Complementation

<DOCINFO
AUTHOR ""
TITLE "Complementation: Cognitive and functional perspectives"
SUBJECT "CELCR, Volume 34"
KEYWORDS ""
SIZE HEIGHT "220"
WIDTH "150">
Complementation
Converging Evidence
in Language and
Communication Research
Linguists have taken a broad view of language and are borrowing methods and findings from
other disciplines such as cognitive and computer sciences, neurology, biology, sociology,
psychology, and anthropology. This development has enriched our knowledge of language
and communication, but at the same time it has made it difficult for researchers in a
particular field of language studies to be aware of how their findings might relate to those in
other (sub-)disciplines.
celcr seeks to address this problem by taking a cross-disciplinary approach to the study of
language and communication. The books in the series focus on a specific linguistic topic and
offer studies pertaining to this topic from different disciplinary angles, thus taking
converging evidence in language and communication research as its basic methodology.
Works in the series are designed to appeal to linguists and non-linguists alike. Each work in
the series is written and/or coordinated by an expert in a particular field.
Editors
Marjolijn Verspoor Wilbert Spooren
University of Groningen Vrije Universiteit Amsterdam
m.h.verspoor@let.rug.nl w.spooren@let.vu.nl
Advisory Board
Colin Brown (Max Planck Institute of Psycholinguistics, Nijmegen)
Walter Daelemans (University of Antwerp)
Cliff Goddard (University of New England)
Roeland van Hout (Tilburg University and University of Nijmegen)
Leo Noordman (Tilburg University)
Martin Pütz (University of Koblenz-Landau)
Elizabeth Traugott (Stanford University)
Volume 1
Complementation: Cognitive and functional perspectives
Edited by Kaoru Horie
Complementation
Cognitive and functional perspectives
Edited by
Kaoru Horie
Tohoku University
John Benjamins Publishing Company

Amsterdam / Philadelphia
TMThe paper used in this publication meets the minimum requirements
8
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
Complementation : cognitive and funtional perspectives / edited by Kaoru Horie.

p. cm. (Converging evidence in language and communication research, issn
1566–7774 ; v. 1)
Proceedings from the 5th International Cognitive Linguistics Conference, held in July 1997
at Vrije Universiteit, Amsterdam, Netherlands.
Includes bibliographical references and index.
1. Grammar, Comparative and general--Complement--Congresses. 2. Cognitive
grammar--Congresses.
3. Functionalism (Linguistics)--Congresses. I. Horie, Kaoru. II. International Cognitive
Linguistics Conference (5th : 1997 : Amsterdam, Netherlands).
III. Series.
P299.C59.C66 2000
415--dc21 00-039744
isbn 90 272 3886 3 (Eur.) / 1 55619 211 8 (US) (Hb; alk. paper)
© 2000 – John Benjamins B.V.
No part of this book may be reproduced in any form, by print, photoprint, microfilm, or any
<TARGET
<REF
</REF
< /R/TREARGET
"pol
"po
"de
E"del
F"toc"
F DOCINFO
AUTHOR ""
"hor1">
"hor2">
"suz">
"ach">
"bec">
">
l">
"ver">
"toc">
TITLE "Table of Contents"
SUBJECT "Converging Evidence in Language and Communication Research, Volume 1"
KEYWORDS ""
SIZE HEIGHT "220"
WIDTH "150"
VOFFSET "4">
Table of contents
1. Introduction 1
Kaoru Horie and Bernard Comrie
2. Complementation in Japanese and Korean:
A contrastive and cognitive linguistic approach 11
Kaoru Horie
3. De dicto complementation in Japanese 33
Satoko Suzuki
4. Variation in Complementation Constructions:
Long-Distance Agreement in Tsez 59
Maria Polinsky
5. Construal and Complementation in French:
The Perspective Dimension 91
Michel Achard
6. Nominalization as complementation in Bella Coola and
Lushootseed 121
David Beck
7. Cognitive constraints on complement clause cliticization in
Spanish 149
Nicole Delbecque
8. Iconicity in English complement constructions:
Conceptual distance and cognitive processing levels 199
Marjolijn Verspoor
Glossary 227
Name index 235
Subject index 237
<LINK "hor1-r24">
"hor1-n*">
"hor1-r38">
"hor1-r4">
"hor1-r2">
"hor1-r23">
"hor1-r21">
"hor1-r3">
"hor1-r39">
"hor1-r36">
"hor1-r18">
<TARGET "hor1" DOCINFO
AUTHOR "Kaoru Horie and Bernard Comrie"
TITLE "Introduction"
KEYWORDS ""
SIZE HEIGHT "220"
WIDTH "150"
VOFFSET "4">
Chapter 1
Introduction*
Kaoru Horie and Bernard Comrie

Tohoku Univ. and Max Planck Inst. for Evolutionary Anthropology
This collected volume originated as a theme session entitled “Complementa-

tion: Interface of Cognitive, Functional and Typological Perspectives”, which
Kaoru Horie organized at the 5th International Cognitive Linguistics Confer-
ence (Vrije Universiteit, Amsterdam, The Netherlands) on July 18, 1997. Most
of the papers included in this volume were presented in this session. The papers
by Delbecque and Beck, which were presented elsewhere at that conference,
have been incorporated.
Complementation, i.e. predication manifested in argument slots, has
intrigued linguists of various theoretical persuasions, from cognitive/functional
linguists to typologists to generativists. After the initial stage of concentrated
generative grammatical research on the largely syntactic aspects of English
complementation in the 1960s to1970s (e.g. Rosenbaum 1967; Bresnan 1979),
linguists started to take the semantic aspects of complementation seriously.
An important body of research on the semantics of English complementa-
tion emerged in the 1960s to 1970s, e.g. Bolinger (1968, 1972, 1974), Kiparsky
and Kiparsky (1970), Karttunen (1971), Borkin (1973), Ross (1973), Riddle
(1974), Hooper (1975), Kirsner and Thompson (1976). Particularly noteworthy
is the series of works by Dwight Bolinger, who investigated the semantic
contributions made by different complement constructions in English, such as
the following pair of sentences (Bolinger’s original semantic characterization is
given in parentheses):
(1) a. John started to get angry. (potentiality)
b. John started getting angry. (reification)
(2) a. I saw John being obnoxious. (percept)
b. I saw John to be obnoxious. (concept)
<LINK "hor1-r41">
"hor1-r15">
"hor1-r31">
"hor1-r35">
"hor1-r29">
"hor1-r44">
"hor1-r12">
"hor1-r14">
"hor1-r19">
"hor1-r11">
"hor1-r7">
"hor1-r37">
"hor1-r1">
2 Kaoru Horie and Bernard Comrie
In the 1980s, the scope of research on complementation was expanded and

enriched by the work of functional-typological linguists, e.g. Givón (1980),
Noonan (1985), Ransom (1986), Lehmann (1988), Wierzbicka (1988). Especial-
ly important is Givón’s “binding hierarchy,” a typologically based hierarchy of
correlation between meaning and form in complementation. According to this
hierarchy, the degree of “semantic binding” (i.e. the matrix subject’s influence
on the event expressed in the complement clause) is closely correlated with the
degree of morpho-syntactic independence of the complement clause. For
example, the following sentences show that the stronger semantic binding of
causation, as compared to cognition, accounts for the lesser syntactic indepen-
dence of the causative verb complement clause from the matrix clause:
(3) a. She made him do the dishes (causative verb complement; lesser syn-
tactic independence in terms of tense-aspect-modality)
b. She thought that he had done the dishes. (cognition verb complement;
more syntactic independence in terms of tense-aspect-modality)
Building on this legacy, the 1990s have seen great theoretical enhancement and
enrichment of the research on the semantics of English complementation (e.g.
Dixon 1991; Frajzyngier and Jasperson 1991; Duffley 1992), as well as of the
typologically oriented research on complementation (e.g. Frajzyngier 1991;
Horie 1993; Dixon 1995; cf. also Horie, to appear). It must be noted, for the
sake of fairness, that formal linguists have continued to investigate complemen-
tation phenomena and have produced some important work in this area (cf.
Chomsky 1986; Rizzi 1990; Authier 1991, 1992).
Since the late 1980s, two differing but closely related research programs
have started to emerge, i.e. Cognitive Linguistics and Functional Linguistics.
Their different historical backgrounds and commonly shared assumption are
aptly summarized in the following quotation from Tomasello (1998: viii–ix):
Historically, Cognitive Linguistics grew out of the Chomskyian paradigm, as
several of its practioners, especially George Lakoff and Ronald Langacker,
became more and more dissatisfied with its narrow scope and its neglect of the
cognitive and social dimensions of linguistic communication. Functional Lin-
guistics (or Functional-Typological Linguistics) arose somewhat independently
based on studies of linguists such as Joseph Greenberg and Talmy Givón, who
were concerned with comparing different languages with one another typo-
logically and historically and who found that Chomskyian linguistics was of
little or no help in this enterprise (also important were linguists of the Prague
School who emphasized the importance of processes of communication in
<LINK "hor1-r32">
"hor1-r28">
"hor1-r26">
"hor1-r10">
"hor1-r40">
"hor1-r34">
"hor1-r6">
"hor1-r27">
"hor1-r20">
"hor1-r43">
"hor1-r8">
"hor1-r9">
"hor1-r5">
"hor1-r15">
Introduction 3
discourse and language). These two approaches have not yet congealed into
one coherent scientific paradigm, but there is widespread agreement among
both cognitive and functional linguists that language is not an autonomous
“mental organ,” but rather that it is a complex mosaic of cognitive and communi-
cative activities closely integrated with the rest of human psychology.
Both approaches, while sharing the assumption that human language capacity
is not autonomous, differ in terms of the relative emphasis they place on two
factors closely related to language, i.e. Cognition and Communication.
Cognitive linguists take most seriously the interaction of general human
cognitive ability and language, and employ such explanatory apparatus as
imagery, metaphor, conceptualization, and prototype (cf. Langacker 1987, 1991;
Lakoff 1987; Dirven and Verspoor 1998). Cognitive Linguistics, especially as
conceived of by Lakoff and Langacker, is based on the view that “language express-
es mental representations or coherent ideas, whether these be called gestalts,
schemas, images, or Idealized Cognitive Models (ICMs)” (Palmer 1996: 29).
Taylor (1998) presents the defining feature of Cognitive Linguistics, especially
Langacker’s version of Cognitive Linguistics (Cognitive Grammar) as follows:
By emphasizing the symbolic nature of language, Cog Ling is profoundly Saus-
surian in conception. For Saussure, the distinctively linguistic approach to the
study of language lay in the recognition that language is a symbolic system. (…)
Langacker’s programme, essentially, extends the Saussurian notion of the sign
to all formal categories of language; morphemes, words, syntactic construc-
tions, fixed expression, are all signs, linking a phonological and a semantic
representation. (ibid: 159–160, emphasis in the original)
Cognitive Linguistics allies closely with Linguistic Anthropology, obviously due
to their common interests in exploring symbolic and image-schematic organi-
zations of language (Palmer 1996). Cognitive Linguistics can also work col-
laboratively with such related disciplines as Cognitive Psychology, Neurosci-
ence, and Natural Language Processing, although such interdisciplinary joint
research possibilities have not yet been fully explored.
Functional linguists take most seriously the dynamic relation between
human communicative activities and language, and are concerned with the
manner in which linguistic structure is constrained by information structure
(cf. Chafe 1994; Lambrecht 1994; Kamio 1997; Van Valin and LaPolla 1997).
Functional linguistics closely allies with Linguistic Typology (cf. Comrie 1989;
Croft 1990; Bybee, Perkins, and Pagliuca 1994; Givón 1995) and Discourse
Studies (cf. Ochs, Schegloff and Thompson 1996). Functional Linguistics is less
easy to delineate than Cognitive Linguistics as the former encompasses more
<LINK "hor1-r15">
"hor1-r9">
"hor1-r30">
diverse theoretical stances than the latter. Croft (1995: 91) presents a typology
of differing functional linguistic positions as follows (cf. Newmeyer 1998 for a
criticism of Functional Linguistics):
a. syntax is arbitrary and self-contained (Autonomist Functionalism)
b. syntax is arbitrary, but not self-contained (Mixed Formal / Functionalism,
Typological Functionalism).
c. syntax is not arbitrary or self-contained (Extreme Functionalism)
Among these different functional linguistic stances, what Croft calls “Typo-
logical Functionalism” is most closely related to Cognitive Linguistics. In fact,
typological functionalists like Talmy Givón use the term “functionalism”
broadly enough to include some theoretical premises which may also be
claimed by Cognitive linguists, as shown below:
Functionalists in linguistics recite, vigorously and often, a set of cherished
premises:
– language is a social-cultural activity
– structures serve cognitive or communicative function
– structure is non-arbitrary, motivated, iconic
– change and variation are ever-present
– meaning is context-dependent and non-atomic
– categories are less-than-discrete
– structure is malleable, not rigid
– grammars are emergent
– rules of grammar allow some leakage (Givón 1995: 9)
The two disciplines, each having different traditions and theoretical inclina-
tions, can and should benefit more from each other’s strengths than has so far
been the case. Cognitive Linguistics probes deeply into highly abstract organiz-
ing principles of human cognition that relate to language use, such as Categori-
zation, Symbolization, Imagery and Metaphor. It excels in analyzing linguistic
phenomena in a single language based on the intuition and introspection of the
researcher, supplemented by other kinds of data (e.g. psycholinguistic). It must
be noted that the results achieved on the basis of a single language within the
Cognitive Linguistic paradigm can often have wider cross-linguistic implica-
tions. However, the primary concern of Cognitive Linguistics is with exploring
and explicating cognitive foundations of human language use; cross-linguistic
generalizations are desirable, but not necessarily intended as a primary goal.
Functional Linguistics is primarily concerned with explicating extralinguistic
communication-related factors such as Topichood, Givenness, Iconicity, and
<LINK "hor1-r13">
"hor1-r17">
"hor1-r33">
"hor1-r5">
"hor1-r9">
"hor1-r22">
"hor1-r42">
"hor1-r16">
"hor1-r25">
Introduction 5
Discourse organization that influence and help shape linguistic structure.

Closely allying with Discourse Studies and Linguistic Typology, Functional
Linguistics is concerned with obtaining spontaneous spoken data and aims at
presenting cross-linguistically applicable generalizations, e.g. implicational
hierarchy. Cognitive Linguistics can enhance its explanatory potentials by more
pronouncedly incorporating the orientations of Functional Linguistics toward
cross-linguistic generalizations and naturally occurring spoken data. Functional
Linguistics, in turn, can widen its scope of inquiry by employing theoretical
notions of Cognitive Linguistics and examining their validity against cross-
linguistic and discourse-oriented data.
In fact, we have started to find increasing convergence of Cognitive
Linguistics and Functional Linguistics in descriptions and explanations of such
complex linguistic phenomena as grammaticalization processes (cf. Heine,
Ulrike and Hünnemeyer 1991; Bybee, Perkins, and Pagliuca 1994; Horie 1997;
Ohori 1998) and various morpho-syntactic constructions (Bybee 1985; Croft
1991; Kemmer 1993; Uehara 1998). One of the symbolic events attesting the
convergence of these two research programs is the establishment of an annual
conference on Conceptual Structure, Discourse and Language, where Cognitive
and Functional linguists present and discuss their recent findings together (cf.
Goldberg 1996; Koenig 1998; Fox, Jurafsky, and Michaelis 1999).
This is a very natural and healthy progression for the two closely related
disciplines since, after all, Cognition and Communication are equally important
factors in accounting for language. As Joan Bybee states, “Cognition shapes
communication and communication shapes cognition and grammar” (July 25,
1998; Kansai Linguistic Society summer seminar, Kobe University, Japan).
With this historical background, the current volume presents converging
evidence on the Cognitive and Functional foundations of complementation.
The two most important foundational issues on which evidence converges are
the following:
i. Cognitive/Functional motivations for form-meaning (function) alignment
in complementation;
ii. Cognitive/ Functional factors governing the speaker’s choice of different
complement constructions/complementizers.
Each paper presents evidence in support of (i) and/or (ii) from a Cognitive or
Functional linguistic viewpoint.
<LINK "hor1-r14">
Kaoru Horie’s “Complementation in Japanese and Korean: A contrastive

and cognitive linguistic approach” investigates the nature of the contrast
between nominalized complement constructions in Japanese and Korean. Horie
argues that complement constructions in Japanese and Korean, in spite of their
superficial morpho-syntactic similarity, show sensitivity to different semantic
parameters. He proposes that different semantic parameters prioritized in the
complementation of Japanese and Korean are also manifested in other parts of
their lexico-grammatical structures, thereby showing a difference in the overall
contrastive linguistic profiles of the two languages.
Satoko Suzuki’s “De dicto complementation in Japanese” examines the non-
nominalized complement construction in Japanese headed by the complementizer
to. Suzuki’s study constitutes one of the first applications of the notion of psycho-
logical distance, which has been applied to other linguistic phenomena, to the
analysis of a Japanese complementizer. Suzuki links the notion of psychological
distance to the notion of “de dicto domain” (i.e. domain of speech) proposed by
Frajzyngier (1991). She argues that the association of “de dicto” characteristics with
a Japanese complementizer to and its variants is universal rather than unique to
Japanese. Furthermore, Suzuki relates the function of complementizers to other
uses of the same linguistic forms (e.g. commonly expressing the speaker’s
psychological distance), thereby supporting the notion of isomorphism (i.e.,
identical forms tend to reflect similarity in communicative function).
Maria Polinsky’s “Variation in complementation constructions: Agreement
climbing in Tsez” presents an information-structural account of variation in the
agreement pattern of the matrix verb which takes an object complement clause
in Tsez, a Nakh-Daghestanian language spoken in the Caucasus region.
Polinsky argues that the choice between the two patterns of agreement, i.e.
Agreement Climbing (the matrix verb agreeing with the absolutive nominal
inside the complement clause) and Matrix Agreement (the matrix verb agreeing
in grammatical noun class with the entire sentential complement), is deter-
mined by the information structure of the utterance, particularly the topicality
of the absolutive NP.
Michel Achard’s “Construal and complementation in French: The perspec-
tive dimension” applies the Cognitive Grammar notion of “perspective” to the
analysis of various facets of French complementation. Perspective is defined as
subsuming the vantage point from which the complement clause is conceptual-
ized, as well as the kind of viewing arrangement that exists between the con-
ceptualizer and the scene conceptualized. Achard argues that the notion of
perspective is crucial in explicating the issues of French complementation such
Introduction 7
as the characterization of French modal verbs and the semantic (cognitive)

distinction between infinitival and finite complements.
David Beck’s “Nominalization as complementation in Bella Coola and
Lushootseed” presents a cognitive linguistic analysis of the nominalized
complement constructions characterized by the prefix s’ in the Salishan
languages Bella Coola and Lushhootseed. Beck argues that while the two
languages share a lexical nominalization process that affixes the s’ prefix to a
bare stem, they differ in the scope that the nominalizer has over clause level
nominalizations. He further presents a comparison between the three-way
distinction in scope found in Bella Coola and Lushootseed nominalizations and
a Langackerian model of three types of nominalization based on English data.
Nicole Delbecque’s “Cognitive constraints on complement clause cliticiza-
tion in Spanish” explores from a cognitive linguistic viewpoint, particularly
drawing upon Fauconnier’s Mental Space theory, the complex interplay of
factors which govern the cliticization of a clausal complement in Spanish.
Delbecque argues that the attitude of the speaker with respect to the content of
the complement clause is particularly crucial in accounting for variable cliticiz-
ability of a clausal complement.
Marjolijn Verspoor’s “Iconicity in English complement constructions: Con-
ceptual distance and cognitive processing levels”, which concludes this collected
volume, argues that to account for English complement use, a conceptual
distinction needs to be made which is based on recent findings in neurobiology.
Specifically, Verspoor argues that English complement constructions involve
four parallel clines, i.e. perception, the need for motor movement, attributive
value, and categorization, along which their form-meaning correlation can be
mapped. She proposes that differing conceptual distance between subject and
event/ state of affairs expressed in the complement may be related to the
differing level of cognitive interaction involved, i.e. direct, indirect, or no direct
interaction. Verspoor’s paper shows how the array of different terms used in the
different traditions of complementation research (and also used by the authors
in this volume) are related to each other, and thus serves as a brief but informa-
tive overview of the papers collected in this volume.
Before concluding this brief introductory chapter, we would like to draw
attention to two prominent features of the papers included in this volume.
One noteworthy characteristic of this volume is “data-orientedness”. Each
researcher takes great pains in analyzing the rich and complex complementa-
tion phenomena in the language(s) being investigated without subjecting the
data to a particular theoretical inclination. In sum, all the papers in this
<DEST "hor1-r5">
"hor1-n*">
"hor1-r1">
"hor1-r2">
"hor1-r3">
"hor1-r4">
volume represent the “bottom-up” approach to complementation phenomena.

Another remarkable characteristic of this volume is the diversity of the
languages investigated, which include not only familiar languages such as
English, French, Japanese, Korean, and Spanish, but also not so widely studied
languages such as Bella Coola, Lushootseed and Tsez. This diversity of languag-
es analyzed in the volume certainly enriches our understanding of complemen-
tation phenomena across languages.
It is clear that this collected volume cannot cover the whole spectrum of
theoretical research on complementation within the Cognitive and Functional
linguistic frameworks. However, it is hoped that it will enable readers to
appreciate converging cross-linguistic evidence on cognitive and functional
foundations of complementation phenomena.
Notes
* The authors thank Andrew Barke, Debra Occhi, Maria Polinsky, Marjolijn Verspoor, and
an anonymous reviewer, for penetrating comments. An earlier version of this paper was
presented at the Workshop on Functionalist Approaches to Grammar (Organizer: Kaoru
Horie) at the 16th English Linguistic Society of Japan (November 7, 1998, Tohoku Uni-
versity, Sendai, Japan). Kaoru Horie was partially supported by the Grant-in-Aids from the
Ministry of Education, Science and Culture (No. 09710365, No.10610522) and a grant from
the Mishima Kaiun Memorial Foundation.
References
Authier, J.-Marc. 1991. “V-governed expletives, case theory, and the projection principle”.
Linguistic Inquiry 22.4: 721–740.
Authier, J.-Marc. 1992. “Iterated CPs and embedded topicalization”. Linguistic Inquiry 23.2:
329–336.
Bolinger, Dwight. 1968. “Entailment and the meaning of structures”. Glossa 2.2: 119–127.
Bolinger, Dwight. 1972. That’s That. The Hague: Mouton.
Bolinger, Dwight. 1974. “Concept and percept: two infinitive constructions and their
vicissitudes”. In Phonetic Society of Japan (ed.), World Papers in Phonetics: Festschrift for
Dr. Onishi’s Kizyu. Tokyo: Phonetic Society of Japan, 65–91.
Borkin, Ann. 1973. “To be and not to be”. Papers from the 9th Regional Meeting of the Chicago
Linguistic Society: 44–56.
Bresnan, Joan W. 1979. Theory of Complementation in English syntax. New York : Garland.
Bybee, Joan. 1985. Morphology. Amsterdam: John Benjamins.
<DEST "hor1-r9">
"hor1-r6">
"hor1-r7">
"hor1-r8">
"hor1-r10">
"hor1-r11">
"hor1-r12">
"hor1-r13">
"hor1-r14">
"hor1-r15">
"hor1-r16">
"hor1-r17">
"hor1-r18">
"hor1-r19">
"hor1-r20">
"hor1-r21">
"hor1-r22">
Introduction 9
Bybee, Joan, Revere Perkins, and William Pagliuca. 1994. The Evolution of Grammar. Tense,
Aspect, and Modality in the Languages of the World. Chicago: The University of Chicago
Press.
Chafe, Wallace. 1994. Discourse, Consciousness, and Time. Chicago: The University of
Chicago Press.
Chomsky, Noam. 1986. Barriers. Cambridge, MA: MIT Press.
Comrie, Bernard. 1989. Language Universals and Linguistic Typology. 2nd ed. Oxford:
Blackwell.
Croft, William. 1990. Typology and Universals. Cambridge: Cambridge University Press.
Croft, William. 1991. Syntactic Categories and Grammatical Relations: The Cognitive
Organization of Information. Chicago: The University of Chicago Press.
Croft, William. 1995. “Autonomy and functionalist linguistics”. Language. 71(3): 490–532.
Dirven, Rene and Marjolijn Verspoor. 1998. Cognitive Explorations into Language and
Linguistics. Amsterdam: John Benjamins.
Dixon, Robert W. 1991. A New Approach to English Grammar, on Semantic Principles.
Oxford: Oxford University Press.
Dixon, Robert W. 1995. “Complement clauses and complement strategies.” In F. R. Palmer
(ed.), Meaning and Grammar. Cambridge: Cambridge University Press, 174–220.
Duffley, Patrick J. 1992. The English Infinitive. London: Longman.
Fox, Barbara A., Dan Jurafsky, and Laura A. Michaelis. (eds). 1999. Cognition and Function
in Language. Stanford: CSLI.
Frajzyngier, Zygmunt. 1991. “The de dicto domain in language”. In Traugott, Elizabeth Closs
and Bernd Heine (eds), Approaches to Grammaticalization, Vol. 1. Amsterdam: John
Benjamins. 219–251.
Frajzyngier, Zygmunt and Robert Jasperson. 1991. “That-clauses and other complements”.
Lingua 83: 133–153.
Givón, Talmy. 1980. “The binding hierarchy and the typology of complements”. Studies in
Language 4(3): 333–377.
Givón, Talmy. 1995. Functionalism and Grammar. Amsterdam: John Benjamins.
Goldberg, Adele E. (ed.). 1996. Conceptual Structure, Discourse, and Language. Stanford: CSLI.
Heine, Bernd, Ulrike Claudi and Friederike Hünnemeyer. 1991. Grammaticalization: A
Conceptual Framework. Chicago: The University of Chicago Press.
Hooper, Joan B. 1975. “On assertive predicates”. In Kimball, John (ed.), Syntax and Se-
mantics 4. New York: Academic Press, 91–124.
Horie, Kaoru. 1993. A cross-linguistic study of perception and cognition verb complements: A
cognitive perspective. Unpublished doctoral dissertation. University of Southern
California. Los Angeles.
Horie, Kaoru. 1997. “Form-meaning interaction in diachrony: A case study from Japanese”.
English Linguistics 14: 428–449.
Horie, Kaoru. to appear. “Complement clauses”. In Haspelmath, Martin, Ekkehard Koenig,
Wulf Oesterreicher, and Wolfgang Raible (eds), Handbook of Language Typology and
Language Universals. Berlin: Walter de Gruyter.
Kamio, Akio (ed.). 1997. Directions in Functional Linguistics. Amsterdam: John Benjamins.
Karttunen, Lauri. 1971. “Implicative verbs”. Language 47: 340–358.
Kemmer, Susanne. 1993. The Middle Voice. Amsterdam: John Benjamins.
<DEST "hor1-r44">
"hor1-r23">
"hor1-r24">
"hor1-r25">
"hor1-r26">
"hor1-r27">
"hor1-r28">
"hor1-r29">
"hor1-r30">
"hor1-r31">
"hor1-r32">
"hor1-r33">
"hor1-r34">
"hor1-r35">
"hor1-r36">
"hor1-r37">
"hor1-r38">
"hor1-r39">
"hor1-r40">
"hor1-r41">
"hor1-r42">
"hor1-r43">
</TARGET "hor1">
Kiparsky, Paul and Carol Kiparsky. 1970. “Fact”. In Bierwisch, Manfred and Karl Erich
Heidolph (eds), Progress in Linguistics. The Hague: Mouton, 143–173.
Kirsner, Robert S. and Sandra A. Thompson. 1976. “The role of pragmatic inference in
semantics: A study of sensory verb complements in English”. Glossa 10(2): 200-240.
Koenig, Jean-Pierre. (ed.). 1998. Discourse and Cognition. Bridging the Gap. Stanford: CSLI
Lakoff, George. 1987. Women, Fire, and Dangerous Things. What Categories Reveal about the
Mind. Chicago: The University of Chicago Press.
Lambrecht, Knud. 1994. Information Structure and Sentence Form. Topic, Focus, and the
Mental Representations of Discourse Referents. Cambridge: Cambridge University Press.
Langacker, Ronald W. 1987. Foundations of Cognitive Grammar. Vol 1. Theoretical Pre-
requisites. Stanford: Stanford University Press.
Langacker, Ronald W. 1991. Foundations of Cognitive Grammar. Vol 2. Descriptive Appli-
cation. Stanford : Stanford University Press.
Lehmann, Christian. 1988. “Towards a typology of clause linkage”. In Haiman, John and
Sandra A. Thompson (eds), Clause Combining in Grammar and Discourse. Amsterdam,
John Benjamins. 181–225.
Newmeyer, Frederick J. 1998. Language Form and Language Function. Cambridge, MA: MIT
Press.
Noonan, Michel. 1985. “Complementation”. In Shopen, Timothy. (ed.). Language Typology
and Syntactic Description, Vol. 2. Complex Constructions. Cambridge, 42–140.
Ochs, Elinor, Emanuel A. Schegloff and Sandra A. Thompson (eds). 1996. Interaction and
Grammar. Cambridge: Cambridge University Press.
Ohori, Toshio. (ed.). 1998. Studies in Japanese Grammaticalization: Cognitive and Discourse
Perspectives. Tokyo: Kurosio Publishers.
Palmer, Gary B. 1996. Toward a Theory of Cultural Linguistics. Austin: University of Texas Press.
Ransom, Evelyn N. 1986. Complementation: Its Meaning and Form. Amsterdam: John
Benjamins.
Riddle, Elizabeth. 1974. “Some pragmatic conditions on complementizer choice”. Papers
from the 11th Regional Meeting of the Chicago Linguistic Society: 467–474.
Rizzi, Luigi. 1990. Relativized Minamality. Cambridge, MA: MIT Press.
Rosenbaum, Peter. 1967. The Grammar of English Predicate Complementation. Cambridge,
MA: MIT Press.
Ross, John Robert. 1973. “Nouniness”. In Fujimura, Osamu (ed.), Three Dimensions of
Linguistic Theory. Tokyo: TEC, 137–258.
Taylor, John R. 1998. “Back to basics”. Conference Handbook of the 16th National Conference of
the English Linguistic Society of Japan. Tokyo: English Linguistic Society of Japan, 158–163.
Tomasello, Michael. (ed.). 1998. The New Psychology of Language. Cognitive and Functional
Approaches. New York: Lawrence Erlbaum.
Uehara, Satoshi. 1998. Syntactic Categories in Japanese: A Cognitive-Typological Introduction.
Tokyo: Kurosio Publishers.
Van Valin Jr., Robert D. and Randy J. LaPolla 1997. Syntax: Structure, Meaning, and
Function. Cambridge: Cambridge University Press.
Wierzbicka, Anna. 1988. “The semantics of English complementation in a cross-linguistic
perspective”. In Wierzbicka, Anna. The Semantics of Grammar. Amsterdam: John
Benjamins, 23–168.
<LINK "hor2-n*">
<TARGET "hor2" DOCINFO
AUTHOR "Kaoru Horie"
TITLE "Complementation in Japanese and Korean"
KEYWORDS ""
SIZE HEIGHT "220"
WIDTH "150"
VOFFSET "4">
Chapter 2
Complementation in Japanese and Korean

A contrastive and cognitive linguistic approach*
Kaoru Horie
Tohoku University
1. Introduction
Japanese and Korean, whose relationship has been debated by a number of

linguists (e.g. Martin 1991), share a number of intriguing typological common-
alities in the domain of syntax such as case-marking, word order, and complex
clause formation.
A number of works on the syntax and semantics of Japanese and Korean
produced in the 1960s and 1970s examined complementation phenomena of
each language under the influence of the transformational generative approach,
notably Rosenbaum 1965 (e.g. Nakau 1973; Kuno 1973; Josephs 1976 and N.
McCawley 1978 on Japanese complementation, and Lee 1968 and Kim 1974,
among others, on Korean complementation).
In spite of recent interest in contrastive Japanese and Korean linguistics as
represented by the annually held Japanese/ Korean linguistic conferences (cf.
Hoji 1990; Clancy 1993; Choi 1993; Akatsuka 1994; Akatsuka, Iwasaki, and Strauss
1996; Sohn and Haig 1997; Akatsuka, Hoji, and Iwasaki 1998; Silva 1998), only a
few systematic contrastive linguistic studies on Japanese and Korean comple-
mentation have been produced (i.e. Song 1981; Horie 1991a, 1993a).
Building on cognitive/functional linguistic works on complementation, as
well as on language-specific descriptive works on complementation of each
language in particular, I explore the semantic basis of complementation in
Japanese and Korean from a contrastive linguistic and cognitive linguistic
perspective. Specifically, I argue that differing semantic orientations shown by
complementation in Japanese and Korean respectively are not accidental, and
that the difference in semantic orientations are manifested in other areas of the
12 Kaoru Horie
lexico-grammatical structures of these languages.

