a whole functions as an auxiliary verb.
Maynard (1992) treats no and no da as variants of the same commentary predicate.

232 SHIGEKO OKAMOTO
(26) Maa, o-niwa ga kiree da koto.

oh PPX garden SM pretty COP
'Oh, how pretty the garden is!'
(26a) O-niwa ga kiree na koto ni odoroita.

PPX garden SM pretty COP COMP be amazed PST
T was amazed that the garden is pretty.9
Such an analysis, however, cannot be maintained, as explained below.
Syntactically, koto as a COMP must take the prenominal form, not the
final form, of a so-called adjectival noun phrase (e.g. kiree na/*da in (26a)),
whereas the exclamatory koto takes the final form (e.g. kiree ?nalda in
(26)). The same distinction also applies to nouns phrases. This suggests
that the exclamatory koto is not a COMP. Furthermore, the final koto can
be preceded by the polite honorific desu (or masu in the case of a verb), as
in (27), whereas koto as a COMP cannot, as in (27a):
(27) Ara, suteki desu koto.
oh nice COP
'Oh, how nice!'
(27a) Suteki na /*desu koto to ittara nai.

COP COMP indescribable
Tt is indescribable how nice it is.'
The honorific desu, for the addressee of the utterance, is normally used
sentence-finally; it cannot occur in a complement clause. This indicates that
koto in a sentence like (27) is not a COMP, but an SFP.
Semantic and pragmatic evidence also supports the treatment of the
Exclamatory koto as a true SFP. Below I will examine two issues: Paraphra-
sability and inferability of meaning. As mentioned above, one might
'paraphrase' koto in (26) as koto ni odoroita, koto to ittara nai, etc.
However, all these 'paraphrases' are more descriptive and explanatory, and
lack the spontaneity of real exclamations. Accordingly, when the speaker
suddenly notices the situation in question, the use of the SFP koto (e.g. (26),
not (26a)) is appropriate (Martin 1975: 945); and an interjection such as
ara 'oh' can cooccur with the exclamatory koto, but not with a descriptive
phrase. Moreover, stylistically, utterances ending with koto are definitely
feminine (see section 2.2.), but the corresponding descriptive expressions
are neutral Thus, the exclamatory koto and its expansions cannot be
considered paraphrases.
With regard to the issue of inferability, ellipsis is, in general, used when
the meaning is assumed to be inferable from the context without an explicit
mention. For example, in (28), the elliptical utterance is based on the
inferability of the intended meaning:
(28) Kiyoki ippyoo o.
honest one vote OM
'(Cast) an honest vote.'
In the case of the exclamatory koto, however, the criterion of inferability
does not seem to be met. For example, a beginning student of Japanese,
who knows only the 'basic' use of koto as a COMP, would have difficulty
interpreting the utterance O-niwa ga utukusii koto 'How pretty the garden
is!', particularly in light of the fact that koto occurs without a case marker.
Even in an appropriate context, a listener might wonder about the presence
of koto or be left with many possibilities of interpretation: e.g., koto ga
wakatta '(I) realized that', koto o kiita '(I) heard that'.
Based on these observations, we can conclude that koto in question is
not a COMP, but an SFP used to express exclamation in a feminine manner.
In other words, the exclamatory S koto is not an elliptical expression.
Rather, it should be considered a grammatical construction used as a main
clause, in which koto serves as a specific modality marker. (Note, however,
that the SFP koto reflects to some extent the function of the corresponding
COMP, or nominalizer, in that the situaion being described is an object of
amazement (see section 4).)
3.2.2. Koto for giving an order or direction

In this usage, koto can be used only in a sentence that expresses a situation
controllable by the addressee, as shown in the following examples:
(29) Oobosya wa daigakusotu de aru koto.

applicant TM college graduate COP
'The applicants must be college graduates.'
234 SHIGEKO OKAMOTO
(30) *Hana ga hayaku saku koto.

flower SM quickly bloom
The flowers must bloom quickly.'
(* indicates that the utterance is unacceptable as a directive/
order.)
Furthermore, the verb cannot be in the so-called potential form (-(rar)e)}2
For example, (4c) does not have a directive force:
(4c) *Saku no naka ni hair - e - nai koto.
fence GEN inside in enter can NEG
'(lit) that you cannot go inside the fence.'
It can only be a fragmental utterance or a complement in a sentence like
(4d):
(4d) Saku no naka ni hair-e-nai koto o siraseru.
OM notify
'I notify you that you cannot go inside the fence.'
Similarly, (11a), in which koto is preceded by the auxiliary nakereba naranai
'must', is unacceptable as a directive utterance:
(11a) *Oya o daizi ni sinakereba naranai koto.
parent OM take good care of must
'(lit) that you must take good care of your parents'
These examples show that a sense of duty or rule is associated with koto,
and that koto imparts the sentence a directive force.
One might argue that (4) and (11), for example, could be paraphrased
as (4a) and (11b), respectively:
(4a) Saku no naka ni hairanai koto o meeziru.
COMP OM order
'I order you not go inside the fence.'
(11b) Oya o daizi ni sum koto o moositukeru.

COMP OM order
'I order you to take good care of your parents.'
I owe this observation to Charles Fillmore.

However, (4a) and (11b) are explicit performative sentences; they are
authoritative, hence stronger orders than (4) and (11). These performative
sentences are formal and not commonly used, whereas directives such as
(4) and (11) are used in ordinary situations. The directive koto might also
be thought equivalent to the auxiliary koto da.13 However, koto da is used
to make a suggestion or recommendation. Thus, in (31), for example, the
use of koto, not koto da, is appropriate, because the teacher is giving an
order, not making a suggestion:
(31) (a teacher to students in class)
a. Syukudai wa asita dasu koto,

homework TM tomorrow turn in
'You must turn in your homework tomorrow.'
b. *Syukudai wa asita dasu koto da.

'It is important to turn in your homework tomorrow.'
These examples show that the SFP koto cannot be appropriately para
phrased by expansion. The problem of inferability is also relevant here. If
an interpreter knew only the 'basic' use of koto as a COMP, he or she might
be puzzled by the presence of koto in (31a) or might come up with
incorrect interpretations: e.g. koto ga dekiru '(you) can', koto ga daizi da 'it
is important that'.
From the foregoing discussion, we can conclude that the directive koto
is not a COMP, but an SFP used to give an order or direction in a more
indirect and informal manner than an exclit performative expression. In
other words, the sentence-type with the directive koto is not an elliptical
expression. Rather, it should be regarded as an independent grammatical
construction used as a main clause, in which koto functions as a modality
The phrase koto da (e.g. (31b)) is sometimes analyzed as consisting of a COMP

(koto) and a COP (da). However, its syntactic and semantic properties suggest that it be
treated as a modal auxiliary verb (see Teramura (1984)). Note also that the usage of koto
as a COMP has its source in the noun koto, which means 'thing/matter/fact' (e.g., Koto ga
zyuudai da. 'The matter is important.'). Although it is beyond the scope of the present
study, the morpheme koto provides many interesting instances of grammaticization.
236 SHIGEKO OKAMOTO
marker. 14 (Note, however, that the directive koto reflects to some extent
the function of the corresponding COMP, or nominalizer, in that the
proposition is a rule/duty or an object of abidance (see section 4).)
3 3 . To
3.3.1. To for defiance

The SFP to for defiance is not interchangeable with tte, as indicated in (5a):
the substitution of the particle tte would strip the utterance of its masculine,
crude tone and turn it into a neutral quotative question (or statement) (see
also example (22) in 2.4.). In contrast, to as a COMP can be replaced by tte,
though with a difference in formality (e.g. (5b)):
(5a) Nan da to/*tte. Tumaranai da to/*tte.
what COP boring COP
'What?! Boring?!'
(NB: Tte is possible in (5a) but lacks the masculine, defiant
force that to has. See below.)
(5b) Nan da to/tte itta. Tumaranai to/tte itta.

COMP say PST COMP say PST
'What did you say? Did you say it's boring?'
Furthermore, the SFP to for defiance, but not the COMP to, must be
preceded by da, as shown in (5c) and (5d):
(5c) *Nan to. * Tumaranai to.
'What?! Boring?!'
(5d) Nan to itta. Tumaranai to itta.

'What did you say? Did you say it's boring?'
The morpheme da is normally used as a copula for a noun (e.g., Kamera
da. '(It)'s a camera.') or for a so-called na-adjective or adjectival noun
(e.g., Kiree da. '(It)'s pretty.'); it cannot be used for an i-adjective (e.g.,
*Tumaranai da. '(It)'s boring.') or a verb (e.g., *Wa$ureta da. '(I) forgot.').
14
Interestingly, Korean kes has a directive usage similar to koto. Ransom (1988: 373)
points out that the COMP kes is used interchangeably with the imperative mood marker,
and it could evolve into a mood marker.
However, to for defiance must be preceded by da, even when it involves an

/-adjective (e.g. the second sentence in (5a)) or a verb (e.g., Wasureta da
to. '(Did you say you) forgot?').15
From the semantic and pagmatic point of view, also, to for defiance
cannot be regarded as a COMP. For example, the sentences in (5a) express
the speaker's defiance in a blunt and crude manner; they cannot be used
as neutral questions. By contrast, those in (5d) are most likely to be neutral
questions. Thus, for example, (5a) can be introduced by a masculine
vocative particle oi 'hey', but not by a neutral nee 'hey' (i.e., Oi/*Nee, nan
da to.); (5d) can be introduced by either particle (i.e., Oi/Nee, nan to itta.).
The problem of inferability is also relevant here: A listener who is unaware
of the usage of to for defiance might interpret the utterance Nan da to, for
example, as a neutral question (e.g., 'What did you say?).
Based on these observations, we can assume that the particular nuance
of defiance is conventionally associated with the final to (or perhaps, with
da to as a whole). That is, the final to is not a COMP, but an SFP used to
express defiance in a masculine, crude manner. In other words, the defiant
S to is not an elliptical expression. Rather, it should be considered a
grammatical construction used as a main clause, in which to serves as a
specific modality marker.
3.3.2. To for self-affirmation

The SFP to for self-affirmation cannot be replaced by tte (e.g. (14a)), but to
as a COMP can, except for a difference in formality (e.g. (14b)):
(14a) Are wa moo yatta to/*tte.
'Let me see, that, I've already finished.'
(14b) Are wa moo yatta toltte itta.

COMP say PST
'I said: That, I've already finished.'
It is also the case that the SFP to, as in (14), cannot be considered an
abbreviation of a 'full' expression such as to omou, as in (14c):
15
It is unclear to me why da is required in the use of to for defiance.
238 SHIGEKO OKAMOTO
(14) Are wa moo yatta to.

'Let me see, that, I've already finished/
(14c) Are wa moo yatta to omou.

COMP think
'I think that I have already finished that.'
As mentioned earlier, (14) is used for soliloquy or thinking aloud in making
a self-affirmation. (14c), on the other hand, is normally addressed to
someone else, and the truth of the proposition is qualified by the verb
omou 'think'. Thus, for example, (14c), but not (14), can be introduced by
a vocative particle such as nee 'hey', and can end with the particle kedo
'but', indicating hesitation. One might also 'paraphrase' to in (14) by other
expressions such as to ieru 'can say' and to minaseru 'can assume'. However,
they each pose the same paraphrasability problem as to omou 'think'.
Moreover, it seems difficult to interpret sentences such as (14) appropriate
ly unless the speaker knows the usage of to for self-affirmation.
Based on these observations, we can conclude that to in question is not
a COMP, but an SFP used to make a self-affirmation in an informal soliloquy
style. In other words, S to for self-affirmation is not an elliptical expression.
Rather, it should be regarded as an independent grammatical construction
used as a main clause, in which to serves as a modality marker.
33.3. To for casusal declaration

As mentioned, in this usage of to, the form is tto, not to or tte (e.g. (16a)).
The form of the corresponding COMP, however, is not tto, but to or tte (e.g.
(16b)):
(16a) Moo neyoo ttol*tol*tte

now go to bed will
'I'll go to bed now.'
(16b) Moo neyoo *tto/to/tte itta.

COMP say PST
T said that I would go to bed now.'
It is also the case that the SFP tto, as in (16a), cannot be regarded as an
abbreviation of a 'full' expression such as to omou, as in (16c):
(16c) Moo neyoo to omou.
COMP think
'I think that I will go to bed now.'
As mentioned earlier, (16a) is a declaration of the speaker's thought and
the speaker is not interested in what the addressee thinks about it. (16c),
on the other hand, is a qualified statement. Thus, for example, (16c), but
not (16), can be introduced by a vocative particle such as nee 'hey', and
followed by a question such as Ii. 'Is it all right?' Further, the inferability of
the correct meaning is also problematic here, as in the case of the other
SFPs discussed earlier.
These observations suggest that to in question is not a COMP, but an
SFP used to make a declaration in a casual soliloquy style. In other words,
S to for declaration is not an elliptical expression. Rather, it should be
considered as an independent grammatical construction used as a main
clause, in which to serves as a modality marker.
3.4. Tte
3.4.1. Tte for reporting

The SFP tte for reporting is not interchangeable with to (e.g. (32)), whereas
tte as a COMP is, except for stylistic nuance (e.g. (32a)):
(32) Niti-yoobi no ni-zi goro da tte/*to.
Sunday GEN two o'clock around COP
Tt's is around two o'clock on Sunday, I hear.'
(32a) <from natural conversation>

Niti-yoobi no ni-zi goro da ttelto itte ta.
COMP say PRG PST
'She was saying that it is around two o'clock on Sunday.'
Syntactically, the COMP tte cannot be followed by an SFP (e.g. (19a)),
while the reportive tte may be followed by another SFP (e.g. sa and yo in
(19b)):
(19a) *Kanozyoy kekkon enki sitai n da tte salvo itte ta.
she marriage postpone want AUX COMP SFP say PRG PST
'She was saying that she wants to postpone the marriage.'
240 SHIGEKO OKAMOTO
(19b) Kanozyo, kekkon enki sitai n da tte sa/yo.

Examples (18) and (18b) further illustrate the difference between the
SFP tte and the COMP tte:
Yama tte abunai n desu tte.
mountain dangerous AUX
(18b) Yama tte abunai n desu tte iimasita.

COMP say PST
T said mountains are dangerous.'
In (18), the polite honorific desu refers to the addressee of the utterance,
while in (18b), desu refers to the addressee of the quoted remark. If the
addressee of (18) is a close friend of the speaker, a plain form da is
normally used, as shown in (18c):
(18c) Yama tte abunai n da tte.
That is, desu in (18) is not part of the quoted remark, hence tte is not a
simple quotative COMP. In contrast, (18b) is a direct quotation that, as such,
includes desu regardless of the addressee of the top-level utterance; thus tte
is a COMP for the quotation.
The SFP tte may be paraphrased with some form of the verb yuu 'say',
as in (19c) for (19):
Kanozyo, kekkon enki sitai n da tte.
she marriage postpone want AUX
(19c) Kanozyo, kekkon enki sitai n da tte itte ta.

COMP say PRG PST
'She was saying: She wants to postpone the marriage.'
However, (19) and (19c) differ in that in (19c) the information source is
specified as kanozyo 'she', while in (19) it is unspecified and could be
kanozyo or someone else. In other words, when tte is used as an SFP the
information source and the act of saying or hearing are backgrounded. This
point is particularly well illustrated in an earlier example (20), in which
speaker A uses the SFP tte consecutively in three utterances. As pointed out
in section 2.4., the function of the reportive SFP tte is very similar to that of
the evidential auxiliary soo da T hear/they say'.
Analyzing the reportive tte as a COMP may also raise the question of
inferability. If the addressee did not know the convention, (19), for example,
could be interpreted incorrectly: e.g. tte omou 'I think that'. Note also that
tte in this usage has a distinct sentence-final intonation rather than a
lingering intonation typical of an elliptical utterance. The foregoing
arguments show that tte in question is not a COMP, but an SFP used as a
reportive evidential marker . In other words, S tte for reporting should be
regarded as a grammatical construction used as a main clause, in which tte
serves as a modality marker.
3.4.2. Tte for insistence

As in the case of the reportive tte, the SFP tte for insistence is not inter
changeable with to (e.g. (23a)), but tte as a COMP is (e.g. (23b)) except for
sylistic nuance:
(23a) <from a TV drama>

Hontoo da tte/*to. Ore puropoozu nanka sitenai ttel*to.
true COP I propose such a thing do NEG
'It's true, really. I didn't propose marriage to her, really.'
(23b) Hontoo da ttelto itte ru daro.

COMP say PRG TQ
'I am telling you that it's true, aren't I?'
Again, it might be argued that the first sentence in (23a), for example, is an
abbreviation of (23b), but the former is weaker than the latter in assertive
force. It is also questionable whether the sense of insistence is inferable if
the listener does not know the convention. For example, one might be
puzzled by the presence of tte in (23a) or interpret it inappropriately.
Thus, it is most appropriate to treat the final tte as an SFP for
expressing the speaker's insistence in an informal style. That is, S tte for
insistence is not an elliptical expression. Rather, it should be regarded as
a grammatical construction used as a main clause, in which tte serves as a
modality marker.
242 SHIGEKO OKAMOTO
4. Discussion and conclusion
As the preceding analysis indicates, it is most appropriate to regard the

final morphemes no, koto, to, and tte as SFPs rather than COMPs. The
sentence-types S no/koto/to/tte are not elliptical expressions, consisting of a
complement clause, a COMP, and an implicit main clause. Rather, they are
functionally specialized and can be regarded as independent grammatical
constructions used as main clauses, in which these SFPs serve as modality
markers.
As COMPs, no, koto, and to/tte serve two functions: To mark a
complement clause or a clause boundary, and to indicate the epistemic
status of the proposition in the complement clause. The first function is
syntactic as well as textual, and essential to these COMPs; the second
function is largely redundant in that the epistemic status is also indicated
by the main verb (in this respect, the COMPs no, koto, to, tte differ from
modal auxiliaries, even though they concern epistemicity).16 In the case of
no, koto, to, and tte as SFPs, on the other hand, their essential functions are
pragmatic in that they serve as modality markers in main clauses. In
particular, these SFPs indicate speech act types and the speaker's attitudes
toward the proposition: i.e. no for explanatory comment, koto for exclama
tion and ordering or directing, to for defiance, self-affirmation, and
declaration, and tte for reporting and insistence. These SFPs are also
expressive in that they convey various pragmatic effects or stylistic nuances,
such as femininity, masculinity, spontaneity, indirectness, informality, and
soliloquy.
To categorize these morphemes as SFPs, however, is not to imply that
these SFPs are totally unrelated to, or discrete from, the corresponding
COMPs. Although I have argued that the constructions S no/koto/to/tte and
their 'paraphrases' by 'full' forms are not semantically and pragmatically
equivalent, we cannot at the same time ignore the closeness in meaning
between the two kinds of utterances: e.g. Saku no naka ni hairanai koto.
'You must not go inside the fence.' vs. Saku no naka ni hairanai koto o
meeziru. 'I order you not to go inside the fence'. The comparison of these
The choice of a COMP is largely controlled by the verb, while the choice of a
modal auxiliary depends on the speaker's subjective judgment about the situation being
described.
two kinds of expressions suggests that the constructions S no/koto/toltte have

evolved from elliptical expressions consisting of complements, and that
these SFPs have their origin in the corresponding COMPs.17 As noted
previously, it is also the case that the functions of these SFPs reflect to some
extent the functions of the corresponding COMPs: no for objectification of
the situation being described, koto for presenting an object of amazement
or abidance, and to and tie for indicating the quotative aspect of the
utterance. It is thus logical to assume that the usages of these SFPs are
instances of reanalysis of morphological functions.
In particular, the reanalysis of these SFPs can be characterized as
involving pragmaticization or pragmatic strengthening. Traugott (1988,
1989) and Traugott and Konig (1991) claim that semantic change in
general, including the process of grammaticization, may involve pragmatic
strengthening, by which the expression becomes more informative and
expressive with increased speaker involvement. Further, they characterize
the process of pragmatic strengthening as a type of metonymy or semantic
transfer through contiguity, in particular, a shift from a conversational
implicature to a conventional one (ibid.). The usages of no, koto, to, and tte
as SFPs are not prototypical instances of grammaticization, but it provides
an interesting illustration of pragmatic strengthening through contiguity in
the semantics and pragmatics of expressions. That is, the usage of these
SFPs can be postulated to be the result of conventionalization of pragmatic
meanings: The elliptical expressions S no/koto/to/tte in certain uses have
come to be "grammaticized" as functionally specialized independent
constructions, and in this process certain implicatures - i.e. particular speech
act types and stylistic nuances - have come to be conventionally associated
with the final morphemes while the (original) 'complements' have become
main clauses.
The motivations for the reanalysis of these final morphemes are
twofold: (1) to bring about certain expressiveness, that is, to perform a
given speech act with particular stylistic nuances (e.g. indirectness,
femininity, masculinity, informality, soliloquy) and (2) to foreground the
information in the (original) 'complement'. The first point may be related
to Martin's explanation that Japanese hate to end a sentence on a tone of
finality, and that this attitude has encouraged the development of sentence-
17
Note that in the case of the SFP no, it is related to the COMP via the auxiliary no
da (see section 3.1.).
244 SHIGEKO OKAMOTO
final particles in general (as well as other linguistic devices): A final particle
"imparts some additional hint of the speaker's attitude toward what he is
saying" (Martin 1975: 914) without explicitly articulating it. Although the
processes of reanalysis are quite different, the shift of S no/koto/to/tte from
complements to main clause constructions may be also compared to the
functional reanalysis of clauses involved in the grammaticization of
epistemic parentheticals in English (e.g. I think) discussed by Thompson
and Mulac (1991) - the process in which the head element becomes an
epistemic phrase or a dependent element, while the 'complement' becomes
the main element.
In this study, I have demonstrated how the constructional view of
grammar enables us to consider subtle but significant differences in the
semantic-pragmatic properties of certain formally similar linguistic
expressions as well as the contiguous nature of semantics and pragmatics
that underlies the usages of these expressions.
References
Cook, Haruko M.
1990 "An indexical account of the Japanese sentence-final particle no." Discourse
Processes 13.401-439.
Du Bois, John W.
1985 "Competing motivations." In J. Haiman (ed.), Iconicity in syntax. Amsterdam:
John Benjamins Publishing Company, 343-365.
1985 "Syntactic intrusions and the notion of grammatical construction." BLS
11.73-86.
1988 "The mechanisms of "construction grammar."" BLS 14.35-55.
Fillmore, Charles J., Paul Kay, & Mary C. O'Connor

alone" Language 64.501-538.
Heine, Bernd, Ulrike Claudi, & Friederike Hunnemeyer

1991 Grammaticalization: A conceptualframework.Chicago: University of Chicago
Press.
Hopper, Paul
1987 "Emergent grammar." BLS 13.139-157.
Josephs, Lewis S.
1976 "Complementation." In M. Shibatani (ed.), Syntax and semantics 5: Japanese
generative grammar. New York: Academic Press, 307-369.
Kuno, Susumu
1973 The structure of the Japanese language. Cambridge: MIT Press.
Lakoff, George
1987 Women, fire, and dangerous things. Chicago: University of Chicago Press.
Lambrecht, Knud
1984 "Formulaicity, frame semantics, and pragmatics in German binominal
expressions." Language 60.753-796.
1988 "Presentational cleft constructions in spoken French." In J. Haiman & S.