The organization of this chapter is as follows. Section 2 presents two types
of semantic distinctions encoded in the complement clauses of English, which
were originally proposed by Dwight Bolinger. Section 3 inquires into the
differing semantic bases of Japanese and Korean complementation and explores
their manifestations in the other areas of their lexico-grammatical structures.
Section 4 summarizes the discussion and presents a prospect for future study.
2. Semantic bases of complementation: A case study from English
As briefly discussed in Horie and Comrie’s introductory chapter, there have

been numerous studies on the semantics of complementation since the 1960s
to the present, focusing on a single or a few selected languages, or on typological
variation. As with many other grammatical phenomena, more studies seem to
have been produced on the complementation of English than on that of any
other language.
English in fact has a fairly complex system of finite and non-finite comple-
ment clauses, as shown in Table 1.
Table 1. Finite and non-finite complement clauses in English

Finite Non-finite
that-clause to-infinitive
wh-clause bare infinitive
participle
gerund
Each of the finite and non-finite complement clauses in English is illustrated

below:
(1) I learned that Mary broke up with her girlfriend. (that-clause)
(2) Bill wondered whether he should attend the meeting. (wh-clause)
(3) The army forced the civilians to leave the town. (to-infinitive)
(4) Why don’t you let her drive ? (bare infinitive)
(5) Have you heard him singing ? (participle)
(6) I regret buying the book. (gerund)
Complementation in Japanese and Korean 13
A variety of semantic distinctions have been proposed on the basis of co-

occurrence of the English complement clause with differing semantic classes of
predicates, e.g. “factive” vs. “non-factive” (Kiparsky and Kiparsky 1970),
“implicative” vs. “non-implicative” (Karttunen 1971), and “assertive” vs. “non-
assertive” (Hooper 1975). However, it was Dwight Bolinger who was most
keenly aware of the subtle semantic distinctions manifested when the two types
of complement are in contrast.
As briefly noted in Section 1, Bolinger (1968, 1974) argued for the following
two types of semantic contrast exhibited by two differing combinations of non-
finite complements in English, as shown below:
(7) a. John started getting angry. (gerund: reification)
b. John started to get angry. (to-infinitive: potentiality)
(8) a. I saw John being obnoxious. (bare-infinitive: percept)
b. I saw John to be obnoxious. (to-infinitive: concept)
The first semantic distinction, i.e. “reification” vs. “potentiality”, closely

corresponds to the distinction between realis and irrealis, which has been
widely applicable to various types of grammatical phenomena other than
complementation (e.g. mood, modality) by linguists like Noriko Akatsuka and
Talmy Givón (cf. Akatsuka 1985; Givón 1995). The second semantic distinc-
tion, i.e. “percept” vs. “concept”, has also been referred to as event vs. proposi-
tion (Vendler 1967) and second-order entities vs. third-order entities (Lyons
1977). The two types of semantic distinctions, though not unrelated to each
other, are different. The first distinction concerns whether a state-of-affairs
actually happens in the real world or is arrested in potentiality. The second
distinction relates to whether a state-of-affairs is perceived by the senses or is
held in mind. In what follows, I will use the more familiar latter sets of terms,
i.e. “realis” and “irrealis”, and “event” and “proposition”, to refer to each
semantic distinction.
English non-finite complements are thus capable of encoding two types
of semantic distinctions. From cognitive linguistic and contrastive linguistic
viewpoints, it is an intriguing question whether Japanese and Korean, which
allegedly show much typological similarity, behave similarly or differently in
terms of the semantic distinctions encoded in their complement systems. I will
pursue this question by employing the two types of semantic distinctions
discussed above.
14 Kaoru Horie
3. Semantic bases of complementation in Japanese and Korean
This section contrasts complementation phenomena in Japanese and Korean

and examines whether the two languages behave similarly or not in terms of the
semantic distinctions encoded therein.
3.1 Nominalization as a prominent syntactic feature characterizing

Japanese and Korean complementation
Languages that have complementation differ considerably in terms of the extent

to which nominalization (cf. Ross 1973) is involved in forming complement
clauses. Many European languages possess two types of complement clauses, i.e.
a less nominal finite clause complement and a more nominal infinitive comple-
ment. Complementation in Japanese and Korean can be characterized by the
extensive use of nominalization as both languages productively use “nominal-
izers” in forming complement clauses:
(9) a. [Asobu] koto-to onazi-kurai-ni, [hito-ni
[play nmlzr-with same-about-cop:conj [person-to
doozyoosuru] koto-ga suki-de aru.
sympathize nmlzr-nom liking-cop:pol
b. [No-nun] kes-kwa machan-kacilo, [nam-ul
[play-adn:pres nmlzr-with same-likewise person-acc
tongcengha-nun] kes-ul cohaha-n-ta.
sympathize-adn:pres nmlzr-acc like-pres-decl
‘(She) likes having fun about as much as (she) likes to sympathize
with others.’
(Ilpontayekmwunko 9: 116; English translation by KH)
In (9a) and (9b), nominalizers koto and kes form the complement clauses of the
matrix clause predicates suki-da and cohahan-ta.
In fact, as shown in Table 2, the majority of complementizers in Japanese
and Korean involve nominalization.1
This paper draws particular attention to the nominalized complements in
Japanese and Korean, a prominent typological commonality between the two
languages. The next subsection presents the results of an earlier attempt I made at
exploring the semantic bases of Japanese and Korean nominalized complements.
Table 2. Complementizers in Japanese and Korean

Nominalized Non-nominalized
Japanese koto, no, tokoro to
Korean kes, ki, (u)m 2 ko
3.2 An earlier exploration into the semantic bases of Japanese

and Korean complementation
In Horie (1991a, 1993a), I presented a preliminary contrastive semantic analysis

of a set of Japanese and Korean nominalizers. Specifically, I argued that
Japanese and Korean radically differ in terms of the way each language encodes
the distinction between what I called “Directly Perceived Event” and “Indirectly
Perceived Event,” or in more familiar terms, “event” and “proposition,” as
shown in Table 3.3
Table 3. Linguistic coding of “event” and “proposition” in Japanese and Korean by

means of nominalizers (Horie 1991a, 1993a)
Event Proposition
Japanese no koto
Korean kes kes/ki/(u)m
What the figure represents is the following. Japanese distinguishes between

“event” and “proposition” by using different nominalizers, i.e. no and koto,
whereas the same nominalizer kes is capable of encoding both “event” and
“proposition” in Korean. Korean additionally has two other nominalizers for
encoding “proposition,” i.e. ki, and (u)m. The following sentences illustrate the
coding difference between the two languages:
(10) (Japanese)
a. Mary-wa [John-ga toori-o wataru] no-o mi-ta. (event)
Mary-top [John-nom street-acc cross nmlzr-acc see-past
‘Mary saw John cross the street.’
b. Mary-wa [John-ga toori-o watat-ta] koto-o sit-ta.
Mary-top [John-nom street-acc cross-past nmlzr-acc know-past
‘Mary learned that John had crossed the street.’ (proposition)
16 Kaoru Horie
(11) (Korean)
a. Mary-nun [John-i kil-ul kenne-nun] kes-ul
Mary-top [John-nom street-acc cross-adn:pres nmlzr-acc
po-ass-ta. (event)
see-past-decl
b. Mary-nun [John-i kil-ul kenne-n] kes-ul
Mary-top [nom street-acc cross-adn:past nmlzr-acc
al-ass-ta. (proposition)
know-past-decl
‘Mary learned that John had crossed the street.’
c. Mary-nun [John-i kil-ul kenne-ss] um-ul
Mary-top [John-nom street-acc cross-past nmlzr-acc
al-ass-ta. (proposition)
know-past-decl
d. Mary-nun [John-i sihem-ey hapkyekha]-ki-lul pala-n-ta.
Mary-top [John-nom exam-in pass-nmlzr-acc hope-pres-decl
‘Mary hopes that John will pass the exam.’ (proposition)
While the descriptions presented above are factually correct, it has become
increasingly obvious that the following two points must also be taken into
consideration:
1. Do Japanese and Korean complementation encode different types of
semantic distinctions?
2. How does complementation in Japanese and Korean cohere with the
overall semantic typology of the languages?
Point (1) was inspired by Dixon (1995), who argued for the universality of the
“realis/irrealis” distinction, as mentioned in Section 2. Dixon’s work led me to
reexamine the coding differences noted between Japanese and Korean in my
earlier works (e.g. Horie 1990, 1991a, 1993a), which may have centered on the
distinction which exists in Japanese.4
Point (2) was inspired by Hawkins (1986), who proposed a “comparative
semantic typology” of English and German by pointing out that English
consistently tends toward greater surface ambiguities than German, as shown
for instance by the English “subject” encoding far more varied semantic roles
(e.g. Agent, Location, Instrument) than its German counterpart (typically
restricted to Agent). This perspective led me to examine the complementation
phenomena in Japanese and Korean, not in isolation, but in relation to the

overall semantic typology of the languages (cf. also Horie 1998a, b, c, Horie and
Sassa, in press; Horie and Kang, in press)
From a contrastive linguistic perspective taking these two points into
account, I suggest that while Japanese and Korean both use nominalization
extensively in forming complement clauses, each language can exhibit a
different preference for the type of cognitive distinction it prioritizes and
encodes by means of nominalization. Furthermore, these different “preferenc-
es”, or semantic orientations, are reflected in other areas of their lexico-gram-
matical structures as well.
In the following subsection, I will analyze sentential nominalizers in
Japanese and Korean from this contrastive perspective and explore their
different semantic bases.
3.3 Nominalizers in Japanese and Korean complementation

and the type of semantic distinctions preferentially encoded
This section inquires into the semantic bases of Japanese and Korean comple-
mentation respectively in 3.3.1 and 3.3.2 from a contrastive linguistic perspective.
3.3.1 The semantic basis of Japanese complementation

Modern Japanese has three sentential nominalizers which can participate in
complementation, i.e. koto, no, tokoro. The three nominalizers are not evenly
distributed. No and koto are the most commonly used sentential nominalizers
in forming complementation in Modern Japanese.
There is agreement among linguists of Japanese that no and koto respective-
ly encode “concreteness/ directness” and “abstractness/ indirectness” (Kuno
1973; Josephs 1976; N. McCawley 1978; Makino 1983; Horie 1990, 1991a, b,
1993a, 1997; Shinzato 1996). These characterizations correspond to what we call
“event” and “proposition,” as shown below:
(12) a. Mary-wa [John-ga toori-o wataru] no-o mi-ta. [=10.] (event)
b. Mary [John-ga toori-o watat-ta ] koto-o sit-ta. (proposition)
No is also capable of encoding “proposition” when it co-occurs with some

complement-taking predicates like siru (“learn”), but koto doesn’t normally
encode “event”, as shown below:
18 Kaoru Horie
(12) b¢. Mary-wa [John-ga toori-o watat-ta ] no-o sit-ta. (proposition)

a¢. *Mary-wa [John-ga toori-o wataru] koto-o mi-ta. (event)
This suggests that no is semantically less specific than koto. No, which arguably
originated as a genitive marker, later developed an indefinite pronominal
function comparable to that of English one, and started to function as a general
sentential nominalizer relatively recently (cf. Horie 1993b, 1995, 1999). Another
less widely supported view is that no is a truncated form of the lexical noun
mono meaning “concrete, tangible thing.” In either historical scenario, it is
understandable why the sentential nominalizer no is associated with such
semantic properties as “event/concreteness/directness” if it developed from a
pronoun used to replace words for concrete things or from a lexical noun itself
meaning “concrete, tangible thing.” Since no is strongly associated with
concreteness/directness because of its historical origin, it does not completely
encroach upon the more abstract “propositional” domain of the other nominal-
izer koto. In fact, many complement-taking predicates of propositional attitude
do not accept no or accept it only marginally:
(13) [Kono-ba kara suguni tatisaru] {koto/?no}-o tuyoku yookyuusuru.
[this-place from soon leave {nmlzr-acc strongly demand
‘(We) strongly demand that (you) leave here immediately.’
Koto originated as a lexical noun meaning “abstract matter,” so it is very natural
that it has developed into a sentential nominalizer encoding a “propositonal”
complement rather than an “event” complement.
Finally, sentential nominalizer tokoro is severely restricted in terms of its
distribution as a complementizer. It basically encodes a subset of “event”
complements encodable by no, i.e. those complements denoting visually
perceived events such as the following:
(14) [Ressya-ga dassensuru] {tokoro/no}-o mokugekisi-ta.
[train-nom get derailed {nmlzr-acc witness-past
‘I witnessed the train get derailed.’
In fact, when the complement denotes an event perceived by the senses other
than vision, e.g. auditorily perceived event, tokoro doesn’t fare well, unlike no:
(15) [Ane-ga huruuto-o huku] {no/?tokoro}-o hazimete kii-ta.
[elder sister-nom flute-acc play {nmlzr-acc first time hear-past
‘For the first time I heard my elder sister play the flute.’
Tokoro originated as a lexical noun meaning “place, location,” which is per-

ceived and always identified by vision. Therefore, it is very natural that tokoro
developed into a sentential nominalizer encoding a “visually perceived event”
complement. Presence of the nominalizer tokoro, which is specialized to encode
only a visually perceivable subset of events, underlines the importance attached
to the semantic distinction between event and proposition in Japanese. I will
discuss this issue in Section 3.4.
Table 4 illustrates the sentential nominalizers in Japanese and their most
typical semantic functions.
Table 4. Sentential nominalizers in Japanese and their most typical semantic functions
in complementation
Event Proposition
no koto
tokoro (visually perceived event)
The following subsection explores the semantic basis of Korean complementation.
3.3.2 The semantic basis of Korean complementation

Modern Korean has three sentential nominalizers, kes, (u)m, and ki. Korean
nominalizers (u)m and ki appear to be very different from Japanese nominalizers
morpho-syntactically as well as semantically. Morpho-syntactically, (u)m and ki
are directly attached to the verb stem (e.g. po of pota “see”) and produce derived
nouns. Chang (1996) provides a few illustrative examples of derived nouns:
(16) a. el “freeze” Æ elum “ice”; nol- “play” Æ nolum “gambling”;
sey “count” Æ seym “calculation”
b. po “see” Æ poki “sample”; kopha “double” Æ kophaki “multiplica-
tion”(ibid: 33; emphasis added)
Chang notes that “such derived nouns are fossilized or frozen and the original
verbal nature becomes defunct” (ibid: 33). None of the Japanese sentential
nominalizers mentioned above is involved in a noun derivation process similar
to (u)m and ki.5
(U)m and ki also differ considerably from the Japanese nominalizers in
terms of the semantic distinctions encoded. Following Kiparsky and Kiparsky
(1970), many Korean linguists (e.g. Kim 1984; Sohn 1994) characterized the
semantic contrast between (u)m and ki in terms of “factive” and “nonfactive.”
20 Kaoru Horie
Kim (1984), for instance, states that “factive predicates require that noun-
phrase complements be headed by the complementizer um, while nonfactive
predicates require that complement clauses be nominalized by the nominalizer
ki” (ibid: 45). However, it seems more appropriate to adopt the terms “realis
(actuality)” and “irrealis (potentiality)” instead to characterize the comple-
ments headed by each nominalizer if we closely examine the semantic class of
predicates co-occurring with each nominalizer.6 Consider the following list of
predicates provided by Kim (1984: 38):
(17) a. Predicates co-occurring with (u)m
alta “to know,” icta “to forget,” thamcihata “to detect,” mwusihata
“to ignore,” kangcohata “to emphasize,” allyecwuta “to inform,”
hwaksilhata “certain,” pwunmyenghata “obvious,” isanghata
“strange,” kekcenghata “to worry about,” selmyenghata “to explain,”
hwuhoyhata “to regret,” sulphehata “to feel sorrowful,” mianhata
“to feel sorry,” uymihata “to mean,” nathanayta “to reveal,” kayepsta
“to feel pitiful,” kkulta “to attract”
b. Predicates co-occurring with ki
wenhata “to want,” palata “to want,” kkelita “to abhor,” memchwuta
“to stop,” kyeysokhata “to continue,” kwenhata “to urge,” yokwuhata
“to demand,” myenglyenghata “to command,” ceyanhata “to pro-
pose,” swipta “easy,” elyepta “difficult,” pwulpheynhata “uncomfort-
able,” konlanhata “difficult,” silhta “to dislike,” kepwuhata “to re-
ject,” yaksokhata “to promise,” huymanghata “to hope”
Typical examples of complement clauses nominalized by ki and (u)m are shown

below (examples 18 and 19 and their translations are from Chang 1996: 145, 144):
(18) [Ku phikoin-i kyelpaykha]-m-i cungmyentoy-ess-ta. (realis)
[the defendant-nom innocent-nmlzr-nom be proved-past-decl
‘It has been proved that the defendant is innocent.’
(19) [Nayil pi-ka o]-ki (-ka) swipta. (irrealis)
[tomorrow rain-nom come-nmlzr-nom easy
‘It is likely that it will rain tomorrow.’
In contrast, as shown in Section 3.2, the two nominalizers do not encode the
distinction between “event” and “proposition” overtly, as they both encode
proposition
(11) c. Mary-nun [John-i kil-ul kenne-ss]-um-ul al-ass-ta. (proposition)
d. Mary-nun [John-i sihem-ey hapkyekha]-ki-lul pala-n-ta.

‘Mary hopes that John will pass the exam.’ (proposition)
Finally, kes is least semantically restricted in that it can encode both “realis” and
“irrealis” complements:
(20) a. Mary-nun [John-i sihem-ey hapkyekha-n]-kes-ul
Mary-top [John-nom exam-in pass-adn:past-nmlzr-acc
al-ass-ta. (realis)
know-pres-decl
‘Mary learned that John passed the exam.’
b. Mary-nun [John-i sihem-ey hapkyekha-l]-kes-lul
Mary-top [John-nom exam-in pass-adn:fut-nmlzr-acc
pala-n-ta. (irrealis)
hope-pres-decl
‘Mary hopes that John will pass the exam.’
Likewise, as shown in Section 3.2, kes encodes both “event” and “proposition”:
(11) a. Mary-nun [John-i kil-ul kenne-nun] kes-ul po-ass-ta. (event)
b. Mary-nun [John-i kil-ul kenne-n] kes-ul al-ass-ta. (proposition)
Kes is similar to Japanese nominalizers koto and tokoro in that it originated as a

lexical noun. Especially interesting from a historical comparative viewpoint is
the similarity between Korean kes and Japanese koto in terms of morphological
shape, a point which Martin (1991: 286) lists as one of the grammatical corre-
spondences between the two languages which might suggest a common origin.
However, the crucial difference between Korean kes and Japanese koto is the fact
that Korean kes encodes both “concrete object” and “abstract matter”, whereas
Japanese koto encodes “abstract matter” only. This semantic flexibility of kes
enables it to encode both “event” and “proposition” complements as in (11a)
and (11b), unlike Japanese koto, which encodes only “proposition” (cf. 12b).
The discussion presented in this section strongly suggests that the semantic
distinction between “realis” and “irrealis” is more overtly encoded in the
complementation of Korean than the distinction between “event” and “propo-
sition” by means of the pair of nominalizers (u)m and ki. Unlike Japanese,
however, Korean lacks a pair of nominalizers such as no and koto, which
prototypically encode the latter distinction.
Table 5 below summarizes the semantic distinctions encoded by each of the
22 Kaoru Horie
four nominalizers in Korean (NB. “event” and “proposition” are put in

parentheses as this distinction is not overly encoded).
Table 5. Sentential nominalizers in Korean and their most typical semantic functions
in complementation
realis irrealis (event) (proposition)
kes kes kes kes
ki (u)m ki,(u)m
The following subsection explores a correlation between the differing semantic

bases of Japanese and Korean complementation and the other areas of the
lexico-grammatical structures of these languages.
3.4 How do the semantic bases of Japanese and Korean complementation

correlate with their lexico-grammatical structures?
When we contrast the overall coding tendencies of nominalizers in Japanese

and Korean presented in Tables 4 and 5, we notice the following strikingly
different coding preferences between the two languages:
1. Japanese nominalizers encode the semantic distinction between “event”
and “proposition” by employing two nominalizers, i.e. semantically more
specific tokoro and more general no in the domain of “event,” as contrasted
with the more abstract “propositional” nominalizer koto;
2. In contrast, Korean nominalizers encode the semantic distinction between
“realis” and “irrealis”, which presents itself in the choice between the “realis”
nominalizer (u)m and the “irrealis” nominalizer ki. The other nominalizer
kes, which is semantically less specific, neutralizes the distinction.
This difference between Japanese and Korean is intriguing and certainly calls
for explanation, because the two languages are known to have remarkably
similar grammatical structures yet reveal fundamental semantic differences in
the area of sentential nominalization.
I argue that it is no accident that each language has a coding preference for
a particular semantic distinction in its sentential nominalization. In the case of
Japanese, the semantic distinction between “concrete object” and “abstract
matter” is a very basic one lexically manifested by means of two nouns mono
“concrete, tangible thing” and koto “abstract matter”. McGloin (1989:110) gives
the following semantic characterizations of the two nouns (cf. also Ohori 1998):
When used as a noun, koto still represents an abstract entity, such as subject
matter, knowledge, etc., and thus should be clearly distinguished from mono
which would represent concrete objects.
The following examples illustrate these semantic characterizations (ibid, glosses
by KH):
(21) a. Iroirona {koto/*mono}-o narai-masi-ta.
various {koto/*mono}-acc learn-pol-past
‘I learned various things.’
b. Iroirona {mono/*koto}-o kai-masi-ta.
various {mono/*koto}-acc buy-pol-past
‘I bought various things.’
It is therefore not surprising that Japanese complementation encodes the distinc-

tion between “event” and “proposition”, which respectively corresponds to the
distinction “concrete objects” and “abstract matter” maintained at the lexical level.
The parallelism between the two types of semantic distinctions maintained in
the lexicon and the grammar of Japanese can be illustrated in Table 6.
Table 6. Parallelism between the semantic distinctions maintained in the lexicon and
grammar of Japanese
Lexicon: mono (“concrete object”) vs. koto (“abstract matter”)
Grammar (complementation): no (“event”) vs. koto (“proposition”)
Superficially, Korean also seems to make a similar concrete-abstract distinction

at the lexical level as it possesses two lexical nouns kes and il. These nouns
apparently correspond to the Japanese nouns mono and koto respectively. The
Korean noun kes, the definitions of which include “a one, the one (=person,
thing)” and “the one that” (Martin and Lee 1967: 103), appears to cover a
similar range of “concrete object” meaning as does the Japanese noun mono.
However, as noted by Hamada (1983: 172), unlike Japanese koto, it seems that
il is more regularly used as a noun encoding “job, work” rather than as a noun
encoding highly generalized “abstract matter”. This observation is also con-
firmed by Martin (1992: 560), whose semantic characterizations of il started
with “a job, a task, a piece of work, business; a matter, a deal”. This suggests that
it is less common to contrast and distinguish “concrete object” and “abstract
matter” in Korean than it is in Japanese.
The higher degree of attention that Japanese pays to “event”, or “directly
24 Kaoru Horie
perceptible phenomena” in the Japanese language was a topic pursued by

Makino (1983). By contrasting Japanese with English, Makino adduced various
kinds of linguistic evidence in favor of his claim that Japanese is more sensitive
to directly perceptible phenomena than English. Not surprisingly, the evidence
includes the distinction between nominalizers no and koto, as well as such
phenomena as abundance in sound symbolisms, frequent use of metaphorical
expressions evoking concrete images, and existence of emotive and desiderative
predicates felicitously used only by the speaker (e.g. kanasii “(I’m) sad” vs.
kanasi-garu “(s/he) looks sad”).
Many of the linguistic phenomena presented by Makino are also present in
Korean (e.g. abundance in sound symbolisms, existence of predicates felicitous-
ly used only by the speaker such as -ko siphta “(I) want to (do)…” vs. -ko
siphehanta “(s/he) wants to (do)…”). This means that Korean grammar also
shows sensitivity to “directly perceptible phenomena”. However, when con-
trasted with Japanese, where both lexicon and grammar encode the semantic
distinction between “concrete object/event” and “abstract matter/proposition”,
Korean appears to pay relatively less attention to the distinction.
Now that a possible association has been identified between the particular
semantic distinction encoded in Japanese complementation and other lexico-
grammatical features of the language, I will investigate whether the semantic
distinction between “realis” and “irrealis” encoded in Korean complementation
has any manifestations in other areas of its lexico-grammatical structure.
It is not entirely clear at this stage why Korean nominalizers (u)m and ki
respectively have a coding preference for “realis” and “irrealis,” because the
lexical origins of these nominalizers, if any, are hard to determine. However, if
one examines different areas of Korean grammar, one can find an indication of
the Korean preference for the “realis” and “irrealis” distinction. One of the
areas of grammar where Korean differs markedly from Japanese is the tense and
mood marking of predicates. Unlike Japanese, Korean distinguishes matrix
predicate endings and adnominal predicate suffixes. In the case of matrix
predicates, Korean has a set of suffixes attached directly to their stems to
indicate tense and mood categories as shown in Table 7. Crucially, these tense
and mood suffixes appear to represent the semantic distinction between “realis”
and “irrealis”, which is also manifested in the Korean complementation.
Particularly noteworthy in Table 7 is the presence of various “realis” tense
suffixes and the “irrealis” mood suffix -keyss-. The grammatical function of
-keyss- is defined by Chang (1996: 128) as follows:
The volitional marker keyss, which is historically a contraction of kes ‘fact’ and
iss ‘exist’ (keyss < kes iss), denotes the speaker’s or (the addressee’s, in inter-
rogatives) volition, intention, or prediction in the future. Semantically, the
volitional keyss is akin to the English modal ‘will.’
Table 7. Korean matrix predicate tense and mood suffixes

(based on Umeda 1991: 88–93)
Realis Irrealis
Present Past Remote Past Mood (Intention/Prediction)
Verb -n- -ess- -ess ess- -keyss-
Adjective zero -ess- -ess ess- -keyss-
Existential predicate zero -ess- (nonexistent) -keyss-
Copula zero -ess- -ess ess- -keyss-
Korean has a similar elaborate system of adnominal predicate tense and mood
suffixes, as shown in Table 8.
Table 8. Korean adnominal predicate tense/mood suffixes

(based on Umeda 1991: 22)
Realis Irrealis
Present Past Future/Probability
Verb -nun -(u)n (Perfective) -(u)l
-ten (Imperfective)
-essten (Past experience)
Existential predicates -nun -ten -ul
-essten
Adjective -(u)n -ten -(u)l
-essten
Copula -n -ten -l
-essten
In contrast, Japanese, which largely lacks the distinction between matrix

and adnominal predicate forms (cf. Horie 1993b, 1995) except for the nominal
adjectives and the copula, does not have as elaborate a system of tense and
mood suffixes as Korean, as shown in Table 9.
Japanese matrix and adnominal predicate suffixes are not as elaborate as
their Korean counterparts and do not encode the semantic distinction between
26 Kaoru Horie
Table 9. Japanese matrix and adnominal predicate tense suffixes