Thompson (eds.), Clause combining in grammar and discourse. Amsterdam:
John Benjamins Publishing Company, 135-179.
Martin, Samuel E.
1975 A reference grammar of Japanese. New Haven: Yale University Press.
Matsumoto, Yo
1988 "From bound grammatical markers to free discourse markers: History of
some Japanese connectives." BLS 14.340-351.
Maynard, Senko K.
1984 "Functions of to and koto-o in speech and thought representation in
Japanese written discourse." Lingua 64.1-24.
1992 "Cognitive and pragmatic messages of a syntactic choice: The case of the
Japanese commentary predicate n(o) da." Text 12.563-613.
McCawley, Noriko A.
1976 "Another look at no, koto, and to: Epistemology and complementizer choice
in Japanese." In J. Hinds & I. Howard (eds), Problems in Japanese syntax
and semantics. Tokyo: Kaitakusha, 178-212.
Okamoto, Shigeko
1993 "Nominal repetitive constructions in Japanese: The 'tautology' controversy
revisited." Journal of Pragmatics 20.433-466.
246 SHIGEKO OKAMOTO
Okamoto, Shigeko
1994 "Augmentative verbal repetitive constructions in Japanese."
Cognitive Linguistics 5.381-404.
Ransom, Evelyn N.
1988 "The grammaticalization of complementizers." BLS 14364-373.
Teramura, Hideo
1984 Nihongo no shintakusu to imi II. (Japanese syntax and semantics II.) Tokyo:
Kuroshio Shuppan.
Thompson, Sandra A. & Anthony Mulac

1991 "A quantitative perspective on the grammaticization of epistemic parenthet-
icals in English." In E. Traugott & B. Heine (eds.), Approaches to
grammaticalization (II). Amsterdam: John Benjamins, 313-329.
Traugott, Elizabeth C.
1988 "Pragmatic strengthening and grammaticalization." BLS 14.406-416.
1989 "On the rise of epistemic meanings in English: An example of subjectifi-

cation in semantic change." Language 65.31-55.
Traugott, Elizabeth C. & Ekkehard König

1991 "The semantics-pragmatics of grammaticalization revisited." In E. Traugott
& B. Heine (ed.), Approaches to grammaticalization (I). Amsterdam: John
Benjamins, 189-218.
Recasting the deictic foundation,
using physics and Finnish
Jan-Ola Östman
University of Helsinki
0. Preamble
One of the most characteristic features in the works of Charles Fillmore is

surely his ability to (re)open up fascinating new areas of research in the
Fillmorean manner that immediately awakes everybody's interest. This
holds true for his insights that led to the development of Case Grammar in
the 1960's as well as for his present-day interest in idiomaticity and the
intricacies of grammatical contructions.
In the early 1970s Fillmore brought deixis into linguistics; not in the
sense that nobody had noticed this area of research earlier, but his
approach was more systematic, and more linguistic than previous ones. A
huge number of articles and books have appeared as a direct or indirect
consequence of his Santa Cruz lectures on deixis in the summer of 1971.
As in his other work, so too in his research on deictic phenomena,
Fillmore has always been interested in how non-western languages work.
Thus, in this paper I want to add some Finnish examples to the expanding
pool of data on deixis. Even though I will suggest some quite drastic
changes of direction for future research in the field of deixis and
demonstration, these suggestions could not have been made without the
pioneering work of Charles Fillmore.1
1
My first serious encounter with deixis came a long time ago in Charles Fillmore's
lectures on semantics at UC-Berkeley. In fact, the present paper has grown out of an
essay I once wrote for Chuck, and which he - faithful to his habit - commented on
extensively. In addition, I would like to thank Erik Andersson and Farrell Ackerman for
248 JAN-OLA ÖSTMAN
1. Introduction
The paper attempts to travel some distance towards challenging the

prevalent view of seeing basic deictic distinctions as necessarily grounded
in some objective, locationally definable space. I want to suggest that other
parameters, usually referred to as 'metaphorical extensions' of primary
deictic distinctions, should not a priori be given a less important status than
location-in-a-physical-world,
I will set about my task via two routes that are going to be somewhat
intertwined in the paper: On a theoretical level I want to remind the
reader both of the relative status of concepts and cultures, and of the
fragility of the Newtonian, traditional-physics, way of seeing and
interpreting the world around us. On a more empirical note, I will give a
brief account of the system of demonstrative adverbs in Finnish — focussing
on non-locational aspects.
2. Data: Demonstrative adverbs in Finnish
I will start with a partial presentation of the subsystems that together

constitute the paradigm of demonstrative adverbs in Finnish. There are
three morphologically encoded distinctions which have to be taken into
account in any characterization of the system. These will be dealt with in
turn in the following.
2.1. Location and direction
There is a three-way distinction according to whether the demonstrative

indicates a location - i.e. 'at', 'In' - or a movement - 'from', 'out of, or
'to', 'into'. The distinction is marked with the local case suffixes: For
instance, the forms for 'here' ('at here'), '(to) here', and 'from here' are
given in table 1.
comments on very early versions, and the audience at the 1990 International Pragmatics
Conference in Barcelona for their response to a brief presentation of the major ideas in
this paper - I particularly remember that Jens Allwood, Dan Slobin and Auli Hakulinen
gave me very useful suggestions. I am also very grateful to Sandra Thompson for her
valuable comments on the penultimate version of this paper.
DEIXIS IN FINNISH 249
'at' täällä [Adessive]

'to' tänne ([Allative])
'from' täältä [Ablative]
Table 1.
Thus, if the object or activity referred to is seen as being located at, or in,
a deictic center in space and time, Adessive or Inessive case markings,
respectively, are used on the demonstrative adverb: täällä has the Adessive
ending -llA.2 If the object is seen as moving/having moved toward the
deictic center, the Illative or the suffix -nne (which functionally corresponds
to the Allative case) are used: tänne has the suffix -nne. And if the
movement is away from a deictic center, the case forms for Elative or
Ablative are used: -ltA in täältä marks the Ablative case.
This system is fairly straightforward, and although there are some very
interesting, even problematic, features of this subsystem, too, they are of a
fairly detailed nature, and not directly relevant to the purpose of this
paper.
2.2. Inner vs. Outer case marking
The second distinction is a two-way distinction, indicated by what in Finnish

grammar is referred to as Inner vs. Outer Local case marking (cf. e.g.
Hakulinen and Karlsson (1979); Karlsson (1982) talks about 'inside
location' and 'outside location'). The paradigm given in table 1 exemplifies
Outer Local case marking as related to the demonstrative adverb meaning
'here'. Below are the corresponding Inner Local case forms for 'here'.
'in' tässä [Inessive: -ssA]
'into' tähän [Illative: -(h)Vn, where V equals the stem-
final vowel]
'out of tästä [Elative: -stA]
Table 2.
Although there have been a great many discussions about the semantics of
2
Capital-letters are used in suffixes to indicate variations in accordance with the
rules for vowel harmony in Finnish; e.g. -A- stands for -a- or -ä- depending on the
characteristics of the stem.
250 JAN-OLA ÖSTMAN
Inner vs. Outer case marking in Finnish in general (for a recent, cognitive
approach, ef. Leino (1989)), the distinction has received hardly any
systematic attention with respect to its manifestation in the deictic system.
One fairly typical, almost intuitive, kind of suggestion is represented by
Lehtinen's (1967) sugggestion that the size of the area referred to is
crucial: If the area referred to is relatively small, as in (1),
(1) Tässä on makuuhuone.
here-INESSIVE is bedroom
'Here's the bedroom'
Inner cases are used, whereas Outer cases are used if the area is of a
relatively larger size; cf. (2).
(2) Sataako täällä paljon?
rain-QUESTION here-ADESSIVE much
'Does it rain a lot here ( = in this area) V
This explanation is not, however, uncontroversial. In fact, I would
rather support the view that the Inner vs. Outer distinction is directly tied
to the role of Local case marking in general in Finnish. Inner cases refer to
a Bounded location of some sort; in the terminology of Lakoff and Johnson
(1980), the space with respect to which something is metaphorically viewed
as a container. The Outer, cases, in contrast, are used in reference to
Unbounded areas. These are issues on which Talmy in particular has done
invaluable research (cf. e.g. Talmy 1978).
The behavior of the postposition päin is particularly revealing for a
proper interpretation of the function of the Inner-Outer distinction. Päin is
a lexical hedge (G. Lakoff 1972) which refers to a very unspecified area,
something like 'hereabouts' when it is added to a demonstrative which
indicates proximity:
(3) Hän asuu täälläpäin.
He/She lives at-here-PÄIN
'He (or She) lives around here somewhere.'
This postposition can only be added to locations expressed with Outer case
marking. It is not, for instance, possible to say (4), with the demonstrative
adverb in the Inessive case.
(4) *Hän asuu tässäpäin.
If we know the specific location where somebody lives, it is feasible to view
that location as a Bounded area. However, non-specificity (i.e. speaker's

indeterminacy, or lack of knowledge) of the precise location would seem in
principle to imply an Unbounded, or at least a less Bounded, area. Thus,
only the latter, the Outer-case marked, but not expressions involving Inner
case marking, may occur together with the vagueness postposition päin.3
The Bounded-Unbounded distinction seems to be very basic in Finnish.
Thus, some words have inherent Boundedness properties, e.g.4
(5) maaseudulla 'in the country(side)' [Outer: Adessive]
*maaseudussa
(6) puutarhassa in the garden' [Inner: Inessive]

*puutarhalla
Similarly, the use of different case markings can differentiate between
homonymie or polysemic interpretations of a word, since the different
homonyms can have different boundedness properties:
(7) metsällä '(to be) working/hunting in an
unspecified forest location' [Outer: Adessive]
metsässä 'in the (specific) forest' [Inner: Inessive]
(8) maalla 'in the country(side)' [Adessive]

maassa 'in the ground/earth/nation' [Inessive]
(9) maailmassa 'in the world' (= on planet Earth);

lit. 'in the earth-air' [Inessive]
maailmalla '(here and there) in the world'
(where 'world' does not necessarily
refer to planet Earth) [Adessive]
3
As another case, consider the function of the indefinite pronoun jossain
'somewhere'. When used with an Outer-case demonstrative, as in Täällä (ulkona) jossain,
'Here (outside) somewhere', it indicates an unspecified, Unbounded area in general;
whereas when it occurs with a demonstrative adverb in an Inner case, the unspecified
target area is seen as located within some Bounded area: Tässä (laatikossa) jossain, 'Here
(inside this box) somewhere'.
4
Cf. also Fillmore's (1975: 18) reference to Robert Austerlitz.
252 JAN-OLA ÖSTMAN
The distinction can be nicely illustrated with the following examples:5

(10) Hän lähti maailmalle —
'He went out in the world - '
— hän kävi jopa kuussa.
'- he even visited the moon/
(11) Maailmassa asuu 500 miljoonaa ihmistä,

There are 500 million people in the world/
*josta vain 5 kuussa.
'out of which only 5 live on the moon.'
It is obvious that Boundedness as I have used it here, and as I have
applied it to Local expressions, is not Boundedness in any objective sense.
Boundedness is an interpretation, a filtered perception by the speakers
which has gradually become codified by the linguistic community.
Unexpectedly perhaps, it seems in addition that the Bounded.
Unbounded distinction in Finnish6 is not primarily to be seen as having its
deictic center with the speaker, so that the speaker him/herself necessarily
views the object referred to as being Bounded or Unbounded. Rather, the
distinction is made in accordance with the addressee's orientation and
perspective. This can be illustrated with the following set of examples:
5
There are, of course, also other factors influencing the particular lexicalized form
a language codifies. As Fillmore (1977) showed very early on, and returned to even more
in detail in his discussions on Construction Grammar (1988; cf. also Fillmore, Kay and
O'Connor 1988), lexical items very often have constructional properties, in particular, they
have to be seen in relation to their opposite or converse terms. Cf. on land vs. at sea, but
on the ground vs. in the air.
Thus, in Finnish, the reason why we sometimes say lahdessa and sometimes lahdella
('in the bay') is only partly due to the Boundedness properties of the bay. In particular,
the expressions partake as terms in the following paradigms:
(i) lahdella vs. rannalla vs. merellä
in-the-bay on-the-shore at-sea
vs.
(ii) lahdessa vs. meressä vs. ilmassa
in-the-bay in-the-sea in-the-air
(In addition, there is the town Lahti, which takes the Inessive form Lahdessa, 'in Lahti',
rather than Lahdella.)
I am here restricting my comments to apply only to what the evidence from the
function of demonstrative adverbs indicates.
(12) a. Tässä on makuuhuone.

in-here is bedroom
b. Täällä on makuuhuone.
at-here is bedroom
'Here is the bedroom.'
(13) a. Tässä on mehua.

in-here is juice
b. Täällä on mehua.
at-here is juice
'Here's juice.'
If I say (12a), I either have to be inside the bedroom together with the
addressee, or I might point out the bedroom to him/her on a map. In
either case, we can say that the bedroom is Bounded - both for the speaker
and for the addressee. However, if I say (12b), I am more likely to be in,
or close to, the bedroom myself, shouting (12b) to the addressee, who
might (as a guest or as a potential buyer) be in(specting) some other room.
For the speaker, the room is specific and Bounded, but not necessarily for
the addressee - except, of course in the general sense that the definition of
a room in the western culture usually involves something bounded by walls,
a floor and a ceiling. Similarly, (13b) would be the thing Mommy shouts
from the kitchen - not focussing on any container-bound juice, whereas
(13a) is what we can expect her to say as she places the glass in front of
the addressee-child. Thus, in (13b) täällä is used by the speaker even if she
can experience the juice as Bounded. Similarly, if you are picking
mushrooms in a forest, and your companion shouts a 'where-are-you', the
response you give is Täällä. [Outer], rather than Tässä. [Inner].
The traditional explanation that expressions with the Outer cases refer
to relatively large areas, while Inner cases refer to relatively small areas, is
— as we can see from the examples in (12) and (13) above — definitely
wrong.7 The distinction in terms of size is not even a necessary corollary
of the Boundedness distinction, although admittedly, something Bounded
7
Lehtinen (1967), for instance, takes the large area vs. small area distinction to be
the basic distinction. However, she herself admits that it does not apply to the distinction
between siellä and siinä. The suggested distinction in terms of Boundedness, however,
covers all three terms in what I will shortly deal with under the heading of the Proximity
system.
254 JAN-OLA OSTMAN
tends more easily to be interpreted as smaller in size than something

Unbounded. A further factor often reverted to in explaining the use of
demonstratives is whether the object referred to can be seen or not; this
parameter might indeed also be able to offer some explanation for what I
have here called the Boundedness distinction. However, from a theoretical
point of view I am more inclined to see the Visibility factor as a (common-
sense rather than rational8) consequence of the Boundedness distinction.
The Outer cases will typically designate the possibility of non-visibility of the
designated object, and the Inner cases will be typically associated with
visible, pointed-to objects. Tule tähän! 'Come here' [Inner] would be
accompanied by a finger gesture, whereas Tule tänne! 'Come here' [Outer]
is what you would shout if you went out into a black forest, looking for
your run-away dog.9
It is of course arguable whether the Boundedness parameter a priori
is the most basic one. Other possibilities closely related to the Boundedness
distinction also suggest themselves as primary, in particular, those of
Specificity, and Givenness. I have already used the Specific-Nonspecific
distinction as a secondary attribute when characterizing the distinction
between täällä [Adessive] and tässä [Inessive]: täällä indicates that the area
referred to is Unspecific (from the point of view of the addressee, we have
to add; cf. example (12b)), and tässä indicates that the area is more
Specific - and known. And this conveniently brings us to the Givenness
parameter. We could argue that tässä in (12a) functions as the theme or
topic of the sentence, whereas täällä in (12b) functions as the rheme, the
8
Cf. Östman (1992).
9
Another distinction that has been suggested (cf. e.g. Lehtinen 1967) to explain the
semantic difference between the use of Inner and Outer cases is in terms of
circumlocutions like 'over here, hereabouts' for Outer cases, and 'right here' for Inner
case markings. I can to some extent go along with the 'right here' vs. 'hereabouts'
distinction, as should be obvious from my discussion above. However, the rendering 'over
here' creates additional problems.
In particular, I would say that 'over here' should be seen as a translation of the
semantics of the Inner, rather than of the Outer, case system. Talmy (Fillmore, p.c.) has
argued that whereas here and there designate areas with a possible border-line between
them, over here and over there rather indicate points with (a considerable amount of)
space between them ~ i.e. between the speaker/designated object and the addressee. Thus,
since the areas referred to by tässä, '(in) here', and tuossa, '(in) there', are Bounded, it is
easy to conceive of them as having space between them.
new information.
In order to decide among these three alternatives, extended text
analyses need to be made. But even without such in-depth analyses, I
suspect that it might not be possible to weigh the factors appropriately.
The Boundedness criterion is a grammar-internal factor — with possible
cognitive repercussions; the Givenness and Specificity criteria are more
pragmatic and interactionally determined. I suspect the criteria work
together (cf. also the discussion in sections 5 and 6 below).
I have preferred to single out Boundedness, because it is a
grammatical feature referred to extensively in Finnish grammar anyway. If
the distinction between Outer and Inner cases in Finnish in general is seen
as an aspectual distinction, where, say, the reason for the use of the
Inessive case in kyyneleissä 'all in tears', is due to the idea that the
designated state of affairs (that of being all in tears) is being characterized
as completed, perfected, an ultimate state, and fulfillment of such a state
of affairs, then the Boundedness feature can be seen merely as a particular
case of a more general aspectual distinction in terms of (im)perfectivity or
(in)completeness.
Whereas the Location/Direction parameter is fairly unproblematic, the
explication of the Inner-Outer, Boundedness parameter is not as
straightforward. Although the problems that arise might at first seem
specific to Finnish grammar, they have important implications both for
deixis and for the semantics of case marking in languages in general.
3. The traditional picture of deixis: Proximity and Distance
In addition to Location/Direction and Boundedness, Finnish also makes a

further distinction in its demonstrative-adverb system. This third parameter
is seemingly made up of items that are traditionally seen as constituting the
'basic' deictic distinction in languages: The distinction between Proximity on
the one hand, and various shades of Distance on the other. Discussions of
this aspect of deixis are explicitly or implicitly based on concepts that are
definable in terms of their scientifically defined correlates in physics, viz. in
terms of space and location, and in terms of time.
A simple distinction between Proximity, 'closeness (in space) to
speaker', and Distance, 'distance (in space) from speaker', is seen as the
primary one in languages that have only two terms in their system of
256 JAN-OLA ÖSTMAN
demonstrative adverbs (English here and there). A three-term distinction

(Proximal, Medial, Distal) is usually said to split up the 'distance from
speaker' into 'shorter physical distance from speaker = closeness to
addressee' and longer physical distance from speaker (and addressee)'
(English there vs. yonder).
The following general picture of how languages tend even further to
split up their deictic terms has been suggested by Fillmore (1980; cf. also
1975: 42-43; and 1982: 47ff.).10
Figure 1.
This suggestion is based wholly on physical characteristics:

Closeness/distance from speaker/addressee, long/short distance, and
visibility. This is a typical starting point for investigations into the semantics
of demonstrative adverbs cross-linguistically (cf. e.g. Weissenborn and Klein
(eds.) 1982). The general strategy in explaining more complicated deictic
systems which are not as readily understandable in terms of their physics
correlates, is to take the distinction based on physical space and location as
the primary one, and treat other aspects as derivatives or metaphorical
extensions of this primary distinction.
However, although different kinds of deixis are accepted as deictic
10
Key:
D demonstrative term expressing:
Dl 'closeness to speaker' [Sp. Proximal]
D2 'closeness to addressee' [Addr. Proximal]
Dx 'at a distance from speaker' [Medial/Distal]
Dsd 'short distance from speaker' [Medial]
Did 'long distance from speaker' [Distal]
D-v 'non-visible' [- Visible]
types, and given detailed treatment by scholars (cf. e.g. Levinson 1983, ch.
2), as we move from Place, Time, and Person deixis to Discourse and
Social, not to mention Affective deixis, we also gradually (and qualitatively)
move away from a prototypical deictic situation as defined in terms of
physics correlates. For instance, and in particular, the use of this vs. that
affectively, to express one's positive or negative attitude, respectively, as in
(14) and (15)
(14) This Bush isn't such a bad guy after all.
(15) That Bush really made a fool of himself again yesterday.
is often seen as a spatial metaphor, as a metaphorical extension of the
Proximity-Distance distinction in the demonstrative pronouns in English:
this is proximal, thus positive; that is distal, thus more negative. (Cf. further
R. Lakoff 1974; and Collinson 1937.)
My purpose is not to argue against a description in terms of
metaphorical extension with respect to the use of this and that (in 14-15)
in English; in fact, the suggestion seems highly plausible. Rather, I want to
argue that not all languages work in the same way, and that other types of
deixis — even Affective deixis — might have to be considered (cognitively)
'basic' (and not 'metaphorical extensions') in other languages, and thus in
language in general.
In order to show this convincingly, I need to make a brief digression of
a more theoretical nature.
4. A theoretical challenge
In this section I want to show how I think the direction of argumentation

that seems to be taken for granted in explanations and characterizations of
deictic systems in language might be theoretically challenged. I am not
suggesting that the 'traditional' view is all wrong. But a revaluation of the
most basic underlying suppositions about the essence of deixis is called for.
Such a revaluation will be necessary in order to complete our picture of,
and our understanding of, deictic phenomena in language — perhaps
especially in non-IE languages.
It is not uncommon to find that when a journalist attempts to give a
report in a newspaper of some issues and phenomena that have been
discussed at a linguistics conference, the readers are served both
258 JAN-OLA ÖSTMAN
overgeneralizations and misinterpretations. In fact, the same happens

across subfields within linguistics; unintentionally giving slightly misguided
representions of other scholars' views and arguments is an unfortunate
occupational disease. Similarly, those of us who keep up with developments
in other sciences are sure to have found instances of how linguistic findings
get misinterpreted and misused in these fields. Still, this state of affairs
should not warrant us to behave in a similar fashion in regard to other
disciplines, and neglect recent advances in those disciplines.
This point is both epistemologically and methodologically relevant for
our present discussion. A reasonable assumption is that one can hardly
reach a holistic understanding of the world unless, ideally, one keeps track
of advances' in as many different disciplines as possible; at least one would
hope that directions of research in one field should not be based on
inadequate knowledge of some other field. In particular, although there is
no direct one-to-one correspondence between what we know about the
world in terms of physics, and our conceptualization of the world, we
cannot overlook the fact that the status given to space in the traditional
picture of debris as represented in figure 1 has its basis in 'neo-traditional'
physics. This view does not take into account more recent findings in
physics, according to which place and time have a very artificial status.11
Thus, physics neither forces us nor licences us to take concrete physical
space for granted.
From the discussion above it does not follow that we can therefore, in
a Bloomfieldian fashion, forget about semantics and pragmatics until we
know where super strings, chaos and quantum theory in physics lead us.12
Still, if one aspect of linguistic adaptation and adaptability (cf. Verschueren
1987) is that languages adapt to circumstances in the world outside
language, to the 'objects of adaptation9, then it is obvious that research
11
In this connection, cf. R. Lakoff (1989: 966): "... the paradigm 'science' social
scientists like linguists are prone to take as a model is Newtonian physics, with its
dichotomies, objectivity, and certainty. But quantum physics has cast doubt on all these
vaunted desiderata, and we might ponder the dubious advantages of modeling our own
theory and method on those of an obsolescent field."
12
Still, for those of us who go for both cognitively and socially realistic descriptions
of phenomena in language, who in some way or other work on Yngve's (1986) assumption
that all things do ultimately hang together, a sane Bloomfieldian-kind-of scepticism is
often only healthy.
within pragmatics needs to take into account as much information as

possible about these circumstances - i.e. about recent advancements in
other sciences, and physics in particular, when we want to explain deictic
phenomena in terms of proximity and distance.
It is nevertheless obvious that in pragmatics the focus has to be placed
not on phenomena in any 'objective' world, but on humans and their
perceptions, categorizations and attitudes of that world more or less
irrespective of advancements in physics. But if this is the case, it is
reasonable to ask why we should worry about physics in the first place. The
point is that knowledge of and advancements in other disciplines still
matter both from a theoretical and from a methodological point of view,
and thus indirectly, also empirically. The simple insight that things might be
equally well looked at in a very different perspective is often enough. What
we need to realize, or remind ourselves of, is that place and time — as
conceptualizations - are only two abstract parameters; other parameters
may be just as important for human conceptualization, perception and
categorization, and thus just as codified or codifiable in the deictic system
of a language.13
5. A challenge from within language: Finnish 'Proximity'
The theoretical points brought forth in section 4 can be seen as constituting

a challenge of the traditional deictic foundation from the outside, from
outside the realm of linguistics. The second type of challenge is a
substantiation of the methodological one, a challenge from within language,
on the basis of data from one particular language, viz. Finnish.
The third subsystem in Finnish has three terms:
forms on
tä(ä)-
tuo- (in informal style some forms have toi- and others to- as
their stems)
si(e)- (the demonstrative pronoun form is se)
13
An aspect which will not be dealt with in this paper, but which can be deduced
from Östman (1992) would be the status of physics (as a science very closely connected
to mathematics) vis-à-vis linguistics, which has at least in some quarters attempted to
break free from the bonds of mathematics during the last decades.
260 JAN-OLA ÖSTMAN
(When referring to these as lexemes, the abbreviations TÄÄ, TUO, and SIE,
respectively, are used in the ensuing discussion.)
The whole system of demonstrative adverbs in Finnish is given in table
3; the demonstrative pronouns 'this', and 'that', as well as the temporal
adverbs, 'now' and 'then', are included for ease of reference.14
TÄÄ TUO SIE