Nonpast Past
Verb -(r)u -ta
Adjective -i -katta
Nominal adjective, Copula -da (matrix) -datta
-na (adnominal)
“realis” and “irrealis” since the so-called “non-past” predicate forms, especially
in the case of verbs, merge present tense (“realis”) and future tense/ mood
(“irrealis”).7 In contrast, the semantic distinction between “realis” and “irrealis”
is very crucial to Korean and is therefore encoded in complementation and
predicate tense and mood suffixes.
In sum, this subsection has shown that the different semantic bases of
Japanese and Korean complementation are not accidental, and are intimately
correlated with the other areas of their lexico-grammatical structures.
4. Conclusion
This paper contrasted complementation phenomena in Japanese and Korean

from a cognitive linguistic perspective, and concluded that each language has a
different coding preference for the type of semantic distinction encoded in
complementation. Crucially, the differing semantic bases of complementation
systems in Japanese and Korean are also manifested in the other areas of their
lexico-grammatical structures. Further research is necessary to explore historical
linguistic reasons, if any, for the Korean preference for a “realis”–“irrealis”
distinction. Also, a wide-scale cross-linguistic study of the semantics of com-
plementation will be indispensable in confirming what may be fundamental
semantic distinctions universally encoded in complementation.
This contrastive study of Japanese and Korean sentential nominalizations
poses an interesting challenge to the language universal for “realis” and “irrealis”
in complementation proposed by Dixon (1995). While it is largely true that
“realis”–“irrealis” is a most fundamental semantic distinction encoded in
complementation strategies of the languages of the world, there are languages
like Japanese which are more or at least equally sensitive to the coding of another
cross-linguistically important semantic distinction, i.e. the distinction between
“event” and “proposition,” in complementation.
<DEST "hor2-n*">
Abbreviations
acc Accusative neg Negative

adn Adnominal nmlzr Nominalizer
conj Conjunctive nom Nominative
cop Copula pres Present
decl Declarative sfp Sentence Final Particle
fut Future top Topic
gen Genitive
Notes
* I thank the participants of the theme session on complementation held at the 5th Inter-
national Conference on Cognitive Linguistics, particularly Bernard Comrie, Susanne Schüle,
and Marjolijn Verspoor for invaluable comments. Thanks also go to Andrew Barke, Chun
Chonghoon, Atsuko Kikuchi, Debra Occhi, Kaori Taira, and two anonymous reviewers for
their insightful comments on the earlier version of this paper. I particularly thank an
anonymous reviewer for constructive criticism, which has greatly enhanced the argumentation
in this paper. The usual disclaimer applies. The current research project was partially
supported by the Grant-in-Aids from the Ministry of Education, Science and Culture (No.
09710365, No.10610522) and a grant from the Mishima Kaiun Memorial Foundation.
1. Japanese and Korean each have a complement clause which doesn’t involve nominalization,
i.e. to-clause in Japanese and ko-clause in Korean:
a. “Asita-wa ko-nai yo,” to (*-o) it-ta.
“tomorrow-top come-neg sfp comp(*-acc say-past
b. “Nayil-un o-ci an-nun-ta,” ko (*-lul) hay-ss-ta.
“tomorrow-top come-nmlzr neg-adn:pres-decl comp (*-acc say-past-decl
‘(He) said, “I won’t come tomorrow” .’
That these complement clauses do not involve nominalization can be observed from the
non-occurrence of the accusative markers o and lul, which are obligatory when preceded by
sentential nominalizers such as koto and kes in (9a) and (9b). Finiteness is not a
straightforward notion easily definable in languages like Japanese and Korean compared to
English, but it is safe to say that the italicized complements in (a) and (b) are more finite than
the nominalized complements shown in (9a) and (9b). As such, finite complements headed
respectively by to and ko in Japanese and Korean are used to encode complements to utterance
predicates and verbs of cognition, a widely observed cross-linguistic phenomenon which is
predicable under Givón’s binding hierarchy (cf. Givón 1980). In what follows, I will not
discuss finite complementation in Japanese and Korean. Suzuki (this volume) partially
complements the current paper as it extensively discusses several types of non-nominalized
complements in Japanese, particularly to-complement.
2. Many Korean grammatical morphemes have two phonologically conditioned variants
depending on whether they are immediately preceded by a vowel or a consonant, e.g. the
28 Kaoru Horie
nominalizer m/um, the nominative case particle ka/i, the accusative case particle lul/ul, the
adnominal predicate suffix n/un, and the future adnominal predicate suffix l/ul.
3. Horie (1993a) presents a cross-linguistic semantic analysis of perception and cognition verb
complements in terms of the notions of “directly perceived events” (events) and “indirectly
perceived events” (propositions), and makes some cross-linguistic coding generalizations of
these notions, e.g. the tendency for events and propositions to be encoded respectively by non-
finite and finite complements.
4. The notions of “realis” and “irrealis,” the analogues of which are proposed by various
linguists such as Dwight Bolinger (Bolinger 1968), are admittably not always easy to identify
cross-linguistically and therefore need to be clearly defined (Bernard Comrie, personal
communication), although space limitations do not allow me to do so here. Suzuki (this
volume) applies a variant of the “realis”/ “irrealis” distinction, i.e. the “de re”/ “de dicto”
distinction proposed by Frajzyngier and Jasperson (1991) to Japanese complementation,
particularly to-complement which is not discussed in this study. Verspoor (this volume)
includes a useful discussion and assessment of the various ontological distinctions proposed
in this volume, including “realis”/“irrealis” and “event”/“proposition”.
5. Unless we consider a nominalizer -i, no longer productive in Modern Japanese, which
produces derived nouns such as yomi-kaki “reading and writing” and hasiri “running.”
6. My use of these terms follows Dixon (1995) in spite of the caveats in note 5 above.
7. It should be noted that Japanese has various mood suffixes, e.g. -(y)oo (speaker’s intention,
propositive), daroo (prediction), which partially correspond to the Korean irrealis suffix
-keyss-. However, it seems that the Korean -keyss- is more frequently used and hence is more
fully integrated into the mood and tense system of Korean than its Japanese counterparts.
References
Akatsuka, Noriko. 1985. “Conditionals and epistemic scale”. Language 62: 625–639.
Akatsuka, Noriko. (ed.) 1994. Japanese/Korean Linguistics 4. Stanford: CSLI. [distributed by
Cambridge University Press]
Akatsuka, Noriko, Shoichi Iwasaki, and Susan Strauss. (eds) 1996. Japanese/Korean Linguistics
5. Stanford: CSLI. [distributed by Cambridge University Press]
Akatsuka, Noriko, Hoji Hajime, and Shoichi Iwasaki. (eds) 1998. Japanese/Korean Linguistics
7. Stanford: CSLI. [distributed by Cambridge University Press]
Bolinger, Dwight. 1968. “Entailment and the meaning of structures”. Glossa 2(2): 119–127.
Bolinger, Dwight. 1974. “Concept and percept: two infinitive constructions and their
vicissitudes”. In Phonetic Society of Japan (ed.), World Papers in Phonetics: Festschrift for
Dr. Onishi’s Kizyu. Tokyo: Phonetic Society of Japan, 65–91.
Chang, Suk-Jin. 1996. Korean. Amsterdam: John Benjamins.
Choi, Soonja. (ed.). 1993. Japanese/ Korean Linguistics 3. Stanford: CSLI. [distributed by
Clancy, Patricia. (ed.).1993. Japanese/Korean Linguistics 2. Stanford: CSLI. [distributed by
Dixon, Robert W. 1995. “Complement clauses and complement strategies”. In F. R. Palmer

(ed.), Meaning and Grammar. Cambridge: Cambridge University Press, 174–220.
Frajzyngier, Zygmunt and Jasperson, Robert. 1991. “That-clauses and other complements”.
Lingua 83: 133–153.
Language 4(3): 333–377
Givón, Talmy. 1995. Functionalism and Grammar. Amsterdam: John Benjamins.
Hamada, Atsushi. 1983. Zoku tyoosen siryoo ni yoru nihongo kenkyuu. (A Study of Japanese
based on Korean Materials, Part II.) Kyoto: Rinkawa Syoten.
Hawkins, John A. 1986. A Comparative Typology of English and German. London: Croom
Helm.
Hoji, Hajime. (ed.) 1990. Japanese/ Korean Linguistics. Stanford: CSLI. [distributed by
Hooper, Joan B. 1975. “On assertive predicates”. In Kimball, John (ed.), Syntax and Se-
mantics 4. New York: Academic Press, 91–124.
Horie, Kaoru. 1990. “How languages encode the cognitive notion of directness and indi-
rectness: A typological study”. In Hoji (1990), 61–77.
Horie, Kaoru. 1991a. “Event nominalizations in Korean and Japanese”. In Kuno, Susumu,
et al. (eds), Harvard studies in Korean Linguistics 4. Seoul: Hanshin publisher, 503–512.
Horie, Kaoru. 1991b. “Cognitive motivations for event nominalizations.” Papers from the 27th
Regional Meeting of the Chicago Linguistic Society, Part I: 233–245.
Horie, Kaoru. 1993a. A cross-linguistic study of perception and cognition verb complements: A
cognitive perspective. Unpublished doctoral dissertation. University of Southern
California.
Horie, Kaoru.1993b. “From zero to overt nominalizer no: A syntactic change in Japanese.”
In Choi (1993), 305–321.
Horie, Kaoru.1995. “What the choice of overt nominalizer no did to Modern Japanese syntax
and semantics.” In Henning Andersen. (ed.), Historical Linguistics 93. Amsterdam:
John Benjamins, 191–203.
Horie, Kaoru. 1997. “Three types of nominalization in Modern Japanese: No, koto, and
zero”. Linguistics 35 (5): 879–894.
Horie, Kaoru. 1998a. “On the polyfunctionality of the Japanese particle no: From the
perspectives of ontology and grammaticalization”. In Ohori, Toshio. (ed.), Studies in
Japanese Grammaticalization. Tokyo: Kurosio Publishers, 169–192.
Horie, Kaoru. 1998b. “Functional duality of case-marking particles in Japanese and its
implications for grammaticalization: A contrastive study with Korean” In Silva (1998),
147–159.
Horie, Kaoru. 1998c. “Komyunikeesyon ni okeru gengoteki bunkateki yooin: nikkan taisyoo
gengogaku no kanten kara”. (Linguistic and cultural factors in communication: from a
Japanese-Korean contrastive linguistic perspective) Nihongogaku, 17(11), 118-127.
Horie, Kaoru. 1999. “From core to periphery: A study on the directionality of syntactic
change in Modern Japanese”. In Fox, Barbara A., Dan Jurafsky, and Laura A. Michaelis.
(eds). Cognition and Function in Language. Stanford: CSLI, 1–14.
Horie, Kaoru, and Yuko Sassa. in press. “From place to space to discourse: A contrastive
linguistic analysis of Japanese tokoro and Korean tey”. In Nakayama, Mineharu, and
30 Kaoru Horie
Charles J. Quinn (eds), Japanese/Korean Linguistics 9. Stanford: CSLI. [distributed by

Horie, Kaoru and Bongshik Kang. in press. “Action/ state continuum and nominative–
genitive conversion in Japanese and Korean”. In Kikusawa, Ritsuko, and Kan Sasaki
(eds), Modern Approaches to Transitivity. Tokyo: Kuroshio Publishers.
Ilpontayekmwunko 9. 1991. Seoul: Talakwen.
Ilpontayekmwunko 17. 1994. Seoul: Talakwen.
Josephs, Lewis S. 1976. “Complementation”. In Shibatani, Masayoshi (ed.), Syntax and
Semantics, Vol. 3. Japanese Generative Grammar. New York: Academic Press, 307–370.
Karttunen, Lauri. 1971. “Implicative verbs”. Language 47, 340–358.
Kim, Nam-Kil. 1974. The grammar of Korean complementation. Doctoral dissertation.
University of Washington.
Kim, Nam-Kil. 1984. The Grammar of Korean Complementaion. Honolulu: Hawaii Uni-
versity Press.
Kiparsky, Paul. and Carol Kiparsky. 1970. “Fact”. In Bierwisch, Manfred and Karl Erich
Heidolph (eds), Progress in Linguistics. The Hague: Mouton, 143–173.
Kuno, Susumu. 1973. The Structure of the Japanese Language. Cambridge, MA: MIT Press.
Lee, Mae-Sung. 1968. Nomianalization in Korean. Seoul: Language Research Institute.
Lyons, John. 1977. Semantics. Vol. 2. Cambridge University Press.
Makino, Seiichi. 1983. “How sensitive is the Japanese language to directly perceivable
phenomena?” In Kazuko Inoue (ed.), Issues in Syntax and Semantics, Festschrift for
Masatake Muraki. Tokyo: Sansyuusya, 127–144.
Martin, Samuel E., and Yang Ha Lee. 1967. A Korean-English Dictionary. New Haven: Yale
University Press.
Martin, Samuel E. 1991. “Recent research on the relationships of Japanese and Korean”. In
Lamb, Sidney M. and E. Douglas Mitchell (eds), Sprung from Some Common Source.
Stanford: Stanford University Press, 269–292.
Martin, Samuel E. 1992. A Reference Grammar of Korean. Tokyo: Tuttle.
McCawley, Noriko Akatsuka. 1978. “Another look at no, koto and to: Epistemology and
complementizer choice in Japanese”. In Hinds, John, et al. (eds), Problems in Japanese
Syntax and Semantics. Tokyo: Kaitakusha, 178–212.
McGloin, Naomi Hanaoka. 1989. A Student’s Guide to Japanese Grammar. Tokyo: Taishukan
Publishing Company.
Nakau, Minoru. 1973. Sentential Complementation in Japanese. Tokyo: Kaitakusha.
Ohori, Toshio. 1998. “Close to the edge: A commentary on Horie’s paper”. In Ohori, Toshio.
(ed.), Studies in Japanese Grammaticalization. Tokyo: Kurosio Publishers, 169–192.
Rosenbaum, Peter. 1967. The Grammar of English Predicate Complementation. Cambridge,
MA: MIT Press.
Ross, John Robert. 1973. “Nouniness”. In Osamu Fujimura (ed.), Three Dimensions of
Linguistic Theory. Tokyo: TEC, 137–258.
Shinzato, Rumiko. 1996. “A cognitive analysis of structural dichotomies”. Gengo Kenkyu 109:
1–23.
Silva, David. (ed.) 1998. Japanese/ Korean Linguistics 8. Stanford: CSLI. [distributed by
Sohn, Ho-min. 1994. Korean. London: Routledge.
</TARGET "hor2">
Sohn, Ho-min. and John, Haig. (eds) 1997. Japanese/Korean Linguistics 7. Stanford: CSLI.
[distributed by Cambridge University Press]
Song, Zino. 1981. “The abstract nominalizers in Korean and Japanese”. Linguistic Journal of
Korea. 6 (2):157–193.
Suzuki, Satoko. This volume. “De dicto complementation in Japanese”.
Umeda, Hiroyuki. 1991. Sutandaado hanguru kooza 2. Bunpoo to goi. (Standard Korean series
2. Grammar and vocabulary) Tokyo: Taisyuukan.
Vendler, Zeno. 1967. Linguistics in Philosophy. Ithaca, N. Y. : Cornell University Press.
Verspoor, Marjolijn. This volume. “Iconicity in English complement constructions: Concep-
tual distance and cognitive processing levels”.
<LINK "suz-n*">
<TARGET "suz" DOCINFO
AUTHOR "Satoko Suzuki"
TITLE "De dicto complementation in Japanese"
KEYWORDS ""
SIZE HEIGHT "220"
WIDTH "150"
VOFFSET "4">
Chapter 3
De dicto complementation in Japanese*
Satoko Suzuki
Macalester College
1. Introduction
In his article titled “The De Dicto Domain in Language” (1991) Frajzyngier

shows that in many languages there exists a distinction between the domain of
speech (de dicto) and the domain of reality (de re). The term de dicto refers to a
semantic domain in which reference is made to the elements of speech. This
notion includes those elements that are sometimes referred to as ‘linguistic’ in
the literature. Vendler’s (1967) notion of ‘propositions’ belongs to the domain
of speech.
In contrast, the term de re refers to a semantic domain in which reference
is made to the elements of the real world. Vendler’s notion of ‘events’ belongs
to the domain de re. To illustrate the distinction, within the system of reference
deictic pronouns are considered to be in the domain de re whereas anaphoric
pronouns (in the classical understanding of the term) are considered to be in
the domain de dicto. All anaphoras belong to the domain of de dicto by defini-
tion since they refer to an element that was mentioned in speech or that is going
to be mentioned in speech.1
Frajzyngier (1991) demonstrates how this distinction is encoded in the
system of complementation (as well as in the system of reference) in several
languages ranging from English to Mupun. Frajzyngier and Jasperson (1991)
focus on the English complementation systems, arguing that the gerundive -ing
and the infinitive clause to + verb belong to the domain de re whereas that-
clauses belong to the domain de dicto.
In Japanese there are three complementizers that are frequently used: no,
koto, and to.2 The first two complementizers function as nominalizers while the
latter does not. In Suzuki (to appear) I observed that in the system of Japanese
complementation the complementizers no and koto belong to the domain of
34 Satoko Suzuki
reality while the complementizer to belongs to the domain of speech.3 I chose

to use the framework of the contrast between de re and de dicto to describe the
differences between these complementizers instead of other notions that have
been applied to Japanese complementation, such as concrete vs. abstract (Kuno
1973), direct vs. indirect (Josephs 1976), and event vs. proposition (Horie, this
volume). This is because these latter notions capture the differences between no
and koto better than the differences between no and koto on one hand and to on
the other.
In my past work (Suzuki 1996a) I argued that no, koto, and to form a
continuum in proportion to how much the speaker has incorporated the
information expressed in the complement into his/her system. However, it has
become clear to me that there is a distinct boundary between no and koto, and
to. This is manifested in the structural differences. While no and koto are
nominalizers and thus incorporate their complement as the object of the matrix
verb, to does not nominalize the complement and thus does not incorporate the
complement into the rest of the sentence as well as the nominalizers do. The
significance of this non-incorporation will be discussed later.
Since I wanted to highlight the boundary between no and koto versus to on
the other, the contrast between de re and de dicto was chosen as the framework.
Another contrast that has been associated with this boundary, namely that
between factivity and nonfactivity (Kuno 1973), was not used since, as dis-
cussed in Suzuki (1994, 1997), the notion of factivity is not entirely adequate to
account for instances of complementation in actual discourse. The framework
of de re vs. de dicto was also chosen because the notion of the domain of speech
fits well with the character of to, which was originally used only for reporting
another speaker’s statement and later became a complementizer with a wider
application (Coulmas 1986: 164).
In this chapter I would like to focus on the functions of the de dicto comple-
mentizer, to, and other morphologically related expressions. In Section 2 I will
recapitulate why to should be considered to be a de dicto complementizer. Section 3
illustrates that in using to the speaker shows psychological distance from the
proposition that is expressed in the complement clause. Section 4 shows that the
association of to and the speaker’s psychological distance can be observed in
other uses of to. In Section 5 other complementizers that are morphologically
related to to (tte, toka, and nante) are discussed. After illustrating that these
complementizers are also de dicto complementizers, I will explain how the
notion of psychological distance is useful in elucidating characteristics of tte,
toka, and nante even when they fulfill functions other than marking complements.
De dicto complementation in Japanese 35
With this paper I hope to make contributions to the study of complement-

izers in the following aspects. First, the paper applies the notion of psychological
distance, which has been applied to other linguistic phenomena, to the analysis
of complementizers. Second, it links the notion of psychological distance to the
de dicto characteristics of certain complementizers and suggests that the
association is universal rather than unique to Japanese. Third, it takes into
account contextual factors in the analysis of complementizers. Fourth, it relates
the function of complementizers to other uses of the same linguistic forms. For
example, both tte as a complementizer and tte as a topic marker have the similar
function of expressing the speaker’s psychological distance. This observation
supports the notion of isomorphism (i.e., identical forms tend to reflect
similarity in communicative function).
Most of the data are taken from naturally occurring discourses. The
exceptions are (4a) and (4b), which were generated in order to highlight the
contrast between two complementizers. When an example is obtained from
written texts, a reference is given at the end of the example. No reference means
that the example is taken from taped conversations. Other researchers’ exam-
ples are cited as well.
2. To as a de dicto complementizer
To is considered to belong to the domain de dicto because it is the most com-

mon complementizer that occurs with verbs of saying and thinking, which
clearly belong to the domain de dicto. Other than this obvious reason, to is
regarded as a complementizer of the domain of speech because it shares several
characteristics with de dicto complementizers in other languages.
Frajzyngier (1991: 225) notes that “if a language has a complementizer
derived from a demonstrative, such a complementizer will mark the comple-
ments of the verbs of saying and thinking, while the sentential complements of
other verbs may be marked by other complementizers or by other means.” In
other words, there is a cross-linguistic tendency that complementizers derived
from demonstratives are de dicto complementizers.4 In his Kokugo Gogen Jiten
(‘National Language Etymology Dictionary’) Yamanaka states that to was
originally a demonstrative which meant ‘that,’ and developed from the same
morpheme as that from which so of sore ‘that’ and sa of sareba ‘then’ are derived
(Yamanaka 1976: 378).
Another characteristic which to shares with de dicto complementizers of
36 Satoko Suzuki
other languages is that it may be used as an independent marker of the de dicto

category. Frajzyngier observes that in colloquial Mupun it is possible to omit
the verb of saying as in (1).
(1) wur n6 i
he comp5 yes
‘He [said] yes.’ (Frajzyngier 1991: 232)
Even though the verb sat ‘say’ is missing from (1), the addressee is able to
assume that the meaning of (1) is ‘he said yes’ because of the presence of the
complementizer n6. In this sense n6 is used as an independent marker of the de
dicto category. Similarly, in colloquial Japanese if a sentence with the comple-
mentizer to lacks a verb, the missing verb is understood to be a verb of saying as
illustrated in (2).
(2) Konyakuchuu ni ichiban taisetsu na no wa, jibun no
during-engagement in most important link pn top self link
hontoo no koe o kiku koto da to yoshihiro-san.
real link voice obj listen nmlzr is comp Ms.-Yoshihiro
‘“The most important thing to do while being engaged is to listen to your
real voice,” [says] Ms. Yoshihiro.’ (More, Dec. 1995: 360)
To is also similar to de dicto complementizers of other languages in its expres-
sion of epistemic modality. Kirsner and Thompson (1976) note that the
complementizer that in English, which Frajzyngier considers a de dicto comple-
mentizer, indicates indirect evidence (i.e. inference) when it occurs with verbs
of perception. Frajzyngier proposes that de dicto complementizers are used to
indicate inference “because of the strong sense that information obtained
through speech is not as reliable as information obtained through direct
observation” (Frajzyngier 1991: 227). The following examples from English
illustrate the contrast.
(3) a. I saw that he was sleeping.
b. I saw him sleeping. (ibid.: 226)6
The same contrast can be seen in the Japanese examples, (4a) and (4b).
(4) a. Watashi wa kare ga nemutte-iru to mita.
I top he subj is-sleeping comp saw
‘I saw (judged) that he was sleeping.’
b. Watashi wa kare ga nemutte-iru no o mita.
I top he subj is-sleeping comp obj saw
‘I saw him sleeping.’
The two sentences are identical except for the complementizers. In (4a) to is
used while in (4b) the combination of the complementizer/nominalizer no and
the object marker o is used. The complement marked by to is considered to
refer to inferred information whereas the complement marked by no is inter-
preted to represent an actual event.
In this way to shares various characteristics with de dicto complementizers
of other languages. It may thus be considered to be a representative of de dicto
complementizers in Japanese. In the next section I will argue that this status of
to as a de dicto complementizer is responsible for its main function, the marking
of the speaker’s psychological distance from the information expressed in the
complement clause.
3. Marking of the speaker’s psychological distance
I would like to propose that the main function of to (and other expressions
discussed later) is to suggest the speaker’s psychological distance from the
information expressed in the complement clause. The notion of psychological
distance has been discussed by Inoue (1978), Slobin and Aksu (1982), and
Kamio (1994, 1995, 1997), among others. It remains elusive. In this study the
following types of information collectively define from what the speaker is
psychologically distanced:
1. Information which the speaker regards (or presents) as that belonging to
others;
2. Information which the speaker does not entirely consider to be true;
3. Information which the speaker disapproves of in content;
4. Information which is viewed by the speaker objectively.
The last type of information (information which is viewed by the speaker
objectively) may need to be explained. Even when we are familiar with or close
to something, it is possible for us to separate ourselves from that entity in order
to describe or analyze it. For example, although we feel affinity to our native
language, we look at it with a detached observers’ viewpoint when we conduct
linguistic analyses. Such separation is regarded as a type of psychological
distance in this study.
The list is based on the information I have gained through my investiga-
tions on complementation (Suzuki 1994, 1996a, 1997, to appear) and discourse
markers (Suzuki 1996b, 1998a, 1998b). This notion of psychological distance
38 Satoko Suzuki
overlaps with the notions of evidentiality (Chafe and Nichols 1986; Willett
1988), territory of information (Kamio 1994, 1995, 1997), and alienation
(Haiman 1989, 1990, 1995, 1998). Note that these latter notions are also
concerned with the speaker’s psychological distance in some way or another
even though some of them do not overtly state so. In this sense this paper
contributes to the broad notion of psychological distance which is relevant to a
full range of linguistic expressions from evidentials to sarcastic language.
Although notions of evidentiality, territory of information, and alienation
are effective in explaining the linguistic phenomena with which the authors
mentioned are concerned, these notions are not independently sufficient for the
purpose of this paper. The speaker’s psychological distance as defined in the
above listing is broad enough to account for all the functions of the different
expressions belonging to the domain of speech discussed in this paper.
The speaker’s psychological distance can be observed when to is in its most
typical environment — with a verb of saying. To marks the complement which
is a (direct or indirect) quotation as in the following example.7
(5) Karera wa ikudooon ni kaaru ga genkaku-sugiru to itta.
they top unanimous adv Karl subj too-strict comp said
‘They unanimously said that Karl was too strict.’
(Fujiwara 1981: 184)
Vološinov (1986: 116) observes: “reported speech is regarded by the speaker as

an utterance belonging to someone else” [Emphasis in original]. Maynard
(1996: 210) says: “quoting someone’s words, involves, by definition, the voice
of another, which results in a double-voiced discourse.” Tannen (1989: 133)
observes that the creation of voices “occasions the imagination of alternative,
distant, and others’ worlds by linking them to the sounds and scenes of one’s own
familiar world” [Emphasis provided by SS]. What is quoted (i.e. what is in the
complement ), thus, represents what the speaker finds distant from him/herself.
Macaulay (1987) and Maynard (1996) argue that this distancing effect of
quotation is present even in self-quotation such as (6).
(6) Ima, taibatsu wa kanarazushimo zettai ni warui
now physical-punishment top necessarily absolute adv bad
wake dewanai to iimashita.
reason is-not comp said
‘I just said that physical punishment is not necessarily absolutely bad.’
(Kishida 1993: 65)
Self-quotation is the speaker’s strategy of presenting what is quoted “as if it

represents someone else’s voice” (Maynard 1996: 224). In other words, the
speaker is presenting the information as that which belongs to others (cf. Item 1
in the listing above).
The preceding discussion concerning to’s occurrence with verbs of saying
shows that the speaker’s psychological distance is closely connected to the very
essence of the de dicto character of to. Examination of data indicates that this
inherent trait of to is present even when it occurs with verbs other than verbs of
saying. The psychological distance that the speaker feels is most evident when
to accompanies counterfactual verbs as shown in the following examples.
(7) Honnin mo sono ki ni natte-iru to
person-in-question also that feeling adv is-becoming comp
shuui wa omoikonde-ita.
surrounding top were-thinking
‘The people around her were thinking (wrongly) that she herself was also
getting serious.’ (Kanai 1995: 10)
(8) Kare wa watashi no kuchi ga fujiyuu na no daroo to
he top I link mouth subj not-well link nmlzr must-be comp
hayagatenshite, itamashisoo na kaotsuki ni naru.
jumping-at-conclusion pitying link facial-expression adv become
‘Jumping to the conclusion that I must be a mute, he shows pity on his
face.’ (Sawaki 1986: 91)
Both the verb in (7), omoikonde ‘thinking wrongly,’ and the verb in (8),
hayagatenshite ‘jumping to the conclusion’ indicate that the speaker is convinced
that the proposition expressed in the complement is false. It is expected that the
speaker is detached from information which s/he does not believe to be true.
To also occurs with verbs such as kaishakusuru ‘interpret’ and handansuru
‘judge’ as shown below.
(9) Korera no joshi wa doko kara arawareta ka to yuu-to,
these link particle top where from appeared fp comp say-when
teidai “wa” no “wa” no kage kara arawareta to kaishakusuru
theme “wa link “wa link shadow from appeared comp interpret
hoka wa arimasen.
other top not-exist
‘When we consider from where these particles came, there is no
other way but to interpret that they appeared from the shadow of
wa that is the thematic wa.’ (Mikami 1960: 11)
40 Satoko Suzuki
(10) Nabeno-ishi wa nagurareta bubun ga i dearu koto o

Dr.-Nabeno top beaten part subj stomach is comp obj
kangae, osoraku suizoo-haretsu ni chigainai to
thinking probably pancreas-rupture adv not-different comp
handanshi …
judging
‘Dr. Nabeno thought (of the fact) that the beaten part was the stomach
and judged that it must be rupture of the pancreas…’
(Miyamoto 1988: 10)
Unlike the cases with verbs which suggest counterfactuality, the speaker of these
sentences is not convinced of the falsity of the proposition. However, the
speaker is not convinced of the truth of the proposition, either. Even when the
judgment is attributed to the speaker as in (9), the speaker is still not entirely
committed to the truth of the information expressed in the complement. S/he
is, in this sense, detached from the information.
The above observations of to’s occurrence with verbs of counterfactuality
and judgment illustrate the second type of information associated with the
speaker’s psychological distance — information the truth of which the speaker
is not entirely convinced (cf. Item 2 in the listing). In the cases which involve
counterfactuality, the speaker is taking the falsity of the information for granted.
In the cases of judgment verbs the speaker is noncommittal regarding the truth
value of the information. Even when the speaker is the agent of the judgment, the
speaker is not prepared to claim the truth of the statement. In this sense the
speaker is detached from the content of the complement clause in both cases.
This analysis, which associates to with information the truth of which the
speaker is not strongly convinced, appears to encounter problems when we
consider cases in which to accompanies verbs of thinking and believing in the
present tense with the speaker as the subject. Observe the following examples.
(11) Wakare o tsugeru kurushimi ga wakaru-to, kono saki,
good-bye obj tell anguish obj understand-when this future
dareka kara seijitsu ni wakare o tsugerareta baai wa, sono
someone from sincere adv good-bye obj be-told case top that
hito no kunoo o omoiyaru kokoro no fukami ga umareru
person link agony obj consider heart link depth subj be-born
to omoimasu.
comp think
‘When you understand the anguish associated with saying goodbye, I

think that in case somebody breaks up with you in a sincere manner,
from this time on you will have the depth in your heart to show
consideration for that person’s agony.’ (Matsumoto 1994: 212)
(12) Sore dake wa uso dewanakatta, to, boku wa ima demo
that at-least top lie not-was comp I top now even
shinjite-iru.
am-believing
‘I believe even now that at least that was not a lie.’ (Miyamoto 1988: 115)
In both of the above examples, since the subject of thinking or believing is the
speaker and the present tense is used, the speaker seems to be committed to the
truth of the information expressed in the complement. However, note that the
speaker is not taking the truth of the information for granted. As I argued
elsewhere (Suzuki 1994, 1996a, 1997), the degree of the strength of conviction
is relevant in the selection of to as opposed to other complementizers. The
strength of the speaker’s conviction may be shown in gradation as follows:
(13) a. The speaker’s conviction is so strong that the truth is taken for granted
b. The speaker is convinced of the truth of the information to
some extent, but the conviction is not strong enough that the truth
is taken for granted
c. The speaker is noncommittal regarding the truth of the information
d. The speaker takes the falsity of the information for granted
In cases of (13b), which are exemplified with verbs of thinking and believing,
(13c), which are exemplified with verbs of judgment, and (13d), which are
exemplified with verbs of counterfactuality, to is chosen to be the complement-
izer since, relatively speaking, the speaker is not entirely convinced of the truth
of the information. On the other hand, in cases of (13a), in which the speaker
takes the truth of the information for granted, complementizers other than to
are used. (14) and (15) illustrate this.
(14) Choo-kun wa, tagai ni jikoshookai o shita atode,
Mr. Cho top each-other adv self-introduction obj did after
watashi ga honkon ni tsuita bakari na no o shiruto, …
I subj Hong-Kong in arrived just link comp obj when-learn
‘When he learned that I had just arrived in Hong Kong after we ex-
changed introductions, Mr. Cho … ’ (Sawaki 1986: 73)
42 Satoko Suzuki
(15) Kore ga honkon no yado no sooba toshite takai no

this subj Hong-Kong link hotel link market as expensive nmlzr
ka yasui no ka wakaranakatta ga, fundakuru to yuu kanji no
fp cheap nmlzr fp knew-not but rip-off comp say sort link
nedan denai koto dake wa wakatta.
price being-not comp at-least top understood
‘I did not know if this was expensive or cheap in the Hong Kong hotel
market, but I understood at least that it was not a rip-off sort of price.’
(ibid.: 56)
(14) and (15) show that the complementizer no or koto is used when the speaker
assumes the truth of the information expressed in the complement.
The correlation among the degree of the speaker’s conviction, the typical
verb types, and the complementizer used is summarized in Table 1.
Table 1. Correlation among speaker conviction, verb types, and complementizer

Degree of conviction Typical Verb Type Complementizer
Truth taken for granted V of learning no, koto
Truth convinced, but not taken for granted V of thinking, believing to
Noncommittal V of judgment to
Falsity taken for granted V of counterfactuality to
To simplify the illustration, I have discussed the typical types of verbs to

correlate with the degree of conviction. However, the character of a verb alone
does not reflect the strength of the speaker’s conviction (See Suzuki 1994, 1996a
for detailed discussions on this matter). Observe the contrast between koto and
to in the following example.
(16) Kare ga amerika ni ryuugakushite ita koto o
he subj America to studying-abroad was comp obj
kangae-awasereba kanari no shiryoku o yoosuru to
if-think-match quite link money obj need comp
omowareru no da ga…
be-thought nmlzr is but
‘When I think (of the fact) that he was studying in the United
States, I would think that quite a bit of money is needed, but…’
(Sawaki 1986: 74)
In the nonfiction book from which (16) is taken, the discourse prior to this
sentence mentions that the person referred to as kare ‘he’ was studying in the
United States. Thus, the speaker of (16) assumes the truth of the complement
clause marked by koto, i.e. kare ga amerika ni ryuugakushite-ita ‘he was studying
in the United States’ to be uncontroversial. On the other hand, the information
expressed in the to-marked clause, kanari shiryoku o yoosuru ‘quite a bit of
money is needed,’ is not assumed to be true. Even though a verb of thinking is
the predicate for both of the complements (kangae for the koto-marked clause
and omowareru for the to-marked clause), the speaker is trying to differentiate
information depending on the degree of his conviction on the truth value of the
information.
(16) shows that verbs of thinking can take another complementizer, koto.
Verbs of saying also can take koto in certain contexts as shown in the following
examples.
(17) Nobuko wa tsure no matte-iru koto o itte kotowatta.
Nobuko top companion subj is-waiting comp obj saying declined
‘Nobuko declined, saying that her companion was waiting.’
(Kudoo 1985: 46)
(18) Ikura watashi ga aishite iru koto o itte mo
how-much I subj loving am comp obj saying even
wakatte moraenai no yo.
understanding not-receive nmlzr fp
‘No matter how often I tell him that I love him, he doesn’t get it.’
(More, Dec., 1992: 330 )
The difference between examples (5) and (6), in which to is used with verbs of
saying, on one hand, and (17) and (18), which have koto with verbs of saying,
on the other, is that in the latter the speaker is taking for granted the truth of the
information expressed in the complement. The selection of koto, rather than to,
shows that the speaker has fully incorporated the information expressed in the
complement into his/ her belief system and feels little psychological distance
from it. The use of to as in (5) and (6), on the other hand, indicates that the
information is not completely or at all incorporated into the speaker’s system
and that s/he feels distanced from it.
To recapitulate what has been discussed, to is a de dicto complementizer as
defined by Frajzyngier (1991). The main function of to, which is to express the
speaker’s psychological distance from the information it marks, is closely
connected to the status of to as a de dicto complementizer. As the last two
44 Satoko Suzuki
paragraphs make clear, however, the presence of a verb of saying or thinking does
not automatically mean that the complementizer will be to. Even when a verb of
saying or thinking is used, if the information expressed in the complement is fully
incorporated into the speaker’s system and the speaker does not feel distanced
from the information, another complementizer koto is likely to be used.8,9
The speaker’s sense of psychological distance is reflected in grammar. It is
well known that a to-marked complement shares many characteristics with an
independent sentence. For example, the particle wa may fulfill the topic-
marking function in a to-marked clause just like it does in a matrix clause, while
it does not mark the topic in an embedded clause. A to-marked clause can have
the shape of an incomplete sentence, which is not an option for a regular
embedded clause. It may also contain modal expressions which are usually
reserved for sentence-final positions. These characteristics, along with the fact
that to is not followed by the object marking particle o (no and koto, nominaliz-
ers which encode de re domain, have to be followed by o), indicate that to-
marked clauses constitute a part of a sentence which is somewhat separated
from the rest of the sentence. In this way, to-marked complements are not
grammatically well-incorporated into the rest of the sentence (See Suzuki
(1996a, 1998b) for more detailed discussions on the independence of to-marked
complements). This grammatical non-incorporation reflects the psychological
distance (i.e. psychological non-incorporation) that the speaker feels toward the
information in the to-complement. Partee (1973), Munro (1982), and Haiman
(1989) observe similar phenomena in quotations in other languages (e.g.
English, Pima). In Japanese the grammatical separation is observed in all
to-marked clauses, not just in quotative complements.
In the next section I will show that the analysis in which to as the comple-
mentizer expresses the speaker’s psychological distance is compatible with the
ways in which to is used in functions other than the marking of complements.
4. Other uses of to
In her article about the Japanese self and society Bachnik (1994: 241) writes as
follows:
For Japanese, beliefs and attitudes can be indexed along a “certainty scale,” the
“certain” pole of which is ‘inside,’ uchi knowledge: firsthand, personally experi-
enced, detailed, and individuated. […] This contrasts with knowledge of a
more removed, or soto ‘outside’ kind: secondhand, detached, communicated
from others, and more generic. (Bachnik 1994: 241)
Quinn (1994) observes that there are a number of grammatical devices in

Japanese that index such uchi/ soto distinctions. Included in his list of such
grammatical devices is the contrast between particles ni and to. Both ni and to
have various functions. The former may indicate location, goal, manner, and
source while the latter may indicate accompaniment, manner, and condition.
To also functions as a complementizer as we have seen. The two particles are
contrasted in their copular use (n(i)ari vs. t(o)ari) as well as in expressions such
as X ni shite mo (‘even given [its] being X’) vs. X to shite mo (‘even supposing
[it] to be X’). Quinn argues that the uses of ni can be seen to cluster around its
basic locative meaning ‘in/ at’ while the uses of to are related to its central
meaning of accompaniment ‘with.’10 If one entity “Y” is related to another “X”
in the manner characterized as “Y is in/at X”, it is likely that Y and X share the
same outer boundary. On the other hand, if “Y is related to X in a relation
mediated by to, […], Y does not co-occupy an identical locus but rather takes
a position alongside, that is outside X” [Emphasis in original] (Quinn 1994:275).
In this way, Quinn interprets ni to represent uchi ‘inside’ while characterizing
to as representing soto ‘outside.’ Thus, to quote from the above passage of
Bachnik, to is associated with knowledge that is “secondhand, detached,
communicated from others, and more generic.” This characterization of to as
representing outside, detached knowledge is compatible with the analysis
presented in this study in which to’s main function as the complementizer is
interpreted to express the speaker’s psychological distance. The information
expressed in the complement is presented as information which belongs to the
outside and thus, information from which the speaker is distanced.
In the next section I will discuss other de dicto complementizers and explain
how the speaker’s psychological distance is the common factor in these expres-
sions even when they fulfill functions other than marking complements.
5. Other de dicto complementizers
There are several expressions that are morphologically related to to: tte,11 nante,
and toka. These expressions also function as de dicto complementizers and
consequently express the speaker’s psychological distance toward the informa-
tion expressed in the complement clause.
Tte is considered to have derived form tote (Matsumura 1988). Tote is an
archaic expression and has the meaning of to itte ‘saying that’ and to omotte
‘thinking that’ among other things (Shinmura 1993). Many dictionaries assume
46 Satoko Suzuki
that tote is a combination of to and the connective particle te (e.g. Wada et. al.
1983; Matsumura 1988; Kindaichi et. al. 1990). However, Nihon Daijiten
Kankookai (1975) observes that since connective particles are usually attached
to predicates, tote may be thought of the contracted form of the combination
[to+predicate+te] such as to iite, which is an archaic form of to itte.
In its function, tte is largely regarded as an informal variant of to. It is
interchangeable with to in most cases although there are certain contexts in
which tte is preferred to to (See Suzuki (1996c) for a discussion of such contexts).
Tte typically occurs with verbs of saying and thinking (as shown in (19) and (20),
respectively) and thus may be considered to belong to the domain de dicto.
(19) Are de ankeeto yooshi tsukutte tte ittara yokujitsu moo
that with survey form make comp when-say next-day already
sudeni meirubokkusu ni irerareteta.
already mailbox in was-put
‘When I said (to him) “Make a survey form with that,” the form was
already put in the mail box the next day.’
(20) Sugee hazukashii. Ore konna koto mo shiranai tte
very embarrassing I this-kind thing even not-know comp
minna ni omowarechau yo.
everybody by be-thought fp
‘Really embarrassing. Everybody will think that I don’t even know about
this kind of thing.’
Tte is considered to belong to the domain de dicto also because when it occurs
without a verb, the missing verb is usually assumed to be a verb of saying as in
the following example.12
(21) Yoku niau tte, okaasama ga.
well suit comp mother subj
‘It suits me well, Mother (said).’ (Kokuritsu Kokugo Kenkyuujo
1991:75)
Nante is considered to have developed from the combination of nado ‘the likes
of; and the like’, and to. Nado is described to have derived from nani ‘what’ and
to.13 It also occurs with verbs of saying and thinking as in the following examples.
(22) Uwaki wa suru kedo, demo kaette-kuru no wa kimi no tokoro
affair top do though but return-come pn top you link place
da nante itta n janai no.
is comp said nmlzr tag fp
‘He said something like, I have affairs, but the place for me to
return to is your place, didn’t he?’ (Kanai 1995: 86)
(23) Nihon josei wa nintaizuyoi nante kangaete-iru n dattara …
Japan woman top patient comp are-thinking nmlzr if-is
‘If you are thinking something like, Japanese women are patient…’
(Kanai 1995: 10)
Toka is a combination of either to and ka (the question marker) or to and ka

‘or.’ The following examples show toka with verbs of saying and thinking.
(24) Anna ricchi na yatsu inee yo da toka itte-ru no
that-kind rich link guy not-exist fp is comp is-saying nmlzr
miru-to ore wa sugoku kokkei.
when-see I top very funny
‘When I see them saying something like, there is nobody as rich as
that guy, I find it very funny.’
(25) Sore dake benkyooshite-reba kitto ooyoo kiku yo toka
that much if-is-studying surely application work fp comp
omou.
think
‘I think something like, if you are studying that much, you should be
able to apply (that knowledge to other things).’
When a verb is missing in a sentence with toka, the missing verb is understood
to be a verb of saying as in the following example.
(26) A: A, demo sono judooteki no saitaru mono wa Kobayashi.
oh but that passive link most person top Kobayashi
‘Oh, but the most passive person is Kobayashi.’
B: Aa tashika ni.
yeah certain adv
‘Yeah, certainly.’
A: Iyaa nanika saikin ibento sukunaidesu ne toka.
well somehow recently event few fp comp
‘(He says) something like, “Well, these days there are only a few
events.”’
Unlike to and tte, which are not considered to have lexical meanings, both nante
and toka are associated with meanings, ‘the likes of; and the like.’ When one
says ‘the likes of X’ or ‘X and the like,’ these expressions suggest lack of specifi-
cation. In sentences with nante and toka the suggestion of lack of specification
48 Satoko Suzuki
often adds an additional element to the speaker’s psychological distance which

is already present because of their being de dicto complementizers. In the above
examples with nante and toka ((22), (23), (24), and (26)) the speaker disap-
proves or is contemptuous toward the information expressed in the comple-
ment. In this sense the speaker is distanced from the information (cf. Item 3 in
the listing in Section 3). This connotation can be detected from the discourse
preceding the utterances shown in these examples. For example, prior to the
discourse shown in (26) speakers A and B were talking about how new students
who recently entered the club to which they belong are passive and how that is
negatively affecting the atmosphere of the club. In this context it is clear that
speaker A disapproves of the comment from the most passive student, a com-
plaint about the lack of activities in the club. The comment is marked by toka.14
This connotation of contempt may be attributed to the suggestion of lack
of specification. Lack of specification suggests lack of the speaker’s willingness
to commit to an entity which is not specified. This suggestion of the speaker’s
non-committal attitude may be interpreted as the speaker’s contempt toward
that entity. Note that expressions in English which suggest lack of specification
such as the likes of such and such also connote the speaker’s contempt in some
contexts. Suzuki (1998a) discusses this matter in more detail.
In the preceding paragraphs it has been shown that complementizers
related to to (tte, nante, and toka) are de dicto complementizers. As discussed in
relation to to, the most typical use of de dicto complementizers, the marking of
quotations, is associated with the speaker’s psychological distance. The speaker
is detached from what s/ he quotes since by quoting s/ he is indicating (or
“presenting” in the case of self-quotation) that the information expressed in the
complement belongs to others. In the case of nante and toka the speaker is even
more detached from the information in the complement since s/ he feels
contempt toward it.
That tte, nante, toka are associated with the speaker’s psychological distance
is also supported by observations of other uses of these expressions. For
example, tte may be used as a topic marker. Tte as a topic marker is associated
with several contexts (See Suzuki (1998b) for a discussion of these contexts),
but perhaps the most common and well known context is the one in which the
speaker repeats what the addressee has just said as in (27).15
(27) A: Kinoo Nori to okonomiyaki tabete-tee.
yesterday Nori with okonomiyaki was-eating
‘Yesterday I was eating okonomiyaki with Nori.’
B: Nori tte ano ano kite-ru hito?

Nori top that that is-coming person
‘Nori, is that the person who is coming here?’
Since it follows what the addressee has just said, this type of tte as a topic marker
retains the quotative character of the de dicto complementizer tte. Furthermore,
what follows this type of tte is often a question. In the phrase, X tte, X represents
information of which the speaker does not have all the facts. The speaker
assumes that the addressee is more familiar with X. Thus, we can expect that the
speaker is detached from X not only because it refers to someone else’s (the
addressee’s) words, but also because it represents information of which the
addressee is more knowledgeable. X is regarded as information that belongs to
others (cf. Item 1 in the listing in Section 3).
Another type of tte as a topic marker involves redefinition of an entity.
Observe the following examples.
(28) Otoosan te hontoo wa ii hito na no yo.
father top really top good person link nmlzr fp
‘Your father is actually a nice person, you know.’
(Yoshida 1971: 443)
(29) Okosan te angai karee zuki desu ne.
children top unexpectedly curry like comp fp
‘Contrary to expectations, children are fond of curry, you know’
(Martin 1975: 940)
In these examples the entities marked by tte are redefined.16 As the adverbial
phrases (hontoo wa ‘actually’ in (28) and angai ‘contrary to expectations’ in
(29)) indicate, surprising or unexpected qualities about the entities marked by
tte are discussed. In both of these examples the entity/information marked by
tte is redefined or characterized in a new light. As mentioned earlier, when one
wants to characterize or describe something, s/he places distance between him/
herself and the entity to be characterized and views it objectively (cf. Item 4 in
the listing in Section 3). In other words, the speaker is detached from the entity.
Thus, the notion of psychological distance is relevant in the analysis of this use
of tte as well.
Nante may be used as a sentence-final expression as illustrated in the
following example.
50 Satoko Suzuki
(30) Hito no inochi o daiji ni suru isha ga, itaike na

people link life obj important adv do doctor subj innocent link
Tatsuhiko ni teokakeru nante.
Tatsuhiko to murder fp
‘(I can’t believe that) a doctor, who is supposed to value people’s
lives, killed innocent Tatsuhiko.’
(Yomiuri Newspaper, March 13, 1996)
As Martin (1975: 945) observes, the sentence-final nante such as the one in the
above example indicates the speaker’s surprise or incredulity. Slobin and Aksu
(1982) say that when one is surprised, one’s mind is not prepared to process the
information and thus the speaker “feels distanced from the situation he is
describing” (Slobin and Aksu 1982: 198). Furthermore, incredulity is often
linked with the speaker’s disapproval of the information (See Item 3 in the
listing in Section 3). Thus, in both accounts the notion of psychological distance
is pertinent in analyzing nante as a sentence-final expression.
Martin (1975) mentions the use of toka as a sentence-final expression. His
example is given below.
(33) Byooki ni shite-iru toka.
ill adv is-doing fp
‘I hear he is ill or something.’ (Martin 1975: 1021)
He observes that toka “can be added at the end of a sentence to make a vague
synonym of the hearsay conversation” (ibid.). He also notes that toka “can be
tacked on to a quoted noun to add a touch of vagueness” (ibid.) as illustrated in
the following example.
(32) Tanaka toka yuu hito
Tanaka or-something say person
‘A person called Tanaka or something.’ (ibid.)
The fact that these instances of toka are used to mark “hearsay conversation”
and “quoted noun” shows that the essence of toka as the de dicto complement-
izer is also pertinent. Furthermore, the vagueness associated with toka in these
examples indicates either the speaker’s uncertainty regarding the accuracy of the
information toka marks or his/ her lack of commitment to it. This may be
interpreted to be his/her strong lack of identification with, or in some contexts
contempt for, the information. In other words, the speaker is distanced from
the information.
In this way, when tte, nante, and toka have functions other than functioning
as complementizers, they indicate that the speaker is detached from the

information that these expressions mark. This observation reinforces the
analysis in which the main function of these expressions as de dicto comple-
mentizers is to signify the speaker’s psychological distance from the information
expressed in the complement. This is because identical forms tend to reflect
similarity in communicative function (Haiman 1985).
6. Conclusion
The preceding discussion demonstrated that the complementizer to and its

morphologically related complementizers, tte, nante, and toka, are what
Frajzyngier (1991) calls de dicto complementizers and that their main function
is to indicate that the speaker is detached from the information expressed in the
complement. Not only in their function as complementizers, but also in their
other functions (such as to as a marker of accompaniment and nante as a
sentence-final particle), these expressions show that the notion of the speaker’s
psychological distance is the essence of their uses.
Implicit in the discussion is the cross-linguistic relevance of the association
between de dicto complementizers and the speaker’s psychological distance.
Given the quotative character of de dicto complementizers and the consequent
presentation of the information expressed in the complement as that which
belongs to others, it is apparent that in any language the nature of de dicto
complementizers is tied to the speaker’s detachment. For example, the de dicto
complementizer in English that is associated with inference and abstraction as
mentioned earlier. Future studies should be conducted regarding the associa-
tion between the speaker’s detachment and de dicto complementizers in other
languages as well, especially those languages in which linguistic forms that work
as complementizers have other functions (e.g. English). Following the assump-
tion of isomorphism, those forms are likely to have a similar communicative
function of expressing the speaker’s psychological distance.
Another direction that a future study may take is to explore the relationship
between the notion of the speaker’s psychological distance discussed here and
the notions of alienation and sarcasm discussed in Haiman (1989, 1990, 1995,
1998) and Adachi (1996). In both English and Japanese (and probably in many
other languages) sarcastic comments are often punctuated with orthographic
quotation marks. In English such a use of quotation marks is so prevalent that
even a hand gesture exists to express the same idea in the spoken language.
<DEST "suz-n*">
52 Satoko Suzuki
Note that the symbol or representation of quotation is involved with both

sarcasm and de dicto complementation. At the core of sarcasm is the speaker’s
detachment from his/her own words, which is not unlike the speaker’s psycho-
logical distance from the information discussed in this study.
Notes
* I would like to thank Kaoru Horie for organizing the theme session on complementation
at the ICLC 97 and for providing me with comments and suggestions for this paper. I am
also grateful to the anonymous reviewers for their suggestions.
1. The way Frajzyngier uses the terms de dicto and de re is different from the way they are
used in philosophy. He says: “The terms de dicto and de re are used here with the meaning
such expressions have in Latin, and independently of various interpretations they have had
since middle ages in philosophical literature” (Frajzyngier 1991: 220). In the philosophical
tradition the terms most likely refer to ambiguity that results in interpretation of certain
expressions. For example, in the sentence ‘Oedipus wanted to marry his mother’ the phrase
‘his mother’ is ambiguous. Construed de dicto, the speaker is reporting Oedipus’s own words.
Construed de re, the speaker is referring to Iocaste. I owe this example to one of the
anonymous reviewers. For general descriptions of the terms used in philosophy, see Audi
(1995) and Craig (1998).
2. See Horie (1991a; this volume) for typological discussions of the system of the Japanese
complementation and comparisons with the system in Korean.
3. For discussions of other complementizers in Japanese, which have more limited distri-
bution than no, koto, and to, see Horie (1997; this volume). For historic perspectives on
Japanese complementation, see Horie (1993, 1997).
4. Not all de dicto complementizers are derived from demonstratives. Frajzyngier (1991) says
that in many languages de dicto complementizers are derived from verbs of saying.
5. Note the abbreviations used in the literal glosses: adv (adverbial marker), comp
(complementizer), fp (final particle), link (linker), nmlzr (nominalizer), obj (direct object
marker), pn (pronoun), subj (subject marker), tag (tag expression), top (topic marker).
6. This contrast is characterized as the contrast between propositions and events. See Schüle
(1997) for a typological discussion of the notions and their relevance in Mayan languages.
The contrast in Japanese may be realized as the opposition between two nominalizers koto
and no (as discussed in Horie (1990, 1991b; this volume)) as well as the opposition between
the complementizer to and the nominalizer no as shown in (4).
7. Issues regarding direct vs. indirect quotations in Japanese have been discussed in Kamada
(1983), Maynard (1984, 1986), and Coulmas (1986). They are beyond the scope of this study.
8. Comparing koto and no, koto is considered to be closer in function to to than no (Suzuki
1996a, to appear).
9. Verbs such as chikau ‘swear; pledge’ and yakusoku-suru ‘promise’ can thus be expected to
occur with both to and koto. In a dictionary entry of chikau (Masuda 1974: 136), I find
examples with both to and koto. To accompanies them because they are verbs of saying. Koto
may also occur with them if the context is such that the speaker wants to emphasize that s/
he does not feel distanced from the information expressed in the complement. For example,
when the speaker is the subject of the sentence and at the moment of speech s/he is fully
committed to the content of the pledge or promise, koto is likely to be used.
10. For additional work which locates the notion of accompaniment at the core of the
meanings of to, see Makino and Tsutsui (1986: 482–483).
11. Tte is realized as te after a nasal consonant.
12. Okamoto (1995) argues that tte which is used sentence-finally has been grammaticalized
as a reportive evidential marker. She compares the following two sentences.
a. Kanojo, kekkon enkishitai n da tte itte-ta.
she marriage want-postpone nmlzr is comp was-saying
‘She was saying that she wants to postpone getting married.’
b. Kanojo, kekkon enkishitai n da tte.
she marriage want-postpone nmlzr is fp
‘I hear that she wants to postpone getting married.’
In (a), in which tte occurs as the complementizer, the source of the information expressed in
the tte-clause, kekkon enkishitai n da ‘she wants to postpone getting married,’ is specified as
kanojo ‘she.’ On the other hand, in (b), in which tte is used sentence-finally, the source of the
information is unspecified. It could be kanojo or somebody else. Okamoto (1995: 240)
observes that when tte is used as a reportive evidential marker, the act of saying or hearing is
backgrounded.
13. Martin (1975: 163) considers nante to have developed from the combination of nani and
to, the same source as for nado.
14. In (25) in which toka is also used, the connotation of the speaker’s contempt does not
appear to exist. However, toka is used here as a hedge. As discussed in Suzuki (1998a),
hedging and belittling functions are closely related to each other since they are both evoked
from the suggestion of the lack of the speaker’s commitment.
15. Takubo (1989) calls this use of tte ‘meta-use.’
16. Niwa (1994) connects the redefining function of tte to the meta-use of tte. He explains
that by deliberately using tte, the linguistic form that quotes a name (the meta form), the
speaker adds a new meaning that should accompany the name, or redefines the meaning of
the name.
References
Adachi, Takanori. 1996. “Sarcasm in Japanese”. Studies in Language 20 (1): 1-36.

Audi, Robert (ed.). 1995. The Cambridge Dictionary of Philosophy. Cambridge: Cambridge
University Press.
54 Satoko Suzuki
Bachnik, Jane M. 1994. “Uchi/ soto: Authority and intimacy, hierarchy and solidarity in
Japan”. In Jane M. Bachnik and Charles J. Quinn (eds), Situated Meaning: Inside and
Outside in Japanese Self, Society, and Language. Princeton: Princeton University Press,
224–243.
Chafe, Wallace and Nichols, Johanna (eds). 1986. Evidentiality: The Linguistic Coding of
Epistemology. Norwood, N. J.: Ablex.
Coulmas, Florian. 1986. “Direct and indirect speech in Japanese”. In Florian Coulmas (ed.),
Direct and Indirect Speech. Berlin; New York; Amsterdam: Mouton de Gruyter, 161–178.
Craig, Edward (ed.). 1998. Routledge Encyclopedia of Philosophy, Vol. 2. London: Routledge.
Frajzyngier, Zygmunt. 1991. “The de dicto domain in language”. In Elizabeth C. Traugott and
Bernd Heine (eds), Approaches to Grammaticalization, Vol. 1. Amsterdam: John
Frajzyngier, Zygmunt and Jasperson, Robert. 1991. “That-clauses and other complements”.
Lingua 83: 133–153.
Haiman, John. 1985. Natural Syntax. Cambridge: Cambridge University Press.
Haiman, John. 1989. “Alienation in grammar”. Studies in Language 13 (1): 129-170.
Haiman, John. 1990. “Sarcasm as theater”. Cognitive Linguistics 1 (2): 181–205.
Haiman, John. 1995. “Moods and metamessages: Alienation as a mood”. In Joan Bybee and
Suzanne Fleischman (eds), Modality in Grammar and Discourse. Amsterdam: John
Haiman, John. 1998. Talk is Cheap. Oxford: Oxford University Press.
Horie, Kaoru. 1990. “How languages encode the cognitive notion of directness and indi-
rectness: A typological study”. In Hajime Hoji (ed.), Japanese/ Korean Linguistics.
Stanford: Center for Study of Language and Information, 61–77.
Horie, Kaoru. 1991a. “Event nominalaizations in Japanese and Korean: A cognitive per-
spective”. In Susumu Kuno, Ik-Hwan Lee, John Whitman, Joan Maling, Young-Se
Kang, and Young-Joo Kim (eds), Harvard Studies in Korean Linguistics IV Seoul:
Hanshin Publisher, 503–512.
Horie, Kaoru. 1991b. “Cognitive motivations for event nominalizations”. In Lise M. Dobrin,
Lynn Nichols, and Rosa M. Rogrigues (eds), Papers from the 27th Regional Meeting of the
Chicago Linguistic Society. Chicago: Chicago Linguistic Society, 233–245.
Horie, Kaoru. 1993. “From zero to overt nominalizer no: A syntactic change in Japanese”. In
Soonja Choi (ed.), Japanese/ Korean Linguistics, Vol. 3. Stanford: Center for Study of
Language and Information, 305–321.
Horie, Kaoru. 1997. “Three types of nominalization in Modern Japanese: No, koto, and
zero”. Linguistics 35: 879–894.
Horie, Kaoru. This volume. “Complementation in Japanese and Korean: A contrastive and
cognitive linguistic approach”.
Inoue, Kyoko. 1978. “Speaker’s perspectives and temporal expressions”. University of
Michigan Papers in Linguistics 2 (4): 105–115.
Josephs, Lewis S. 1976. Complementation. In Shibatani, Masayoshi (ed.), Syntax and
Semantics, Vol. 5: Japanese Generative Grammar. New York: Academic Press, 307–369
Kamada, Osamu. 1983. “Nihongo no kansetsu wahoo”. Gengo 12 (9): 108–117.
Kamio, Akira. 1994. “The theory of territory of information: The case of Japanese”. Journal
of Pragmatics 21: 67–100.
Kamio, Akira. 1995. “Territory of information in English and Japanese and psychological
utterances”. Journal of Pragmatics 24: 235–264.
Kamio, Akira. 1997. Territory of Information. Amsterdam: John Benjamins.
Kindaichi, Haruhiko et. al. 1990. Nihongo Daijiten. Tokyo: Koodansha.
Kirsner, Robert and Thompson, Sandra A. 1976. “The role of pragmatic inference in
semantics: A study of sensory verb complements in English”. Glossa 10 (2): 200–240.
Kuno, Susumu. 1973. The Structure of the Japanese Language. Cambridge, MA: Harvard
University Press.
Kokuritsu Kokugo Kenkyuujo. 1991 [1951]. Gendaigo no Joshi, Jodooshi.. Tokyo: Kokuritsu
Kokugo Kenkyuujo.
Kudoo, Mayumi. 1985. “No, koto no tsukaiwake to dooshi no shurui”. Kokubungaku Kaishaku
to Kanshoo 3: 45–52.
Macaulay, Ronald. K. S. 1987. “Polyphonic monologues: Quoted direct speech in oral
narratives”. IPRA Papers in Pragmatics 1 (2): 1–34.
Makino, Seiichi and Tsutsui, Michio. 1986. A Dictionary of Basic Japanese Grammar. Tokyo:
The Japan Times.
Martin, Samuel E. 1975. A Reference Grammar of Japanese. New Haven: Yale University Press.
Masuda, Koh (ed.). 1974. Kenkyusha’s New Japanese-English Dictionary. Tokyo: Kenkyuusha.
Matsumura, Akira (ed.). 1988. Nihon Bunpoo Daijiten. Tokyo: Meiji Shoin.
Maynard, Senko K. 1984. “Functions of to and koto-o in speech and thought presentation in
Japanese written discourse”. Lingua 64: 1–24.
Maynard, Senko K. 1986. “The particle -o and content-oriented indirect speech in Japanese
written discourse”. In Florian Coulmas (ed.), Direct and Indirect Speech Berlin; New
York; Amsterdam: Mouton de Gruyter, 179–200.
Maynard, Senko K. 1996. “Multivoicedness in speech and thought representation: The case
of self-quotation in Japanese”. Journal of Pragmatics 25: 207–226.
Munro, Pamela. 1982. “On the transitivity of ‘say’ verbs”. In Paul J. Hopper and Sandra A.
Thompson (eds), Syntax and Semantics, Vol. 15: Studies in Transitivity. New York:
Academic Press, 301–318.
Nihon Daijiten Kankookai. 1975. Nihon Kokugo Daijiten, Vol. 14. Tokyo: Shoogakkan.
Niwa, Tetsuya. 1994. “Shudai teiji no tte to inyoo”. Jinbun Kenkyuu: Osaka Shiritsu Daigaku
Bungakubu Kiyoo 46 (2): 27–57.
Okamoto, Shigeko. 1995. “Pragmaticization of meaning in some sentence-final particles in
Japanese”. In Masayoshi Shibatani and Sandra A. Thompson (eds), Essays in Semantics
and Pragmatics. Amsterdam: John Benjamins, 219–246.
Partee, Barbara H. 1973. “The syntax and semantics of quotation”. In Stephen R. Anderson
and Paul Kiparsky (eds), A Festschrift for Morris Halle. New York: Holt, Rinehart and
Winston, 410–418.
Quinn, Charles J. 1994. “Uchi/soto: Tip of a semiotic iceberg? ‘Inside’ and ‘outside’ knowl-
edge in the grammar of Japanese”. In Jane M. Bachnik and Charles J. Quinn (eds),
Situated Meaning: Inside and Outside in Japanese Self, Society, and Language. Princeton:
Princeton University Press, 249–294.
Schüle, Susanne. 1997. “Perception verb complements in Mayan languages”. Paper presented
at the 5th International Cognitive Linguistics Conference at Vrije Universiteit, Amster-
dam.
56 Satoko Suzuki
Shinmura, Izuru (ed.). 1993. Koojien. Tokyo: Iwanami Shoten.