'here' 'there' ('there')15
Inner Outer Inner Outer Inner Outer
tässä [In] täällä [Ad] tuossa tuolla siinä siellä

tähän [111] tänne [All] tuohon tuonne siihen sinne
tästä [El] täältä [Ab] tuosta tuolta siitä sieltä
'this' 'that' ('that')
tämä tuo se
'now'/'then' 'then' 'then'
tällöin tuolloin silloin
Table 3.
[In = Inessive; Ad = Adessive; El = Elative; 111 = Illative;
All = Allative; Ab = Ablative]
According to the general scheme presented in figure 1, a three-way

distinction would typically reserve one term for closeness to speaker (Dl;
'here'; TÄÄ), another for 'closeness to addressee and/or short distance from
speaker' (D2, Dsd; 'there'; TUO, or: SIE), and the last term for 'long
distance from speaker and/or non-visible' (Did, D-v; 'yonder'). This type of
description has also been applied to the Finnish system - often with little
14
Only those temporal adverbs that are formed from the same base form as the
demonstrative adverbs are given; for instance, the default word for 'now' is nyt.
15
The Inessive and Elative forms of the SIE-term are exceptional, but functionally
their use is comparable to what would be regular forms.
success. It is obvious, for instance, that neither the SIE nor the TUO term
is used in Finnish as an equivalent to the English yonder, as we can see
from example (16).
(16) A: Mitä tuossa laatikossa on? (pointing to a box on the kitchen
table, as A and B are in the kitchen making dinner)
'What is in that [=TUO] box?'
B: Siinä on käärme.
There's a snake (in) there [=SIE].'
Instead of starting by going through other people's suggestions in
detail, and presenting counterexamples to their characterizations of the
TÄÄ-TUO-SIE distinction, I will proceed by taking into account what I have
said about other possible ways of looking at deictic systems in languages,
and criticize alternative analyses of the Finnish 'Proximity' system as I go
along. In particular, I want to suggest what an alternative analysis of the
Finnish demonstrative system might look like, an analysis which is not
primarily based on physical aspects like location.
The greatest problems with the Finnish system are centered around the
interpretation of the third term in the system, the SlE-forms. At least in an
initial analysis we come close to understanding the TÄÄ and TUO terms if
we keep to the characterization in figure 1. But the third term does not fit
in neatly. There are at least three alternative strategies to follow in this
situation:
Strategy 1 would simply say that the general scheme is valid, and if there
are uses of, in this case, the third term that do not fit, such
uses are due to metaphorical extensions/interpretations of
what the third term in a three-term system typically stands for.
Strategy 2 would say that the proximity distinction is not a major or basic
one in Finnish at all. The primary distinction is in terms of
something else, and only as a consequence of this other
distinction can the uses of TÄÄ, TUO, and SIE be talked about
in proximity terms.
Strategy 3 would say that the first two terms make up a basic, simple
proximity system - just as in English; but that the third term
is either
(3a) completely outside the system that the other two terms
262 JAN-OLA ÖSTMAN
partake in; or,

(3b) a general-deictic term, something like the initial D, or Dx in
figure 1, but on a different, complementary plane.
Both versions of Strategy 3 seem somewhat dubious. Strategy 3b is
questionable on empirical grounds, since, if taken to its logical conclusion,
the third term should then also cover, or include the meanings of the two
other terms — if seen as the initial D - or of the other Dx term; in
particular, one should not be able to say things like
(17) Älä sinne mene - mene tuonne, tai jää tänne.
'Don't go there [=SIE], go there [=TUO], or stay here [=TÄÄ].'
which is a perfectly good sentence, referring to three different spatial
directions/locations.
Strategy 3a seems methodologically dubious as a general strategy: If
taken to its extreme it would be equivalent to saying that the SIE-term is
not a demonstrative adverb at all. This is of course a possible solution, but
it does not do much in helping us to understand examples like (17). In
connection with this, a very interesting kind of suggestion made by
Larjavaara (1985; cf. also Larjavaara 1990) is worth mentioning. He
suggests that there are two (or actually four) sub-subsystems superimposed
on each other, which together explain the use of demonstrative adverbs in
Finnish. Very briefly, Larjavaara's parameters are:
Proximity-Distance to differentiate between TÄÄ and TUO;
Speaker-centeredness to differentiate between TÄÄ and TUO on the one
hand (as Speaker-centered) and SIE on the other hand (as not
Speaker-centered).
In addition, in order to be able to explain all the data, he differentiates
between SIE-terms as used when the referent is Old/Given and TUO-terms
when the referent is New. On top of all this, the meanings and functions of
the three terms with respect to the three parameters already given, will be
differentiated on the basis of whether the situation is Polarized or not, i.e.
whether the speakers have perceptual access to each other or not.
Even if I have certain reservations about some of Larjavaara's points,
this is not the proper place to go into a detailed discussion of these. I have
presented Larjavaara's suggestion in order to give the reader some idea of
the complexity of the system. I will leave his system as it stands, confident
that something along the lines of his analysis would allow a computer to
produce appropriate demonstrative adverbs in particular, well-specified

situations.
Strategy 1 tends to be treated as a default strategy when describing
deixis. But this is where the discussion in section 4 becomes crucial.
Linguists have learnt to be sceptical about analyses in other parts of
grammar that utilize too many concepts borrowed from some classical
model of description, e.g. analyses making use of concepts like Gerundive
and Subjunctive, taken over in toto and being applied to, say, English, not
to mention to non-western languages. That is, we need to be aware of the
general problems in trying to fit categories and distinctions into ready-made
'moulds', especially if such moulds cannot be shown a priori to have any
more (nor any less) value than some other possible system. Thus, it should
be obvious that my first choice for Finnish would not be an adherence to
Strategy 1. Furthermore, in this case the data (cf. example (16)) is simply
not in agreement with a 'here-there-yonder' system. Cf. also example (18).
(18) Mitä sinulla siinä on? Tuo likainen tahra paidassasi.
'What have you got there [SIE]? That [TUO] dirty stain on your
shirt.'
However, there have been some fairly interesting suggestions that
ultimately rely on space and location, but go beyond these in a number of
ways. A fairly early example is Itkonen's (1966) suggestion of a concept of
'sphere of perception', such that TÄÄ-forms refer to the speaker's, SIE-
forms to the addressee's, and TUO-forms to both the speaker's and the
addressee's spheres of perception. He runs into difficulties, however, when
he has to posit secondary spheres of perception in order to account for all
the data. In addition, Larjavaara (1985: 19) includes a number of
counterexamples to Itkonen's distinction.
A more recent proposal has been made by Wiik (1975), who suggests
that SIE indicates closeness to addressee, and TUO closeness to a third
person or thing. However, to be able to account for all the relevant data,
he has to stipulate two functions of SIE, as indicating
(a) closeness to addressee; and
(b) Old/Given information (as opposed to TUO, which is said to indicate
New/unknown information). (This, as we saw, was also a parameter in
Larjavaara's system.)
But, first, if there is a distinction between SIE and TUO in terms of
making reference to a second or a third person, respectively, this is at most
264 JAN-OLA ÖSTMAN
a loose tendency. TUO can very well be used of something close to the
addressee, especially if the speaker and addressee are further apart than a
couple of feet:
(19) Tuossa sinun vieressäsihän veitsi on!
there-INESS your beside-INESS:2sg-RELEV knife is
T h e r e [=TUO] next to you is where the knife is!'
Secondly, the idea of SIE indicating Givenness of referent is not as
straightforward as one would like it to be. (And in addition, Givenness is
of course inherently linked up with Discourse deixis.16) Wiik gives the
following example:
(20) Tuolla näet tuomiokirkon. Siellä on kaunis alttaritaukh
T h e r e [=TUO] you see the Dome. There's a beautiful altar
painting there [=SIE]/
However, I can perfectly well introduce the topic of the Dome with a
sentence like (21).
(21) Siellä tuomiokirkossa on kaunis alttaritaukh
T h e r e [-SIE] in the Dome is a beautiful altar painting.' 17
And thirdly, it would surely be more satisfying if we were able to
stipulate one basic function for all the uses of the deictic SIE. In Wiik's
analysis we would have to say that there are in fact two homonymous
terms: SIE1 and SIE2. Here my stand is a cognitive-linguistic one in line with
the notion that languages tend to adhere to a basic one-meaning-one-form
principle (cf. Bolinger 1977).
16
It may indeed be worth while investigating to what extent the SIE term has a basic
Discourse deictic function in present-day Finnish. The use of se (which is both the third
person singular neuter pronoun, and the third-term demonstrative pronoun) for cohesive
purposes in discourse is well-known.
One can of course here argue that the difference is that in (21) the deictic SIE-
term is used attributively, and that the new referent is thus introduced in the same phrase.
Still, the difference between SIE and TUO in this respect is that both can occur
attributively, but if they occur independently TUO is accompanied by some ostensive
gesture, whereas SIE is not. Thus in the first sentence of example (20), TUO can be said
to be used attributively to the necessary gesture of pointing, and SIE in the second
sentence has a Discourse deictic function. The difference between New and Given is thus,
it seems to me, secondary, and dependent on the presence of an ostensive gesture.
Strategy 2 invites the stipulation of a completely different 'basic'
parameter behind the semantics and pragmatics of the demonstrative
system in Finnish, a parameter which would be capable of accounting for
the data more homogeneously than the Proximity distinction.
Such a parameter — I propose — can be set up in terms of Affective
deixis. As I have mentioned in section 4 and above, a distinction on Spatial,
Temporal, and Personal grounds cannot a priori be taken as more basic
than one on Discourse or Social grounds, and I would like to suggest that
Affective deixis, or Emotional deixis as it is sometimes called, is also a
potentially primary way of perceiving and thus codifying the world of deixis.
According to this suggestion, the SIE-term is the neutral one, the TÄÄ-
term is speaker-oriented, and the TUO-term is non-speaker oriented.18
The distinction is in terms of what scholars refer to as Involvement, Affect,
or Empathy: The TÄÄ-term is used for referents the speaker has empathy
with, or is involved with. Thus, as a consequence, it will also be used for
things proximal to the speaker. The TUO-term is used for cases where the
speaker does not have empathy with the referent, or where s/he even has
antipathy towards the referent — thus, as a consequence it is used of distal
things, and of referents that are new, because they are unknown. The SIE-
term, finally, is basically used when empathy, involvement, and affect is not
at issue in the communicative situation.
Before going through this suggestion in some detail, I want to make
the general comment that I see my own approach as complementary to
Larjavaara's very elaborate one. Mine is an approach that tries to get at
the (unconscious) cognitive categorization that Finnish speakers may be
involved in as they make decisions of what demonstrative to use.
Furthermore, I take for granted that since the three terms TÄÄ, TUO, and
SIE do pattern together, we should try to find some common denominator,
some prototype definition of what semantico-functional parameter the
TÄÄ-TUO-SIE system in Finnish is based on.
One kind of counterargument against the suggestion of Affect and
Empathy would be to say that, since the other two subsystems in the
18
A somewhat similar distinction might be appropriate for German hier - dort - da,
with da as the neutral term (cf. e.g. Ehrich 1982: 62); and for Japanese sore - kore - are,
with are being the neutral term (cf. Fillmore 1982: 54). Compare also the -va, -sa, -ny
distinction in Yuma, where according to Halpern (1946) the third term indicates location
unspecified'. Perhaps also the Spanish ahi can be seen as a neutral term in the sense I
have in mind here.
266 JAN-OLA ÖSTMAN
Finnish deictic system have to do with space, why should not the third
subsystem have this feature in common with the other two? However, as I
have shown in the discussion above, it is not at all clear that both of the
other two subsystems have to do with space as a basic parameter. It is true
that the Location/Direction subsystem definitely has to do with physical
space, but the Inner-Outer system proved to be based on an aspectual
distinction. (And we even found - cf. examples (12) and (13) - that
Boundedness, too, is relative to the addressee's orientation.) That is, space
is not as such of primary importance in all aspects of the Finnish
system.19
In the spirit of the generative argumentation of the 60s, I will now give
a number of arguments for, and instantiations of, the distinction I have
suggested. I will pay most attention to the SIE-term, since its use has been
notoriously difficult to pin down. Also, the SIE-term is crucial in that if we
accept the challenge from cognitive linguistics of finding some unifying
parameter with whose help we can explain the TÄÄ, TUO, and SIE terms
and relate them to one another, then focussing on the functions of the TÄÄ
and TUO terms might become counterproductive: In both the figure-1
approach and in my alternative, the TÄÄ-term will be linked to the
speaker, and the TUO-term to the addressee, or to some form of distance
from the speaker. Against this background, it will be predictably difficult to
say which of the suggested interpretations is more basic and primary than
the other. However, if the SIE-term can be shown to fit one system better
than another, we can obviously decide in favor of the system it better fits.
As we have already seen, the use of SIE is not easily amenable to the kind
of model presented in figure 1. The points below attempt to show the
feasibility of the alternative model. The first two arguments establish the
SIE term as more 'neutral' than the other two. The rest of the arguments
address the issue of Affective deixis more directly.
1. First of all, although Finnish does not usually make use of any
grammaticalized deictic form like English existential there in sentences like
There's a snake in the garden, on the few occasions that Finns do need such
This being the case, we could even be so bold as to use this counterargument for
our benefit: Finnish makes one distinction in terms of space, and another in terms of
Boundedness. There is thus no a priori reason why the third distinction should be in
terms of space. The default expectation in this situation would be that the distinction is
not based on location.
a grammatically neutral term, siellä or siinä (i.e. SIE) are the ones that are
given this neutral status.20
The inherent neutrality of the SIE-term also shows up in that it is the
term that is used neutrally without necessary reference to any particular
location. Thus, if I say (22),
(22) Ostatko sen kirjan?

buy-you-QUESTION SIE-ACC book-ACC
'Are you going to buy that [=SIE] book?'
the deictic element refers to something contextually known, but is
otherwise, with respect to its location, unspecified. In contrast, if I say (23),
(23) Ostatko tuon kirjan?
'Are you going to buy that [=TUO] book?'
I would have to have the book perceptually present, and the question
would typically be accompanied by a pointing gesture. (Cf. fn. 17.)
2. Secondly, expressions like here and there and related 'freezes' do not
have the ME-First expected TÄÄ-term first (see Cooper and Ross 1975 for
a detailed discussion of the ME-First principle); rather, the SIE-term comes
first, with the general pattern being 'SIE (&) TÄÄ':
siellä (sun)21 täällä 'here and there'
sinne (sun) tänne 'here and there' [directional 'to']
sitä (sun) tätä 'this and that'
silloin tällöin 'now and then'
In the Affective-deictic interpretation of the system, we can now find an
explanation for why the SIE-term is the first term (this is a fact that has
created some uneasiness among Finnish linguists, cf. e.g. Kalliokoski 1986):
The SIE-term, being empathy-neutral, is simply the one that gets codified
20
We can here note that the grammatical neutrality of German da, Japanese are, and
Finnish SIE is motivated in a sense that English there is not.
21
Expressions more similar to English 'here and there', with the ordinary
conjunction ja 'and', have recently begun to occur; thus, we may find siellä ja täällä
instead of the more 'archaic' siellä sun täällä or siellä täällä,
268 JAN-OLA ÖSTMAN
as the first (and neutral) element in lexicalizations.22

Larjavaara points out that speakers of Finno-Ugric languages in
general are not as speaker-centered as speakers of western IE languages.
Thus, in the case of Finnish demonstratives, TÄÄ cannot be taken as the
basis for explicit lexicalizations, because it is too self-centered; TUO cannot
be used, because it is too negative, and thus too impolite. (See below.) But
SIE is just perfect,23
Admittedly, the frequency of the second formant in the first syllable of
the first term (-i-) is higher than that of the second term (-ä-), and could
thus, in accordance with the ME-First principle, be seen as the reason why
the SIE-term comes first. This suggestion does not, however, explain why
we do not have, for instance * Täällä (sun) tuolla as meaning 'here and
there524 25
22
Kalliokoski (1986) has argued that the order of elements in freezes of the siellä
täällä kind is one indication that the ME-First principle cannot be upheld cross-
linguistically. Alternatively, I would say that the reason Finnish does not follow the
principle is not because the western-culture based principle as such is wrong, but only
that speakers of Finnish, and no doubt of other languages too, do not necessarily
categorize the world in the same terms (of physics) as speakers of Indo-European, western
languages have grown up to do.
23
In general, Finnish seems to be more apt to use what I have called the YOU-
Importance principle (Östman 1981a), rather than the ME-First principle. Thus, in
Finnish one typically says Marja ja minä 'Marja and me', and sinä ja minä 'you and me'.
The opposite order of the terms is almost unthinkable.
24
Neither suggestion explains why täällä and not tuolla is chosen as the second term
in siellä täällä, 'here and there'. My suggestion is that since both SIE and TUO, according
to almost all analyses, are viewed as distal in relation to the TÄÄ term (but cf. below, for
additional explanations), having two distal terms to stand for an expression that means 'a
little bit all over', would exclude the area in close vicinity to the speaker. Thus, since siellä
has to be chosen because of its neutrality, täällä is the only, and obvious, choice as the
second term.
25
I understand, though, that täällä tuolla is indeed an expression that in the very
recent years has started to creep into spoken Finnish - probably due to influence from
the format of equivalent expressions in English. (Kari Pitkänen, p.c.) I have also found an
instance in a newspaper of the expression tuolloin tällöin 'now and then' (or rather, 'then
and now'), but I presume this is word play rather than a general tendency.
3. Thirdly, when these terms are used in an emotional deictic manner
(what I have called Affective deixis in this paper; cf. R. Lakoff 1974;
Collinson 1937; Lyons 1977), TUO is the most negatively loaded:
(24) Tuo Vayrynen on sitten maailman suurin älykääpiö.
That [TUO] Väyrynen is really the world's biggest pighead.'
TÄÄ is more positively loaded:
(25) Tä(m)ä Fillmore on sitten fiksu tyyppi.
This [TÄÄ] Fillmore is a really sharp guy.'
whereas SIE takes an intermediate position, and is used when no definite,

negative or positive, stand is taken or implied26:
(26) NO, se Donner on nyt sellainen kun se aina on ollut.
'Well, that [SIE] Donner is the way he's always been.'
From an interactional point of view, we could say27 that SIE, by being
the neutral default term, imputes a (real or apparent) feeling of common
experience between the interactants. This also suggests that SIE is not a
general, non-affective term in the system, but rather a term that
Note also that se is the third person neuter singular pronoun, 'it', and in many
dialects of Finnish - and very frequently in informal speech - it is often used without
derogatory implication of people.
The distinction that Fillmore (1975: 71) mentions for English between 'this' as being
used when only the speaker knows what is referred to, and 'that' when both speaker and
addressee know, seem not to work for Finnish; cf.
(i) Tapasin eilen poliisin Espalla; no, se/tää/tuo kaveri/kundi/tyyppi oli ihan
umpikännissä.
'I met a policeman on the Espa boulevard yesterday; well, this/?that guy was
really stoned.' '
(ii) Muistatko (sitä/tätä/tuota) kaveria, joka myi meille lippuja matsiin? No,
se/tää/tuo kundi kertoi minulle, että...
'Do you remember the guy who sold us tickets to the game? Well that/?this guy
told me that...'
I am not completely convinced that this holds as a tendency even in English, unless
reinterpreted in terms of Affective deixis.
21
I am here following a suggestion made by Auli Hakulinen (p.c.).
270 JAN-OLA ÖSTMAN
presupposes a shared experience, a shared perspective on things.28 TÄÄ
(as in (25)) involves a process of the speaker 'placing' the addressee next
to him/her and imposing his/her perspective on the addressee in order for
both of them eventually to end up sharing the speaker's perspective. Thus,
as a consequence, TÄÄ will be used to introduce something the
addressee/audience is not familiar with from before, as in Sit tää koira
yhtäkkiä ilmestyy ja ... Then this [TÄÄ] dog suddenly appears and ...'
whereas some other specifier has to be used if the speaker is not interested
in pulling in the addressee into his/her story: Sit yks' koira ilmestyi ja ...
Then a [='one'] dog appeared and ...'. In a Cognitive-Affective
perspective, if a referent is viewed as important enough by the speaker for
min/her to pull in the addressee, the speaker (possibly inadvertently)
displays a potentially positive attitude to this referent. Dialogically
speaking, TUO (as in (24)) would primarily focus on the speaker's relation
to the object referred to. Here, at the same time as the speaker (almost
accidentally) pulls in the addressee, the speaker's negative attitude towards
the referent under discussion is clearly displayed. In this way, the function
of Affective deixis can be supported interactionally.29
4. Fourthly, we need to account for a frequently occurring suggestion that

SIE is addressee-oriented. The reason for this view, I suggest, has to do
with the importance of politeness in interaction. For something situated at
a distance from the speaker, the use of the (at least potentially) negative-
empathy term TUO would not be appropriate if that referent is
simultaneously close to the addressee (no matter how this is interpreted,
e.g. as being in the addressee's sphere of perception). As we already saw
above, SIE presupposes a shared perspective with the addressee. Since, in
terms of Affect, SIE is neither directly negative nor positive, it can be seen
as a useful euphemism when orienting towards the addressee without
having to express any further aspect of involvement.
28
Cf. Fillmore's (1982: 54) discussion of are in Japanese.
29
The primacy of cognition vs. interaction is a notorious chicken-and-egg issue. In
this paper whose main empirical task is to find the parameter according to which the
demonstrative adverb system in Finnish is codified, I have preferred to treat them as
cooperating.
5. Fifthly, why is SIE often felt to be used for old information?30 Old
information is common to both speaker and addressee. Thus, the neutral
SIE is obviously the default choice. If any of the speakers are particularly
involved with the matter talked about, this can be negotiated in other ways
- explicitly or implicitly. If implicitly (cf. Östman 1986), then TUO is indeed
one alternative (Bl, B2, and B3 are alternative responses to A):
(27) A: Tää Jörn osaa sitten olla herttainen.
This [=TÄÄ] Jörn really knows how to be sweet.'
Bl: Kuules, tuota Donneria et kyllä tuo tähän taloon yöksi!
Tisten, don't you even think of bringing that [=TUO] Donner
into this house for the night.'
B2: Sinähän taidat olla vähän rakastunut siihen Donneriin!
'You seem to be a little bit in love with that [SIE] Donner.'
B3: Olkoonkin herttainen, kunhan muistat, että tää söpöliini, se on
mun!
'Call him sweet if you like, as long as you remember that this
[TÄÄ] cutie is mine.'
6. Sixthly, with the suggested interpretation of the function of the third
term, SIE, we can also see why one might not so readily want to use TUO
when the speaker and addressee are close together (cf. Wiik's suggestion
above):
(i) TUO is non-speaker oriented (thus TÄÄ could just as well be used);
and,
(ii) SIE, being neutral with respect to Proximity, can thus be used in all
situations that do not explicitly require TÀÂ or TUO.
7. Seventhly, we can now also, referring back to example (20) discussed
earlier, explain why we do not so readily use siellä to accompany an
ostensive act (cf. also fn. 17). The reason is the neutral character of siellä:
If one points to something, one makes a marked choice of indicating one
phenomenon over and above all others.
30
Saukkonen's (1967) findings rather suggest that se is neutral with respect to Given
and New. Cf. also fh. 16.
272 JAN-OLA ÖSTMAN
These arguments will have to do for the time being.31 The important
point to note is that the parameter of Affective deixis that is needed in
order to adequately describe the Finnish demonstrative-adverb system is
not derived from the place-time parameter, but is an independent, basic
parameter. And since this is so for Finnish, it is most likely a parameter to
be taken seriously for debris in general.
As we know, our views of place and time are based on, and filtered
through, our perception. And if as a corollary to this, we can in addition
say that time and place a priori do not have any more justified status in the
world according to physics than does any other product of our perception,
then there is no reason to assume that place and time are the only
concepts with cognitive status which partake as determinants of
demonstrative-adverb systems.
5, Further implications
Having a member of a system as the default member - the way SIE can be
seen as a default member in the demonstrative system — is nothing
extraordinary in Finnish grammar. As I pointed out in Östman (1981b)
there seems to be a clear tendency to use the nominative case as a default
case where it should not really appear according the the rules of Finnish
grammar. This happens in cases where there is either just, say, one NP in
the sentence, or where there for other reasons can be no possibility of
misunderstanding. Thus, the impersonal construction (sometimes
erroneously called a 'passive') takes a nominative form as object case
where one would expect an accusative (cf. Östman 1981b), Also, in
imperative sentences, not only do objects occur in the nominative, as in
(28) Ota auto!
take-IMP:2sg car-NOM:sg
Take the car.5
31
For instance, it is not inconceivable that the uses of sitä and t(u)ota as pragmatic
particles could also be explained with reference to the discussion in this paper. Basically,
t(u)ota is a hesitational device; hesitation marks indeterminacy, which might be connected
with negative attitudes. Sitä has been extensively discussed by Hakulinen (1975); perhaps
my suggestion of its euphemistic use for politeness purposes might come in handy here.
These are, obviously, fairly speculative suggestions and matters that need to be discussed
more fully.
Cf.
(29) Otan auton.
take-PRES:lsg car-ACC:sg
T m taking the car.'
but also adverbial phrases in imperative constructions can take the default
nominative case instead of the accusative; cf.
(30) Juokse reippaasti koko matka!
run-IMP:2sg briskly whole journey-NOM:sg
'Run briskly all the way.'
(31) [A recent ad in Finland:]

Hoida ihoasi kunnolla
treat-IMP:2sg skin-PART:sg-POSS:2sg properly
koko talvi!
whole winter-NOM:sg
T a k e good care of your skin all winter.'
A very different question that needs to be addressed is why I prefer
something like my Affect-based system to the extremely elaborate system
suggested by Larjavaara (1985). If Larjavaara's suggestion works, then why
bother with a different - and what is worse, a not yet fully worked-out -
attempt to describe the demonstrative system in Finnish? The answer of
course has to do with what kind of explanations one searches for in
linguistics. I see Larjavaara's suggestion as basically language-internal,
whereas mine is more cognitively based.
It is true that with respect to the Inner-Outer subsystem in Finnish, I
stopped with the concept of Aspect as a feasible explanatory factor. But in
the case of the 'Proximity' subsystem I had to go further, into the
behaviorally even more basic concepts of Affect and Emotion. This is
bound to create an uneasiness as regards my general view on explanations
in linguistics: How does the researcher know that s/he has reached a
satisfactory level of explanation? My answer is that s/he stops when s/he
has found a common denominator that can not only describe the system at
work, but also ultimately explain this system as motivated on cognitive
and/or socio-cultural grounds. In addition, following DuBois (1985),
linguistic systems and subsystems also take on explanatory power, and - as
we have also seen in regard to the Finnish system (cf. fn. 25) - at this
level, re-grammaticalizations are always possible, with stages of competing
274 JAN-OLA ÖSTMAN
systems being constantly operative.
6. Conclusion
My main point in this paper has been a methodological one. In particular,

I have wanted to show that even if some aspects of a linguistic system (in
this case the Finnish demonstrative adverb system) can be explained in
terms of ready-made distinctions found useful in other languages and
cultures (here, the proximal-medial-distal system), we might get a more
coherent view of interrelating linguistic phenomena if we allow ourselves
the luxury of breaking away from such ready-made moulds, A distinction
we may have thought to be primary may prove to be a secondary
distinction. In this paper I have wanted to stress the importance of an
Affective parameter to explain the workings of the Finnish system of deixis.
Linguistics has deep-rooted conceptions of the nature of deixis, which
are partly based on metaphorical material borrowed from other disciplines.
Using these metaphors for the description of deixis with inadequate regard
for their original context of use is doing violence to both disciplines and
results in distortion. Thus, it seems to me that we need a multidisciplinary
approach to lead us to new ways of conceptualizing the human deictic
system and to remind us of the fact that the concepts of our own field are
metaphorical and subject to change.32
There are a number of residual issues that I have not had the
possibility to go deep into in this paper. For instance, I have not explicitly
brought in Sapir and Whorf, although it is obvious that the reader can
easily tie what I have said to issues generally discussed within linguistic
relativity. Much more should, for instance, have been said about the status
of objectivity and objective knowledge in different cultures. The discussion
also casts doubt on any a priori reliance on concrete relations as being
more fundamental for cognition than other relations, as is often assumed
in cognitive "linguistics. Other related issues are those of determinism, and
the feasibility of localist approaches to language. I have used physics as an
example of how advances in one science outside linguistics can and should
affect the choices and perspectives we take on linguistic descriptions.
Basically any other science could have done the job (for a brief sojourn
32
I am grateful to Mauri Ronimus for lengthy discussions on these issues.
into biology, see Östman 1988).

I have given an interpretation of the Finnish demonstrative adverb
system in cognitive and in particular, in affective terms. I have consciously
wanted to take one line of thought to see how far it can be extended. I
have, however, mentioned a number of other parameters that I have
interpreted as being secondary to the Affective deixis parameter for
Finnish. I would not see it as a catastrophe if it turned out one day that,
say, a dialogic/interactive parameter, or a discourse parameter proved to be
more basic than the one I have set up as Affective deixis. I am convinced,
though, that it is not enough to be able to roughly simulate the intricacies
of complex systems like that of place deixis in Finnish; ultimate
explanations have to be not only cognitively plausible, but also interactively
and culturally feasible. It is in this spirit that I have offered a somewhat
drastic reanalysis of the 'Proximity' system in Finnish in this article.
References
Bolinger, Dwight L.
1977 Meaning and form. London & New York: Longman.
Collinson, W.E.
1937 Indication. (Language Monographs 17.) Linguistic Society of America.
Cooper, William E. & John Robert Ross

1975 "World order". In Robin E. Grossman, L. James San & Timothy J. Vance
(eds.), Papers from the parasession on functionalism. Chicago, IL: Chicago
Linguistic Society, 63-111.
Ehrich, Veronika
1982 "Da and the system of spatial deixis in German". In Jürgen Weissenborn &
Wolfgang Klein (eds.), Here and there: Cross-linguistic studies on deixis and
demonstration. Amsterdam & Philadelphia: John Benjamins, 43-63.
DuBois, John W.
1985 "Competing motivations". In John Haiman (ed.), Iconicity in syntax.
Amsterdam & Philadelphia: John Benjamins, 343-365.
1975 Santa Cruz lectures on deixis 1971. Bloomington, IN: Indiana University
Linguistics Club.
276 JAN-OLA ÖSTMAN
1977 "Topics in lexical semantics". In Roger W. Cole (ed.), Current issues in

linguistic theory. Bloomington, IN: Indiana University Press, 76-138.
1980 Lectures on linguistic semantics, University of California, Berkeley, spring

1980. [Unpublished]
1982 "Towards a descriptive framework for spatial deixis". In Robert J. Jarvella

& Wolfgang Klein (eds.), Speech, place, and action: Studies in deixis and
related topics. Chichester, etc.: John Wiley & Sons, 31-59.
1988 "The mechanisms of 'Construction Grammar'". BLS 14.35-55.

alone" Language 64:3.501-538.
Hakulinen, Auli
1975 "Finnish sitä - An instance of the interplay between syntax and
pragmatics". Scandinavian Conference of Linguistics 2.147-163.
Hakulinen, Auli & Fred Karlsson

1979 Nykysuomen lauseoppia. [Modern Finnish syntax] Helsinki: Suomalaisen
Kirjallisuuden Seura.
Halpern, A.M.
1946 "Yuma III: Grammatical processes and the noun". UAL 12.204-212.
Itkonen, Terho
1966 "Tutkimus suomen asyndetonista". [Investigating asyndeton in Finnish]
Virittäjä 70.
Kalliokoski, Jyrki
1986 "Rinnastusfraasin jäsenten järjestyksestä". [On the order of units in
coordinate phrases] Kieli 1.73-102.
Karlsson, Fred
1982 Finnish grammar. Porvoo, Helsinki, Juva: WSOY.
Lakoff, George
1972 "Hedges: A study in meaning criteria and the logic of fuzzy concepts". CLS
8.183-228.

1980 Metaphors we live by. Chicago, IL: University of Chicago Press.
Lakoff, Robin
1974 "Remarks on this and that". Berkeley Studies in Syntax and Semantics, XVII.
Berkeley, CA: Department of Linguistics, UC-Berkeley.
1989 "The way we were; or, The real actual truth about generative semantics: A
memoir". Journal of Pragmatics 13:6.939-988.
Larjavaara, Matti
1985 "Suomen demonstratiivisysteemin rakenne". [On the structure of the
demonstrative system in Finnish] Sananjalka 27.15-31.
1990 Suomen deiksis. [Deixis in Finnish] Helsinki: Suomalaisen Kirjallisuuden

Seura.
Lehtinen, M.
1967 Basic course in Finnish. Indiana University Press.
Leino, Pentti
1989 "Paikallissijat ja suhdesääntö: Kognitiivisen kieliopin näkökulma". [Local
cases and the relation rule: An approach from cognitive grammar] Virittäjä
93:2.161-219.
Levinson, Stephen C.
1983 Pragmatics. Cambridge: Cambridge University Press.
Lyons, John
1977 Semantics. Cambridge: Cambridge University Press.
Östman, Jan-Ola
1981a "You know: A discourse-functional approach". Amsterdam & Philadelphia:
John Benjamins.
1981b "The Finnish 'passive' and Relational Grammar". CLS 17.286-294.
1986 Pragmatics as implicitness. An analysis of question particles in Solf Swedish,

with implications for the study of passive clauses and the language of
persuasion. University of California, Berkeley PhD dissertation. Ann
Arbor, MI: University Microfilms International, 86-24885.
1988 "Adaptation, variability, and effect: Comments on IPrA Working

Documents 1 & 2". IPrA Working Document 3.5-40.
1992 "Rationality and, and versus, common sense as pragmatic explanators".

Paper presented at LACUS 19, 1992, Montreal, Canada.
278 JAN-OLA ÖSTMAN
Saukkonen, Pauli
1967 "Persoonapronominien hänise, he:ne distinktiivi oppositio". [The distinctive
opposition between the personal pronouns hän he:ne] Virittäjä 71:3.286-
292.
Talmy, Leonard
1978 "The relation of grammar to cognition - A synopsis". TINLAP 2.
Verschueren, Jef
1987 Pragmatics as a theory of linguistic adaptation. (IPrA Working Document
1). Antwerp: International Pragmatics Association.
Weissenborn, Jürgen & Wolfgang Klein (eds.)

1982 Here and there: Cross-linguistic studies on deixis and demonstration.
Amsterdam & Philadelphia: John Benjamins.
Wiik, Kalevi
1975 "Ruotsinkielisten tuolla-siellä -vaikeudesta". [On the difficulties that
Swedish speakers have with tuolla-siellä] Kielitieteellisiä lastuja 1.36-37.
Turku: Linguistic Association of Finland.
Yngve, Victor H.
1986 "To be a scientist". LACUS Forum 13.1-21.
Frame semantics and the lexicon:
Nouns and verbs in the body frame
Miriam R.L. Petruck

University of California, Berkeley
1. Introduction
Fillmore (1975, 1976a) introduced the FRAME idea into linguistics in part
as a response to what has been called "checklist theories of meaning". In
that approach to semantics, the meaning of a linguistic form is represented
in terms of a checklist of conditions that have to be satisfied in order for
the form to be appropriately or truthfully used. In FRAME SEMANTICS,
word meaning is characterized in terms of experience-based schematizations
of the speaker's world - i.e. frames. It is held that understanding any
element in a frame requires access to and understanding of the whole
structure. Later on, Fillmore (1978) characterized the frame as the most
central and powerful kind of domain structure, thus alluding to a
frame-based organization of the lexicon. In a frame-based organization of
the lexicon, it is the frame which provides the conceptual underpinnings for
related senses of a single word and semantically related words. This would
necessarily include all categories of words, including nouns, verbs, adjec
tives, adverbs, prepositions, and conjucntions, as well as phrases and
expressions.1
In my own work (Petruck 1986, forthcoming), I have attempted to
validate and extend the frame semantics approach to the study of meaning
- word meaning, word structure, and semantic structure in the lexicon - by
demonstrating its applicability to one area of the lexicon in one language,
1
Of course, Fillmore's other writings on frame semantics (1977, 1982, 1985, 1992)
have influenced my work on the subject as well.
280 MIRIAM R.L. PETRUCK
body part terminology in Modern Hebrew. I proposed an experience-based

schematization of the BODY FRAME and described the meanings of Hebrew
body part words in terms of what speakers know or believe about the
entities designated by those words. I characterized the internal structure of
polysemous words and the semantic structure of the domain in terms of this
schematization. While the frame semantics approach could be demonstrated
with any lexical semantic domain, it seems especially appropriate for body
part terminology because its referential domain, the body, is one with which
all people have primary experience.
In this paper, I will be dealing with certain kinds of nouns and verbs in
Hebrew based on body part terms. I claim that these lexical items are part
of the body frame as well because it is experience of and knowledge about
the body that provides the conceptual underpinnings for their formation
and understanding. By analyzing compound and complex forms and
uncovering the kinds of experience and knowledge involved in their
formation, this work provides further specifications of the body frame and
further support for the frame semantics approach to meaning. Ultimately,
it contributes to the development of a theory of the lexicon based on the
frame.
Metaphor and metonymy play an important role in the development of
new senses for existing words and the formation of new words (Lakoff and
Johnson 1980; Claudi and Heine 1986; Heine and Claudi 1986; Sweetser
1988, 1990; Traugott 1988, 1989). While a metaphoric transfer occurs
between two discontinuous domains on the basis of (perceived) shared or
analogous characteristics, a metonymic transfer occurs within spatio-tempor-
ally adjacent or contiguous domains (Jakobson 1971).
Metaphor and metonymy occur within the body part domain itself. For
example, in the Hebrew body part domain, similarity of relative position
and structure between fingers and toes along with the known special
function of one finger allow for the different body part senses of the
polysemous term 'ecba - 'finger, index finger, toe5. Comparable principles
apply for non-body part senses of 'ecba - i.e. for extensions to other
domains. Thus, for instance, in the measure word sense of "ecba - 'finger's
breadth', the relevant factors are the width of the finger and the practice
FRAME SEMANTICS AND THE LEXICON 281
of measuring distances with certain parts of the body.2 And, as apparent

from 'ecba'ot ha-kfafa - 'fingers of the glove', 'ecba is also used in the
domain of clothing terms. In addition, the denominal body part verb hicbia
- 'raise finger, vote' lexicalizes the purposeful functioning of the body part
as instrument. Metonymic extensions of these kinds are not particularly
surprising. Parts of the body are used to determine units of measure; parts
of the body are intimately associated with the clothing that covers them;
and parts of the body have functions and perform actions. The extension
of body part terms to measure words, clothing words, or body-part-as-instru-
ment verbs seems quite natural in part because of the physical proximity of
the extension to its source.
The morphological structure of the words in the Hebrew body part
domain are noteworthy because for the most part they are simple terms,
unmotivated and unanalyzable, and often serve as the basis of compound
and complex forms. The description and analysis of compound and complex
words based on body part terms testify to the important role of metaphor
and metonymy in the development of new senses for existing words and the
formation of new words. In section 2 of this paper, which follows, I will be
talking about two categories of extension, the one exemplified by compound
nouns and the other exemplified by morphologically complex nouns and
verbs derived from body part terms. Having determined the different
semantic factors which motivate the metaphoric or metonymic extensions,
I appeal to the body frame to facilitate an account of such extensions.
2. Analysis of the data
First, in section 2.1,I present a set of compound noun data to illustrate the
phenomenon with Hebrew body part terms; a subset of the compound noun
data is discussed in detail. Next, in section 2.2.1, I cover a set of complex
nouns, more specifically, names for clothing and clothing-like items, which
are derived .from body part terms. Then, in section 2.2.2, I turn to a set of
denominal body part verbs whose parent nouns are instruments. With each
set of data, I claim that knowledge and experience of the body and its parts
2
There are other instances of body part terms used as measure words in Hebrew
including zeret - 'pinky', 'span' and 'ama - 'forearm', 'cubit'. In English we find foot and
finger.
are integral to the creation or formation of the compound or complex
forms.
2.1. Compound nouns
Some extensions of body part terms are far from their source - that is, the
extended sense does not refer to something literally so close to the body.
For example, in English we have head of the department, body of work,
mouth of the river, eye of a needle, and many more. Data of this sort
illustrate the non-corporeal sense of polysemous body part terms and
indicate the domains with which the body part domain interacts. A
complete specification of the body frame would have to include all such
domains and an explanation of the interaction.
One way of illustrating this sort of extension in Hebrew is with
compound nouns, smixut forms, where the first of the two nouns is a body
part term. Compounding is a process available in the language for the
creation of new words; in particular, compounding is a naming device
(Zimmer 1971; Downing 1977). Objects and entities or parts of objects and
entities named in this way are given in Tables la and lb. The claim is that
compound forms of this sort are linked to body part terms in the body
frame because it is experience of the body and knowledge about the body
that has been drawn upon for their formation and understanding.
The data show that there are a number of characteristics of a body part
which can serve as the basis of the extension, including function, size, shape,
and position. Other factors such as configuration and orientation may need
to be considered as well. For example, pi ha-me'ara - 'mouth of the cave'
and pi ha-be'er - 'mouth of the well' are based on the shape of the mouth
in a particular configuration. Forms such as pney ha-binyan (face of the
building) - 'surface of the building' and pney ha-yam (face of the sea) -
'surface of the sea' show the need for different orientations: The former
assumes the canonical vertical orientation of human beings and the latter
assumes a horizontal orientation. In some cases, more than one factor plays
a role in the extension. For instance, in ragley ha-ŝulxan (legs of the table)
- 'table legs', while the function of the legs may be viewed as the major
factor, the extension obtains in part because of their shape.
FUNCTION
šeled ha-binyan (skeleton on the building) - frame of the building

ragley ha-šulxan (legs of the table) - (the) table legs
lešon xaxamim (language of wise ones) - Mishnaic Hebrew
sfat 'em (lip of mother) - mother tongue/native language
SIZE3
guf ha-'avoda (trunk of the-paper) - main part of the paper

guf ha-'oniya (trunk of the-ship) - hull of the ship
guf ha-'inyan (trunk of the-matter) - essence of the matter
SHAPE
roš šum (head of garlic) - bulb of garlic

roš kruv - head of cabbage
roš bacal - (head of) onion
roš dla'at - (head of) pumpkin
pi ha-me'ara - mouth of the cave

pi ha-be'er - mouth of the well
šiney ha-masrek - teeth (of the comb)

šen šum (tooth of garlic) - clove of garlic
Sen sela (tooth of rock) - reef
yad 'avšalom4 (arm of/for Absalom) - Absalom's Monument
lešon yam (tongue of sea) - fjord
beten ha-mandolina (belly of the mandolin) - body of the mandolin
Table la: Compound Nouns
3
As a body part word guf has two senses - it refers to the whole body or a large part
of the body, the trunk. It is the latter sense which figures into the meaning of the
compounds listed here.
4
Theuse of yad - 'arm' for 'monument' shows up in other constructions which are
not smixut forms - e.g. yad la-banim (arm to-the-sons) - 'Monument to the Sons', a
monument to fallen soldiers, and yad va-šem (arm and-name) - 'Yad VaShem', a memorial
monument to the martyrs of the Holocaust.
284 MIRIAM R.L, PETRUCK
POSITION
roš ha-har (head of the mountain) - (the) summit

roš ha-šulxan - head of the table
roš' ha-šura - head of the line
roš ha-mišpaxa - head of the family
roš ha-'ir.(head of the city) - mayor
roš ha-memšala (head of the government) - prime minister
pney ha-binyan (frace of the-building) - surface of the building

pney ha-yam (face of the-sea) - surface of the sea
pney ha-'adama (face of the-land) - surface of the land
pney ha-'ir (face of the-city) - appearance of the city
sfat ha-yam (lip of the sea) •- seashore

sfat ha-nahar (lip of the river) - riverbank
gav ha-har » back of the mountain

gav ha-sefer (back of the book) - spine of the book
gav ha-kise - back of the chair
(katef ha-har - shoulder of the mountain)5

ragley ha-har (feet of the mt.) - foot of the mountain
lev ha-'ir (heart of the city) - center of the city

lev ha-'inyan - heart of the matter
'ecem ha-'inyan (bone of the-matter) - essence of the matter