Slobin, Dan I. and Aksu, Ayhan A. 1982. “Tense, aspect and modality in the use of the
Turkish evidential”. In Paul J. Hopper (ed.), Tense-aspect: Between Semantics and
Pragmatics. Amsterdam: John Benjamins, 185–200.
Suzuki, Satoko. 1994. “Is that a fact? Re-evaluation of the relationship between factivity and
complementizer choice in Japanese”. In Susanne Gahl, Andy Dolby, and Christopher
Johnson (eds), Proceedings of the 20th Annual Meeting of the Berkeley Linguistics Society.
Berkeley: Berkeley Linguistics Society, 521–531.
Suzuki, Satoko. 1996a. “Incorporation of information and complementizers in Japanese”.
Pragmatics 6 (4): 511–551.
Suzuki, Satoko. 1996b. “Functions of tte- and to yuu no wa-marked phrases in Japanese
discourse”. Word 47 (3): 321–337.
Suzuki, Satoko. 1996c. “The discourse function of the quotation marker tte in conversational
Japanese”. In Jan Johnson, Matthew L. Juge, and Jeri L. Moxley (eds), Proceedings of the
22nd Annual Meeting of the Berkeley Linguistics Society. Berkeley: Berkeley Linguistics
Society, 387–393.
Susuki, Satoko. 1997. “The relevance of factivity to complementizer choice in Japanese”.
Studies in Language 21(2): 287–311.
Suzuki, Satoko. 1998a. “Pejorative connotation: A case of Japanese”. In Andreas H. Jucker
and Yael Ziv (eds), Discourse Markers: Descriptions and Theory. Amsterdam: John
Suzuki, Satoko. 1998b. “Tte and nante: Markers of psychological distance in Japanese
conversation”. Journal of Pragmatics 29: 429–462.
Suzuki, Satoko. To appear. “Japanese Complementizers: Interactions between Basic
Characteristics and Contextual Factors”. Journal of Pragmatics.
Takubo, Yukinori. 1989. “Meishiku no modaritii”. In Yoshio Nitta and Takashi Masuoka
(eds), Nihongo no Modaritii. Tokyo: Kuroshio Publishers, 211–33.
Tannen, Daborah. 1989. Talking Voices: Repetition, Dialogue, and Imagery in Conversational
Discourse. Cambridge: Cambridge University Press.
Vendler, Zeno. 1967. Linguistics and Philosophy. Ithaca, NY: Cornell University Press.
Vološinov, V. N. 1973. Marxism and the Philosophy of Language. (Trans. by Ladislav Matejka
and I. R. Titunik) New York: Seminar Press.
Wada, Toshimasa et. al. 1983. Kogo Daijiten. Tokyo: Shoogakkan.
Willett, Thomas. 1988. “A cross-linguistic survey of the grammaticalization of evidentiality”.
Studies in Language 12 (1): 51–97.
Yamanaka, Joota. 1976. Kokugo Gogen Jiten. Tokyo: Azekura shoboo.
Yoshida, Kanehiko. 1971. Gendaigo Jodooshi no Shiteki Kenkyuu. Tokyo: Meiji Shoin.
Data
Fujiwara, Masahiko. 1981. Wakaki Suugakusha no Amerika. Tokyo: Shinchoosha.

Kanai, Mieko. 1995. Ren’ai Taihei ki. Tokyo: Shuueisha.
Kishida, Shuu. 1993. Shitto no Jidai. Tokyo: Bungei Shunjuu.
</TARGET "suz">
Matsumoto, Yuuko. 1994. Wakare no Bigaku. Tokyo: Kadokawa Shoten.

Mikami, Aira. 1960. Zoo wa Hana ga Nagai. Tokyo: Kuroshio Shuppan.
Miyamoto, Teru. 1988. Hishochi no Neko. Tokyo: Koodansha.
More, December 1992. Tokyo: Shuueisha.
More, December 1995. Tokyo: Shuueisha.
Sawaki, Kootaroo. 1986. Shinya Tokkyuu , Vol.1. Tokyo: Shinchoosha.
Yomiuri Newspaper, March 13, 1996. Tokyo: Yomiuri Shuppansha.
<LINK "pol-n*">
<TARGET "pol" DOCINFO
AUTHOR "Maria Polinsky"
TITLE "Variation in complementation Constructions"
KEYWORDS ""
SIZE HEIGHT "220"
WIDTH "150"
VOFFSET "4">
Chapter 4
Variation in complementation
constructions
Long-Distance Agreement in Tsez*
Maria Polinsky
University of California, San Diego
1. Introduction
Studies of variation in syntactic constructions have long focused on two

separate but interrelated questions. First, if a language provides its speakers with
a choice of several syntactic structures (A, B, C…) to express a particular
proposition P and the truth conditions on proposition P do not vary, what is
the syntactic relationship between the individual structures? Second, still
assuming that the truth conditions on proposition P do not vary, what elements
of meaning, if any, distinguish A from B from C, etc.?
In this paper I will address mainly the latter question, and I will approach
it through the grammar of complementation in Tsez, a Nakh-Daghestanian
language of the Caucasus. In a Tsez sentence with a sentential complement in
the absolutive position, the matrix verb can agree in noun class with that entire
complement or it can agree with the absolutive NP inside the complement
clause. At first blush, the variation seems random, and this provides a linguist
with a typical situation when the truth conditions do not vary. I will examine
that variation in detail and argue that the choice between two structures can be
predicted by the information structure of the proposition expressed by the
complement clause.
Arguing for an information-structural account of a linguistic phenomenon
presupposes a certain view of grammar, and a note on that is in order here. The
view adopted here is informed by the observation that there is a principled
60 Maria Polinsky
connection between the formal structure of a clause and the pragmatic struc-
ture of the proposition expressed by that clause (Prince 1981, 1992; Vallduví
1992; Lambrecht 1994; Polinsky 1998). Information structure is understood as
the notional structuring of a proposition along several orthogonal dimensions,
namely those of: presupposition(s) and assertion; topic(s) and focus, and
identifiable and activated components (see Lambrecht 1994: 4–13). For the
purposes of this paper, the major dimension of interest is going to be that of
topic/focus distinctions. The definitions of the relevant notions will be given
below (Section 3).1
The paper is organized as follows: In Section 2, I will present the relevant
data and state the main problem addressed in the paper. The examples of
variable agreement will be given in Subsection 2.2, after I have introduced the
preliminary grammatical information on Tsez. In Section 3, I will present
several generalizations concerning an unusual agreement pattern in Tsez,
including restrictions on it. In Section 4, I will propose an information-struc-
tural account of variation in agreement. In Section 5, I will compare the
information structure account with a configurational account and will discuss
theoretical implications of the proposed analysis.
2. Data
2.1 Preliminaries
The data analyzed in this paper come from Tsez, a Nakh-Daghestanian lan-
guage spoken by about 12 thousand people in the mountains of the Northeast
Caucasus. The data were collected in the course of my own fieldwork and were
also drawn from a corpus of yet unpublished Tsez narrative texts collected by
Ramazan Rajabov and Issa Abdulaev.
To be able to follow the discussion below, the reader should be familiar
with some basic facts of Tsez grammar.
Since the main issues addressed in this paper revolve around agreement, let
me start by outlining how agreement works in Tsez. In the singular, Tsez nouns
divide into four noun classes (genders), each noun belonging to one and only
one of the four classes.
The assignment of nouns to classes is subject to several principles, some of
them semantic, others formal. For a detailed account of class assignment, see
Polinsky (in press); Polinsky and Jackson (1999); Comrie and Polinsky (1998a).
Variation in complementation constructions 61
For the purposes of this paper, it is sufficient to know that inanimate complex
NPs, and compound or derived nouns are assigned to class iv.2 All nouns
denoting males belong to class i, nouns denoting females belong to class ii, and
nouns denoting animals are in class iii. In the glosses below, grammatical noun
class is indicated by Roman numerals. For example, girl:ii means that the noun
‘girl’ belongs to grammatical class ii.
In the plural, only two grammatical noun classes are distinguished — the
plural male class (i.pl in the glosses) and the elsewhere class (ii–iv.pl in the
glosses).
Agreement is determined by the absolutive nominal and is marked by
agreement prefixes on verbs, some adverbials, and some postpositions (Comrie
et al., in press). Tsez prefixal agreement markers are shown in Table 1.
Table 1. Agreement prefixes in Tsez

Singular Plural
Class i Ø- b-
Class ii y-
¸
Class iii b- ˝ r-
Class iv r- ˛
To illustrate the operation of agreement prefixes, let’s examine the examples in

(1), where the verb -ik’i- changes to reflect the noun class of the absolutive
argument:
(1) a. aAo Ø-ik’i-s.
shepherd:i i-go-past.evid
‘The shepherd went.’
b. baru y-ik’i-s.
wife:ii ii-go-past.evid
‘The wife went.’
c. ’omoy b-ik’i-s.
donkey:iii iii-go-past.evid
‘The donkey went.’
d. bix r-ik’i-s.
grass:iv iv-go-past.evid
‘Grass went (spread).’
e. aAo-bi b-ik’i-s.
shepherd-pl i.pl-go-past.evid
‘The shepherds went.’
62 Maria Polinsky
f. baru-bi/’omoy-bi/bix-bi r-ik’i-s.
wife-pl/donkey-pl/grass-pl ii–iv.pl-go-past.evid
‘The wives/donkeys/grasses went.’
Agreement prefixes appear only on vowel-initial verbs and adjectives. Even

some of the seemingly vowel-initial verbs do not take agreement prefixes (e.g.,
esay- in (2)), possibly because they had an initial laryngeal at an earlier stage in
the language (Polinsky, in press).
Tsez is morphologically ergative, distinguishing between transitive subjects,
which appear in the ergative, and the single argument of intransitive verbs and
direct objects, both of which appear in the absolutive case. The examples in (2)
illustrate the ergative-absolutive morphology; (2a) features an intransitive verb,
(2b, c) a transitive one.
(2) a. kid y-ay-si.
girl:ii:abs ii-arrive-past.evid
‘The girl arrived.’
b. kid-bā ged esay-si.
girl-erg dress:abs wash-past.evid
‘The girl washed the dress.’
c. kid-bā esay-si ged.
girl-erg wash-past.evid dress:abs
‘The girl washed the dress.’
Examples (2b) and (2c) illustrate word order variation in a transitive clause.
Overall, Tsez is a head-final, left-branching language which respects head-
finality at the level of morpheme ordering and phrases but has a rather free
order of clause constituents.
Let us now examine a transitive clause with both NPs modified by participi-
al clauses (one of which is formed from (2a)). Examples (3a–c) show just a
small number of various possibilities of the order of clause constituents,
whereas (3d) shows that Tsez does not permit discontinuous constituents.
(3) a. [Au‡ newaz-āy y-āy-ru] kid-bā
[yesterday Mokok-from ii-arrive-past.part [girl:ii-erg
[aba-ni arbašuni-‡-ãy y-āq-ru] ged žaq’u‡
black-restr silk-cont-abl ii-make-past.part dress:ii:abs today
esay-si
wash-past.evid
‘Today the girl who arrived yesterday from the village of Mokok
washed the dress made of black silk.’
b. žaq’u‡ esaysi [qabani arbašuni‡ãy yāqru] ged

today washed [black of silk made dress
[Au‡ newazāy yāyru] kidbā
[yesterday from Mokok arrived girl
‘Today the girl who yesterday arrived from Mokok washed the dress
made of black silk.’
c. žaq’u‡ [qabani arbašuni‡ãy yāqru] ged esaysi
today black of silk made dress washed
Au‡ newazāy yāyru kidbā
yesterday from Mokok arrived girl
or: ‘The girl who yesterday arrived from Mokok washed the dress made
today of black silk.’
d. *Au‡ žaq’u‡ esaysi qabani arbašuni‡ãy yāqru
yesterday today washed black of silk made
ged newāzāy yāyru kidbā
dress from Mokok arrived girl
Complement clauses are treated as complex NPs, and when appearing in the
position of the absolutive argument, are treated as class iv nouns for agreement
purposes, e.g.
(4) [t’ek’mo-bi t’et’r-a] r-igu yo‡
[book-abs:pl read-inf iv-good is
‘It is good to read books.’
Finally, let me introduce different case frames of Tsez verbs. Intransitive verbs
invariably take their single argument in the absolutive, as illustrated by exam-
ples (1) and (4). For two-place verbs, there are three major case frames:
[erg, abs], as in (2); [dat, abs], occurring with the majority of psychological
predicates and illustrated in (5), and [poss-ess, abs], often used in the expres-
sion of involuntary actions, as shown in (6).3
(5) xexbi-r kanpitya-bi r-eti-x.
children-dat candy-abs:pl ii–iv.pl-like-pres
‘Children like candy.’
64 Maria Polinsky
(6) uži-q-ø cikay y-exu-s.

boy-poss-ess glass:abs ii-break-past.evid
‘The boy accidentally broke the glass.’
We are now ready to move to the crucial set of Tsez data, namely, variable or
transparent agreement.
2.2 Variable agreement in Tsez
In those cases where the absolutive argument of a verb is expressed by a

sentential complement, Tsez offers two agreement possibilities — either the
matrix verb agrees with the sentential complement as a single complex NP, thus
assigning it to class iv, or the matrix verb agrees with the absolutive nominal
inside the complement clause. These two possibilities are illustrated by (7) and
(8) respectively (the square brackets in the examples do not imply any syntactic
analysis but rather serve to indicate the constituents of the embedded proposi-
tion). In (7), the sentential complement ‘that the boy ate bread’ determines
class iv agreement in the matrix verb (here and below the relevant agreement
prefixes are given in boldface):
(7) eni-r r-iy-xo [už-ā magalu
mother-dat iv-know-pres [boy-erg bread:iii:abs
b-āc’-ru-‡i]
iii-eat-past.part-nmlzr
‘The mother knows that the boy ate bread.’
In (8), the matrix verb shows class iii agreement; this agreement can be
triggered only by the absolutive ‘bread’ within the sentential complement:
(8) eni-r b-iy-xo [už-ā magalu
mother-dat iii-know-pres [boy-erg bread:iii:abs
b-āc’-ru-‡i]
iii-eat-past.part-nmlzr
To distinguish between the agreement in these sentences, I will refer to agreement

with the entire sentential complement — as in (7) — as Matrix Agreement and to
agreement with the absolutive in the sentential complement — as in (8) — as
Long-Distance Agreement.4 Long-Distance Agreement is possible only when the
relevant sentential complement is in the absolutive position and cannot occur
out of a sentential complement which is in a position other than absolutive.5
Variation between Local Agreement and Long-Distance Agreement is

relatively independent of word order, as shown by (9) and (10), which differ
from (7) and (8) only in the linear order of constituents:
(9) eni-r [už-ā magalu b-āc’-ru-‡i]
mother-dat [boy-erg bread:iii:abs iii-eat-past.part-nmlzr
r-iy-xo
iv-know-pres
‘The mother knows that the boy ate bread.’ (Local Agreement)
(10) eni-r [už-ā magalu b-āc’-ru-‡i]
mother-dat [boy-erg bread:iii:abs iii-eat-past.part-nmlzr
b-iy-xo
iii-know-pres
‘The mother knows that the boy ate bread.’ (Long-Distance Agreement)
A number of questions arise as we compare (7) and (9) on the one hand and (8)
and (10) on the other. First, are there any restrictions on the types of matrix
verbs which permit Long-Distance Agreement? The answer to the first question
is obscured by Tsez morphophonemics. Only those verbs which subcategorize
for a proposition as their argument can show variable agreement. Given that
only some Tsez verbs are marked for agreement, this effectively limits the
occurrence of variable agreement to verbs ‘know’, ‘expect’, ‘believe’, ‘want’,
‘find’, which are all experiential (subcategorizing for the dative and absolutive
argument), and to such verbs as ‘be good’, ‘be old’, which consist of an agree-
ment-marking adjective and the copula yo‡ /zow-. Whether or not such limita-
tions are an accident of Tsez morphology or a derivative of some global cross-
linguistic principle remains to be seen.6 At this point, it would be honest to
admit that Long-Distance Agreement predicates do not form a coherent class.
The second question is the question that I am going to explore, namely:
what is the overall grammar of Long-Distance Agreement, and what determines
the choice between Local Agreement and Long-Distance Agreement? I will
show that the choice between Local Agreement and Long-Distance Agreement
is motivated by information structure.
66 Maria Polinsky
3. Long-distance agreement in detail
3.1 Long-distance agreement and clause boundaries
Local Agreement seems to be an unproblematic agreement option in Tsez

because it follows a simple rule, mentioned above — the clausemate absolutive
determines agreement on the matrix verb. The sentential complement is treated
as a complex absolutive NP, and the matrix verb agrees with it. Local Agree-
ment is invariably in class iv, which is the class to which most abstract nouns
and composita are assigned. The sentential complement is treated as an abstract
noun, which explains the class assignment.
Long-Distance Agreement is more unusual in that the absolutive seems to
determine class agreement across clausal barriers. Long-Distance Agreement
would be a fairly straightforward phenomenon if it turned out that an embed-
ded absolutive NP undergoes Raising into the matrix clause. In a sense, this
would make the alternation between Local Agreement and Long-Distance
Agreement similar to the well-known English pairs without Raising (~Local
Agreement) and with Raising (~Long-Distance Agreement), e.g.,
(11) a. Sherlock Holmes found that the butler was visibly agitated.
b. Sherlock Holmes found the butler to be visibly agitated.
Even if we ignore a number of theoretical problems associated with the raising

analysis (cf. Authier 1991; Langacker 1995; Achard, this volume), the evidence
against it under Tsez Long-Distance Agreement is fairly straightforward. If the
absolutive raised from the embedded clause, it would be able to participate in
main clause scrambling (cf. the scrambling possibilities illustrated in (1a–c)
above). However, this prediction is not borne out, as shown by the ill-formed
(12a, b), which should be compared with the well-formed (10):
(12) a. *magalu eni-r [už-ā b-āc’-ru-‡i]
bread:iii:abs mother-dat [boy-erg iii-eat-past.part-nmlzr
b-iy-xo
iii-know-pres
b. *eni-r [už-ā b-āc’-ru-‡i]
mother-dat [boy-erg iii-eat-past.part-nmlzr
b-iy-xo magalu
iii-know-pres bread:iii:abs
Thus we have to conclude that the absolutive triggering agreement in the matrix
verb is indeed part of the embedded clause.
So far, the complement clauses which can trigger Long-Distance Agreement
have uniformly been of one type, namely, nominalizations in -‡i. It is generally
known that non-finite clauses and clauses without a complementizer have more
transparent boundaries than tensed clauses and clauses with a complementizer.
As many other head-final languages, Tsez does not allow tensed embedded
complements, but it still permits Long-Distance Agreement in the presence of
an overt complementizer. Examples (13a) and (13b) show the Local Agree-
ment/Long-Distance Agreement contrast in the presence of an overt comple-
mentizer.7 In (13a), the complementizer clause requires Local Agreement, and
(13b) illustrates Long-Distance Agreement.
(13) a. eni-r [už-ā magalu
mother-dat [boy-erg bread:iii:abs
b-āc’-ru-†in] r-iy-xo
iii-eat-past.part-comp iv-know-pres
?/
b. *eni-r [už-ā magalu
mother-dat [boy-erg bread:iii:abs
b-āc’-ru-†in] b-iy-xo
iii-eat-past.part-comp iii-know-pres
Although (13b) is not completely ruled out, it is judged to be worse than (10),
and I will not pursue the analysis of (13b) here. Accordingly, I will concentrate
only on non-finite complements, which have no complementizer (and thus are
IPs, not CPs).
At this point, we can establish that agreement can be determined across the
non-finite, non-tensed head. This is consistent with cross-linguistic facts on
Long-Distance Agreement, which also point to the transparency of non-finite
complements (Spencer 1991: 389).
3.2 Long-distance agreement and fronted material
Let me now introduce an intriguing fact concerning Long-Distance Agreement

which will eventually lead us to a possible solution. It turns out that Long-
Distance Agreement is impossible if the relevant embedded clause contains a
clause-initial adverbial or a fronted adjunct clause. Compare (14a), which is
68 Maria Polinsky
well-formed, and the ill-formed (15a):8 The only difference between these pairs
is in the position of the adverbial Au‡ ‘yesterday’ — in (15), it occupies the
initial position in the embedded clause, whereas in (14) it is clause-medial.
(14) a. eni-r b-iy-xo [už-ā Au‡
mother-dat iii-know-pres [boy-erg yesterday
magalu b-āc’-ru-‡i]
bread:iii:abs iii-eat-past.part-nmlzr
‘The mother knows that the boy ate bread yesterday.’
b. eni-r r-iy-xo [už-ā Au‡
mother-dat iv-know-pres [boy-erg yesterday
‘The mother knows that the boy ate bread yesterday.’
(15) a. *eni-r b-iy-xo [Au‡ už-ā
mother-dat iii-know-pres [yesterday boy-erg
‘The mother knows that yesterday the boy ate bread.’
b. eni-r r-iy-xo [Au‡ už-ā
mother-dat iv-know-pres [yesterday boy-erg
‘The mother knows that yesterday the boy ate bread.’
The contrast between (16) and (17) is similar to that observed in (14) and (15),
but we are now dealing with the position of an adjunct clause within the
sentential complement. In (16a), the adjunct clause is not in the initial position
within the embedded complement, and Long-Distance Agreement is possible.
However, in (17a), the adjunct clause is in the initial position in the embedded
structure, and Long-Distance Agreement is blocked:
(16) a. eni-r [magalu [uži ay-nāy]
mother-dat [bread:abs [boy:abs come-when
nesi-r b-eti-xosi-‡i] b-iy-xo
he-dat iii-want-pres.part-nmlzr iii-know-pres
‘The mother knows that when the boy comes home he will want bread.’
b. eni-r [magalu [uži ay-nāy]
mother-dat [bread:abs [boy:abs come-when
nesi-r b-eti-xosi-‡i] r-iy-xo

he-dat iii-want-pres.part-nmlzr iv-know-pres
(17) a. *eni-r [[uži ay-nāy] magalu
mother-dat [[boy:abs come-when bread:abs
nesi-r b-eti-xosi-‡i] b-iy-xo
he-dat iii-want-pres.part-nmlzr iii-know-pres
b. eni-r [[uži ay-nāy] magalu
mother-dat [[boy:abs come-when bread:abs
nesi-r b-eti-xosi-‡i] r-iy-xo
he-dat iii-want-pres.part-nmlzr iv-know-pres
Thus, the descriptive generalization is that Long-Distance Agreement is impossible

in the presence of a fronted adverbial or adjunct clause. Note, however, that
Long-Distance Agreement is not blocked by any clause constituent which is at
the left edge of the complement clause and which represents ‘old information’
— this is made obvious by (14a), where the ergative NP is at the left edge and
Long-Distance Agreement still occurs. This suggests that only the fronting of
adjuncts creates a barrier to Long-Distance Agreement. In the next subsection,
I will examine another categorical restriction on Long-Distance Agreement.
3.3 Long-distance agreement and wh-words
Another syntactic restriction on Long-Distance Agreement is observed in its

interaction with wh-words. Wh-words in Tsez have a rather complex distribu-
tion. Argument wh-words occur in situ although they can occasionally be front-
ed. Adjunct wh-words have to be fronted. Let us illustrate this with a multiple
wh-question (the absolutive wh-word šebi belongs to class iv and can be inter-
preted as ‘who’ or ‘what’, which is why it is glossed as wh below):
(18) nā ’al-ā šebi r-ok’ek’-ā
when Ali-erg wh :abs iv-steal-past:interr
‘When did Ali steal what ?’
The interaction between Long-Distance Agreement and wh-questions can be

summarized by two descriptive generalizations.
70 Maria Polinsky
i. Wh-words in the matrix clause block Long-Distance Agreement. If the matrix

clause is a wh-question, regardless of the linear or structural position of the wh-
word, Long-Distance Agreement is impossible. For example,
(19) a. ‡u-r r-iy-xo [c’o>or-ā micxir
wh-dat iv-know-pres [thief-erg money:iii:abs
b-ok’āk’-ru-‡i]?
iii-steal-past.part-nmlzr
‘Who knows that the thief stole the money?’
b. *‡u-r b-iy-xo [c’o>or-ā micxir
wh-dat iii-know-pres [thief-erg money:iii:abs
b-ok’āk’-ru-‡i]
iii-steal-past.part-nmlzr
‘Who knows that the thief stole the money?’
ii. Wh-words in the complement clause block Long-Distance Agreement. If the

embedded clause contains a wh-word, then Long-Distance Agreement is
impossible, as in (20):
(20) eni-r r-iy-xo/*b-iy-xo [‡u
mother-dat iv-know-pres/*iii-know-pres [who:erg
micxir b-ok’āk’-ru-‡i]
money:iii:abs iii-steal-past.part-nmlzr
‘The mother knows that who stole the money?’
If the wh-word is in the absolutive, it is impossible to tell whether the matrix

verb agrees with the entire complement clause or just with šebi, because in
either case the agreement class is noun classiv. In other words, the agreement
patterns are indistinguishable:
(21) eni-r r-iy-xo [c’o>or-ā šebi
mother-dat iv-know-pres [thief-erg what:iv:abs
r-ok’āk’-ru-‡i]
iv-steal-past.part-nmlzr
‘What does the mother know that the thief stole?’
To summarize this section, I have shown that Long-Distance Agreement does

not result from the placement of the relevant absolutive NP in the matrix
clause. The absolutive does stay in the embedded clause but it triggers agree-
ment in the matrix verb. I have also shown here that Long-Distance Agreement
is categorically restricted under several conditions. Long-Distance Agreement
cannot occur if the embedded clause contains fronted adjunct material (adverbial
or clausal). Long-Distance Agreement also cannot occur if either the matrix

clause or the embedded clause contains a wh-word. Is it possible to tie these
seemingly unrelated generalizations together? This is the goal of the next section.
4. Long-distance agreement and topicality
4.1 Basic notions
I will attempt to explain Tsez variable agreement in terms of information

structure categories. In what follows, I will pursue a distinction between
topicalization as a syntactic phenomenon (cf. Müller and Sternefeld 1993;
Rudin 1988) and topicality as a semantico-pragmatic status (see Kiss 1995: 10
for a similar contrast).
In theory-neutral terms, topicalization is the overt marking of a linguistic
expression as a topic. This marking can be achieved by placing the expression in
a linear position associated with topics (often, though not always, the front of
the relevant clause) and/or by morphologically marking it as such, as in the use
of the well-known Japanese topic marker wa.
Topic is understood as an expression subject to the ‘aboutness condition’
(Reinhart 1982). More specifically, “a constituent is a topic expression if the
proposition expressed by the clause with which it is associated is pragmatically
construed as being about the referent of this constituent” (Lambrecht 1994: 131).
A linguistic expression can but does not have to be encoded as a topic expres-
sion, and this optionality will be crucial for the analysis below.
It is important to consider the issue of multiple topics. There is no princi-
pled way of ruling out the presence of several referents that are linked by the
aboutness relation to the proposition expressed by the clause with which they
are associated,9 and in fact there is no grammatical evidence in favor of the
topic uniqueness claim (Lambrecht 1994: 149–50; Culicover 1996). Although
multiple argument topics are possible (Lambrecht 1991: 149–50; Givón 1984;
Polinsky 1998), it is more common for multiple topics to have different
constituent status. For instance, a topic which serves as a scene-setting expres-
sion and which corresponds to a syntactic adjunct and a topic which is an
argument often co-occur, cf. (22), based on Culicover (1996: 453, ex. (20b)):
(22) They insisted that on Sundays, the promises you make you always have
to keep.
72 Maria Polinsky
Now that the relevant concepts have been explained, we can move on to the
main hypothesis concerning Long-Distance Agreement.
4.2 Main hypothesis
Recall (15a) and (17a), where the fronting of an adverbial or an adjunct clause
interferes with Long-Distance Agreement. In (15), the word Au‡ ‘yesterday’ is
a sentential adverbial, which can occupy different linear (and syntactic)
positions (Jackendoff 1972; Potsdam 1996). We can hypothesize that the fronting
of the adverbial or the more deeply embedded clause presumably indicates their
topicality (on the topicalization of adverbials, see Gundel 1974/1988: 152–4).
Assuming the definitions here, there are two pieces of evidence showing
that the fronted adverbial is indeed topicalized. First, the fronted adverbial is
outside the scope of sentential negation (23a), a typical feature of topics
(Lambrecht 1994: 153–55), whereas an adverbial in any other position is under
the scope of negation (23b):
(23) a. Au‡ už-ā magalu b-ac’-n-ānu
yesterday boy-erg bread:abs iii-eat-past-neg
‘Yesterday, the boy did not eat bread.’
*‘It is not the case that the boy ate bread yesterday.’
b. už-ā Au‡ magalu b-ac’-n-ānu
boy-erg yesterday bread:abs iii-eat-past-neg
‘The boy did not eat bread yesterday.’
(=‘It is not the case that the boy ate bread yesterday.’)
Second, Tsez has information-structure signaling particles, both for topic-

marking and for focus-marking (for lack of space, I will not present a detailed
analysis of these particles). The fronted adverbial can occur with a topicalizing
particle but not with a focus particle. Compare:
(24) a. Au‡-no už-ā magalu b-ac’-no
yesterday-top boy-erg bread:abs iii-eat-past.non.evid
‘As for yesterday, the boy ate bread.’
b. *Au‡-kin už-ā magalu b-ac’-no
yesterday-foc boy-erg bread:abs iii-eat-past.non.evid
‘It was yesterday that the boy ate bread.’
c. už-ā Au‡-kin b-ac’-no magalu
boy-erg yesterday-foc iii-eat-past.non.evid bread
‘It was yesterday that the boy ate bread.’
The hypothesis I would like to propose is that Long-Distance Agreement occurs

when the referent of the absolutive NP is the main topic. (Note that the referent
of the absolutive has no claim to topichood in the main sentence because that
sentence has its own topic.) The reason fronted adverbials or embedded clauses
block Long-Distance Agreement is that they are main topics. Further, Long-
Distance Agreement is incompatible with wh-words because for a wh-word the
rest of the proposition is topical and the selection of a specific topic within that
proposition creates a conflict (see Section 5.1 for details). Thus:
(25) If the referent of the absolutive NP of the complement clause is the main
topic of that clause, then Long-Distance Agreement must take place.
In the next subsection, I will present the evidence supporting (25). In Section 5,
I will discuss the issue of multiple topicality and the possibility of a syntactic
account of Long-Distance Agreement.
4.3 Evidence for topic effects
There are several arguments supporting the hypothesis in (25).
Sentence focus constructions