'acm-o (bone-his) - himself
beten ha-'adama (stomch of the-land) - bowels of the earth

beten ha-'oniya (stomach of the ship) - hold of the ship
Table lb: Compound Nouns
T h e predominance of physical over functional characteristics as

providing the bases of extensions is noteworthy and suggests the i m p o r t a n c e
of perception in m e t a p h o r of this sort (Andersen 1978). In addition, these
data provide evidence for the grounding of metaphorical concepts in
5
katef ha-har - shoulder of the mountain is listed in parentheses because it is a
military topographic term, not necessarily used by the ordinary speaker in everyday
conversation.
physical experience (Lakoff and Johnson 1980a, 1980b) and offer further
justification for the experience-based schematization of word meaning
(Fillmore 1982, 1985; Lakoff 1987). If words acquire new senses through
speakers' experience of and interaction with the world, then it is most
appropriate to invoke experience in the characterization of word meaning.
A closer examination of this phenomenon is warranted in cases where
the semantic motivation for the extension is based on similarity of position.
The data show that roS - 'head' is used for the top part of a variety of
objects and entities; panim - 'face' names the outside surface of different
things; safa - 'lip' can be used for an edge; gav - 'back' for the back of other
things, lev - 'heart', 'ecem -'bone', and beten - 'stomach' name different
kinds of inside parts of other entities. Extensions based on similarity of
position are particularly interesting. When parts of objects are named in
this way, we see that there are entitites and objects which can be viewed as
structured in ways which parallel the structuring of the human body and, in
fact, it is relative position that matters. Some examples from the present
data are given in Table 2.6 The metaphors obtain because of a perceived
similarity between the structure of the body and the structure of some other
object, along with the ability to conceptualize one object in terms of
another.
The clearest illustration of this is har - 'mountain' where a schematiza
tion of the human body has been applied to a mountain to name its parts.
Thus, we findrošha-har (head of the mountain) - '(the) summit', gav ha-har
- 'back of the mountain', katef ha-har - 'shoulder of the mountain', which
is a military/topographic term, and ragley ha-har (feet of the mountain) -
'foot of the mountain'. Note that this schematization assumes the canonical
upright and forward-facing orientation of the human body (Clark 1973).
Note also that har - 'mountain', which has no inherent front/back orient
ation, has been schematized as facing the speaker, as evidenced by the use
of gov - 'back' for the distant part. Thus, knowledge of the way the body is
carried and oriented also figures into the way entities structured in terms
of the human body are named.
6
Except for lešon yam (tongue of the sea) - 'fjord', listed in parentheses, all of the
extensions in Table 2 are based on similarity of position.
har - mountain
roš ha-har (head of the mountain) - (the) summit

(katef ha-har - shoulder of the mountain)
ragley ha-har (feet of the mountain) - foot of the mountain
gav ha-har - back of the mountain
hinyan - building
pney ha-binyan (face of the building) - surface of the building

geled ha-binyan (skeleton of the building) - frame of the building
'adama - land, earth
pney ha-'adama (face of the land) - surface of the land

beten ha-'adama (stomach of the earth) - bowels of the earth
'ir - city
roš' ha-'ir (head of the city) - mayor

pney ha-'ir (face of the city) - appearance of the city
lev ha-'ir heart/center of the city
yam ° sea
pney ha-yam (face of the sea) - surface of the sea

sfat ha-yam (lip of the sea) - (the) seashore
(lešon yam (tongue of the sea) - fjord)
Table 2: Parallel Structuring
Some of the cases in Table 2 involve an external surface and an internal

part (or set of parts) of the body, where it is clear that the structuring of
these entities parallels the structuring of the body. This is so with hinyan -
'building' and 'adama -'land,earth', where we have the term panim - 'face'
used in the sense of 'surface'. In each we also find an internal part of the
body used to name an inside part of the object or entity. Med ha-binyan
(skeleton of the building) - 'frame of the building' names an inside set of
parts of a binyan. The horizontal parts of the mid and upper sections of a
building are comparable to the ribs of a skeleton. Here, function is an
important factor as well: A skeleton supports the body and frames support
buildings, beten ha-'adama (stomach of the earth) - 'bowels of the earth'
names an inside container-like part of 'adama. In both of these cases, it
seems that it is first necessary to establish the extension from 'face' to

all-over area 'surface' for the parallel structuring to obtain. The 'face' to
'surface' schematization also figures into the naming of parts of 'ir andyam.
Thus, it seems that it is the face/surface which determines the general
appearance of the body/object.
The so-called "parts" of 'ir - 'city' require some further explanation if we

are to accept the claim that the structure of 'ir parallels that of the body.
In a certain sense,pney ha-'ir the 'appearance of the city' is how a city looks
from the outside, its all-over surface. Although lev ha-'ir - 'heart/center of
the city' doesn't have to be at the geographical center of a city, it is unlikely
to designate an area too near the outside of the city, where one would view
its appearance. So, it is easy to see lev ha-'ir as an inside "part" of a city, in
particular a vital part where things happen. Regarding roš ha-'ir (head of
the city) - 'mayor', we need to refer to the many roS examples in Table 1
to see evidence of the schematization of the body as top-to-bottom
hierarchy and the transfer of ros as the top part of that hierarchy to other
hierarchies.7 Consider roS ha-mišpaxa - 'head of the family' and roS
ha-memšala (head of the government) - 'prime minister' which show that
artificially constructed bodies also have heads. The upward orientation of
the human body serves as a model for representing a hierarchic organiza
tion with a roS at the top.
yam - 'sea' also requires a bit of work and a bit of imagination. If the
face is the surface of the sea, as suggested by pney ha-yam (face of the sea)
- 'surface of the sea', then the lips are the edge of the sea, as suggested by
sfat ha-yam (lip of the sea) - '(the) seashore', when the mouth is open wide
enough to extend over the entire face. The tongue, part of the mouth, and
hanging out of it, is the fjord. With Mon yam - 'fjord', the extension is
based on shape. Schematization of the face in a particular configuration is
required. Thus, knowledge of the body and the functioning of different
parts of the body plays a role in the naming of entities outside of the body.
Compound nouns exemplifying the semantic factor of position have
been discussed at length because they provide the clearest illustration of
mappings, albeit partial in some cases, from one domain, the body, to other
7
Infact ros ha-'ir is not a part of 'ir in the same way that, for example, ros ha-har is
part of har, although roS ha-'ir may well be "attached", emotionally or psychologically to
'ir.
concrete domains. More generally, the compound noun data are impor
tant because they exemplify an extremely basic level of metaphor which
serves as the basis of further extensions. The conceptual underpinnings of
these extensions are provided by and structured in the body frame. As we
have seen, different characteristics can serve as the basis of an extension
and other factors such as orientation and configuration need to be
considered. Characterizing these compound nouns in terms of the body
frame makes it possible to incorporate all of this information in the
description of their meanings.
2.2. Complex nouns and verbs
Now let us turn to complex nouns and verbs derived from body part terms.
The formal apparatus of Hebrew derivational morphology is such that the
creation of vocabulary from native material is a relatively simple and
straight-forward matter. In principle, once the consonantal structure of an
existing word is determined it can be used in any of the miškalim, nominal
patterns, or hinyanim, verbal conjugations. For the present discussion, a new
word is created as a result of the body part term having undergone one of
the derivational processes in the language. Two kinds of examples - derived
nouns based on body part terms and derived verbs - are presented.9
2.2.1. Complex nouns

The complex nouns listed in Table 3 name articles of clothing or clothing
like entities and can be characterized as consisting of a body part term and
a suffix, where there are four possible suffixes: -a, -on, -it, -iya.10 11
8
InPetruck (1988), there is a discussion of mappings from the body to the domain
of location, providing a detailed example of mapping from a concrete domain to an
abstract domain.
9
I have limited the discussion in this paper to nouns and verbs. It should be noted
that there are also adjectives, adverbs, prepositions and conjunctions derived from body-
part terms in Hebrew.
10
Petruck (1987) covers other groups of complex nouns derived from body part terms
including names for diseases, (e.g. damernet - 'hepatitis' from dam - 'blood'), instruments
(e.g. yadit - 'handle' from yad - 'arm/hand', and abstract nouns (e.g. resit - beginning from
roš - 'head').
As already mentioned, it is not particularly surprising to find that there

are many clothing words derived from body part terms given the purpose
of clothing, to cover and protect the body, and given the physical proximity
of clothing to the body. Obviously, the size, shape, and structure of the body
part that an article of clothing covers determines a lot about that article of
clothing.
bitna - lining beten - belly, stomach

kfafa - glove kaf - hand
cavaron - collar cavar - neck
karsolit - gaiter karsol - ankle
gufiya - undershirt guf - body, trunk
xaziya - bra xaze - chest, breasts
motniya - vest moten - waist
ktefiya - shoulder trap katef - shoulder
micxiya - visor mecax - forehead
'ecba'on - thimble 'ecba - finger
'ozniya - earphone 'ozen - ear
Table 3: Clothing and Clothing-Like Items
The important role of metaphor an metonymy in the creation of new

words has been noted above. The general principle on which the extension
from body part to clothing is based is metonymy; within this general
principle, we also see metaphoric transfer. Thus, for example, the transfer
from cavar - 'neck' to cavaron - 'collar' is metonymic, as is the transfer from
kaf - 'hand' to kfafa - 'glove'. However, the transfer of beten - 'stomach' to
bitna - 'lining' is metaphoric. Both name inside parts; the shared character
istic of "inside" allows the metaphoric transfer. Note that bitna - 'lining'
names something which isn't necessarily worn on or near the body part.
Kfafa - 'glove' is noteworthy since it reflects an older stage of Hebrew
when the monolexeme kaf -'hand', as opposed to the compound form kaf
Hebrew morphophonemics is such that in certain circumstances the addition of a

suffix can result in a change of the internal vowel or vowels of the base word. This is the
case with a number of the clothing words.
yad (spoon of arm) - 'hand', named a part of the body.12 Today, the
monolexeme kaf only means 'spoon'; kfafa is a cover for the hand, not for
a spoon, cavaron - 'collar' and karsolit - 'gaiter' are straight-forward in that
they name clothing worn on or near the body part from whose name the
term is derived; in both cases the garment is worn around the body part.
Like hitna - lining', gufiya - 'undershirt'and xaziya - 'bra' are based on
polysemous body part terms. With both guf - 'body, trunk' and xaze - 'chest,
breasts', the two senses of each word can be characterized as whole and
part of that whole. In each case, the derived form is based on the "part"
sense of the word. In principle, gufiya could have meant 'wetsuit' and xaziya
could have meant 'undershirt', motniya - 'vest' (worn on the chest) doesn't
seem to have anything to do with moten - 'waist', although a typical vest
reaches to the waist.13 Although micxiya - 'visor' is worn on the forehead,
its purpose is to protect the eyes, ktefiya - 'shoulder strap' seems straight
forward as it names the part of a garment worn on the shoulder. However,
shoulder pads and epaulettes are also worn on the shoulder and the term
ktefiya does not refer to either of these items. Note that kotfot - 'epaulettes'
is heard in the army.
While they are not really articles of clothing, 'ecba'on - 'thimble' and
'ozniya - 'earphone' are also things which are worn - 'ecba'on is worn for
protection; it covers the body part all around. Given the size and shape of
a thimble, it seems possible to analyze 'ecba'on comparable to cavaron -
'collar', where -on is one of the clothing word suffixes, 'ozniya is different
because it goes inside the body part, in contrast with, say, gufiya and xaziya.
Clothing terms are linked to body part terms in the body frame because
it is knowledge about the body - the physical characteristics of its parts -
and what is done with the body - that it is customary to cover the body in
a particular way - that is drawn upon in the formation or creation of such
words. The data demonstrate this in a profound way: In Hebrew names for
clothing and parts of clothing are derived from body part terms. I am not
saying that all clothing terms have to be morphologically related to body
part terms. Rather, the fact that there are such clothing terms provides
support for the connection to the body frame.
See, for example, Psalms 71:4 and Job 29:9.
Perhaps motniya is a caique from the British English waistcoat.

2.2.2. Denominal verbs

Hebrew denominal body part verbs whose parent nouns are instruments are
listed in Table 4.14 This small set of data includes verbs from three
different conjugations, which can be characterized by the following
morphological patterns: C1aC2aC3; C1iC2(C2)Ce3; hiC1C2iC3, where C stands
for any consonant. In these data, the consonants of the body part term have
been extracted and used in one of the verbal patterns.15 Each of the verbs
in Table 4 designates an action performed by the possessor of the body
part. The kind of action performed can depend upon what the anatomical
or biological function of the body part is, what the body part can do given
its size, shape, and structure, as well as what people believe the body part
does.
naxar - snore nxir - nostril

pana - turn panim - face
'inpef - talk nasally 'af - nose
'iyen - look into 'ayin - eye
'izen - balance 'ozen - ear
ziyen - fuck zayin - penis
yida - throw yad - arm, hand
rigel - spy regel - leg, foot
'if ef - blink 'afaf - eyelid
girger - gargle gargeret - Adam's apple
he'ezin - listen 'ozen - ear
hicbia - riase finger, 'ecba - finger
hilšin - tell on lašon - tongue
Table 4: Body Part as Instrument
There is a metonymic relationship between the body part as instrument
14
Petruck (1987) also covers other groups of complex verbs derived from body part
terms including those where the body part is an affected object (e.g. 'akav - 'follow' from
'akev - 'heel') or a product (e.g. hištin - 'urinate' from Men - 'urine').
Apparent deviations from these patterns are a result of predictable changes. Thus,
for example, the second vowel in yida, whose underlying consonantal root is YDH, is a
rather than the expected e. Although the glottal fricative is not realized in contemporary
Hebrew pronunciation, there are still surface phonetic manifestations of its underlying
existence. The underlying form /yideh/ is realized as [yida] because in word-final position,
a glottal lowers the preceding vowel.
and the work to which the instrument can be put, which is indicated in the
denominal verb. Some of the verbs designate actions which are the body
part's anatomical function. Examples of this are ziyen - 'fuck' and 'if'ef -
'blink'. Other verbs designate actions that are more conscious, deliberate or
purposeful than their regular/typical function. For example, 'iyen - 'look
into' is not just to use the eye for seeing, but rather to see consciously, with
a purpose in mind. Similarly, yida - 'throw', which is restricted in use,
involves the use of the hand for an action with some force, throwing, and
not for example, giving, grasping, or handling. Likewise, with hicbia - 'raise
finger, vote' and hilšin - 'tell on', the actions are purposeful. A finger is
raised to get attention, as for example, in a class; the purpose is especially
clear in the 'vote' sense of the word. The tongue is used for talk, but telling
on (someone) is talk with a specific purpose and seems to include a value
judgement. The distinction can best be seen with 'izen - 'balance' and he
he'ezin - 'listen'. The verb 'izen reveals folk knowledge of the anatomical
connection between the inner ear and a person's sense balance, the inner
ear being the instrument which controls balance. 16 For the person whose
ear is doing the balancing, this act is involuntary. In contrast, listening is a
deliberate and purposeful act. In addition, these verbs differ from each
other in that izen designates an action physically outside of the body.
naxar - 'snore', 'inpef - 'talk nasally', and rigel - 'spy' are particularly
interesting because the action designated doesn't really involve the body
part. It is not the nostrils (or even the nose) that "snores", but rather the
soft palate. The vibrations of the soft palate, which can also create some
movement of the nostrils, cause the sound that we associate with snoring.
Physiologically, talking nasally only involves the nose in a negative way - i.e.
air doesn't pass through the nose, usually because the nasal passages are
blocked. The verbs suggest that people believe that the body part
mentioned is the instrument performing the action, more specifically, that
the body part is the point of origin of the action. And it is that belief which
16
For a detailed discussion of the bodily basis of balance, see Johnson 1987: 74-96.
The physiological and anatomical relationship between the inner ear and balance is known
by ordinary speakers. Indeed it was a linguistically sophisticated native speaker of Hebrew
who pointed out the relationship between 'ozen - 'ear' and the verb 'izen - 'balance'.
However, it should be noted that the histories of these two words show that they are from
different sources, which becomes apparent when considering the comparable terms in
Arabic. Whereas the Arabic body part term 'u un - 'ear' is cognate with the Hebrew term
'ozen, the Arabic verb wazana - 'weigh' is not cognate with the Hebrew verb 'izen.
is lexicalized in the verb. Although the verb rigel - 'spy' incorporates the
notion of going on foot, spying involves using one's eyes much more so than
one's legs or feet. In this case, the history of the word is important.
Consider the Biblical phrase le-ragel 'et ha'arec (to-spy OBJ the-land) - 'to
spy the Land', where the spying actually included more physical involvement
of legs/feet.
girger - 'gargle', related to gargeret - 'Adam's apple' seems to incorporate
the fact that the Adam's apple is the visible external manifestation of the
throat which is an internal part. In addition, for a person who has a
protruding Adam's apple, it is possible to see its vibration while gargling.
Finally, with pana - 'turn', which is clearly related to partim - 'face', the verb
incorporates the idea that turning one's face counts as turning. If someone
says Turn to me, turning one's face toward them is appropriate. It is
possible to comply with the request by turning your face without turning
your body, but not vice versa.
These denominal verbs are linked to body part terms in the body frame
because their formation and understanding depend upon knowledge and
beliefs about the body - its physical structure, how it functions, and how
speakers experience it altogether. Again, not all verbs invovlving an action
of a body part are derived from a body part term. However, the existence
of these denominal body part verbs provide support for the connection to
the body frame.
3. Concluding remarks
We have seen that body part terms are used in the naming of (1) that
which is conceived of in terms of the body, as in the compound nouns used
to designate parts of objects or entities outside the body and (2) that which
is close to the body - either actual physical contact, as with clothing words,
or physical actions that parts of the body perform, as with denominal verbs
whose parents are instruments. The claim is that these words are linked to
the body frame not simply because of the obvious lexical or morphological
relationship between the compound or complex forms and the body part
terms from which they are derived. They are linked to the body frame
because it is knowledge and experience of the body and its parts which
provide the conceptual underpinnings for their formation and understand
ing.
In this paper, I have demonstrated the feasibility of appealing to the

same apparatus, the frame, to account for certain types of compound and
complex words based on body part terms. Such data, which draw upon
knowledge and experience of the body and its parts, highlight different
parts of the body frame. Additional data of various kinds will have to be
considered to provide a full account of other relevant lexical sets, thus
revealing other aspects of the body frame. For example, in Hebrew some
function words are derived from body part terms. For instance, the
prepositions 'al-pney- (on-the face-of) - 'on', bi-fney- (in/with-face-of) -
'against' and li-fney- (to-face-of) - 'before' are derived from the body part
word partim - 'face'. What aspects of the body frame are brought into focus
with these function words? There are also phrases, such as roš xalul (head
hollow) - 'empty-headed' (figuratively, 'stupid'), roš tov (head good) -
'smart', and 'al ha-roš šel-i (on the-head of-mine) - 'on my head'/Tm
responsible' that make use of body part terms. What knowledge and
experience of the body and its parts are drawn upon in such data?
Hebrew body part terminology has been shown to be a particularly rich
source for the investigation of the characterization and representation of
meaning in terms of knowledge and experience. In-depth studies of
individual domains in many different languages will lead to a better
understanding of lexical structure and semantic structure in the lexicon
more generally.
References
Andersen, Elaine S.
1978 "Lexical universals of body-part terminology". In J.H. Greenberg, CA.
Ferguson, & E.A. Moravcsik, (eds.), Universals of human language, Vol. 3
(Word Structure). Stanford: Stanford University Press, 335-368.
Clark, Herbert
1973 "Space, time, semantics, and the child". In T. Moore (ed.), Cognitive
development and the acquisition of language. New York: Academic Press, 27-
63.
Claudi, Ulrike & Bernd Heine

1986 "On the metaphorical base of grammar". Studies in Language 10.297-335.
Downing, Pamela
1977 "On the creation and use of English compound nouns". Language
53:4.810-842.
1975 "An alternative to checklist theories of meaning". Proceedings of the First
Annual Meeting of the Berkeley Linguistics Society, 123-131.
1976a "Frame semantics and the nature of language". Annals of the NY Academy
of Sciences 280.20-32.
1976b "The need for frame semantics within linguistics." Ms. Typescript of lecture
given at Computational Linguistics (COLING) meeting. Ottawa.
1977 "Topics in lexical semantics". In Roger Cole (ed.), Current issues in linguistic
theory. Bloomington: Indiana University Press, 76-138.
1978 "On the organization of semantic information in the lexicon". In Donka

Farkas et al. (eds.), Papers from the Parasession on the Lexicon. Chicago:
Chicago Linguistic Society, 148-173.
1982 "Frame semantics". In Linguistics in the morning calm ed. by The Linguistic
Society of Korean. Soeul: Hanshin, 111-137.
1985 "Frames and the semantics of understanding". Quaderni di Semantica

6:2.222-254.
Fillmore, Charles J. & B.T. Atkins

1992 "Towards a frame-based organization of the lexicon: The semantics of RISK
and its neighbors". In Adrienne Lehrer & Rva Kittay (eds.), Frames, fields,
and contrasts: New essays in semantics and lexical organization. Hillsdale:
Lawrence Erlbaum, 75-102.
Heine, Bernd & Ulrike Claudi

1986 "The metaphorical base of grammatical categories in Ewe (West Africa)". In
Joshua A. Fishman et al (eds.), The Fergusonian impact. Vol. 1, From
phonology to society, 367-375.
Heine, Bernd, Ulrike Claudi & Friederike Hunnemeyer

1991 Grammaticalization: A conceptualframework. Chicago: University of Chicago
Press.
Heine, Bernd & Mechthild Reh

1984 Grammaticalizaton and reanalysis in African languages. Hamburg: Helmut
Buske.
Jakobson, Roman
1971 "Two aspects of language and two types of aphasic disturbances". In Roman
Jakobson & Morris Halle, Fundamentals of language. The Hague: Mouton,
69-96.
Johnson, Mark
1987 The body in the mind. Chicago: University of Chicago Press.
Lakoff, George
1987 Women, fire and dangerous things. Chicago: University of Chicago Press.

1980a Metaphors we live by. Chicago: University of Chicago Press.
1980b ''The metaphorical structure of the human conceptual system". Cognitive

Science 4.195-208.
Petruck, Miriam R.L.

1986 Body part terminology in Hebrew: A study in lexical semantics. Unpublished
PhD dissertation. University of California, Berkeley.
1987 "Morphologically derived 'extensions' of Hebrew body part terms." Ms.