Topic is typically absent from the so-called “all-new” or “sentence-focus”
constructions (Lambrecht 1994). A sentence-focus construction is such that it
is formally marked as expressing a pragmatically structured proposition in
which both the subject and the intransitive predicate are in focus. Thus, the
focus domain is the entire sentence (Lambrecht 1994; Lambrecht and Polinsky
1997). In Tsez sentence-focus constructions, illustrated in (26a–c), the absolu-
tive nominal either precedes the predicate and cannot be separated from it by
intervening lexical material (Lambrecht and Polinsky 1997) or, occasionally,
follows the predicate:
(26) a. (madaA) isi y-egir-xo.
(outside snow:ii:abs ii-send-pres
‘It is snowing.’
b. *isi madaA y-egir-xo.
snow:abs outside ii-send-pres
c. (madaA) y-egir-xo isi.
(outside ii-send-pres snow:abs
74 Maria Polinsky
If the embedded clause contains no topic, that is, if a sentence-focus construc-

tion is embedded, Long-Distance Agreement is ungrammatical. Compare the
well-formed (27a), which has Local Agreement, and the ungrammatical (27b).
The absolutive argument in (27) is not a topic; rather, the material in the
embedded clause is entirely in focus. Accordingly, Local Agreement is the only
agreement option.
(27) a. eni-r [isi y-egir-xosi-‡i] r-iy-xo
mother-dat [snow:ii:abs ii-send-pres.part-nmlzr iv-know-pres
‘The mother knows that it is snowing.’
b. *eni-r [isi y-egir-xosi-‡i] y-iy-xo
mother-dat [snow:ii:abs ii-send-pres.part-nmlzr ii-know-pres
‘The mother knows that it is snowing.’
Overt focus or topic marking

As I mentioned earlier, Tsez has a number of particles whose function is to
mark a constituent as (contrastive) topic or focus. If the absolutive argument in
the complement clause is morphologically marked as focus, Long-Distance
Agreement is ill-formed. For example,
(28) a. eni-r r-iy-xo [t’ek’-kin y-igu
mother-dat iv-know-pres [book:ii:abs-foc ii-good
yā‡-ru-‡i]
be-past.part-nmlzr
‘The mother knows that the BOOK is good.’
b. *eni-r y-iy-xo [t’ek’-kin y-igu
mother-dat ii-know-pres [book:ii:abs-foc ii-good
yā‡-ru-‡i]
be-past.part-nmlzr
‘The mother knows that the BOOK is good.’
In the opposite instance, where the absolutive argument is overtly marked as

(contrastive) topic, Long-Distance Agreement is the only option:
(29) a. eni-r y-iy-xo [už-ā t’ek’-gon
mother-dat ii-know-pres [boy-erg bookii:abs-top
tetr-āsi yā‡-ru-‡i]
read-res be-past.part-nmlzr
‘The mother knows that the book, the boy is reading.’
b. *eni-r r-iy-xo [už-ā t’ek’-gon
mother-dat iv-know-pres [boy-erg bookii:abs-top
tetr-āsi yā‡-ru-‡i]
read-res be-past.part-nmlzr
‘The mother knows that the book, the boy is reading.’
Quantified indefinite NPs and non-referring expressions

We can further test the topic explanation by capitalizing on the requirement
that only referring expressions be topics (for a detailed discussion of that
requirement and further references, see Lambrecht 1994: 150–60). Topic
expressions can denote only referents, therefore, a topic must be a definite
description. The requirement that topic be a definite description can be stated
in truth-conditional terms, following Keenan (1974). The truth value of a
sentence cannot be evaluated if the entity of which the predicate holds cannot
be identified prior to determining whether the predicate holds of it or not. (For
a pragmatic motivation of the same requirement, see Lambrecht (1994: 156–7).)
Assuming that topics must be definite descriptions or referring expressions
and assuming the hypothesis in (25) above, we can predict that those linguistic
elements which do not fit these criteria will not trigger Long-Distance Agreement.
This prediction is fully borne out.
Tsez is clearly resistant to having quantified indefinite NPs as topics.10
Long-Distance Agreement does not occur when the complement clause has a
quantified indefinite absolutive. Thus:
(30) a. eni-r r-iy-xo [siskin a>i
mother-dat iv-know-pres [some bird:iii:abs
b-āxu-ru-‡i]
iii-die-past.part-nmlzr
‘The mother knows that a bird died.’
b. *eni-r b-iy-xo [sis(kin) a>i
mother-dat iii-know-pres [some bird:iii:abs
b-āxu-ru-‡i]
iii-die-past.part-nmlzr
‘The mother knows that a bird died.’
Reflexive and reciprocal pronouns are not independent referring expressions,

which predicts that they cannot function as topics. Thus, we can expect a
restriction against Long-Distance Agreement triggered by embedded recipro-
cals or reflexives in the absolutive position. This prediction is borne out, as
shown by (31b):
76 Maria Polinsky
(31) a. eni-r r-iy-xo [už-ā nesā že

mother-dat iv-know-pres [boy-erg self:i:abs
žāk’-ru-‡i]
beat-past.part-nmlzr
‘The mother knows that the boy beat up himself.’
b. *eni-r Ø-iy-xo [už-ā nesā že
mother-dat i-know-pres boy-erg self:i:abs
žāk’-ru-‡i]
beat-past.part-nmlzr
‘The mother knows that the boy beat up himself.’
Nominals occurring in N–V compounds represent another subset of non-refer-

ring expressions. These nominals are words, not noun phrases. This suggests
that a nominal constituent of a compound verb, where the N¢ fills the absolutive
position, should not be able to trigger Long-Distance Agreement. Tsez has a
large number of N–V predicates formed with the light verbs -oda ‘do, make’,
-oqa ‘be, stay’, and several others, for instance:
(32) a. tušman ø-oda
enemy:i:abs I-make
‘make an enemy’
b. moxo b-oda
thread:iii:abs iii-make
‘spin yarn’
c. razawot r-oda
divorce:iv:abs iv-make
‘divorce’
When such verbs occur as predicates in complement clauses, Long-Distance

Agreement is impossible (of course, if the noun in the N–V compound is
class iv, as in (32c), there is no way to tell Local Agreement from Long-Distance
Agreement):11
(33) a. obi-r [kidb-ā eni-r kumek
father-dat [girl-erg mother-dat help:iii
b-oy-xosi-‡i] r-iy-xo
iii-make-pres.part-nmlzr iv-know-pres
‘The father knows that the girl is helping the mother.’
b. *obi-r [kidb-ā eni-r kumek
father-dat [girl-erg mother-dat help:iii
b-oy-xosi-‡i] b-iy-xo
iii-make-pres.part-nmlzr iii-know-pres
‘The father knows that the girl is helping the mother.’
Pronominals as topics
It has long been noted that unaccented pronominals are preferred topic expres-
sions (Gundel 1974/1988; Lambrecht 1994: 174, and many others). If an unac-
cented pronominal occurs in complement clauses, it normally triggers Long-
Distance Agreement. Equivalent constructions with Local Agreement are
judged as “artificial”. Compare:
(34) a. dā-r Ø-eti-x [že idu>or ø-aya-ni-x]
me-dat i-want-pres [he:i:abs home i-come-inf-purp
‘I want him to come home.’
b. ??dā-r r-eti-x [že idu>or ø-aya-ni-x]
me-dat iv-want-pres [he:i:abs home i-come-inf-purp
‘I want that he come home.’
Examples such as (34a) are quite rare because Tsez is a pro-drop language, and
unaccented pronominals are dropped. Utterances such as (35) are more natural
than utterances like (34a). As (35) shows, the silent absolutive NP still triggers
Long-Distance Agreement:
(35) dā-r [Ø y-ik’i-xosi-‡i] y-iy-xo
me-dat [pro:abs:ii:(“she”) ii-go-pres.part-nmlzr ii-know-pres
‘I know that she goes.’
Corpus data
Several further arguments supporting (25) are weaker because they represent
tendencies, rather than categorical empirical generalizations. These tendencies
became apparent after the examination of narrative texts, most of which are
fairy tales.
Given the well-known correlation between subject and topic, we can expect
absolutive arguments to be likelier topics in intransitive clauses, where the
absolutive is the subject (as opposed to object in transitive or experiential
clauses). This expectation is corroborated by the text counts, albeit very limited:
out of 15 instances of Long-Distance Agreement, nine were associated with
intransitive clauses.
Next, in narrative texts, Local Agreement consistently occurs when the
referent of the absolutive in the embedded clause is part of the scene that
78 Maria Polinsky
represents new information. An example of such use is given below:

(36) qaA‡i-s zaman b-aj-†’o žedu-r
dawn-gen time:iii:abs iii-come-when they-dat
r-esu-n [sida> Aur-ā gor
iv-find-past.non.evid [somewhere ravine-iness circle:abs
ca†i-n b’e†-no … kecāsi Eswali-n]
throw-past.ger sheep-and sleeping Foundling-and
‘As dawn came, they found somewhere (in the mountains), in a hollow
ravine, the sheep that had gathered in a circle and nearby the Foundling,
fast asleep.’
Here, the proposition expressed by the complement clause constitutes new

information, both for the hearer/reader and for the search expedition, whose
point of view is assumed in the narrative. Let me emphasize that the informa-
tion is new by virtue of establishing relations between terms in the proposition
(Lambrecht 1994: 47–9). This means that the individual terms in the proposi-
tion do not have to be new; for instance, the Foundling and the sheep he was so
careless with are discourse-old, but finding them in a ravine is new. This allows
the narrator to treat the entire complement as new information, and according-
ly, to use Local Agreement.
If the referent of the absolutive is a discourse topic, Long-Distance Agree-
ment occurs, as illustrated by a textual example in (37). Note that the NP ‘king’
is also in the canonical topic position here:
(37) nesi-r [xan q’warid Ø-oq-no Ø-āki-ru-‡i]
he-dat king:I:abs sad I-become-ger I-go-past.part-nmlzr
ø-iy-xo
I-know-past.evid
‘He knows that the king left sad.’
Of course, it is rather hard to predict all the factors operating over narratives,
and the mere fact that a certain agreement pattern does not occur is not an
indication that it is impossible. However, the textual distribution of Long-
Distance Agreement is consistent with (25).
To summarize, there are several pieces of evidence supporting the hypothe-
sis that Long-Distance Agreement occurs where the referent of the absolutive is
the main topic of the embedded clause. The evidence presented here includes
the ill-formedness of Long-Distance Agreement with sentence-focus construc-
tions, with reflexives in the absolutive position, with N–V compounds, and with
absolutives overtly marked for focus. Long-Distance Agreement must occur if

the absolutive is marked for (contrastive) topic and is preferred with pronominal
absolutives, discourse-old information, and with the absolutives of intransitives.
5. Information structure and syntax
In the preceding section, I defended the following hypothesis: Long-Distance

Agreement occurs when the absolutive in the embedded clause is the main topic
within that clause, otherwise, Local Agreement has to be used.
The hypothesis in (25) correctly accounts for all the cases where Long-
Distance Agreement occurs. It also accounts for the obligatory Local Agreement
in the case of wh-words in the embedded clause (compare (20)). In a wh-
question, the wh-word expresses focus and the remaining proposition is topic.
Thus, the referent of the absolutive NP is not the main topic because it consti-
tutes part of a larger propositional topic. The condition in (25) is not met,
hence Long-Distance Agreement fails.
5.1 Problems and (some) solutions for the information structure account
An obvious problem for the information structure account resides in the linear
position of the absolutive argument which determines Long-Distance Agreement.
It is well-known that there is a strong cross-linguistic tendency for a topic expres-
sion to be the first constituent in an utterance (Lambrecht 1994:199–205 and
further references there). In the examples above, the absolutive in the embedded
clause can occur in a position other than the left edge of the complement clause.
One possible way of solving this problem is to pursue the distinction
between those topic expressions that announce a topic and all other topic
expressions. According to Lambrecht (1994: 202), only those expressions that
announce a new topic or make a shift from one topic to another must occur
sentence-initially. It seems that Lambrecht’s criterion applies more to external
topics. Even if it did not, it would make the wrong prediction because the topic
of the embedded clause is most often new (different from the topic of the main
sentence), and should appear in the initial position.
I would like to explore a different possibility, namely, that the linearization
rules operating in an embedded structure are different from those in a main
clause. The difference in word order between the matrix and embedded clause
is a common fact, and some languages, for instance Dutch, have systematic
80 Maria Polinsky
contrasts between matrix and embedded clauses. It is common for embedded

clauses to preserve more conservative patterns of word order or morphosyntax
(Horie 1993; this volume); matrix clauses are more subject to change, probably
due to their higher frequency. Therefore, as a particular case of such differences,
we can expect that the expression of information structure would differ between
a matrix and an embedded clause. The topic-first principle may apply in matrix
clauses and may be unnecessary in embeddings.
Assuming this possibility, how can we then account for the interaction
between adverbial placement and Long-Distance Agreement (see (15) and (17)
above)? If main topic were independent of the linear position, why would the
fronted adverbial block Long-Distance Agreement? This apparent conflict leads
to a very promising result.
Provided that the topicalized adverbial has the scene-setting function
(Lambrecht 1994: 118) of establishing the spatial or temporal framework for the
main predication, we can still expect the presence of another topic element, the
referent of an argument that the proposition is about. This view acknowledges
the existence of multiple topics (Lambrecht 1994; Culicover 1996) as long as
such topics bear different relations to the proposition.
In fact, Tsez particle distribution offers support for the multiple topic idea.
The topicalizing particles -gon and -no can occur both on the adverbial and the
argument which is topicalized, cf. (38).12,13 (Meanwhile, the doubling of
expressions with the focus particle -kin is impossible.)
(38) a. Au‡-no už-ā-gon magalu b-ac’-no
yesterday-top boy-erg-top bread:iii:abs iii-eat-past.non.evid
‘Yesterday, as for the boy, he ate the bread.’
b. Au‡-gon už-ā magalu-n b-ac’-no
yesterday-top boy-erg bread:iii:abs-top iii-eat-past.non.evid
‘Yesterday, as for the bread, the boy ate it.’
But the opposite ordering of topics marked by topic particles is unacceptable —

the topicalized adverbial must occur before the topicalized noun, the opposite
is ungrammatical:
(39) a. *už-ā-gon Au‡-no magalu b-ac’-no
boy-erg-top yesterday-top bread:iii:abs iii-eat-past.non.evid
‘As for the boy, as for yesterday, ate bread.’
b. *už-ā-n Au‡-gon b-ac’-no magalu
boy-erg-top yesterday-top iii-eat-past.non.evid bread:iii:abs
‘As for the boy, as for yesterday, ate bread.’
These examples suggest two conclusions. First, the particle distribution con-
firms that multiple topics are indeed possible (Lambrecht 1994: 149–50;
Culicover 1996). Second, the linear precedence effect observed in Tsez is just a
stronger version of the tendency observed for English (Culicover 1996: 453–5)
whereby topicalized non-NPs precede topicalized NPs, as shown by the contrast
between (22), repeated below, and (40) (Culicover 1996):
(22) They insisted that on Sundays, the promises you make you always have
to keep.
(40) ??They insisted that the promises you make, on Sundays you always have
to keep.
This suggests that topicality, which I defined above as a semantic feature, dis-
tributes differently over adverbials than it does over arguments. For an adverbi-
al to be a topic, it has to be topicalized, which is expressed by fronting. This
obligatory fronting applies both in matrix and in embedded clauses — to verify
this, cf. examples (3b, c) for matrix clauses and (3a), (15b) for embedded ones.
For Tsez arguments, on the other hand, there is no one-to-one correlation
between the main topic status and the position of the respective argument. In
the Tsez embedded clause at least, the main topic does not have to be in the
leftmost linear position. In the matrix clause, main topic is more likely to occur
in the sentence-initial position, but again, this is a tendency, not an ironclad
rule (Comrie et al., in press).
Let us now see what can motivate this discrepancy between arguments and
adjuncts with regard to topic status. We can start with the well-known correla-
tion between topichood and the subject position. Aside from configurational
considerations (which I believe might be secondary), this correlation is driven
by the semantically obligatory status of the argument in the subject position.
Likewise, objects are semantically obligatory as arguments of two-place verbs.
The semantically obligatory status leads to the propensity of arguments to be
associated with ‘the aboutness condition’, essential for the topic status. This
suggests that arguments may need less overt material (morphological marking,
prosody, position) to express their topic status. Adjuncts are optional and as
such are less likely to be topics, and their occurrence as topics calls for more
overt means of marking, of which linear position may be the most effective tool.
The proposal that I have made consists of two parts. First, topic status can
be expressed through topicalization or it can be assumed due to inherent
features of a linguistic expression. Essentially, this means that topichood is not
uniform and can be divided into syntactic and semantic phenomena. Syntactic
82 Maria Polinsky
topichood is a product of a syntactic operation which may be constrained by the

internal grammar of a given language. Semantic topichood is a feature (or a set
of features) relatively independent of the syntactic position assumed by an NP.
In principle, the syntactic and semantic aspects of topichood may not interact
at all. However, as we saw from the Tsez examples above (e.g., (15)), they still
interact, and linguistic theory so far has no mechanism to account for this
interaction. The second part of the proposal states that topicalization principles
may differ across matrix and embedded clauses, but this is the most tentative
part of the proposal, and I can only flag it in hopes of future research investigat-
ing this issue.
Let me now turn to the issues that will arise under a syntactic account of
Long-Distance Agreement.
5.2 Problems for a syntactic account
As I said earlier, there is also a possibility of accounting for Long-Distance

Agreement in purely syntactic terms. Since there is no principled way of ruling
out a syntactic account, I will briefly explain how this account might proceed
and present some problems associated with it.
An obvious way of accounting for Long-Distance Agreement in configura-
tional terms is to demonstrate that the absolutive NP moves to the specifier of
IP and then to explain the agreement facts by the transparency of the IP bound-
ary. We can then consider an analysis in the spirit of Exceptional Case Marking
(ECM). A standard ECM analysis deals with case assignment, under which a
clausemate verb fails to govern its argument and a matrix verb governs it
instead. The governed NP remains in the clause where it belongs and does not
raise. In the case of our paradigm, of course, we are dealing with agreement, not
case government. Developing the general line of reasoning associated with
ECM, one can suggest that if the verb agrees with the entire complement (Local
Agreement), the embedded clause is a CP, even with a null complementizer. If
Long-Distance Agreement occurs, we are dealing with an IP, and a mirror
image of ECM takes place.
This analysis would account for the absence of raising effects illustrated in
(12) above. But there is also evidence against this analysis. First, both Local
Agreement and Long-Distance Agreement can take place in the presence of an
overt complementizer, as shown by (13b), which differs from (9), (10) in the
presence of an overt complementizer. Second, if Long-Distance Agreement
occurred in non-tensed clauses which do not involve complementation, and
Local Agreement occurred under complementation, one would expect stronger

adjacency restrictions under Long-Distance Agreement, also by analogy with
ECM. The examples above show that this is not the case; compare (8) or (14a).
In further violation of adjacency, secondary predicates in the matrix clause can
also be subject to Long-Distance Agreement, for example -’a†o ‘be fast’ in (41):
(41) a. eni-r [pat’-ā muminat
mother-dat [Fatima-erg Muminatii:abs
žāk’-ru-‡i] y-’a†o y-esu-s
beat-past.part-nmlzr ii-fast ii-find-past.evid
‘The mother quickly found out that Muminat, Fatima beat her up.’
b. ?eni-r y-’a†o [pat’-ā muminat
mother-dat ii-fast [Fatima-erg Muminatii:abs
žāk’-ru-‡i] y-esu-s
beat-past.part-nmlzr ii-find-past.evid
‘The mother quickly found out that Muminat, Fatima beat her up.’
If we try to reconcile the effects of adverbial fronting with agreement facts, we

might seek a configurational solution according to which both the adverbial and
the absolutive argument adjoin to IP. There are two problems that this analysis
is unable to handle. First, this analysis faces the problem of the variable linear
position of the absolutive NP, which does not have to be at the left edge of the
complement clause. Under the information structure account, the way to
explain this is by appealing to the inherent topicality difference between
arguments and adjuncts (see 5.1). Under the syntactic analysis, it is unclear how
to deal with this problem. A less serious problem is that there have been
arguments against treating topicalization as IP-adjunction (as opposed to
scrambling, which is IP-adjunction) — cf. Müller and Sternefeld (1993).
However, there are reasons to doubt some of Müller and Sternefeld’s arguments
since their analysis also rules out multiple topicalization, which we have
evidence for (see 5.1 above and see also Culicover 1996: 452).
Finally, since topicalization is not clause-bound, a configurational analysis
predicts that the topical absolutive NP could move out of the complement
clause. This is not the case, as shown by (12) above and by the ill-formed (42):
(42) *muminat-no eni-r [pat’-ā
Muminat-top mother-dat [Fatima-erg
žāk’-ru-‡i] y-esu-s
beat-past.part-nmlzr ii-find-past.evid
‘Muminat, the mother found out that Fatima beat her up.’
84 Maria Polinsky
Overall, a syntactic account seems to run into more problems than an account
which distinguishes between topic features and “induced” topicalization and
which does not insist on the uniformity of syntax and information structure.
5.3 Information structure and representational modularity

Concerning the place of information structure in grammar, there are two
distinct possibilities. The first possibility is that information structure is
reducible to another level of structure, most likely syntactic structure, although
the latter might require some modification to derive information structure
effects. In the Principles and Parameters theory, much of that work is done by
Logical Form, although other mechanisms also exist; see Rochemont (1986),
Kiss (1995). In other words, information structure as a pragmatic construct is
correlated with syntax but does not need to be postulated because syntactic
principles successfully predict the form (and the choice) of the relevant linguis-
tic expression. As I showed in 5.2, Long-Distance Agreement presents serious
problems for a purely syntactic account, and this calls for other solutions.
Under the first possibility, one can also recognize an approach which would
treat both Local Agreement and Long-Distance Agreement as base-generated
structures. On this approach, there is no predictability with regard to word
order. Even the actual choice between Local Agreement and Long-Distance
Agreement becomes rather unpredictable.
If information structure is derived from semantic structure and the analysis
is couched in truth conditional terms, the contrast between Local Agreement
and Long-Distance Agreement becomes totally obscured and unmotivated.
Furthermore, the mechanism used to distinguish between topicalized and non-
topicalized absolutive in the embedded clause is still syntactic and very much a
fact of the internal grammar of Tsez.
One can object to these negatives by invoking a number of very successful
attempts to incorporate the categories of topic and focus into syntax (e.g.,
Aissen 1992; Kiss 1995; Rochemont 1986, and many others) or into semantics
(Rooth 1995; Szabolcsi 1994, and others). This incorporation, however, does
not constitute proof of the first possibility. Rather, it simply shows that there is
a strong correlation between information structure categories and syntax or se-
mantics, just as there is a strong correlation between semantic roles and syntax
or between aspectual categories and syntax.
This leads us to the other possibility, whereby information structure is
separate from syntactic structure or from semantic structure. This view is
consistent with representational modularity (Jackendoff 1996: Ch. 2) which

relies on three core ideas, namely:
i. grammar is divided into several formal systems with their own principles of
combination (Jackendoff 1996: 41);
ii. since the systems are separate, there are legitimate mismatches between
them, and in the case of variable agreement, information structure and
syntax do indeed demonstrate such a mismatch (between the semantic
topic status and the syntax of topicalization by fronting);
iii. the systems interface and the interface is subject to a specific set of con-
straints.
If we adopt representational modularity, at least as a working hypothesis, the
next question arises: what is the ‘division of labor’ between information
structure and other levels of linguistic representation? It is possible that some of
the phenomena conventionally treated in syntactic terms may be reducible to
information structure (Culicover 1996; Williams 1997; Lambrecht and Polinsky
1997; Polinsky 1998). The opposite is also possible, whereby some information
structure effects can be explained in syntactic terms (Rochemont 1986; Kiss
1995). Overall, these questions lead us to issue (iii) above, and indicate that
what is really needed is a theory of interfaces. I am not ready to propose such a
theory, but the Tsez data I analyzed indicate that the interface is wider than one
might expect, including not just word order or case marking, but also the
morphosyntax of agreement.
6. Conclusion
In this paper, I have examined the phenomenon of variable agreement in Tsez,

which distinguishes between Local Agreement (with the complement clause as
a single NP) and Long-Distance Agreement which consists of long-distance
agreement crossing clause boundaries. I have shown that the choice between
Local Agreement and Long-Distance Agreement is not arbitrary but is deter-
mined by the information structure of the complement clause. More specifical-
ly, Long-Distance Agreement occurs where the referent of the absolutive NP in
the complement clause is the main topic of that complement clause.
The results presented above strongly suggest that the information structure
of embedded clauses may differ from the information structure of root clauses.
In particular, linearization principles may be different across these clause types.
86 Maria Polinsky
Another result achieved here is the distinction between arguments and adjuncts
in terms of their activation as main topics. The topic status of arguments is
achieved and encoded differently than the topic status of adjuncts, and at least
for Tsez, this has profound structural consequences. In a nutshell, arguments
have inherent topicality, and this allows them to appear as topics without changing their
linear position in a sentence. Probably due to their semantic optionality, adjuncts
need to be specifically marked as topics, and the left-edge linear position is a strong
topic indicator.
The analysis of Tsez data also confirms the possibility of multiple topics, suggested
in the literature, and shows the cross-linguistic validity of those restrictions on
multiple topic ordering that were earlier observed in English (Culicover 1996).
In the discussion of the data, I presented several possibilities of accounting for
the effects involving information structure, including the absorption of such effects
into a syntactic account. However, as I have attempted to show in 5.2, the syntactic
account fails to explain and correctly predict a number of facts associated with
variable agreement in Tsez. The information structure account, which specifically
separates the two components and does not require an isomorphic relation between
them, is more successful.
Abbreviations
abl ablative nmlzr nominalizer

abs absolutive past.evid past evidential
cont contentive location series past.non.evid past non-evidential
dat dative past.part past participle
erg ergative pl plural
ess essive poss possessive location series
evid evidential pres present
foc focus pres.part present participle
gen genitive purp purposive
ger gerund res resultative
iness inessive restr restrictive
inf infinitive top topic
interr interrogative
<DEST "pol-n*">
Notes
* The work presented here was supported in part by the NSF grant SBR-9220219 and UCSD
Senate grant 960940S. I am grateful to Arsen Abdulaev, Paxrutdin Magomedinov, and
Ramazan Rajabov for their generous consultations on Tsez. I am indebted to Knud Lam-
brecht, who was willing to set aside his own work in order to discuss the arcane Tsez data. I
would also like to thank Farrell Ackerman, Bernard Comrie, Grev Corbett, Kaoru Horie, Dan
Jackson, Robert Kluender, John Moore, Eric Potsdam, David Perlmutter, Andrew Spencer,
Jakov Testelets, Helma van den Berg, and the participants in the Symposium on Comple-
mentation at ICLA-97 in Amsterdam for their comments on various aspects of this project.
I am solely responsible for all errors in the final product.
1. Recognizing information structure as a component influencing linguistic form is only/the
first step towards understanding grammar. Determining the role and place of information
structure in grammar is a crucial next step for any linguistic theory. Notably, there is very
little doubt among the practitioners of any version of grammatical theory that information
structure affects sentence structure, but there is wide-ranging disagreement concerning two
further issues: the place of information structure in grammar and the extent to which
information structure impacts the form of linguistic expressions. To appreciate this
disagreement, the reader may compare the discussion in Lambrecht (1994: esp. 1.2); Vallduví
(1992, 1995); Rooth (1995); Zubizarreta (1998: esp. ch. 1). At the end of this paper, having
presented my account of variable agreement, I will discuss the place of information structure
in the overall structure of language.
2. Since many abstract nouns are formed by morphological derivation, there is an impression
that abstract nouns are indeed assigned to class iv — an impression which is not exactly true
since a large number of Arabic loan-words denoting abstract concepts are in class iii.
3. For a detailed survey of Tsez cases, especially local, see Comrie and Polinsky (1998b).
4. Spencer (1991: 389) uses the term Agr-Climbing in a similar sense, patterning the term
after Clitic Climbing.
5. Agreement Climbing is not unique to Tsez — it is found in other Daghestanian languages,
for example in the closely related Hunzib (van den Berg 1995:190, 211, 240) and in the much
more distant Rutul language. It is also noted for Hungarian and Chukchi (Spencer 1991:389)
as well as for some Indic languages (Davison 1988, 1991).
6. It would be interesting to test the agreement possibilities of speech verbs which have the
ergative/absolutive frame, but none of the relevant verbs take agreement prefixes.
7. The complementizer -†in may historically be a contraction of e†in ‘said’. Regardless of the
historical origins of -†in, it is synchronically unrelated to ‡i.
8. Note that (15) would be uninterpretable if the adverbial were part of the matrix clause
(because of the semantic conflict between the meaning of “yesterday” and that of the present
tense). This again shows that the boundary between the embedded and the matrix clauses is
quite rigid.
9. This is in contrast to the uniqueness of focus, discussed, for example, in Lambrecht (1994:
ch. 6).
88 Maria Polinsky
10. For the discussion of this particular restriction, see Pesetsky (1987); Lambrecht
(1994: 156).
11. The constraint against nominals in compounds differentiates Tsez from some other
languages with Long-Distance Agreement, namely, Hindi. Davison (1988, 1991) shows that
for Hindi no such constraint is needed, and in fact, the referential status of the agreement
trigger is irrelevant.
12. The use of pronominal copies in the English translations does not reflect Tsez structure.
13. Although the two particles have distributional differences, these differences are irrelevant
for this paper.
References
Achard, Michel. This volume. “Construal and complementation in French”.

Aissen, Judith. 1992. “Topic and focus in Mayan”. Language 68: 43–80.
Authier, J.-Marc. 1991. “V-governed expletives, case theory, and the Projection Principle”.
Linguistic Inquiry 22: 721–40.
Comrie, Bernard, and Maria Polinsky. 1998a. “Gender in a historical perspective: Radial
categories meet language change”. In Carol Justus and Edgar Polome, (eds.), Language
Change and Typological Variation. (Journal of Indo-European Linguistics Monograph
Series.), 535–58.
Comrie, Bernard, and Maria Polinsky. 1998b. “The great Daghestanian case hoax”. In Anna
Siewerska and Jae Jung Song, (eds.), Case, Typology and Grammar. Amsterdam: John
Comrie, Bernard; Maria Polinsky, and Ramazan Rajabov. in press. “Tsezian languages”. In Alcie
Harris and Rieks Smeets, (eds.), Languages of the Caucasus. New York: Curzon Press.
Culicover, Peter. 1996. “On distinguishing A¢-movements”. Linguistic Inquiry 27: 445–63.
Davison, Alice. 1988. “Constituent structure and the realization of agreement features”.
Papers from the 24th Regional Meeting of the Chicago Linguistic Society, Parasession on
Agreement, 41–53.
Davison, Alice. 1991. “Feature percolation and agreement in Hindi-Urdu”. Paper presented
at the 1991 South Asia Conference on Agreement in South-Asian Languages.
Givón, Talmy. 1984. “Direct Object and Dative Shifting: Semantic and pragmatic case”. In
Frans Plank, (ed.), Objects. Towards a Theory of Grammatical Relations. London-
Orlando, Fla.: Academic Press, 151–182.
Gundel, Jeanette. 1974/ 1988. The Role of Topic and Comment in Linguistic Theory. New
York: Garland.
cognitive perspective. Ph. D. Dissertation, University of Southern California.
Jackendoff, Ray. 1972. Semantic Interpretation in Generative Grammar. Cambridge: MIT
Press.
Jackendoff, Ray. 1996. The Architecture of the Language Faculty. Cambridge: MIT Press.
Keenan, Edward L. 1974. “The functional principle: Generalizing the notion ‘subject of ’”.
Papers from the 10th Regional Meeting of the Chicago Linguistic Society, 298–309.
Kiss, Katalin E., (ed.). 1995. Discourse Configurational Languages. Oxford: Oxford University
Press.
Lambrecht, Knud. 1994. Information Structure and Mental Representation of Sentence
Constituents. Cambridge: Cambridge University Press.
Lambrecht, Knud, and Maria Polinsky. 1997. “Typological variation in sentence-focus
constructions”. Papers from the 33rd Regional Meeting of the Chicago Linguistic Society,
189–207.
Langacker, Ronald W. 1995. “Raising and transparency”. Language 71: 1–62.
Müller, Gereon, and Wolfgang Sternefeld. 1993. “Improper movement and unambiguous
binding”. Linguistic Inquiry 24: 461–507.
Pesetsky, David. 1987. “Wh-in-situ: Movement and unselective binding”. In Eric Reuland
and Alice ter Meulen, (eds.), The Linguistic Representation of (In)definiteness. Cam-
bridge: MIT Press, 98–129.
Polinsky, Maria. 1998. “A non-syntactic account of some asymmetries in the double object
construction”. In Jean-Pierre Koenig, (ed.), Discourse and Cognition: Bridging the Gap.
Stanford: CSLI, 403–22.
Polinsky, Maria, in press. “Agreement in Tsez: The trivial and the unusual”. Paper presented
at NSL-9, University of Chicago, and submitted to the proceedings.
Polinsky, Maria, and Dan Jackson. 1999. “Noun classes: Language change and learning”. In
Barbara Fox, Dan Jurafsky and Laura Michaelis, (eds.), Cognition and Function in
Language. Stanford: CSLI, 29–52.
Potsdam, Eric. 1996. Syntactic issues in the English imperative. Ph. D. Diss., University of
California, Santa Cruz.
Prince, Ellen. 1981. “Towards a taxonomy of given-new information”. In Peter Cole (ed.),
Radical Pragmatics. New York-San Diego: Academic Press, 223–55.
Prince, Ellen. 1992. “The ZPG letter: Subjects, definiteness, and information status”. In
William C. Mann and Sandra Thompson, (eds.), Discourse Description: Diverse Linguis-
tic Analyses of a Fund-Raising Text. Amsterdam: John Benjamins, 295–326.
Reinhart, Tanya. 1982. Pragmaics and Linguistics: An analysis of Sentence Topics. Indiana
University Linguistics Club.
Rochemont, Michael. 1986. Focus in Generative Grammar. Amsterdam: John Benjamins.
Rooth, Mats. 1995. “Indefinites, adverbs of quantification, and focus semantics”. In Gregory
Carlson and Francis Pelletier (eds.), The Generic Book. Chicago: University of Chicago
Press, 265–99.
Rudin, Catherine. 1988. “On multiple questions and multiple fronting”. Natural Language
and Linguistic Theory 6: 445–501.
Spencer, Andrew. 1991. Morphological Theory. Oxford: Blackwell.
Szabolcsi, Anna. 1994. “All quantifiers are not equal: The case of focus”. Acta Linguistica
Hungarica 42: 171–87.
Vallduví, Enric. 1992. The Informational Component. New York: Garland.
Vallduví, Enric. 1995. “Structural properties of information packaging in Catalan”. In Kiss
(ed.), 1995, 122–52.
</TARGET "pol">
90 Maria Polinsky
Van den Berg, Helma. 1995. A grammar of Hunzib. Munchen-Newcastle: Lincom Europa.
Williams, Edwin. 1997. “Blocking and anaphora”. Linguistic Inquiry 28: 577–628.
Zubizarreta, Maria-Luisa. 1998. Prosody, Focus, and Word Order. Cambridge: MIT Press.
<TARGET "ach" DOCINFO
AUTHOR "Michel Achard"
TITLE "Construal and complementation in French"
KEYWORDS ""
SIZE HEIGHT "220"
WIDTH "150"
VOFFSET "4">
Chapter 5
Construal and complementation in French

The perspective dimension
Michel Achard
Rice University
1. Introduction
The challenge facing researchers in their work on complementation is twofold.

First, it is difficult to treat complementation in a general way, because languages
exhibit a great deal of variation in their coding of complement distinctions. For
example, as documented in this volume by the articles of Horie and Suzuki on
Korean and Japanese, some languages place great importance on complement-
izer choice. In other languages, such as Tsez, Bella Coola and Lushootseed,
respectively investigated in this volume by Polinsky, and Beck, different
complements are expressed by different nominalization patterns. Finally, as
shown in this volume by Delbeque’s and Verspoor’s papers, English and the
Romance languages exhibit different possible verbal forms in their comple-
ments. Second, it is equally difficult to adopt a consistent set of theoretical tools
to explore that diversity, so that the similarities and differences between the
languages can be brought into focus and investigated.
In this paper, I would like to introduce one such tool, namely the Cognitive
Grammar notion of “perspective” (Langacker 1985, 1990, 1991). I show that the
specific perspective imposed on the complement scene determines the syntactic
form of French sentential complements. This paper exclusively focuses on
French. However, I believe that the concept of perspective can be successfully
applied to other languages, even though they might mark complement distinc-
tions differently from French. Perspective therefore importantly contributes to
our understanding of complementation.
In Cognitive Grammar, meaning is equated with conceptualization. A
linguistic expression is characterized semantically relative to one or more
92 Michel Achard
knowledge structures called “cognitive domains” (Langacker 1987: 147–166).

Any knowledge system or conceptualization can function as a domain for the
characterization of a linguistic expression, regardless of its possible complexity
or abstractness. To take an example, the domain of time in general, as well as
the way we socially learn to subdivide it into years, months, and weeks, is
necessarily activated when the word Tuesday is used.
The cognitive domains a linguistic expression evokes are necessary to its
meaning, but they are not sufficient. The meaning of an expression cannot be
derived from the sole observation of characteristics intrinsic to the entity
described, but also needs to include the way in which the conceptualizer
chooses to think about it and represent it. Langacker (1988: 7) expresses that
pivotal claim of the CG framework as follows: “Linguistic expressions and
grammatical constructions embody conventional imagery, which constitutes an
essential aspect of their semantic value. In choosing a particular expression or
construction, a speaker construes the conceived situation in a certain way, i.e. he
selects a particular image (from a range of alternatives) to structure its conceptu-
al content for expressive purposes”. Alternate constructions impose contrasting
images on the conceived situation. The meaning of an expression therefore
includes both the knowledge system it evokes when the expression is activated,
as well as the particular “construal” the conceptualizer imposes on the scene.
In this paper, I show that the specific form of each French complement
construction reflects a specific construal imposed on a conceptual base.
Consequently, variations in the syntactic form of the complements will be
treated as the reflection of variations in the construal of the complement scene.
More specifically, I argue that different perspective configurations of the
complement scene pertaining to (i) the vantage point from which the comple-
ment clause is conceptualized, and (ii) the kind of viewing arrangement which
exists between the conceptualizer and the conceptualized entity account for the
difference between finite and infinitival complements, as well as the presence of
aspect and voice markers on the infinitival complement.
This chapter is organized in the following fashion: Section 2 presents an
overview of sentential complementation in French. Section 3 briefly reviews
some previous cognitive analyses of complementation. Section 4 presents the
notion of perspective, and its relevance to complementation. Section 5 investi-
gates the distribution of complements in several constructions. Section 6
discusses the results. Section 7 concludes the chapter.
Construal and complementation in French 93
2. An overview of French sentential complementation
From a purely descriptive standpoint, French sentential complements can be

syntactically realized as an infinitive, or as a finite (indicative or subjunctive)
clause. This is illustrated in (1)–(3):
(1) a. Il faut parler plus fort
it necessary-3sg.pres.ind speak.inf more loud
‘You must speak louder.’
b. Je veux rester ici
I want-1sg.pres.ind stay.inf here
‘I want to stay here.’
c. Marie aime aller au cinéma
Mary like-3sg.pres.ind go.inf to cinema
‘Mary likes going to the movies.’
d. Jean sait nager
John know-3sg.pres.ind swim.inf
‘John knows how to swim.’
(2) a. Jean force Marie à aller à la fac
John force-3sg.pres.ind Mary to go.inf to the university
‘John forces Mary to go to the university.’
b. J’ai envie de retourner en Californie
I have-1sg.pres.ind envy to return.inf in California
‘I feel like going back to California.’
c. Il est facile de recevoir un procés verbal au Texas
It be-3sg.pres.ind easy to receive.inf a traffic ticket in Texas
‘It is easy to get a traffic ticket in Texas.’
(3) a. Jean sait que Marie est malade
John know-3sg.pres.ind that Mary be-3sg.pres.ind sick
‘John knows that Mary is sick.’
b. Je voudrais que Jean vende sa voiture
I want-1sg.pres.cond that John sell-3sg.pres.subj his car
‘I would like John to sell me his car.’
c. Je crois qu’il a démissionné
I believe-1sg.pres.ind that he have-3sg.pres.ind resigned
‘I believe that he resigned.’
d. Je suis content qu’elle ait compris
I be-1sg.pres.ind happy that she have-3sg.pres.subj understood
‘I am glad that she understood.’
94 Michel Achard
The complement in (1) is a bare infinitive. In (2), the infinitival complement is

preceded by the complementizer de. In (3), it is coded as a finite clause, with a
main verb in the indicative mood in (3a) and (3c) or the subjunctive as in (3b)
and (3d). These different kinds of complements cannot indiscriminately occur
in any construction. It is, for example, well documented in the literature that
the modals cannot be followed by finite complements, and that the verbs of
volition and emotional reaction can only be followed by infinitival comple-
ments when the main and subordinate clause subjects are coreferential (Achard
1993; Farkas 1992; Ruwet 1984). These two examples of distributional con-
straints are illustrated in (4) and (5):
(4) a. Marie peut soulever 100 kilos
Mary is able to-3sg.pres.ind lift.inf 100 kilos
‘Mary is able to lift 100 kilos.’
b. *Marie peut qu’elle soulève 100 kilos
Mary is able to-3sg.pres.ind that she lift-3sg.pres.ind 100 kilos
‘Mary is able to that she lifts 100 kilos.’
(5) a. Jean veut arriver à l’heure
John want-3sg.pres.ind arrive.inf at the time
‘John wants to arrive on time.’
b. *Jean veuti qu’ili arrive à l’heure
John want-3sg.pres.ind that he arrive-3sg.pres.subj at the time
‘John wants that he arrives on time.’
The constructions presented in this paper are quite diverse, but they nonethe-
less exhibit a conceptual core that becomes apparent when they are considered
at a level of abstraction that overlooks their individual differences. This concep-
tual core is represented in Figure 1 below.
Figure 1. Complementation in conceptual terms

Figure 1 will help us to develop a vocabulary to be used throughout this paper.

The outer rectangle CL1 represents the sentence; V1 represents the main verb.
The inner rectangle CL2 represents the complement structure, and V2 represents
the subordinate verb. The dual role of the speaker, as a speaker and a conceptua-
lizer, is illustrated by S and C0 respectively. The dotted arrow going from C0 to
CL1 indicates the direction of conceptualization. It illustrates the fact that the
whole sentence represents the conceptualization of C0. The different comple-
ment constructions under investigation in this paper exhibit variations within
this abstract template in two related areas, namely: (i) the inner structure of the
subordinate clause CL2, which can be in a finite mood (where V2 is in the
indicative or the subjunctive mood) or infinitive (where V2 is in the infinitive
mood), and (ii) the semantic role, syntactic function, and explicit mention of
the trajector of the subordinate structure (left unspecified in Figure 1).
3. Complements in conceptual semantics
The notion of subordination often used to describe multi-clause constructions

requires some explanation. As a starting point, I will follow Givón (1980) in
assuming that the prototypical value of an independent clause is to describe a
single independent event. A subordinate clause by comparison describes an
event that is viewed as conceptually dependent on the main event. There is a
consensus in cognitive/ functional linguistics that the syntactic form of the
complement is not strictly a structural matter, but that it reflects the way in
which the event it codes is viewed (Givón, 1980; Haiman 1985; Langacker
1991). For example, Givón (1980) argues that the degree to which the comple-
ment clause approximates an independent clause is representative of the degree
of independence with which it is construed by some conceptualizer. Givón
(1980: 338) provides the following parameters for evaluating the degree of
independence of the subordinate event with respect to the main event: “(i) The
degree to which the agent/ subject/ topic marking of the embedded clause
reflects the markings in independent/ main clauses; (ii) The degree to which
independent-clause tense-aspect-modality marking of the verb is preserved in
the embedded clause; (iii) The presence or degree-of-presence of predicate-
raising of the complement clause verb onto the main verb; i.e. the degree to which
the complement verb is lexicalized as one word with the main verb.” For example,
in the causative construction Marie fait pleurer Jean ‘Mary makes John cry’, the
subordinate verb does not include the tense-aspect-modality markings characteristic
96 Michel Achard
of verbs in independent clauses, and its logical subject is coded as an accusative.

Consequently, the subordinate event (John’s crying) is construed as highly
dependent on the main verb. When the complement is an indicative clause as
in Marie voit que Jean pleure ‘Mary sees that John is crying’ for example, the
subordinate event is construed with the maximum possible independence given
its position as a complement (conceptually dependent on the main clause).1
In an analysis very much compatible with Givón’s, Langacker (1991) argues
that the functional organization of the verbal group, which closely parallels the
organization of the nominal group, exhibits a layering of semantic functions
(Langacker 1991: 13–95). Just as a nominal stem represents a noun type, a verb
stem represents a process type. Both types need to be instantiated, i.e. consid-
ered with respect to a particular location in their domain of instantiation before
being quantified. For verbs, the non-finite markers (voice and aspect in French)
assume the function of quantification. The quantified instances are finally
grounded. Under the CG analysis, an infinitival complement profiles a type
specification, while a full indicative clause profiles a grounded instance of a
process type.
In the CG analysis, the difference between finite and infinitival comple-
ments is essentially a matter of grounding (Langacker 1991). Finite comple-
ments necessarily include grounding predications, that is to say markers of
person, mood, and tense, which relate the content of the complement clause
directly to the speech situation. In other words, grounding predications provide
the complement clause with its own independent viewing frame, within which
the subordinate process can appear in profile. When the verb is in the infinitive,
the complement clause does not include any grounding predications. It is
therefore not directly related to the speech situation, and can only be accessed
through the main verb.2
Both Givón’s and Langacker’s analyses are crucial to our understanding of
complementation because they assign semantic value to the complement forms,
but they do not tell the whole story. In particular, they do not explain why a
particular scene is construed with a particular degree of independence from the
main event. For example, they rightfully point out that in Jean croit qu’il a gagné
‘John thinks that he has won’, the subordinate event is viewed with more
independence than in Jean croit avoir gagné ‘John believes to have won’.
However, they do not explain what motivates the particular construal of the
subordinate event in both cases.3 In this paper, I present cases where the
syntactic shape and distribution of complements directly result from the
particular perspective imposed on the complement scene. In that light, the
account proposed here is compatible with both Givón’s and Langacker’s, but it
relates their findings to a specific dimension of construal (perspective) which
has been independently shown to be critical to linguistic investigation.
4. Perspective and syntactic realization
The term perspective is used in this paper in the sense of Langacker (1985, 1990,
1991), and I will therefore only briefly present it. The conceptualizer’s relation
to the scene conceptualized is considered metaphorically in terms of perception
(Langacker 1985, 1990). Consequently, an important dimension of construal
concerns the specific perspective imposed on the conceptualized scene. Perspec-
tive subsumes the particular vantage point from which the scene is conceptual-
ized, as well as the viewing arrangement existing between the perceiver and the
perceived entity. Let us first consider the notion of vantage point.
The ground (which contains the speaker, the addressee, and the immediate
speech circumstances) represents the default vantage point from which a
linguistic expression is conceptualized. Langacker (1991: 441) writes: “I suggest
that one should take quite seriously the notion of the ground being — in some
real sense — the vantage point from which a linguistically coded scene is
viewed. In particular, the circumstances of the speech event, together with the
nature of the grounding relationship can be thought of as defining a sort of
viewing frame representing what is immediately accessible for focused observa-
tion.[…] The viewing frame serves as a window on the situation described by
a finite clause, and the clause’s processual profile is by definition the focal point
within the immediate scope thus defined”.
However, the internal complexity of multi-clause constructions provides an
added layer of difficulty pertaining to the vantage point from which the
complement scene is conceptualized. The difference between different con-
structions is illustrated in (6) and (7):
(6) Le professeur fait rire ses étudiants
The professor makes-3sg.pres.ind laugh.inf his students
‘The professor makes his students laugh.’
(7) Le professor croit que ses étudiants
The professor believe-3sg.pres.ind that his students
rient de lui
laugh-3pl.pres.ind of him
‘The professor believes that his students are laughing at him.’
98 Michel Achard
In (6), the sentence is conceptualized from the vantage point of the speaker.
Note that the main subject (the professor) might not even be aware of the effect
she has on his students. In other words, the main subject does not act as a
conceptualizer with respect to the complement scene. In constructions such as
(6), the speaker is the only relevant conceptualizer in the sentence. The situation
is di erent in (7). The communicative purpose of the sentence is to describe the
mental stance of the main clause subject with respect to the complement scene, or
in other words the conceptualization of the main clause subject. The speaker reports
that conceptualization, which, in the case of (7), she might not even share.4
With the verbs whose subjects function as a conceptualizer with respect to
the complement scene, the complement constructions include an additional
conceptualizing relation. The profiled relation existing between the main
subject S1 and the complement clause CL2 can be viewed as equivalent to that
existing between the speaker and her conceptualization CL1. For example, in
(7), the relation profiled by croire between the main clause subject le professeur
and the complement ses étudiants rient de lui is viewed as the equivalent of the
construal relation existing between the speaker and the sentence. In the same
way as the speaker conceptualizes the sentence, the main clause subject is
presented as the conceptualizer of the complement clause. Langacker (1991: 442)
expresses the similarity between the two relations as follows: “The analogy is
strongest when the subject (or another main-clause participant) functions as
conceptualizer with respect to the contents of the subordinate clause, e.g. with
verbs like say, believe, imagine, want, enjoy and realize. The subject’s conceptual-
izing role vis-à-vis the subordinate structure is then comparable to that of the
speech-act participants in conceptualizing an expression’s meaning (the
construal relationship).” When the main clause subject acts as a conceptualizer
with respect to the complement scene, it naturally constitutes the vantage point
from which the complement clause is conceptualized.
The consideration of the vantage point from which an entity is construed is
necessary not sufficient. The relative position of the viewer and the object of
perception in the viewing relation (the viewing arrangement) represents
another important aspect of perspective. Langacker (1985: 120) argues that a
viewing situation can be characterized: “relative to the asymmetry between the
observer in a perceptual situation and the entity that is observed”. This asym-
metry yields two basic types of viewing arrangements, respectively called the
Optimal Viewing Arrangement (henceforth OVA) and the Egocentric Viewing
Arrangement (henceforth EVA). These two arrangements are presented in
Figure 2 below (from Langacker 1985: 121).
Figure 2. Optimal viewing arrangement and egocentric viewing arrangement
Figure 2a depicts the OVA, in which S (the self or observer) strictly acts as
the subject of perception, and is not itself an object of perception. It is thus
characterized as maximally subjective, and O, the object of perception (the
other or observed) as maximally objective (a strict object of perception). The
OVA magnifies the observer/observed asymmetry. The self is so involved in the
perceptual activity that she loses all self-awareness and strictly concentrates on
the object of perception. To be fully objective, the entity observed must occupy
a privileged position of maximum perceptual acuity. It has to be close to the
observer for maximum salience, but not so close as to blur her vision. It must
be a clear, well-defined object, distinct from both the background and the
observer. When it occupies the objective scene or on-stage region, characterized
as the area of maximum visual attention and acuity, the conditions of the OVA
are met. Since the observer has lost all self-awareness in her observing the scene,
her participation in the viewed activity is said to be maximally subjective.
The OVA contrasts with the EVA, which is presented in Figure 2b. The
main difference is the extension of the objective scene. The EVA “accommo-
dates the natural interest that most people have in themselves and the relations
they bear to the entities around them” (Langacker 1985: 122). The locus of
attention goes beyond the region of maximal perceptual acuity to incorporate
the observer. The latter is thus part of the extended viewing scene. She remains
the observer, but she is also, to some degree, a participant in the scene observed.
She is therefore more objectively construed.
From a linguistic standpoint, we are not interested in perception per se, but
in the meaning of linguistic expressions. The perceptual relationship described
above is only a special case of a more general construal relationship between the
conceptualizer and the object of conceptualization. The terms subjectivity/
objectivity will thus be used with respect to the role played by the concept-
ualizer within the conceptualization that constitutes an expression’s semantic
value (its scope of predication). Three possibilities arise, presented in Figure 3
below (from Langacker 1985: 125).
100 Michel Achard
Figure 3. Degrees of subjectivity of the ground
In these diagrams, the dark circles stand for the conceptualized entity. G stands
for the ground, the dotted circle stands for the objective scene and the box
stands for the expression’s scope of predication. In 3a, the conceptualization
makes no reference to the ground. This represents the case for most common
nouns, such as cat, dog, or rock. In 3b, the ground is included in the meaning of
the expression but is not profiled. It is used only as a reference point for the
purposes of isolating the entity described. This configuration describes a noun
such as Tuesday in its deictic sense of I’ll come on Tuesday, for example. The
ground is a reference point because it provides the current point of time from
which Tuesday can be calculated. In 3c, the ground is the profiled entity within
the scope of predication of the expression. The expression designates (profiles)
a grounded element. This configuration would describe a subject pronoun such
as I for instance. The three stages of Figure 3 depict a scale of subjectivity of the
ground. It is maximally subjective in 3a, because it is not invoked as part of the
expression’s meaning. It is construed more objectively in 3b, where it is part of
the expression’s meaning but not profiled, and it is maximally objective in 3c,
where it is the profiled object of conceptualization.
In the remainder of this paper, I show that the notions presented here can
valuably be applied to complement constructions, and that the perspective
imposed on the complement scene is reflected in the distribution of infinitival
and finite complements in French causation, modal, and conceptualizing
subject constructions.
4.1 Perspective and French complements
The two main components of perspective, namely vantage point and viewing
arrangement, can be directly applied to complementation. As was indicated before,
both the speaker and the main clause subject can be possible vantage points for a
complement clause. When the speaker is the only relevant conceptualizer, the
construction can be referred to as a speaker-oriented construction. On the other

hand, when the main subject acts as the conceptualizer of CL2, the construction
is referred to as a conceptualizing subject construction.5 It was already men-
tioned that when the main subject acts as a conceptualizer with respect to her
conceptualization, the sentence contains two relevant conceptualizing relations.
The first one occurs between the speaker and the conceptualizer CL1. The
second one occurs between the subject C1 and her conceptualization CL2. Since C0
and C1 represent the vantage points from which their respective conceptualizations
are structured, these construal relations can be thought of as two axes of viewing
arrangement, along which different viewing configurations will be possible. These
two axes are represented in the diagram in Figure 4 below by the two solid arrows
outside CL1. The relation between the speaker and the sentence defines the
subjective axis, because it contains the main conceptualizing subject. The relation
existing between C1 and the complement CL2 defines the objective axis, because
it occurs as part of the objective scene (the speaker’s conceptualization).
Figure 4. Two axes of viewing arrangement
Figure 4 presents the two axes of viewing arrangement along which different
viewing configurations will be shown to motivate different complement
constructions. Obviously, the respective relevance of the two axes of viewing
arrangement depends on the construction considered. In the next section, I
argue that the presence of grammatical markers on the infinitival complement
of certain modals is directly imputable to variations in viewing configurations
along the subjective axis. I also show that the presence of infinitival and finite
complements in conceptualizing subject constructions (where C1 is the concep-
tualizer of CL2) can be explained by a difference in the viewing configuration
along the objective axis.
102 Michel Achard
5. Configurations and complement forms
This section explores different complement constructions and shows how the
perspective imposed on the complement scene motivates the specific form used
to code that scene.
5.1 Modals and conceptualizing subject verbs