Paper read at Symposium on Hebrew and Arabic Word Formation,
Univeristy of Massachusetts, Amherst, Massachusetts.
1988 "Metaphorical and morphological extensions of PANIM - 'face'." Ms. Paper

read at Linguistics Society of America meeting, New Orleans, Louisiana.
forthcoming
Frame semantics and structure in the lexicon: The case of Hebrew body part
terminology.
Sweetser, Eve E.
1986 "Polysemy vs. abstraction: Mutually exclusive or complementary?" Proceed
ings of the Twelfth Annual Meeting of the Berkeley Linguistics Society, 28-538.
1988 "Grammaticalization and semantic bleaching". Proceedings of the Fourteenth

Annual Meeting of the Berkeley Linguistics Society, 389-405.
1990 From etymology to pragmatics: Metaphorical and cultural aspects of semantic

structure. Cambridge: Cambridge University Press.
Traugott, Elizabeth
1982 "From propositional to textual and expressive meanings: Some semantic-
pragmatic aspects of grammaticalization". In Winfred P. Lehmann & Yakov
Malkiel (eds.), Perspectives on historical linguistics. Amsterdam: John
Benjamins, 245-271.
1988 "Pragmatic strengthening and grammaticalization". Proceedings of the

Fourteenth Annual Meeting of the Berkeley Lingusitics Society, 406-416.
1989 "On the rise of epistemic meanings in English: An example of subjectifica-

tion in semantic change". Language 65:1.31-55.
Zimmer, Karl E.
1971 "Some general observations about nominal compounds". Working Papers on
Language Universals 8:C1-21.
The conceptual basis
of
performativity
Jef Verschueren
Belgian National Fund for Scientific Research
and
University of Antwerp (IPrA Research Center)
1. Introduction.
It was no doubt Charles Fillmore's article on "Verbs of judging" that first

convinced me of the need to approach lexical semantics in pragmatic terms
(where 'pragmatics' carries its fullest interdisciplinary weight as a cognitive,
social, and cultural perspective on language and communication). This
article is intended to suggest a solution to a linguistic-philosophical problem,
the problem of performativity, on the basis of what could ultimately be
characterized as a 'Fillmorean' approach to the meaning of a certain class
of verbs used to describe forms of linguistic action.1
1
It is not only for this reason that the present paper is suitable for insertion in a
Festschrift in honor of Charles Fillmore. In fact it could and should have been written as
early as 1982 when I presented the basic ideas as a lecture in a course entitled
"Metapragmatics: Or, the lexicalization of linguistic action" which I was co-teaching with
Charles Fillmore (at his suggestion), John Gumperz, and John Searle at the University of
California at Berkeley. No doubt, the present formulation owes a lot to discussions with
my co-teachers and the participants in this course, as well as with more or less recent
audiences in Amsterdam, Budapest, Duisburg, Ljubljana, and Vienna (where I was invited
to talk about the subject by, respectively, Rob Grootendorst and Lachlan Mackenzie,
Ferenc Kiefer, René Dirven, Igor Žagar, and Wolfgang Dressier). Special thanks are due
to John Searle whose "How performatives work", the general drift of which I violently
disagree with, made me decide that I should commit my thoughts to paper, and to
Masayoshi Shibatani and Sandra Thompson, whose decision to compile a Festschrift made
300 JEF VERSCHUEREN
The British philosopher John Austin is generally credited with the

'discovery' of a class of utterances containing a verb that names the speech
act type which the utterance is used to perform. Austin (1962) called them
'explicit performatives'. Most of the literature on the subject, including John
Searle's recent contribution (Searle 1989), relies on the simpler expression
'performatives'. The phenomenon to be explained is then referred to as
'performativity'.2
The novelty of the topic, for philosophers, was in the interest it
engendered in matters beyond logic, truth and falsity. For linguists, on the
other hand, it provided a useful point of reference to guide renewed
attention for questions of meaning and usage. Only recently has attention
been drawn to the fact that performative utterances were debated by
linguists as early as the late 19th century.3 Thus Škrabec (1903: 555)
identifies a Praesens effectivum, "which not only names the action, but which
even performs it at the same time by uttering the verb in question."4
In this article I will briefly present the range of phenomena covered by
the label 'performativity' (section 2), describe the 'problem of
performativity9 as stated from åkrabec (1903) to Searle (1989) (section 3),
review Searle's (1989) approach to the problem (section 4), propose an
alternative analysis (section 5), and formulate a hypothesis about the way
in which the alternative could even solve Škrabec's problem (section 6).
me do so. In its final stages, this paper has benefited from comments by Sandra
Thompson, Jan Blommaert, Gino Eelen, and Michael Meeuwis.
2
The extensive philosophical and linguistic literature on performativity, heavily
concentrated in the 1970's, includes: Åqvist 1972, Bach 1975, Bach & Harnish 1979,
Cornulier 1975, 1980, Davis 1964, Fauconnier 1979, Ginet 1979, Grewendorf 1979,
Grodzinski 1980, Heal 1974, Hedenius 1963, Lemmon 1962, McCawley 1977, Récanati
1980, 1981, Sadock 1974, Urmson 1977, Walker 1969, Warnock 1973, Yanovsky &
Maclaran 1980, Zaslawsky 1979, Zuber 1981. For more references, see Nuyts &
Verschueren 1987.
3
See Keck & Stubbs 1984, Hinrichs 1985, Leeuwen-Turnovcová 1986, and Žagar 1988.
4
"Ich habe in meiner eben erwähnten Polemik das Wort 'effectivum' in einem etwas
verschiedenen Sinne gebraucht, nämlich von einem Praesens, das die Handlung nicht nur
bezeichnet, sondern eben durch das Aussprechen des betreffenden Verbums zugleich vollzieht:'
(Italics correspond to the translated portion of the quote in the main body of the text; all
translations in this article are mine.)
THE CONCEPTUAL BASIS OF PERFORMATIVITY 301
2. The phenomenon
The canonical form of an 'explicit performative,' as introduced by Austin

(1962) is as follows:
(1) I (hereby) V (simple present, indicative, active) ...
Typical examples are:
(2) I (hereby) abdicate
(3) I apologize for stepping on your toes
(4) I promise to come on Wednesday
(5) I order you to leave the room
Note that the simple present verb form, in these examples, should refer to
the moment of speaking, and not - as it does in other cases - to a general
or habitual truth.
This canonical form is subject to a certain range of cross-linguistic
variation, depending on specific characteristics of individual languages. Thus
Hungarian not only offers us the completely analogous:
(6) Kér-em a segît-ség-ed-et
[Ask-I the help(V)-Nominalizer-your-Accusative]
(I ask your help/I ask you to help me)
but also:
(7) Kér-lek, segît-s
[Ask-I(Subject)you(Object)? help-Imperative]
The latter form of expression is possible only because of the existence of
a verbal suffix which combines the first person singular as subject with the
second person singular as object. (See Verschueren 1980 for more details.)
There may be no reason at all to regard (7) as less 'canonical' than (6)
for Hungarian. However, most languages also have forms of expression
which are functionally quite close to examples such as (2) through (5),
though they differ from the schema in (1) along one or more significant
parameters. The following examples represent some of the types discussed
in the literature.
A first type of non-central cases of performatives, which we may label
'semi-performatives' for convenience's sake, makes use of different forms
of the verb. A present continuous tense is used in (8), a first person plural
302 JEF VERSCHUEREN
form in (9), a third person singular form in (10), and the passive in (11)
through (13):
(8) I am asking you to do this for me Henry, I am asking you to do it
for me and Cynthia and the children. (Searle 1989)
(9) We pledge our lives, our fortunes and our sacred honor. (Searle
1989)
(10) Le porteur déclare être majeur. (Cornulier 1980)
(The holder [of this ticket] declares that he/she is over 18 years
old.)
(11) Passengers are hereby advised that all flights to Phoenix have been
cancelled. (Searle 1989)
(12) Passengers are warned not to lean out of the window.
(13) You are dismissed.
Of these examples, (8) is the most disputable case of a performative: it
does not even contain the content of the request referred to with 'asking,5
which indicates that the act is a description of the request after it has been
made by other (linguistic) means, rather than the making of a request itself.
This does not mean that the present continuous cannot carry a
performative meaning; but if it does, at least repetition of a foregoing act
seems to be involved. Thus (14):
(14) I'm ordering you to get out!
is likely to occur after the 'primary' performative "Get out!" has failed to
yield the desired result.
A second class of non-central cases of performatives contains verbal
modifiers of various kinds:
(15) Yd guess there were twenty of them.
(16) I would say that this is completely wrong. (McCawley 1977)
(17) I would estimate/reply/etc. that ... (McCawley 1977)
(18) I will offer you the following alternative: ... (McCawley 1977)
(19) I will (simply) answer you that ... (McCawley 1977)
(20) Yd like to ask you to change your plans.
(21) Yd like to announce the engagement of ... (McCawley 1977)

(22) Let me ask you what bothers you. (McCawley 1977)
(23) Let me conclude/respond/etc. that ... (McCawley 1977)
(24) Let us say that these are semi-explicit performatives. (Cornulier
1980)
(25) Allow me to call you a bastard. (Cornulier 1980)
Though such verbal modifiers already add to the semantic and
pragmatic complexity of the performatives, there is a third set of non-
central cases which, for lack of a better description, contains a variety of
'complex constructions':
(26) That's a squirrel, shall we say. (Cornulier 1980)
(27) This is to thank you for ... (Cornulier 1980)
(28) This is to inform you that ... (Cornulier 1980)
(29) (This is) just to say hello [e.g. on a postcard]. (Cornulier 1980)
(30) I am writing/calling to tell you that ... (Cornulier 1980)
(31) If I dare speak here, it is to express my profoundest gratitude ...
(Cornulier 1980)
Finally, even constructions with nouns rather than verbs have been
proposed as (semi-)performatives:
(32) I'll come and see you next week, and that's a promise. (Searle
1989)
(33) My question is, what do you mean by ... (Cornulier 1980)
(34) You will find in this letter my best wishes for the New Year.
(Cornulier 1980)
The least objectionable of these may be the most complex one, example
(34), which is itself a New Year's wish, though literally it refers to a wish
to be found elsewhere in the letter (though it may not be found at all). To
examples (32) and (33) the same comments apply as to (8): In both cases
it can be argued that the nouns describe, rather than perform, acts which
precede (as in (32)) or follow (as in (33)).
304 JEF VERSCHUEREN
3. The problem
The original, pre-Austinian formulation of 'the problem of performativity'

was a logico-grammatical one, an attempt to match grammatical behavior
with functional and logical properties. Thus Škrabec says about the cases
of praesens effectivum which he had observed:
"Durch das Aussprechen der Worte: 'ich danke, verspreche, gelobe, empfehle mich'
sind die betreffenden Akten vollzogen. Das Aussprechen des Wortes mag zwar eine
gewisse, wenn auch noch so kurze Zeitdauer erfordern, aber daran denkt der
Sprechende nicht und es kommt darauf auch gar nicht an, der Akt ist als momentan
aufzufassen. Es ist nun sonderbar, das diesen momentanen Akt die meisten slavischen
Sprachen durch Aussprechen des Praesens eines imperfektiven Verbums vollziehen.
Eine Ausnahme macht das Slovenische; es scheint in demselben nämlich bei den
imperfektiven Verben das Gefühl für das Unvollendete der Handlung zu lebhaft zu
sein, als dass man ohne weiteres sich ebenfalls so ausdrücken dürfte. [...] es ist aber
schwer zu begreifen warum das logisch richtige gegenüber dem logisch falschen ein
Verderbniss sein sollte." (1903: 555)
[By uttering the words 'I thank, promise, pledge, offer my services,' the corresponding
acts are performed. Uttering the word may occupy a certain, even if very brief, time
span, but the speaker does not think of that and it also does not matter. The act is
to be regarded as momentaneous. It is strange though, that this momentaneous act
is in most Slavic languages performed by uttering the praesens of an imperfective
verb. (...)
An exception is Slovenian; it seems as if in this language the connotation of the
incompletedness of the action associated with the imperfective verb is too prominent
for this form of expression to be used as such. [...] but it is difficult to understand
why the logically correct should be a deviation vis-à-vis the logically false."
In other words: explicit performatives can only be viewed as momentaneous
acts, without duration; therefore it seems illogical that in most Slavic
languages, with few exceptions, imperfective verbs are used.
Replacing the term praesens effectivum with Koinzidenzfall5 many
5
"Diesen Fall, in dem die Handlung, von der gesprochen wird, eben im Aussprechen
der betr. Form besteht, habe ich den Koinzidenzfall genannt, weil in ihm Sprechen und
Handeln tatsächlich identisch sind, nicht aber weil beides zeitlich zusammenfiele, denn das
letztere ist ja doch natürlich bei jedem aktuellen Präsens der Fall." (Koschmieder 1935:
288)
[I have called this case, in which the action of which one speaks consists in the
uttering of the corresponding form, Koinzidenzfall because speaking and acting are really
identical, not just because they coincide in time, since that is of course the case with every
'actual present.']
years later, Koschmieder (1935) formulates the same problem from a

different point of view which emphasizes the possibility of the exceptions
to the imperfective rule:
"Warum aber kann in der Koinzidenz die Perfektivität eine Gegenwart bezeichnen?
Die Koinzidenz von Sprechen und Tun ist nur in der Gegenwart denkbar, da in
Vergangenheit oder Zukunft nur über etwas gesprochen werden kann. [294] Wenn
ich aber in der Gegenwart ausdrücken will, dass das Aussprechen meiner Worte eben
die Handlung 'ist', von der ich spreche, so drücke ich das 'Eintreten' dieser Handlung
in der Gegenwart aus, und das kann ich in der Gegenwart, weil eben das Aussprechen
des betr. Wortes das Wikrlichkeitwerden der Handlung ist. Es stehen sich in diesem -
- aber nur in diesem Falle -- 'Währen' und 'Eintritt' einer Handlung in der
Gegenwart gegenüber wie sonst in Vergangenheit und Zukunft in jedem Falle. Wenn
also die Koinzidenz den 'Eintritt' in der Gegenwart bezeichnen soll, so ist die
Perfektivität hier durchaus am Platze. Zu erklären ist hier dann die Imperfektivität,
die in verschiedenen Sprachen auftritt." (1935: 294-295)
[Why is it that in the Koinzidenz the perfective can refer to the present? The
coincidence of speaking and acting is only in the present imaginable, since in the past
or the future it is only possible to speak about something. But when I want to express
in the present that the uttering of my words 'is' at the same time the action, then I
express the 'entering' of this action into the present, and I can do that in the present,
since the uttering of the word in question is itself the realization of the action. In this
case - but not in any other case - the continuation and the entrance of an action are
opposed to each other in the present, just as they always are in the past and the
future. If the Koinzidenz must designate 'entrance' into the present, then the
perfcctive should be the norm. Therefore, what needs explaining is the imperfective
which is used in several languages.]
Again the use of the imperfective is declared to be illogical. Koschmieder's

formulation seems to suggest that the perfective is not really exceptional
and not restricted to one particular Slavic language such as Slovenian. Let
us therefore assume that the perfective form is at least sometimes possible
in an explicit performative formula, even though there is some evidence
that at least in present-day speech it is truly exceptional.
Towards the end of this article we'll come back to the problem as
described by Skrabec and Koschmieder, with a hypothesis that would
predict the predominance of imperfective performatives, thus showing their
formulation to be misguided. But let us first consider a completely different
definition of the problem of performativity, the one offered by Searle
(1989).
For Searle, the basic problem to be explained is that I can promise to
come and see you by saying "I promise to come and see you," whereas one
306 JEF VERSCHUEREN
cannot fry an egg by saying "I fry an egg." In more philosophical terms: how
do we explain the existence of a class of sentences the meaning of which is
such that we can use them to perform the act named by the main verb,
simply by uttering the sentence (which literally says that we are performing
the act in question)? How do we explain the existence of a class of verbs
which lends itself to such usages? How do we explain that many other verbs
which look semantically similar, cannot be used in the same way (so that,
e.g., I cannot boast by saying "I boast that ..." or insinuate by saying "I
insinuate that...")? Is there anything in the meaning of the verbs that makes
the difference?
In our own terms: does the phenomenon of performativity have a
conceptual basis? Are the observed facts merely determined by convention?
Or are there any deeper pragmatic (i.e. cognitive and/or social) constraints
involved? Before venturing upon a response to these questions, we shall
first scrutinize Searle's account.
4. Searle's account of performativity
Two strong points of Searle's (1989) account of performativity deserve to

be mentioned. First, he argues that performatives (contrary to some
accounts given earlier; e.g. Bach & Harnish 1979, Récanati 1980) cannot be
regarded as indirect speech acts. In other words, their performative meaning
is not derivable from an assertion by means of Gricean implicatures. We
agree, since such an explanation is counterintuitive in view of the notion of
'indirectness/ which refers to instances of language use where one means
more than one literally says. Indeed, in explicit performatives one means
exactly what is literally expressed.
Second, we agree that—if one accepts the usefulness of the
classification—explicit performatives can reasonably be regarded as
'declarations/ instances of the class of speech acts which bring about
changes in the world. Searle distinguishes between the following categories:
(35) i. - The meeting is adjourned
- I pronounce you husband and wife
- War is declared
ii. - I promise to come and see you
- I order you to leave the room
- I state that it is raining
iii. - Let there be light

Categories (i) and (iii) are 'extra-linguistic declarations,' creating extra-
linguistic facts, while category (ii) consists of linguistic declarations,'
creating linguistic facts. The validity of this distinction depends, of course,
on the definition of linguistic vs. extra-linguistic reality. Searle further
distinguishes between categories (i) and (ii) on the one hand, which are
performatives, and (iii) which contains non-performative declarations. Thus
not all declarations are performatives, though all of them can be replaced
by them (e.g. "Let there be light" may be expressed as "I decree that there
be light").
So far so good. There are, however, quite a few problems with the
overall pattern of argumentation. A first one has to do with the notion of
ambiguity. One of Searle's ways of clarifying the nature the problem of
performativity is the following:
"Furthermore, how can one and the same unambiguous sentence have both a literal
performative and a literal nonperformative use?" (p. 538)
This question is asked in relation to examples such as (36):
(36) I order you to leave the room
which may be either an order to leave the room or an assertion in response
to a question such as (37):
(37) What do you do if I make too much noise?
Searle's question is fundamentally misguided in that (36) does not have one
unambiguous meaning with two usages. Though the verb to order has one
meaning, one must admit that sentence (36) has different meanings in the
two contexts described, unless one were to exclude temporal deixis from
meaning: in its assertion sense, the simple present in (36) describes a
general truth, whereas in its order sense the moment of speaking is referred
to.
Second, Searle proposes to reverse the relationship between the
performative sense of explicit performatives and their assertive sense, which he
rightly criticized. He suggests that the assertive nature of explicit
performatives can be derived from their performative sense (i.e. from the
'declaration'):
"Declarations, by definition, make their propositional content true. That's what a
successful declaration is. It is an utterance that changes the world in such a way as
to bring about the truth of its propositional content. If I say, "The meeting is
308 JEF VERSCHUEREN
adjourned/' and succeed in my declaration, then I make it the case that what I said
is true; similarly with "I order you to leave the room." But it is important to
emphasize, contrary to the hypothesis that I considered earlier, that the truth of the
statement derives from the declarational character of the utterance and not
conversely. In the case of performative utterances, the assertion is derived from the
declaration and not the declaration from the assertion." (pp. 553-554)
Though 'the truth of the statement' may indeed derive from the
declarational character of the utterance, this does not warrant the step
towards a claim as to a derivation of the assertive nature of performatives
itself. Leaving this unwarranted argumentational step aside, the conclusion
is as counterintuitive as the idea that all performatives would be indirect
speech acts. It is counterintuitive in view of the function of sentence types:
declarative sentence types are usually associated with the force of an
assertion as their basic function; therefore, viewing the assertive nature of
a declarative sentence as derived in the case of performatives stretches the
imagination to an unacceptable extent.
A third problem is associated with the notion of self-reference which is
central to Searle's account. After restating the problem as follows
(repeating the denial of ambiguity),
"How can the literal meaning of an ordinary indicative sentence encode the actual
performance of an action named by the main verb? And how can the literal meaning
both encode the performative and the assertive meaning without being ambiguous?
It is not enough to say that in the one case the speaker intends the utterance as a
performative and in the other as an assertion. The question is: how could one and
the same literal meaning accommodate both intentions?" (p. 551)
he formulates the main elements of his explanation:
"The apparatus necessary for answering these questions includes at least the following
three elements:
First, we need to recognize that there is a class of actions where the

manifestation of the intention to perform the action, in an appropriate context,
is sufficient for the performance of the action.
Second, we need to recognize the existence of a class of verbs which contain the
notion of intention as part of their meaning. To say that a person performed the
act named by the verb implies that he or she did it intentionally, that if it wasn't
intentional, then the agent didn't do it under that description. Illocutionary
verbs characteristically have this feature. I cannot, e.g., promise unintentionally.
If I didn't intend it as a promise, then it wasn't a promise.
Third, we need to recognize the existence of a class of literal utterances which

are self referential in a special way, they are not only about themselves, but they
also operate on themselves. They are both self-referential and executive,
[...] Typically we perform a type of illocutionary act by uttering a type of sentence

that encodes the intention to perform an act of that type, e.g. we perform directive
speech acts by uttering sentences in the imperative mood. But another way to
manifest the intention to perform an illocutionary act is to utter a performative
sentence. Such sentences are self-referential and their meaning encodes the intention
to perform the act named in the sentence by the utterance of that very sentence. Such
a sentence is "I hereby order you to leave." And an utterance of such a sentence
functions as a performative, and hence as a declaration because (a) the verb "order"
is an intentional verb, (b) ordering is something you can do by manifesting the
intention to do it, and (c) the utterance if both self-referential and executive, as
indicated by the word "hereby" [...]." (pp. 551-552)
In this explanation, Searle treats the problem at the level of the complete
class of intentional verbs. This extension of the target problem is completely
superfluous: of course only expressions containing verbs which describe
forms of verbal action can coincide with the action itself. This was already
observed by Koschmieder:
"Wenn nun die Handlung, von der gesprochen wird, eben im Aussprechen der betr.
Form besteht, so kann es sich ganz offenbar nur um Verba dicendi natürlich im
weitesten Sinne des Wortes handeln, d.h. um solche Verba, die eine durch Sprechen
vollziehbare Handlung bezeichnen, oder auch deren Symbolisierung ausdrücken."
(1935: 288)
[When the action about which one speaks consists in uttering the corresponding
form, then we are clearly dealing with verba dicendi though of course in the widest
possible sense of the word, i.e. verbs which describe actions that can be performed
by speaking or which may symbolize them.]
Clearly, what needs explaining is the behavior of a specific type of speech
act verbs, in their capacity as speech act verbs and not in their capacity as
intentional verbs (though they all are intentional verbs). This shift of focus
is, incidentally, symptomatic of the overdependence of Searlean speech act
theory on the notion of intention in general (see Rosaldo 1982; Du Bois
1987). The superfluous widening of the focus is further aggravated by the
(not unrelated) introduction of an equally superfluous distinction: The one
between self-referentiality and executiveness. It is hard to imagine what a
completely self-referential utterance could be if not an instance of the type
of act it describes, and therefore at the same time executive in Searle's
sense.
If we confine the discussion to properties of utterances containing
speech act verbs, and if we start from the thesis that a completely self-
310 JEF VERSCHUEREN
referential utterance is always an instance of the type of act it describes,

then the problem of performativity becomes the problem of how to define
complete self-reference in speech act verbs. The fourth weak point in Searle's
account, however, completely prevents him from formulating the problem
in this way, and ultimately from explaining performativity» We quote:
"Now this whole analysis has a somewhat surprising result. If we ask what are the
special semantic properties of performativity within the class of intentional verbs
which enable a subclass of them to function as performative verbs, the answer seems
to be, roughly speaking, there are none. If God decides to fry an egg by saying, "I
hereby fry an egg" or to fix the roof by saying, "I hereby fix the roof," He is not
misusing English. It is just a fact about how the world works, and not part of the
semantics of English verbs, that we humans are unable to perform these acts by
declaration. But there is nothing in the semantics of such verbs that prevents us from
intending them performatively; it is just a fact of nature that it won't work. [...]
There are a number of semantic features which block a performative occurrence.
So for example, famously, "hint," "insinuate," and "boast" cannot be used
performatively, because they imply that the act was performed in a way that was not
explicit and overt, and performative utterances are completely explicit and overt. But
there is no special semantic property of performativity which attaches to verbs and
thereby enables them to be used performatively." (p. 554)
Searle overlooks the fact that, while saying "I hereby fry an egg" may be
intended as a 'declaration,5 it can never be self-referential except in a
different world which, as a result, also has a different semantics with to fry as
a speech act verb. Hence, the contrast he posits between semantics and
facts of nature is a spurious one.
The result of Searle's account is indeed 'somewhat surprising.' He sets
out to explain the phenomenon of performativity (i.e. to explain how
performativity is possible), and then he comes to a conclusion which makes
any sort of explanation impossible. The basic problem, which makes the
result less surprising than it seems at first sight, may be that Searle has
never shown great respect for verbs and verb meanings (not only in his
attempts at classifying speech acts, as in Searle 1976, but even in work of
which a significant portion is explicitly devoted to them, as in Searle &
Vanderveken 1985). The problem of performativity is fundamentally
concerned with the question as to how it is possible to use certain members
of a specific category of verbs in a specific way. Performativity is a property
of verbs. By denying this fact, an explanation is made impossible.
In what follows I will explore the following hypothesis. The possibility
of being used completely self-referentially is a semantic-conceptual property
of individual verbs, in such a way that a 'performative' use of verbs lacking
this property could not be cognitively processed. Though in a practical

description one cannot avoid, as Searle correctly suggests, concentrating on
restrictions on performativity (i.e. properties that block performative usage),
these restrictions are not ad hoc, given by the limitations of 'nature,' but
restrictions on the cognitive processibility of the lexical information
contained in the verbs. Moreover, the restrictions are predictable on the
basis of a general and coherent descriptive model for the analysis of
metapragmatic terms (i.e. natural language terms which describe verbal
behavior). Hence, performativity can be positively defined in terms of this
model, and it can be shown that there is a semantic property attached to
verbs which enables them to be used performatively.
5. A descriptive model and its explanatory power
Our explanation for the phenomenon of performativity relies on a model

for the description of linguistic action verbs, or metapragmatic terms in
general, which was developed elsewhere (Verschueren 1985: 58-63, 1987a),
extensively used for other research purposes (Verschueren 1989), and which
itself hinges on a wider theory of pragmatics. This wider theory of
pragmatics (which cannot be fully explained in the context of this paper; see
Verschueren 1987b, 1995a and 1995b for details) differs significantly from
orthodox speech act theory in a variety of ways; most importantly, it does
not hinge on intentionality as an absolute point of departure, but it includes
the entire realm of cognitive, social, and cultural phenomena in a general
functional perspective on language and communication which even allows
før an ethnographic approach to word meaning. The descriptive model in
question is graphically represented in Figure 1.
According to this model, a non-circular description of linguistic action
verbs (henceforth LAVs) should specify the conditions under which a verb
V can be appropriately used in a description D (which is itself also a
linguistic action) of a linguistic action A. Those conditions (C) can be
formulated as conditions on or properties of the act to be described (A-
conditions; simply abbreviated as C in Figure 1) and sometimes in terms of
conditions on or properties of the describing act (D-conditions; abbreviated
Cd in Figure 1). Since both A and D involve all the ingredients of a
complete speech event, conditions or properties may be attached to
speakers (Sa and Sd) and hearers (Ha and Hd), their beliefs, desires,
312 JEF VERSCHUEREN
intentions, and social relationships, as well as aspects of the physical and

social context, the communicative intentions and the linguistic choices made
in order to achieve a specific communicative effect. One of the linguistic
choices made in D is the choice of a verb V (a LAV) which describes A as
a form of verbal behavior. Though the descriptions of all LAVs require D-
conditions, most of these bear on beliefs or assumptions held by the utterer
of D (i.e. Sd) which simply reflect conditions on or properties of A. For
instance, if the appropriateness of to promise in a description of A is
goverend by the condition that Sa intends to carry out the action specified
in the propositional content of his/her utterance, then the use of to promise
in description D is also governed by the D-condition that Sd must believe
that Sa had this intention when performing act A. But in addition to such
'reflections9 of A-conditions, some LAVs require for their appropriate use
also the satisfaction of conditions which can only be formulated with
reference to D. Let us take to lie as an example. Appropriate use of the
verb (in its prototypical sense; see Coleman & Kay 1981) involves at least
the following A-conditions:
- the propositional content of A deviates somehow from the 'truth'
- there is a discrepancy between Sa's beliefs and the propositional
content of the utterance
- Sa has an intention to mislead Ha
All of these are 'reflected' in D (in the form of Sd's belief that these A-
conditions are satisfied), but in addition there are the following D-
conditions (i.e. conditions which can only be formulated by making explicit
reference to something that is to be situated on the D-side):
- Sd's negative value judgment about A (a negative value judgment is
always associated with the use of the verb to lie, as emerges from the
fact that whenever there are attenuating circumstances in Sd's mind,
use of the verb will be avoided as much as possible)
- Sd's judgment about the truth value of the proposition in A (which
creates an interpretive distance in a sense to be explained below)
Our hypothesis is the following: complete self-reference in speech act
verbs (one type of LAVs) requires the possibility of making A and D
coincide. In other words, speech act verbs which do not require D-conditions
(beyond those of the 'reflective' type) for an adequate description of their
meanings, can in principle be used performatively. As soon as explicit D-
conditions are required, there is a conceptual distance between A and D
which makes it impossible to interpret (i.e. to process cognitively) an

utterance in which V would occur self-referentially.
The simplest cases in which complete coincidence between A and D is
possible are those in which all C's are defined by the institutional context
in which A belongs: it is not accidental that Austin's most typical examples
of explicit performatives involved acts such as christening and adjourning.
The next category, which includes Searle's prototypical example of a speech
act, a promise, involves acts which are governed by relatively clear (though
mostly culture-specific) social norms so that their status is usually
unambiguous (or at least lends itself to being treated as unambiguous).
Asking, ordering, declaring, warning, announcing, apologizing, thanking, and
many others also do not require distance between A and D. For all speech
act verbs that cannot be used performatively, however, such a conceptual
distance—which can take different forms—can be shown to be required. Let
us present a brief overview of the most-cited examples (which we list in
four groups, which are not meant to represent a taxonomy, since different
aspects of the analysis can be at work in the same verbs).
To hint, insinuate, intimate, allude to all require an interpretive distance
between A and D. Reason: the 'mode of delivery' in A is implicitness;
hence the meaning that is hinted at, insinuated, intimated or alluded to is
fundamentally 'constructed' by Sd. The D-condition involved is Sd's
interpretive intervention which leads him/her to believe that more is meant
in A than is actually said. This is not to deny that interpretation is in other
cases an active process as well. But this process is completely foregrounded
by describing A as an instance of hinting or insinuating, so that such a
description is not possible without creating a distance between A and D.
Therefore A and D cannot conceptually coincide, and such verbs cannot be
used performatively.
To lie, allege, perjure, kid, boast or brag, also require an interpretive
distance, but of a slightly more specific kind. Reason: acts of lying,
perjuring, and the like, have a 'status' which is (at least temporarily) not
meant to be recognized by Ha; hence, in order to be describable with
reference to that status, an interpretive intervention on the part of Sd is
needed which creates a conceptual distance between A and D. The more
specific D-condition (as already mentioned above in connection with to lie)
is the following: Sd's judgment concerning the truth value of the proposition
in A.
To threaten, grumble about (as opposed to protest), swear or curse,
314 JEF VERSCHUEREN
whitewash (as opposed to excuse), slander (as opposed to accuse), but also
to lie and to boast or brag, all require an evaluative distance between A and
D. Reason: as was already anticipated in relation to to lie, the social
acceptability of the described acts is low, so that negative value judgments
on the part of Sd are associated with the choice of the LAVs listed. D-
condition: Sd's negative value judgment. This D-condition makes it
impossible for a speaker to describe his or her own act as an instance of
threatening, slandering, lying or boasting, without passing a value judgment
on it and thereby undercutting the act while it is performed. This is the
classic case of what Vendler (1976) called 'illocutionary suicide' in an article
which contained most of the ingredients needed for the type of explanation
I am trying to formulate for the phenomenon of performativity. That this
is exactly what is happening can best be clarified with reference to forms
of usage in which the negative value judgment is cancelled. Consider (38)
and (39):
(38) I boast myself a patriot
(39) I promise I'll kill you
Sentence (38) was taken from a British dictionary6; since patriotism is
usually evaluated positively, the content of the sentence cancels any
negative associations the verb to boast normally has. Therefore, the D-
condition is overruled, and to boast can be used performatively. Similarly
to promise in (39) does not carry the negative connotations which to threaten
is burdened with, so that if to promise is used as a euphemism one can
perfectly threaten performatively.
Finally, to convince, annoy, persuade, amuse even require a temporal
distance between A and D. Reason: the communicative effect of A must
already have taken place before A can be described as having convinced,
annoyed, persuaded, amused someone. D-condition: Sd's assessment of the
communicative effect of A. It goes without saying that A an D cannot
coincide if they must be consecutive.
6
The fact that this sentence was found in a dictionary, does not make it acceptable for
all speakers of English. In fact, a self-reflective use of to boast is no doubt extremely
exceptional - which underscores the general point we have made about this verb.
Figure 1
316 JEF VERSCHUEREN
Briefly, our hypothesis concerning the conditions under which a

linguistic action verb has performative potential, Le, the conditions under
which it can be used fully self-referentially, seems to be completely
confirmed by the examples found in the speech act literature to illustrate
the 'problem' of performativity.
6. Back to 1903
Let us briefly restate the hypothesis: complete self-reference in speech act

verbs requires the possibility of making A and D coincide; in other words,
the meaning of the verbs should be such that their usage does not require
a conceptual distance between A and D. If the explanations adduced so far
are sufficiently convincing to justify our hypothesis in connection with the
examples we find in the speech act literature, we are still stuck with the
problem of performativity as formulated by Škrabec and Koschmieder. How
can our descriptive framework help us to explain the predominance of
imperfective forms in explicit performative usage? More recently, Hinrichs
(1985: 137) suggested that it was strange that no known language had
developed a special grammatical category for the Koinzidenzfall, in spite of
its occurrence in all present-day European standard languages. And
Leeuwen-Turnovcová (1986: 12) proposes that the issue of 'aspect' may be
completely irrelevant to a discussion of Koinzidenz which should rather be
regarded as a "Spezialfall der Kommunikation," a special case of
communication to be considered in its own right. I believe that both
suggestions are misguided, as well as the formulations by Škrabec and
Koschmieder.
Though I cannot speak of Slavic languages with any degree of authority,
no evidence has ever been offered to me that the perfective choice occupies
a significant proportion of the performative spectrum.7 Therefore, the first
thing that needs to be explained is the use of the imperfective in explicit
performatives. If we replace the earlier logico-grammatical formulation of
7
On the contrary, Žagar (1990) argues about Slovenian (the language in which,
according to Škrabec, a perfective form of the performative would be possible) that only
the imperfective can express a real performative, whereas the use of a perfective expresses
only an intention or a readiness to perform an act Thus the imperfective Obljubljam
would mean "I promise", but the perfective Obljubim means "I am about to promise, I am
willing to promise,"
the problem with a straightforward approach in terms of the descriptive

framework we have sketched, then the explanation is relatively transparent.
The choice between alternative 'aspects' is something that necessarily takes
place on the D-side of our model: Sd has to make a choice on the basis of
his or her judgment concerning the temporal/aspectual status of the act
performed. This inevitable interpretive intervention creates a conceptual
difference between A and D. But Sd's role is focused on more explicitly
when the perfective is used, since the perfective is the marked case in the
contrast between perfective and imperfective verbs in Slavic languages.
Therefore it is predictable that the perfective would be used less—if at
all—in a performative utterance since it draws the attention more strongly
to an interpretive distance between A and D.
Contrary to Searle's claim that the performative occurrence of certain
speech act verbs is more or less accidental and has nothing to do with the
meaning of those verbs, we believe that we have demonstrated how a
coherent model for the description of the meaning of speech act verbs (or
LAVs, or even metapragmatic terms in general) makes it predictable, on
semantic and conceptual grounds (as these are understood within the
context of a combined cognitive, social, and cultural perspective on
language), whether a verb is suitable for performative use or not. Facts of
nature have nothing to do with this. The phenomenon can entirely be
formulated in terms of the cognitive processibility of utterances containing
verbs that are meant to make those utterances self-referential. What is not
predictable, however, is whether a given verb will in fact be used
performativbly in a given language. At the level of factual occurrence, there
are mostly language-specific conventions at work. Thus nothing can block
the use of English to ask (e.g. in its requesting sense) in an explicit
performative formula. But it is a fact of English usage that this is rarely
done, whereas the corresponding Hungarian kér (as in example (6)) is used
in most everyday requests. Similarly, there is a great deal of variation
between languages as to whether the canonical form of the performative
(examples (1) through (7)) will be used with a specific verb, or whether
preference will be given to some non-central performative expression type
or semi-performative formula (as in examples (8) through (34)). Though the
basic conceptual principle seems quite clear, susceptibility for performative
use may itself be a gradable phenomenon (as was already suggested in our
formulation of an explanation for aspect choice in Slavic languages, and as
should be clear from the observation that there may be a fuzzy boundary
318 JEF VERSCHUEREN
between D-conditions that are 'reflective' and those that are not). This
gradability, in turn, may not be unrelated to the preference for a specific
form of expression. But a much more detailed investigation is called for to
make those issues tangible.8
References
Åqvist, Lennart
1972 . "Performatives and verifiability by the use of language". Filosofiska Studier
14.
Austin, John
1962 How to do things with words. Oxford: Oxford University Press.
Bach, Kent
1975 "Performatives are statements too." Philosophical Studies 28.229-36.
Bach, Kent, and Robert M. Harnish

1979 Linguistic communication and speech acts. Cambridge, MA: MIT Press.
Coleman, Linda and Paul Kay

1981 "Prototype semantics: The English verb lie" Language 57:1.26-44.
Cornulier, Benoît de
1975 "La notion d'auto-interprétation." Etudes de Linguistique Appliquée 19.52-
82.
1980 Meaning Detachment (Pragmatics & Beyond 1:7). Amsterdam: John
Benjamins.
Davis, Steven
1964 "'I know' as an explicit performative." Theoria 30.157-65.
Du Bois, John
1987 "Meaning without intention: Lessons from divination." IPrA Papers in
Pragmatics 1:2.80-122.
8
It should be stressed that the descriptive model for linguistic action verbs, as handled
in this article, makes use of only a few research angles provided by the more general
theory of pragmatics with reference to which it was claimed that it has to be situated. The
same theoretical perspective also provides the means for further, more detailed,
investigations of actual patterns of usage.
Fauconnier, Gilles
1979 "Comment contrôler la vérité: Remarques illustrées par des assertions
dangereuses et pernicieuses en tout genre." Actes de la Recherche en
Sciences Sociales 25.3-22.
Fillmore, Charles
1971 "Verbs of judging". In CJ. Fillmore & D.T. Langendoen (eds.), Studies in
Linguistic Semantics. New York: Holt, Rinehart & Winston, 272-289.
Ginet, Carl
1979 "Performativity." Linguistics and Philosophy 3:2.245-65.
Grewendorf, Günther
1979 Explicit performatives and statements. Journal of Pragmatics 3:5.431-45.
Grodzirński, Eugeniusz
Performative utterances. Wroclaw: Ossolineum.
Heal, Jane
"Explicit performative utterances and statements." The Philosophical Quarterly
24.106-21.
Hedenius, Ingemar
1963 "Performatives." Theoria 29.115-36.
Hinrichs, Uwe
1985 "Der Koinzidenzfall in den Balkansprachen." Zeitschrift ßr Balkanologie
1985.136-51.
Keck, Gabriele, and Michael Stubbs

1984 "Koschmieder on speech act theory: A historical note." Journal of
Koschmieder, Erwin
1935 "Zu den Grundfragen der Aspekttheorie." Indogermanische Forschungen
53.280-300.
Leeuwen-Turnovcovâ, Jiřina van

1986 Illokutive Komposita von Verben des Sagens: Eine vergleicehnde Untersuchung
zu pragmatischen Aspekten der Verballexik im Cechischen und Russischen.
Wiesbaden: Otto Harrassowitz.
Lemmon, E.J.
1962 "On sentences verifiable by their use." Analysis 22:4.86-89.
320 JEF VERSCHUEREN
McCawley, James D.
1977 "Remarks on the lexicography of performative verbs." In A. Rogers et al.
(eds.), Proceedings of the Texas conference on performatives, presuppositions
and implicatures. Arlington, Virginia: Center for Applied Linguistics, 13-25.
Nuyts, Jan, & Jef Verschueren

1987 A comprehensive bibliography of pragmatics. (4 vols.)
Amsterdam/Philadelphia: John Benjamins.
Récanati, François
1980 "Some remarks on explicit performatives, indirect speech acts, locutionary
meaning and truth-value." In John Searle, Ferenc Kiefer & Manfred
Bierwisch (eds.), Speech act theory and pragmatics. Dordrecht: Reidel, 205-
20.
1981 Les énoncés performatifs: Contribution à la pragmatique. Paris: Editions de

Minuit.
Rosaldo, Michelle
1982 " "The things we do with words: Ilongot speech acts and speech act theory
in philosophy." Language in Society 11.203-237.
Sadock, Jerrold M.
1974 Toward a linguistic theory of speech acts. New York: Academic Press.
Searle, John R.
1976 "A classification of illocutionary acts." Language in Society 45:1.137-151.
1989 "How performatives work". Linguistics and Philosophy 12.535-558.
Searle, John R. and Daniel Vanderveken

1985 Foundations of illocutionary logic. Cambridge: Cambridge University Press.
Škrabec, Stanislav
1903 "Zum Gebrauch der Verba perfectiva und imperfectiva im Slovenischen."
Archiv für Slavische Philologie 25.554-64.
Urmson, J.O.
1977 "Performative utterances." Mid-West Studies in Philosophy 2.120-27.
Vendier, Zeno
1976 "Illocutionary suicide." In Alfred F. MacKay & Daniel D. Merrill (eds.),
Issues in the philosophy of language (Proceedings of the 1972 Oberlin
Colloquium in Philosophy). New Haven: Yale University Press, 135-45.
Verschueren, Jef
1980 On speech act verbs. Amsterdam/Philadelphia: John Benjamins.
1985 What people say they do with words: Prolegomena to an empirical-conceptual

approach to linguistic action. Norwood, NJ: Ablex Publishing Corporation.
1987a • "Metapragmatics and universals of linguistic action." In J. Verschueren

(ed.), Linguistic action: Some empirical-conceptual studies. Norwood, NJ:
Ablex Publishing Corporation, 125-140.
1987b Pragmatics as a theory of linguistic adaptation. (IPrA Working Document

1) Antwerp: International Pragmatics Association.
1989 "Language on language: Toward metapragmatic universals." IPrA Papers in

1995a "Linguistic pragmatics and semiotics." Semiotica 104:1/2.45-65.
1995b "The pragmatic perspective." In J. Verschueren, J.-O. Östman & J.

Blommaert (eds.), Handbook of pragmatics: Manual.
Amsterdam/Philadelphia: John Benjamins, 1-19.
Walker, Jeremy D.B.

1969 "Statements and performatives." American Philosophical Quarterly 6:3.217-
25.
Warnock. G.J.
1973 . "Some types of performative utterance." In Essays on J.L. Austin, by Sir I.
Berlin et al, 69-89. Oxford: Oxford University Press, 1973.
Yanofsky, Nancy M. & Rose Maclaran

1980 Remarks on performativity. CUNY Forum 7:8.237-47.
Žagar, Igor 2 .
1988 "Aspect et performativité en Slovène." Acta Linguistica Hungarica
38:1/4.275-287.
Zaslawsky, Denis
1979 "Pronoms personnels, performatifs et actes de langage." Langue Française
42.48-59.
Zuber, Ryszard
1981 "Marche modali e performativi espliciti." Materiali Filosofici 4:5.111-20.
Index
Aboutness: 148, 150 Asher: 129, 131

Accusing: 314 Aske: 145
Ackerman: 4, 5, 8, 12, 16, 22, 247 Asking: 313
Act: Aspect: 111, 116-118
Momentaneous — : 304 Assertion: 307
Actor: 157, 183 Pragmatic — : 147
Adaptability: 258 Atkins: 25, 27, 28, 37, 41, 44, 69, 295
Adjourning: 313 Austerlitz: 251
Adverb: Austin: 300, 301, 318
Demonstrative — : 248-255, 260
Temporal — : 260 Bach: 120, 129, 300, 306, 318
Adverbial phrase: 76-77 Barnes: 175, 186
Advising: 65 Beaugrande: 118, 129
Affect: 265, 274 Berk: 44, 69
Agent: 140, 142, 145, 154, 156 Blakemore: 202, 216, 217
Alleging: 313 Blass: 202, 217
Allerton: 151, 186 Blommaert: 300
Allwood: 248 Boasting: 313, 314
Alluding: 313 Bolinger: 185, 186, 264, 275
Altschuier: 46, 69 Borrowing: 47-48
Ambiguity: 307 Boundedness: 251, 252, 254, 255, 266
Andersen: 284» 294 Bowerman: 75, 88
Andersson: 247 Bragging: 313, 314
Announcing: 313 Breal: 12, 22
Annoying: 314 Bresnan: 7, 22, 157, 186
Answer: Brown: 43, 48, 62-68, 69
Elliptical — : 204 Burzio: 164, 186
Antecedent: 74
Clausal — : 76 Case: 1-24, 145-190
Non-clausal — : 75 — frame: 153
Antonym: 26 - marking: 16-21, 249-255
Apologizing: 313 Category:
Appleton: 47, 69 Grammatical — : 177
Åqvist: 300, 318 Chafe: 204, 217
Arabic: 292 Chinese: 43-71, 191-200
Argument selection principle: 11, 12 Chomsky: 91, 108, 152, 187
Argument structure: 85-87
Arluke: 46, 69
324 ESSAYS IN SEMANTICS AND PRAGMATICS
Christening: 313 De Castro Campos: 75, 88

Chu: 69 Decision: 30, 31, 34
Clancy: 61, 69 Declaration: 224, 306, 307
Clark: 193, 199, 285, 294 Declaring: 313
Claudi: 244, 280, 294, 295 Decoding vs. encoding: 39, 40
Clause: Defiance: 224, 236
—linking: 83-85 Definition: 26
Grammaticalization of — : Deixis: 191, 198, 199, 247-275
214-217 Affective - : 265-272
Suspended - : 201-217 DeLancey: 161, 187
Coherence: 118-120 Dictionary: 25-42
Coleman: 312, 318 — example: 36-41
Collinson: 257, 269, 275 Constraints on — : 38, 40
Comment: Focus of — : 36
Explanatory/clarifying — : 221, 228 Function of -- : 37, 40
Complementation: 33-34 Purpose of - : 36, 39
Complementizer: 214-216, 219 Source of — : 38, 40
Complimenting: 65 Direction: 248-249, 266
Comrie: 79, 88, 157, 187 Dirven: 299
Concession: 207 Discourse:
Conditional: 82 — modality strategy: 229
Mono-clausal — : 74 — representation structure: 115
Confucius: 47, 191 Narrative- : 111-132
Connective: Distance: 91-109, 255-257
(Pseudo-)logical - : 207-213 — expression: 94-108
Consequent: 74 — noun: 96
Construction: 216 — -noun adjunct construction: 92
Bi-clausal — : 151 Conceptual - : 312
Grammatical — : 180 Evaluative — : 314
Impersonal — : 272 Interpretive — : 313
Lexically motivated — : 92 Temporal — : 314
Conventionalization: 216 Doi: 60, 61, 62, 63, 69
Convincing: 314 Downing: 282, 294
Cook: 202, 218, 221, 244 Dowty: 2, 6, 8, 9, 10, 11, 12, 13, 15, 22,
Cooper: 267, 275 111, 116, 118, 130, 155, 187
Cornulier: 300, 302, 303, 318 Dressier: 118, 299
Counteridentical: 77-78 Dry: 116
Crick: 198, 199 DuBois: 221, 244, 273, 275, 309, 318
Croft: 120, 129 du Gard: 47
Cross-reference: 26 Dummy-'it': 105
Cruttendon: 151, 186 Dutch: 167
Cursing: 313 Dziwirek: 1, 3, 4, 15, 16, 22
Dative: 1, 3-5 Eelen: 300

Davis: 300, 318 Effector: 156
INDEX 325
Ehrich: 265, 275 Foley: 6, 7, 23, 81, 88, 156, 157, 159,164,
Ellipsis: 83-85 166, 179, 187
Emotive: 127 Formality: 236, 237
Empathy: 265 Frame: 279-294
English: 44-46, 73-90, 145-190 — semantics: 25-42, 279-294
Middle - : 44 Body - : 279-294
Erickson: 45, 69 Theta - : 153
Etymology: 26 Francis: 67, 70
Event: 120-123, 128 French: 46-47, 114-115, 145-190
Exclamation: 222, 231 Fried: 145
Excusing: 314 Fuchs: 151, 166, 168, 187
Experiencer: 106, 154, 156, 158, 173 Fujii: 74, 88
Metaphorical — : 140-142 Functions:
Explicitness vs. implicitness: 37 Grammatical — : 16-21
Extent: 102
Gawron: 2, 6, 9, 12, 23, 109
Faber: 151, 166, 187 Gaze: 133
Face: 43-71 Gensler: 219
—saving: 45 German: 265, 267
Give - : 52-53 Ginet: 300, 318
Lose - : 50-52 Givenness: 254, 255, 262, 263, 264
Negative — : 62, 64-66 Givón: 149, 188, 197, 199
No - : 49 Gloss: 26
Positive - : 62, 66-68 Goal: 30, 31, 102, 103, 104, 107, 108,
Restoring — : 67 140, 142
Thin/thick - : 53-54 Goldberg: 7, 23, 92, 103, 109, 145, 167,
Fauconnier: 73, 76, 77, 88, 300, 319 168, 188
Fillmore: 2, 7, 8, 9, 10, 12, 22, 23, 25, 27, Goodman: 78, 88
28, 41, 42, 44, 69, 73, 75, 78, 82, 83, Government binding theory: 91, 167, 168
85, 88, 91, 93, 104, 109, 111, 112, Grammar: 133-143
123, 130, 133, 136, 137, 140, 141, Case - : 247
153, 154, 155, 156, 157, 167, 173, Construction — : 7, 91-109, 167,
187, 191, 198, 199, 201, 216, 218, 180, 252
219, 221, 234, 244, 247, 251, 252, Head-driven phrase structure — : 94
254, 256, 265, 269, 270, 275, 276, Lexical functional — : 7
279, 285, 295, 299, 319 Relational — : 1,. 15
Finnish: 247-275 Role and reference — : 7
Fleischman: 145 Grewendorf: 300, 319
Focus: 145, 147-153, 159, 164 Grodzinski: 300, 319
— articulations: 148 Grootendorst: 299
— types: 146 Gruber: 7, 23, 155, 188
Argument - : 149, 150, 152 Grumbling: 313
Predicate - : 149, 150, 175 Gumperz: 299
Sentence - : 149, 150, 161
Unmarked — position: 163 Haiman: 217, 218
Hakulinen: 248, 249, 269, 272, 276 Jacobsen: 87, 88

Halliday: 188, 218 Jakobson: 180, 295
Halpern: 265, 276 Japanese: 48, 51, 55, 73-90, 201-217, 219-
Harnish: 300, 306, 318 244, 265, 267, 270
Hasan: 218 Jespersen: 75, 88
Hasegawa: 84 Johnson: 133, 136, 143, 193, 199, 276,
Hawkins: 157, 188 280, 285, 292, 296
Heal: 300, 319 Johnson-Laird: 124, 131
Hebrew: 280-294 Josephs: 220, 245
Hedenius: 300, 319
Heine: 221, 244, 280, 294, 295 Kalliokoski: 267, 268, 276
Hindi: 17 Kamp: 111, 114, 115, 116, 118, 130
Hinrichs: 300, 316, 319 Kannerva: 7, 22, 157, 186
Hinting: 313 Karlsson: 249, 276
Höhle: 148, 188 Kaufman: 197, 200
Honor: 68 Kay: 7, 23, 92, 93, 153, 154, 157, 167,
Honorific: 240 187, 201, 216, 218, 221, 244, 252,
Hopper: 111, 130, 221, 245 276, 312, 318
Hu: 48, 49, 50, 51, 54, 56, 57, 70 Keck: 300, 319
Huang: 67 Kegl: 41
Hungarian: 112, 113, 117, 301, 317 Kidding: 313
Hunnemeyer: 244, 295 Kiefer: 299
Kinship terminology: 194
Idiom: 205-206 Kirita: 60, 70
Ikegami: 87, 88 Kirsner: 167, 188
Illocutionary suicide: 314 Klein: 256, 278
Imperative: 272 Koinzidenzfall: 304, 316
Implicature: 306 König: 221, 243, 246
Inference: 216 Korean: 48, 51, 55, 236
Inflection: 26 Koschmieder: 304, 305, 309, 316, 319
Informality: 224 Kubinski: 21, 23
Information: Kucera: 67, 70
- structure: 146, 159, 170 Kuno: 151, 188, 220, 221, 245
Old - : 271 Kurath: 70
Insinuating: 313 Kuroda, 151, 188
Insistence: 226, 241
Instrument: 154, 156 Lakoff(G.): 137, 139, 143, 193, 199, 221,
Intimating: 313 245, 276, 280, 285, 296
Invariance principle: 137, 138 Lakoff (R.): 257, 258, 269, 277
Inversion: 1-24 Lambrecht: 146, 147, 148, 149, 150, 151,
Isenberg: 119, 130 160, 161, 164, 165, 168, 169, 170,
Itani: 209, 218 175, 179, 184, 185, 188, 221, 245
Itkonen: 276 Language change:
Contact-induced — : 197
Jackendoff: 2, 6, 7, 23, 152, 188 Larjavaara: 262, 263, 265, 268, 273, 277
INDEX 327
Lascarides: 129, 130, 131 Moore: 12, 22
Leeuwen-Turnovcovâ: 300, 316, 319 Morgan: 109
Lehtinen: 250, 253, 254, 277 Morphology:
Leino: 250, 277 Derivational — : 288
Lemmon: 300, 319 Muir: 70
Levin: 5, 23, 41, 46, 70 Mulac: 244, 246
Levinson: 43, 48, 62-68, 69, 257, 277
Lexicalism: 91 Nakane: 62, 63, 70
Lexicon: 111-132, 279-294 Natsuoka: 60, 63, 70
Li: 172, 189 Non-discreteness: 11
Location: 154, 248-249, 266 Noun: 128, 279-294
Locative: 156 Complex — : 288
Longacre: 118, 131 Compound — : 282-288
Lying: 312, 313 Nunberg: 102, 104, 109
Lyons: 189, 269, 277 Nuyts: 300, 320
Mackenzie: 299 Oberländer: 129, 131

Mackie: 79, 88 Object: 145, 163, 164, 183
Maclaran: 300, 321 O'Connor: 92, 93, 167, 187, 201, 216,
Malayalam: 8, 12 218, 221, 244, 252, 276
Markedness: 317 Offering: 65
Martin: 220, 221, 222, 223, 224, 226, 232, Ohara: 219
243, 244, 245 Ohori: 201, 218
Mathesius: 172, 189 Okamoto: 221, 245, 246
Matsumoto (Yo): 219, 221, 245 Order: 222, 233, 313
Matsumoto (Yoshiko): 43, 62, 63, 70 Östman: 254, 259, 268, 271, 272, 275,
Maurois: 47 277
Maynard: 220, 221, 229, 230, 231, 245 Oxford English Dictionary: 44
McCawley: 220, 245, 300, 302, 303, 320
Meaning: Particle:
Pragmaticization of — : 219-244 Clause-linking — : 203
Meeuwis: 300 Sentence-final — : 219-244
Mental space: 73-90 Utterance-final pragmatic — : 203
Metalinguistic indicator: 26 Path: 141
Metaphor: 133-143, 248, 280 Patient: 140, 142, 145, 156
Conduit - : 141, 142 Perfectivity: 112
Metonymy: 280 Performative:
Michaux: 47 Explicit - : 300, 301, 316
Michaelis: 145, 167, 189 Semi— : 301
Miller: 46, 70, 124, 131 Performativity: 299-318
Minami: 81, 88 Perjuring: 313
Moens: 120, 122, 123, 131 Perlmutter: 164, 190
Modality marker: 241 Persuading: 314
Mohanan: 2, 8, 9, 12, 13, 17, 23 Pete: 113, 131
Molendijk: 114, 131 Petruck: 279, 288, 291, 296
Phrase: Reinhart: 148, 190

'from'- - : 95, 134, 135, 136 Relation:
'to'- - : 99 Ancestor-descendant — : 191-200
'of- - : 102 Pragmatic — : 148
Phraseology: 201-217 — changing construction: 185
Pitkänen: 268 Semantic — : 32-33
Plan: 125 Relativity:
Polish: 1-24 Cultural-linguistic — : 191-200
Politeness: 43-71 Reporting: 239
Universals of — : 43 Requesting: 64
Pollard: 94, 109 Ritchie: 118, 131
Polysemy: 108 Robert: 46, 70
Possession: .28, 29, 30, 31, 32, 33, 34 Rohrer: 114, 115, 118, 130
Praesens effectivum: 300, 304 Role:
Predicate: 1-24 — hierarchy:
Prestige: 68 Pragmatic - : 146, 159, 164,
Presupposition: 184
Pragmatic — : 147 Semantic - : 155, 159
Processibility: 311 Frame-specific — : 153
Projection principle: 91, 167 Grammatical - : 145-190
Promising: 312, 314 Macro— : 157
Pronoun:168 Pragmatic — : 145-190
Demonstrative — : 260 Semantic - : 145-190
Pronominal: 165 Thematic — : 153
Proposition: 106 Ronimus: 274
Meaning of — : 159 Rosaldo: 309, 320
Pragmatically structured — : 146, Rosch: 197, 199
148 Ross: 267, 275
Prosody: 185
Protagonist: 28, 29, 30, 33 Sadock: 300, 320
Protesting: 313 Sag: 94, 109
Prototype: 197 Sakamoto: 60, 63
Proximity: 255-257, 259-272 Salience hierarchy: 9
Sapir: 274
Quantifier: 128 Sasse: 149, 151, 168, 190
Question: Saukkonen: 271, 278
Echo - : 203 Saussure: 197, 199
Quotative — : 236 Scene: 111, 154
Schema: 28
Raiffa: 29, 42 Schmerling: 145, 151, 169, 185, 190
Ransom: 236, 246 Selection:
Rappaport Hovav: 5, 23 Lexical — : 185
Reason: 207 Searle: 299, 300, 302, 303, 305, 306-311,
Récanati: 300, 306, 320 313, 317, 320
Reh: 295 Self-affirmation: 224, 237
INDEX 329
Self-reference: 308, 310, 316 Synonym: 26
Semantics: Quasi- — : 34
Constructionist approach to — :7
Grammatical function inducing — : Talmy: 254, 278
6, 12 Target: 134, 137, 138
Lexical — : 191 Temporal structure: 111-132
Lexical categorial approach to — : 7 Temporal discourse interpretation prin
Surface case marking inducing — : 6 ciple: 118
Sense: Tense: 111, 114-118
Distinct — : 31 Teramura: 87, 88, 221, 229, 230, 235, 246
Sentence: Text linguistics: 118-120
Use of - : 159 Thanking: 64, 313
Setting: 28, 30, 33 Theme: 141, 156
Shakespeare: 44 Thomason: 197, 200
Shibatani: 12, 23, 219, 299 Thompson: 73, 172, 189, 190, 244, 246,
Simpson: 70 248, 299
Škrabec: 300, 304, 305, 316, 320 Threatening: 313, 314
Slandering: 314 Ting: 47
Slavic languages: 304, 316, 317 Tjosvold: 45, 70
Slobin: 248 Topic: 145, 147-153, 159, 164
Slovenian: 316 —comment sentence structure: 229
Smith: 120, 131 Unmarked — position: 163
Source: 29, 31, 33, 102, 103, 104, 107, Transcription:
108, 133, 134, 137, 138, 140, 141 Phonetic — : 26
Spanish: 265 Transitivity: 85
Specificity: 254, 255 Translation: 26
Speech act: Traugott: 221, 243, 246, 280, 296
— theory: 309 Tsunoda: 12, 24
— verb: 309, 317
Indirect — : 306 Undergoer: 157, 183
Spelling: 26 Urmson: 300, 320
Spencer-Brown: 7, 23 Utterance: 159
Sperber: 118, 131
Steedman: 120, 122, 123, 131 Vanderveken: 310, 320
Stimulus: 154 Van Dijk: 118, 130
Stubbs: 300, 319 Van Oosten: 146, 190
Subject: 145, 150, 183 Van Valin: 6, 7, 23, 24, 81, 88, 145, 157,
— argument: 152 159, 164, 165, 166, 179, 187, 190
Focal - : 161 Vendler: 120, 131, 314, 320
Sugamoto: 87, 88 Verb: 279-294
Sugiyama Lebra: 61, 62, 63, 68, 71 — agreement rule: 16
Swearing: 313 Denominal - : 291-293
Sweetser: 73, 79, 88, 145, 280, 296 Imperfective - : 304, 305, 316
Synecdoche: Intentional — : 309
Constructional — : 205 Linguistic action — : 311
Resultative change-of-state — : 126

Speech act - : 309, 317
Verkuyl: 112, 131
Verma: 24
Verschueren: 258, 278, 300, 301, 311,
320, 321
Vet: 114, 131
Wacha: 112, 132

Walker: 300, 321
Wandruzka: 153, 161, 184, 190
Warning: 313
Warnock: 300, 321
Watters: 149, 190
Webelhuth: 8, 22
Weiner: 70
Weissenborn: 256, 278
Whitewashing: 314
Whorf: 274
Whorfian hypothesis: 194-198
Wierzbicka: 4, 5, 12, 24
Wiik: 263, 271, 278
Wilson: 119, 131
Yanovsky: 300, 321

Yngve: 258, 278
Young: 50, 51, 64, 66, 70
Yuma: 265
Zaenen: 102, 109

Žagar: 299, 300, 316, 321
Zaslawsky: 300, 321
Zimmer: 282, 297
Zuber: 300, 321
Zwicky: 7, 24
In the PRAGMATICS AND BEYOND NEW SERIES the following titles have been
published thus far:
1. WALTER, Bettyruth: The Jury Summation as Speech Genre: An Ethnographic Study
of What it Means to Those who Use it. Amsterdam/Philadelphia, 1988.
2. BARTON, Ellen: Nonsentential Constituents: A Theory of Grammatical Structure
and Pragmatic Interpretation. Amsterdam/Philadelphia, 1990.
3. OLEKSY, Wieslaw (ed.): Contrastive Pragmatics. Amsterdam/Philadelphia, 1989.
4. RAFFLER-ENGEL, Walburga von (ed.): Doctor-Patient Interaction. Amsterdam/
Philadelphia, 1989.
5. THELIN, Nils B. (ed.): Verbal Aspect in Discourse. Amsterdam/Philadelphia, 1990.
6. VERSCHUEREN, Jef (ed.): Selected Papers from the 1987 International Pragmatics
Conference. Vol. I: Pragmatics at Issue. Vol. II: Levels of Linguistic Adaptation. Vol.
III: The Pragmatics of Intercultural and International Communication (ed. with Jan
Blommaert). Amsterdam/Philadelphia, 1991.
7. LINDENFELD, Jacqueline: Speech and Sociability at French Urban Market Places.
Amsterdam/Philadelphia, 1990.
8. YOUNG, Lynne: Language as Behaviour, Language as Code: A Study of Academic
English. Amsterdam/Philadelphia, 1990.
9. LUKE, Kang-Kwong: Utterance Particles in Cantonese Conversation. Amsterdam/
Philadelphia, 1990.
10. MURRAY, Denise E.: Conversation for Action. The computer terminal as medium of
communication. Amsterdam/Philadelphia, 1991.
11.LUONG, Hy V.: Discursive Practices and Linguistic Meanings. The Vietnamese
system of person reference. Amsterdam/Philadelphia, 1990.
12. ABRAHAM, Werner (ed.): Discourse Particles. Descriptive and theoretical investi
gations on the logical, syntactic and pragmatic properties of discourse particles in
German. Amsterdam/Philadelphia, 1991.
13. NUYTS, Jan, A. Machtelt BOLKESTEIN and Co VET (eds): Layers and Levels of
Representation in Language Theory: a functional view. Amsterdam/Philadelphia,
1990.
14. SCHWARTZ, Ursula: Young Children's Dyadic Pretend Play. Amsterdam/Philadel
phia, 1991.
15. KOMTER, Martha: Conflict and Cooperation in Job Interviews. Amsterdam/Phila
delphia, 1991.
16. MANN, William C. and Sandra A. THOMPSON (eds): Discourse Description:
Diverse Linguistic Analyses of a Fund-Raising Text. Amsterdam/Philadelphia, 1992.
17. PIÉRAUT-LE BONNIEC, Gilberte and Marlene DOLITSKY (eds): Language Bases
... Discourse Bases. Amsterdam/Philadelphia, 1991.
18. JOHNSTONE, Barbara: Repetition in Arabic Discourse. Paradigms, syntagms and
the ecology of language. Amsterdam/Philadelphia, 1991.
19. BAKER, Carolyn D. and Allan LUKE (eds): Towards a Critical Sociology of Reading
Pedagogy. Papers of the XII World Congress on Reading. Amsterdam/Philadelphia,
1991.
20. NUYTS, Jan: Aspects of a Cognitive-Pragmatic Theory of Language. On cognition,
functionalism, and grammar. Amsterdam/Philadelphia, 1992.
21. SEARLE, John R. et al.: (On) Searle on Conversation. Compiled and introduced by
Herman Parret and Jef Verschueren. Amsterdam/Philadelphia, 1992.
22. AUER, Peter and Aldo Di LUZIO (eds): The Contextualization of Language. Amster
dam/Philadelphia, 1992.
23. FORTESCUE, Michael, Peter HARDER and Lars KRISTOFFERSEN (eds): Layered
Structure and Reference in a Functional Perspective. Papers from the Functional
Grammar Conference, Copenhagen, 1990. Amsterdam/Philadelphia, 1992.
24. MAYNARD, Senko K.: Discourse Modality: Subjectivity, Emotion and Voice in the
Japanese Language. Amsterdam/Philadelphia, 1993.
25. COUPER-KUHLEN, Elizabeth: English Speech Rhythm. Form and function in every
day verbal interaction. Amsterdam/Philadelphia, 1993.
26. STYGALL, Gail: Trial Language. A study in differential discourse processing.
27. SUTER, Hans Jürg: The Wedding Report: A Prototypical Approach to the Study of
Traditional Text Types. Amsterdam/Philadelphia, 1993.
28. VAN DE WALLE, Lieve: Pragmatics and Classical Sanskrit. Amsterdam/Philadel
phia, 1993.
29. BARSKY, Robert F.: Constructing a Productive Other: Discourse theory and the
convention refugee hearing. Amsterdam/Philadelphia, 1994.
30. WORTHAM, Stanton E.F.: Acting Out Participant Examples in the Classroom.
31. WILDGEN, Wolfgang: Process, Image and Meaning. A realistic model of the
meanings of sentences and narrative texts. Amsterdam/Philadelphia, 1994.
32. SHIBATANI, Masayoshi and Sandra A. THOMPSON (eds): Essays in Semantics and
Pragmatics. Amsterdam/Philadelphia, 1995.
33. GOOSSENS, Louis, Paul PAUWELS, Brygida RUDZKA-OSTYN, Anne-Marie
SIMON-VANDENBERGEN and Johan VANPARYS: By Word of Mouth. Metaphor,
metonymy and linguistic action in a cognitive perspective. Amsterdam/Philadelphia,
1995.
34. BARBE, Katharina: Irony in Context. Amsterdam/Philadelphia, 1995.
35. JUCKER, Andreas H. (ed.): Historical Pragmatics. Pragmatic developments in the
history of English. Amsterdam/Philadelphia, 1995.
36. CHILTON, Paul, Mikhail V. ILYIN and Jacob MEY: Political Discourse in Transi
tion in Eastern and Western Europe (1989-1991). Amsterdam/Philadelphia, n.y.p.
37. CARSTON, Robyn, Nam SUN SONG and Seiji UCHIDA (eds): Relevance Theory.
Applications and implications. Amsterdam/Philadelphia, n.y.p.
38. FRETHEIM, Thorstein and Jeanette K. GUNDEL (eds): Reference and Referent
Accessibility. Amsterdam/Philadelphia, n.y.p.
39. HERRING, Susan (ed.): Computer-Mediated Communication. Linguistic, social, and
cross-cultural perspectives. Amsterdam/Philadelphia, n.y.p.

Essays in Semantics and Pragmatics

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Essays in Semantics and Pragmatics

Uploaded by

Copyright:

Available Formats

ESSAYS IN SEMANTICS AND PRAGMATICS

Pragmatics & Beyond

Masayoshi Shibatani and Sandra Thompson (eds)

Essays in Semantics and Pragmatics

JOHN BENJAMINS PUBLISHING COMPANY

Library of Congress Cataloging-in-Publication Data

When someone is as energetic and productive as Chuck Fillmore, it is easy

a kind of continuum, the papers range from lexicography (Atkins) to

A predicate construction: Inversion in Polish

The role of the example in a frame semantics dictionary

Shifting face from Asia to Europe

Distance in construction grammar

Lexical information and the temporal interpretation of discourse

Reflections on metaphor and grammar

The pragmatics of case: On the relationship between

Ancestor-descendant and cultural-linguistic relativity

Remarks on suspended clauses:

Pragmaticization of meaning in some sentence-final particles

Recasting the deictic foundation, using physics and Finnish

Frame semantics and the lexicon:

The conceptual basis of performativity

Polish, like many languages, contains constructions in which a nominal with

(lb) (Dobrze) dzisiaj spaïam

(2a) Tç ksiązke czytalo się Jankowi z przyjemnoscią

(2b) Janek czytaï tç ksiązke (z przyjemnoscią)

associated with synthetic morphological expressions, i.e., wordforms

1. The Polish inversion construction2

Dziwirek observes that there are several characteristic properties associated

For purposes of brevity, I will focus primarily on so-called productive inversion

(3a) (Dobrze) dzisiaj spalam

(3b) Dobrze mi się dzisiaj spalo

(4a) Janek czytal tç ksiązkę (z przyjemnoscią)

(4b) Te ksiązkę czytalo się Jankowi z przyjemnoscią

mentally to it." (1989: 427)

2. Mapping theories and constructions

The attempt to establish principled relations between arguments with

An exposition of the specific properties associated with Polish predicates is

standard assumptions concerning mapping within generative grammars.

associated arguments is construed. One approach, which might be referred

2.1. An inversion predicate

entailments of the PREDICATES. Different predicates may entail different

2.1.1. Background for the proposal

a. volitional involvement in event or state a. undergoes change of state

These properties are hypothesized to be entailments following from the

Argument Selection Principle

semantics also sought by Gawron and Fillmore.9 However, these linguists

ARG Argument Level

CASE Case Marking

T. Mohanan suggests that, given the organization of semantic and

2.1,2. The Polish inversion predicate construction

the following representative PREDICATES serving as the semantic heads of

sleep <ARG> sleep rflx ADVERB <ARG> DAT

read <ARG1, ARG2> read rflx ADVERB <ARG1, ARG2> DAT

SUBJ OBJ SUBJ OBJ

NOM ACC ACC

These representations can be read as follows: (1) A PREDICATE, i.e., the

3. Case marking and grammatical functions14

One phenomenon purportedly keyed to the SUBJ status of nominals in

Detailed arguments for these assumptions are developed in Ackerman (1994).

The data in this section are taken from Dziwirek (1991).

condition proved to be empirically supportable. In fact, given the apparent

(5b) niinaa baala-ko uthaaegii

(6a) ravii-ne baala-ko uthaayaa

Another piece of core information about the frame consists of a compre

¤ meaning, including systematic and idiosyncratic polysemy, the differenti

meaning in a closed corpus (for instance, in a scholarly or a terminolog