The first benefit of the recognition of perspective (in particular the distinction
between conceptualizing subject constructions and speaker-oriented construc-
tions) is that it allows us to solve a lingering problem in French linguistics,
namely the difference between the modal verbs from other verbs which take
infinitival complements.
A modal system is typically composed of a small class of verbs that exhibit
certain semantic and structural characteristics. For instance, in English, the
modal verbs do not have infinitive forms, they do not agree with the subject,
and they do not have participial forms. French modals, however do not have
these distinguishing characteristics. The verbs traditionally considered modals
are not structurally or morphologically different from other verbs. They are
usually described as auxiliaries, but the broadness of that category makes it
virtually impossible to precisely isolate a modal class. For example, Cervoni
(1987: 83) gives the following definition of auxiliaries (translation mine): “Au
sens le plus strict du terme, un auxiliaire est une forme verbale dont le rôle est
de pourvoir un autre verbe au participe ou à l’infinitif des catégories formelles
qui lui manquent: temps, mode, personne, aspect.” [‘In the strictest sense of the
term, an auxiliary is a verbal form whose role is to provide another verb in the
infinitival or participial form with the formal categories it is lacking: tense,
person, mood, aspect.’]
Since the modals are not morphologically distinct, other criteria are needed
to differentiate them from any other verb which takes an infinitival comple-
ment. Different solutions have been proposed. For instance, Benveniste
(1974: 177) includes the modals in his study of auxiliary relations (relations
d’auxiliation).6 There are two primary modal verbs: pouvoir ‘can’ and devoir
‘must’. Other verbs, such as vouloir ‘want’, falloir ‘be necessary’, désirer ‘desire’,
espérer ‘hope’, which are followed by infinitival complements can also occasion-
ally provide auxiliary relations of a modal type (1974: 188). Cervoni (1987)
argues that even though pouvoir and devoir are the only indisputable modals,
the system of verbal modality should also include savoir ‘know’ and vouloir.7
The other verbs that can also be followed by an infinitival complement (such as
croire ‘believe’ or espérer ‘hope’) are used as “semi-auxiliaries of modality”.
Benveniste’s and Cervoni’s positions are representative of the aforemen-
tioned difficulty of finding valid criteria to differentiate modals from other
verbs in French, and therefore call into question the validity of positing such a
class for French. In Achard (1996), I propose that the distinctive criterion that
allows us to separate modals from other verbs that take infinitival complements
is the role of the main clause subject with respect to the complement scene. The
subject of CSC verbs acts as the conceptualizer of the complement scene,
whereas the conceptualizing role of the subject of modals (if it exists at all) is
not directly relevant to the construal of the complement scene.8 In Jean doit
partir ‘John must leave’ for instance, the subject Jean may not even be aware
that he needs to leave. Even if he is, his role as a conceptualizer is not directly
relevant, and the complement scene is construed from the vantage point of the
speaker. The recognition of the role of the main clause subject in different
constructions (and thus of the vantage point from which the complement is
conceptualized) enables us to argue that: (i) it is indeed legitimate to refer to a
class of modals in French, and (ii) the modal class should include pouvoir ‘can’,
devoir ‘must’, and savoir ‘know how’ capability. More specifically, it allows us to
include savoir in its ‘know how’ sense, and to reject vouloir as a modal in
French. Consider the examples in (8) and (9) from Achard (1996):
(8) a. Marie sait nager
Mary know-3sg.pres.ind swim.inf
‘Mary knows how to swim.’
b. Marie sait qu’elle nage le matin
Mary know-3sg.pres.ind that she swim-3sg.pres.ind the morning
‘Mary knows that she swims in the morning.’
c. Marie sait que Paul nage à midi
Mary know-3sg.pres.ind that Paul swim-3sg.pres.ind at noon
‘Mary knows that Paul swims at noon.’
(9) a. Jean veut venir
John want-3sg.pres.ind come.inf
‘John wants to come.’
b. *Jeani veut qu’ili vienne
John want-3sg.pres.ind that he come-3sg.pres.subj
‘John wants that he come.’
104 Michel Achard
c. Jean veut que Marie vienne

John want-3sg.pres.ind that Mary come-3sg.pres.subj
‘John wants Mary to come.’
Note that the sense of savoir with an infinitive in (8a) is different from the sense
of the same verb with a finite complement in (8b) and (8c). The sentence in
(8a) merely describes Mary’s capability of performing the process of nager, but
she does not act as a conceptualizer with respect to that process. On the other
hand, in the cognitive sense of savoir in (8b) and (8c), the communicative
purpose of the sentence is precisely to present the main subject’s conceptualiza-
tion. Consequently, savoir needs to be analyzed as a polysemous verb. It has a
cognitive sense, but it is a modal in its capability sense. Importantly, the modal
sense of savoir cannot be followed by an finite complement.
Vouloir on the other hand has the same sense with an infinitival comple-
ment in (9a) and a finite complement in (9b) and (9c). In the three sentences,
the subject has a conceptualizing role with respect to the complement process.
In the three cases, the complement of vouloir is construed from the vantage
point of C1. The verb is therefore excluded from the modal class. Importantly,
it is the vantage point from which the complement scene is conceptualized
which enables us to differentiate between two verbs which have otherwise a
similar syntactic distribution.
I will now consider the effect of the second component of perspective,
namely viewing arrangement, on complement distribution. I will present in
turn the situation along the subjective and objective axis. In Section 5.2, I argue
that the different syntactic form of the complement in causation, movement,
and modal constructions is explainable by different configurations of viewing
arrangement along the subjective axis, that is between the speaker and her
conceptualization CL1. In Section 5.3, I show that the distribution of infinitival/
finite complements in conceptualizing subject constructions reflects different
configurations of viewing arrangement along the objective axis.
5.2 Along the subjective axis: Causation, movement

and modal constructions
Let us begin by presenting the distribution of the complement forms. The

complement of causative and movement constructions, as well as that of certain
modals can only be realized as a bare infinitive. Crucially, it cannot be inflected
with voice or aspect markers. This is illustrated in (10)–(13):
(10) a. Marie a fait rire Jean

Mary has-3sg.pres.ind made laugh.inf John
‘Mary made John laugh.’
b. *Marie a fait avoir ri Jean
Mary has-3sg.pres.ind made have.inf laughed John
‘Mary made John have laughed.’
(11) a. Marie va chercher le journal
Mary go-3sg.pres.ind look for.inf the paper
‘Mary goes to get the paper.’
b. *Marie va avoir cherché le journal
Mary go-3sg.pres.ind have.inf looked for the paper
‘Mary goes to have got the paper.’
(12) a. Marie sait nager
Mary know-3sg.pres.ind swim.inf
‘Mary knows how to swim.’
b. *Marie sait avoir nagé
Mary know-3sg.pres.ind have.inf swum
‘Mary knows how to have swum.’
(13) a. Jean peut soulever la table
John is able to-3sg.pres.ind lift.inf the table
‘John is able to lift the table.’
b. *Jean peut avoir soulevé la table9
John is able to-3sg.pres.ind have.inf lifted the table
‘John is able to have lifted the table.’
With other modals, however (most notably with necessity and epistemic
modality), the infinitival complement can include the markers of voice and
aspect, as illustrated in (14) and (15):
(14) a. Marie doit revenir à 6 heures
Mary must-3sg.pres.ind return.inf at 6 hours
‘Mary must return at 6 o’clock.’
b. Marie doit être revenue à 6 heures
Mary must-3sg.pres.ind be.inf returned at 6 hours
‘Mary must have returned at 6 o’clock.’
c. Marie doit absolument être élue
Mary must-3sg.pres.ind absolutely be.inf elected
‘Mary must absolutely be elected.’
106 Michel Achard
(15) Both examples are intended in their epistemic sense

a. Paul peut passer la prendre, ils habitent à côté
Paul can-3sg.pres.ind pass.inf her.cl take.inf they live to side
‘Paul can come to pick her up, they live next to each other.’
b. Paul peut être passé la prendre, ils habitent
Paul can-3sg.pres.ind be.inf passed her.cl take.inf, they live
à côté
to side
‘Paul can have come to pick her up, they live next to each other.’
The data in (10)–(15) show that we need to explain why voice and aspect
marking is possible with certain modals and impossible with others, as well as
with causation and movement constructions. The hypothesis I would like to
propose is that the marking on the infinitive reflects different configurations of
viewing arrangement between the speaker and her conceptualization CL1. I
argue that the possible presence of the markers on the infinitival process follows
the subjectification of the main relation. According to Langacker (1991: 270),
subjectification “involves some facet of the profiled relationship being reorient-
ed from the objective axis to the subjective axis, so that is no longer anchored by
an objective participant (the subject) but rather by a reference point construed
more subjectively (the default case being the ground itself).” If we apply this
general process to complement constructions, subjectification occurs when the
main relation becomes less and less associated with the main clause subject to
become more and more associated with the speaker.
According to that hypothesis, the presence of a bare infinitive is the sign of
the maximally objective construal of the objective scene, while the presence of
the markers of voice and aspect on the complement represents a more subjec-
tive construal of the complement scene. The key point in the analysis consists
in precisely evaluating the viewing arrangement inherent to each of the afore-
mentioned constructions.
A crucial aspect of the viewing arrangement is determined by the precise
nature of the main relation. In particular, it is important to precisely locate (if
possible) the origin of the force which brings about the main process. For
example, in (10a), (11a), (12a), and (13a), the subject represents the origin of
the main relation. In other words, the origin of the force that brings about the
infinitival process is clearly delineated as the main clause subject, and the
relation clearly occurs between S1 and the complement clause. Importantly, the
speaker is a mere observer of the main process. For example, in (10a), Mary is
the origin of the force that brings about John’s laughing. In (11a), Mary
represents the locus and origin of the movement. In (12a), Mary is the locus of
potency of the modal force, which if unleashed will trigger an occurrence of the
infinitival process.10
Since the main relation is clearly anchored by the main clause subject, and
the speaker is a mere observer of that relation, the main relation is maximally
separate from the conceptualization relation. Another way of expressing this
would be to say that the two relations originate from two clearly delineated and
maximally different sources. The conceptualization relation originates with the
speaker, and the main relation originates with the main subject. Since the
conceptualizing subject (the speaker) is kept outside the object of its conceptu-
alization CL1, I take such constructions as instantiating the OVA presented in
Section 4. The causation and movement constructions, as well as savoir and
pouvoir ability can therefore be said to impose a maximally objective construal
of the objective scene.
Because we are dealing with complement constructions where the comple-
ment clause is conceptually dependent on the main clause, positing the total
separation of the conceptualization and main relation would represent a severe
oversimplification. At the very least, even in the case of a maximally objective
construal of the main relation, the speaker can always evaluate some aspect of
that relation, as illustrated in (16):
(16) Marie peut soulever 100 kilos facilement
Mary is able to-3sg.pres.ind lift.inf 100 kilos easily
‘Mary is able to lift 100 kilos easily.’
The main relation profiled by the modal pouvoir in its ability sense is viewed as
maximally separate from the conceptualization relation, because Mary is the
clearly delineated origin of the modal force. However, the presence of facilement
‘easily’ shows that the speaker can evaluate the main relation. The modal
relation is therefore not entirely separate from the conceptualization relation.
The speaker’s role with respect to the modal relation is minimal, namely the
perception and evaluation of some characteristic of that relation.
The constructions illustrated in (14) and (15), namely obligation (deontic)
and epistemic modality present a different type of viewing arrangement,
because the separation between the main and conceptual relation is blurred.
The subject is not the origin of the force that triggers the occurrence of the
infinitival process. For example, in (14), the source of necessity does not come
from Mary, but it resides elsewhere, possibly with the speaker or in the circum-
stances of the speech situation. The crucial point is not the exact location of the
108 Michel Achard
locus of potency of the force which initiates the infinitival process, but the fact
that that force becomes less and less clearly delineated and associated with the
subject to become more diffuse and associated with the speaker. Consequently,
the difference between the main and conceptualization relations becomes less
distinguishable. The constructions in (14) and (15) illustrate the EVA presented
in Figure 2, because the distinction between the subject and object of conceptu-
alization is blurred, and the conceptualization CL1 is construed from a more
internal or subjective vantage point.
In order to more clearly illustrate the difference between the two types of
constructions, consider the difference between the ability sense of pouvoir
illustrated in (16), and the epistemic pouvoir, illustrated in (17) and (18):
(17) Il peut ne pas avoir compris, répète
He may-3sg.pres.ind neg not have.inf understood repeat.imp
la question
the question
‘He may not have understood, repeat the question.’
(18) Il peut être enfermé, appelle le
He may-3sg.pres.ind be.inf locked up call.imp him.cl
‘He might be locked up, call him.’
The sentence in (16) was already argued to present a maximally objective

construal of the objective scene. Mary is the locus of potency that anchors the
modal relation. That relation is maximally separate from the conceptualization
relation, and the speaker is a mere spectator of the main relation. The clear
delineation of the locus of potency, as well as the maximum separation between
the main and conceptualization relations led us to analyze the construction as
an instance of the OVA.
The analysis of epistemic modality makes use of Langacker’s Evolutionary
Momentum Model (1991). Pouvoir places the complement process in potential
reality, and devoir places the infinitival process in projected reality. The passage
from the possibility sense presented in (16) to epistemic modality involves the
diffusion of the locus of potency to the point where the latter can only be
identified with the evolutionary momentum of reality, that is to say the way
reality can reasonably be expected to evolve (Langacker 1991; Achard 1996). For
example, in (17), the locus of potency is not associated with the subject, but it
resides in the analysis of the elements of reality, which are accessible to the
speaker through observation and reasoning. In other words, because the
appreciation of the evolutionary momentum of reality is speaker internal by
definition, the main relation is no longer separate from the conceptualizing

relation. In fact, epistemic modality represents the endpoint of the subjectific-
ation of the main relation because its realignment onto the conceptualizing
relation is complete (Langacker 1991).
The last piece of the analysis consists in showing why an objective construal
of CL1 precludes the presence of the markers of aspect and voice on the infin-
itival process, while a subjective construal renders their presence possible. I
would like to argue that the presence of grammatical markers reflects the
speaker’s conceptual manipulation of the infinitival process, and that manipula-
tion is only possible in the case of a subjective construal of CL1, because it is
only in that configuration that the speaker has independent conceptual control
of the subordinate process.11
First, the analysis of the presence of grammatical markers in terms of the
speaker’s conceptual control and manipulation of the complement should not
be surprising when one considers the kind of information provided by the
markers. As was pointed out in Section 3, an infinitive provides information
regarding the core content of a process type (Langacker 1991). The addition of
the markers of aspect and voice provides information relative to the internal
structure of that process. The aspectual marker is traditionally analyzed as
indicating, with respect to some temporal reference point, whether the process
is completed or still in progress. Concerning the passive marker, Langacker
(1991: 197) writes that it “overrides the content verb with respect to one
dimension of imagery, namely which processual participant assumes the status
of trajector.” In other words, the passive marker provides information concern-
ing the specific coding of the participants in the complement process, i.e. their
figure/ground organization.
These cursory observations do not do justice to the complexity of the
problem of the semantic characterization of the grammatical markers con-
cerned, but they should be sufficient for the purposes of this paper. The
important point is that the addition of a specific grammatical marker (aspect or
voice) to a process type imposes its own particular meaning on the conceptual
content provided by the infinitival process. To understand the idea of conceptu-
al manipulation as it is used here, we need to reconsider the layering of seman-
tic functions exhibited by the clause. Recall from Section 3 that an infinitive
represents a process type, and that the non-finite markers provide information
relative to the inner structure of a particular instance of that type. That infor-
mation for the verb phrase is viewed as corresponding to the function of
quantification for nominals. I propose that the quantification of an instance of
110 Michel Achard
the process profiled in the complement represents the speaker’s conceptual

manipulation of the complement process. Let us take an example. The stem
nager ‘swim’ represents a process type. That process can be instantiated, that is
considered with respect to a particular situation. It can further be quantified. In
Marie doit avoir nagé, elle a les cheveux mouillés ‘Mary must have swum, her hair
is wet’ for example, the process nager is instantiated because the speaker has a
particular episode of the process in mind, tied to a specific time and place. It is
also quantified, because the instance of the process type is considered with
respect to its inner structure (accomplished at the time of the utterance). The
operations of instantiation and quantification represent the speaker’s conceptu-
al manipulation of the original nager type.
Second, it is the subjective construal of CL1 that provides the speaker with
the necessary control to conceptually manipulate the infinitival process. Since
the difference between the conceptual relation and the main relation is blurred,
the speaker, as the subject of the conceptual relation is also directly associated
with the main relation. She therefore has direct access to the infinitival process,
and can conceptually manipulate it. In order to illustrate this rather abstract
point, consider the example of epistemic modality in (19):
(19) Marie doit être revenue à 6 heures, je vois sa voiture
Mary must-3sg.pres.ind be.inf returned at 6 hours, I see her car
‘Mary must have returned by 6 o’clock, I see her car.’
The meaning of epistemic modality involves evaluating the likelihood of

occurrence of the infinitival process. In (19), the speaker’s interpretation of
reality allows her to place Mary’s return in projected reality. The speaker
necessarily has direct access to the complement process because the whole
evaluative procedure that allows the placement of the complement process with
respect to reality is speaker-internal. To phrase it differently, with epistemic
modals, the speaker has control over the complement process because both
main and conceptualization relations are anchored by the speaker. Consistent
with the hypothesis proposed here, that control is reflected by the presence of
the markers of voice and aspect on the infinitival complement.
Conversely, the speaker’s absence of conceptual control in the cases of
objective construal of CL1 can be explained by the speaker’s lack of independent
access to the infinitival process, because the main relation is anchored by the
subject, and that the speaker’s role is merely that of an observer. The variation
in the marking on the infinitival process in the constructions presented in this
section is thus directly imputable to the kind of viewing arrangement existing
between the speaker and the complement scene, or in other words, along the
subjective axis.
5.3 Infinitival and finite complements in conceptualizing

subject constructions
Another example of the relevance of perspective to the form of complements in
French can be found along the objective axis. In Achard (1996), I argue that the
distribution between infinitival and finite complements in CSC is explainable by
different configurations of viewing arrangement between the main clause subject
and the complement clause. Recall that CSC’s are constructions that describe the
conceptualization of the main subject. They are illustrated in (20)–(22):
(20) a. Pauli voit qu’ili a les cheveux blancs
Paul see-3sg.pres.ind that he has-3sg.pres.ind the hairs white
‘Paul sees that he has gray hair.’
b. *Paul voit avoir les cheveux blancs
Paul see-3sg.pres.ind have.inf the hairs white
‘Paul sees having gray hair.’
(21) a. Pauli croit qu’ili comprend
paul believe-3sg.pres.ind that he understand-3sg.pres.ind
‘Paul believes that he understands.’
b. Paul croit comprendre
Paul believe-3sg.pres.ind understand.inf
‘Paul believes to understand.’
(22) a. *Pauli veut qu’ili parte
Paul want-3sg.pres.ind that he leave-3sg.pres.subj
immédiatement
immediately
‘Paul wants that he leave immediately.’
b. Paul veut partir immédiatement
Paul want-3sg.pres.ind leave.inf immediately
‘Paul wants to leave immediately.’
To briefly summarize the situation, the verbs of perception in (20) can only be
followed by a finite complement when the main and subordinate subjects are
coreferential. Conversely, in (21), the volition verbs can only be followed by an
infinitival complement in coreferential cases. The verbs of cognition illustrated
in (21) can be followed by both types of complements.
112 Michel Achard
The analysis presented in Achard (1996) rests on two main points. First, the
specific form of the complement reflects a particular viewing arrangement
between C1 and CL2. A finite clause is the sign of an objective construal of CL2,
and an infinitival complement reflects a subjective construal of CL2. Secondly,
the distribution of the different complements with the different main verbs is
imputable to the specific viewing arrangement imposed by the semantics of the
main verb, along with the specificity of the situation where the main and
complement clause subjects are coreferential. I will briefly summarize these two
points here, and refer the reader to Achard (1996) for further detail.
With respect to the first point, the crucial element in the difference of
viewing arrangement between finite clauses and infinitival complements
concerns the exact nature of the object of conceptualization, and more specifi-
cally the role of the main participant S2 in the subordinate relation as an object
or subject of conceptualization. When the complement is coded as a finite
clause, it is construed with maximum objectivity from the vantage point of C1.
Crucially, the main participant in the subordinate relation is also construed
objectively, as an object of conceptualization. It is on stage, profiled as the
trajector in the subordinate relation. The clause itself, including its main
participant, is the object of conceptualization of C1. Note that this configuration
maximizes the asymmetry between C1 and S2, since they are respectively treated
as subject and object of conceptualization. The viewing arrangement of a finite
clause is presented in Figure 5 below.
objective axis
scope
C1 S2
predication
V point CL2
Figure 5. Finite clauses: Optimal viewing arrangement
Figure 5 illustrates the OVA. The complement clause is viewed as an object

of conceptualization from the vantage point of C1. Importantly, the subordinate
subject S2 is also on stage, and thus objectively construed.
The viewing arrangement expressed by the presence of an infinitival
complement reflects the blurring of the asymmetry between subject and object
of conceptualization. S2 is not viewed as part of the object of conceptualization,
but as the conceptualizer of the complement clause. More specifically, the
conceptualizer is of course C1, but C1 construes CL2 from the vantage point of
S2 (this is possible since the two are coreferential). The participant in the
subordinate relation is therefore not an object of conceptualization, but an
implicit conceptualizing presence, not itself conceptualized. The object of
conceptualization is the complement process alone, represented by the infin-
itival verb. The viewing arrangement of infinitival complements is illustrated in
Figure 6 below.
objective axis
scope
C1 / S2
predication
V point CL2
Figure 6. Infinitival complements: Egocentric viewing arrangement
Figure 6 illustrates the EVA presented in Figure 2b, because the complement
clause is conceptualized from inside its scope of predication. CL2 is conceptual-
ized from the vantage point of C1 occupying the position of S2.12
With respect to the second point, the distribution of the infinitival/finite
complements with the different verbs is a matter of semantic compatibility. The
verbs of perception and volition impose viewing arrangements representative
of the OVA and EVA respectively, because of their respective meanings. The
other verbs are compatible with both viewing arrangements depending on the
speech situation. The perception verbs perfectly illustrate the OVA because a
perceptual relation maximizes the subject/object asymmetry by profiling the
contact existing between the subject and an external object, that is to say, an
object with its own independent existence. The object of perception is placed at
a comfortable distance (not too close or too far) from the perceiver, so that the
act of perception can be as efficient as possible. Consequently, the object of a
perception verb is always construed objectively, independently of the particular
grammatical construction in which it appears. The Optimal Viewing Arrange-
ment they impose between their subject and object represents an important
semantic/conceptual characteristic of the verbs of perception. It explains why
in a CSC, the whole CL2 clause must be construed objectively from C1’s vantage
point. The participation of S2 in the relation profiled by V2 is also construed
objectively. Importantly, the coreferentiality of C1 and S2 does not change the
type of viewing arrangement existing between subject and object. The conceptual
114 Michel Achard
distance between these two entities that is characteristic of the perception verbs
is always preserved, so S2 cannot be treated as a conceptualizing presence, i.e.
construed subjectively.
The verbs of volition present the opposite case. The type of viewing
arrangement existing between their subject and their complement exemplifies
the EVA. The object of volition is not an element of the outside world, but can
exist exclusively as part of the subject’s conceptualization. For example, in Jean
veut une orange ‘John wants an orange’, the orange may exist exclusively as part
of John’s conceptualization. In a CSC, the complement content is speaker-
internal in the sense that it has no independent existence outside the subject’s
conceptualization of it. Verbs such as vouloir blur the asymmetry between their
subject and object of conceptualization because the conceptual distance
between their subject and complement is considerably shorter than in the case
of perception verbs. The reason for such short distance is that the content of the
complement of vouloir is not external to the main clause conceptualizer but
conceived by her alone. In the same way as the distance between subject and
object introduced by the perception verbs illustrates the OVA, the conceptual
proximity between the two arguments introduced by vouloir can be said to
illustrate the EVA.
Given the particular viewing arrangement of vouloir, coreferentiality
between main and subordinate subjects is quite different from a situation of
non-coreferentiality, because it further reduces the conceptual distance between
subject and object. In both cases, CL2 represents the conceptualization of C1, an
event conceived of by C1 alone. However, when the main participant in CL2 is
different from C1, it cannot be treated as a conceptualizing subject, and must
therefore be objectively construed. The whole clause, including its main
participants is the object of conceptualization. On the other hand, when C1 and
S2 are coreferential, the sentence describes C1’s conceptualization of her
participation in a particular event. Since that event is the sole creation of C1, the
latter’s most natural vantage point is S2’s. S2 is thus most naturally construed as
a conceptualizing presence, not itself conceptualized and not specifically
mentioned. The object of conceptualization is the process alone.13
6. Discussion
In this section, I would like to examine some of the consequences of the results.
First, one could legitimately wonder if what has been called perspective
throughout this paper is truly the same phenomenon, because the subjective/
objective construal of an entity is not necessarily reflected by the same syntactic
behavior. For example, the maximally objective construal of the objective scene
along the subjective axis is reflected by the fixed (uninflected) nature of the
complement, as in Marie sait nager ‘Mary knows how to swim’. A more
subjective construal of the objective scene is accompanied by possible marking
on the complement, which implies an increase in the amount of phonological
elaboration of that complement, as in Marie doit être revenue ‘Mary must have
come back’. Conversely, along the objective axis, the opposite situation seems
to obtain. The objective construal of the complement scene is reflected by the
explicit mention of the subordinate subject, as well as the markers of tense,
mood and person on the verb, as in Jean veut que Paul parte ‘John wants Paul to
leave’. The subjective construal of S2 is reflected by the presence of a bare
infinitive, and thus by a decrease in the amount of phonological elaboration of
the complement as in Jean veut partir ‘John wants to leave’. However, this
seemingly opposite result of a subjective/ objective construal in different
situations should not be surprising because the notion of viewing arrangement
is always relative to a specific situation, and its syntactic reflex depends on
exactly what element the construal applies to. To come back to our example, the
difference in the situation along the subjective and objective axes can be
explained by the inherent complexity of multi-clause constructions, as well as
which entity is construed subjectively or objectively. Along the subjective axis,
the object of conceptualization with respect to which construal exhibits
potential variation is the whole objective scene, and not a clearly delineated
entity within that scene. The shift from an objective to a subjective construal is
characterized by the realignment of some aspect of the main relation onto the
conceptualization relation, but the vantage point from which the clause is
conceptualized is not shifted in a way that can be easily located. Along the
objective axis, the shift in viewing arrangement is easier to delineate. The results
presented in Section 5.3 are imputable to the subjective/objective construal of
the subordinate subject, and its lack of explicit mention when it is construed
subjectively is consistent with similar patterns observed in different construc-
tions in different languages (Langacker 1985, 1990, 1991). In spite of these
differences, I would still claim that the phenomena presented in Sections 5.2
and 5.3 represent instances of the same general principles.
The second point that emerges from this analysis is that it is at the same
time compatible and complementary to the cognitive accounts of complements
found in Givón and Langacker. The recognition of the role of perspective does
116 Michel Achard
not invalidate the considerations of independence and semantic import of specific

complement types. On the contrary, it motivates them. It remains true that a
bare infinitive denotes a complement viewed less independently than a finite
clause, but it also provides a way to explain why a given scene is viewed with a
particular degree of independence. The recognition of perspective also enables
us to differentiate between constructions that exhibit similar syntactic forms.
Consider for example the sentences Jean sait nager ‘John is capable of swim-
ming’ and Jean veut nager ‘John wants to swim’. In both cases, the complement
is coded as an infinitive, but the infinitive coding reflects different conceptual
organizations. With the volition verb, it reflects the subjective construal of the
subordinate subject S2. Since the presence of a finite clause reflects an objective
construal of the complement scene from the vantage point of the main clause
subject, the perspective analysis accounts for the fact that the conceptualizing
subject verbs, but not the modals can be followed by finite complements.
Third, the approach presented here recognizes not only the importance of
the semantics of the main verb but also the properties of the situation the
sentence codes. In other analyses, the main verb is often judged fully responsi-
ble for the kind of complements which follow it. For example, in Givón (1980),
the binding strength of the main verb explains the syntactic form of its comple-
ments. It is worth noticing that the kind of viewing arrangement imposed on
the complement scene can also be viewed as a semantic property of certain
verbs. It is part of the meaning of the perception and volition verbs to respec-
tively impose the OVA and EVA between their subjects and complements. The
opposite nature of these verbs is mirrored by their opposite syntactic behaviors.
However, viewing arrangement is primarily a matter of construal, and is
therefore potentially influenced by the specific pragmatics of the speech
situation. With most verbs of cognition, the complement scene can be con-
strued both objectively and subjectively, both construals respectively reflected
by the syntactic presence of a finite or infinitival complement (Achard 1996).
Finally, we can wonder how important the notion of perspective is to the
complete understanding of complement constructions. In this presentation, I
was merely concerned with presenting the notion and its potential value for
complement research, and I did not pay enough attention to the fine grained
relation existing between perspective and other elements of construal such as
grounding for example. Furthermore, since this paper focussed on a single
language, the usefulness of perspective to a typological study of complements
remains a matter for further research. However, the close connections existing
between the perspective analysis and other accounts presented in this volume
are quite encouraging. For example, Verspoor’s investigation of conceptual

distance nicely complements the account defended here, because the notion of
perspective can only reach its full explanatory power if it is embedded in a more
global theory of conceptual distance. Furthermore, my analysis is fully compati-
ble with Delbeque’s account of the cognitive constraints existing on cliticization
in Spanish, because both emphasize the crucial importance of the respective
role of the speaker and main clause subject in the construal of the complement
scene. Finally, although the coding of the complements is very different in the
two languages, Polinsky’s account of Tsez complementation seems quite
compatible with a perspective analysis. I will take the existence of the connec-
tions between these approaches to different linguistic phenomena in different
languages as a positive sign. Despite differences in the ecology of the different
languages, as well as the methodology used to investigate them, a solid under-
standing of the core meaning of complementation seems to be within our reach.
7. Conclusion
This paper has argued that the notion of perspective can valuably be applied to
the investigation of complement constructions in French. The discussion
centered around the following three points: (i) the vantage point from which
the complement clause is conceptualized enables us to differentiate between the
modal and CSC verb classes, (ii) different viewing configurations along the
subjective axis account for different possible marking on the infinitival comple-
ment for causative, movement and modal constructions, and (iii) the distribu-
tion of infinitival/finite complements with CSC verbs is explainable by different
viewing arrangement configurations along the objective axis.
Abbreviations
cl clitic pl plural
cond conditional pres present
imp imperative sg singular
ind indicative subj subjunctive
inf infinitive
118 Michel Achard
Notes
1. The independence of finite complements is further enhanced by the necessary presence of

the complementizer que. Langacker (1991: 447) argues that its English correspondent that
imposes an “atemporal, maybe even nominal construal on the complement scene”, and a
similar analysis can be adopted for its French counterpart que. This nominal construal
emphasizes the independence of the complement clause, which is construed holistically, and
manipulated as a unitary entity.
2. Although both indicative and subjunctive complements can be said to be grounded, they
differ by the nature of their grounding. In Achard (1993), I argue that the indicative
represents a proposition, or in other words an event manipulated with respect to some
conceptualizer’s conception of reality. It crucially includes tense predications, which provide
the proposition a putative location in reality. By contrast, the subjunctive marks the
consideration of an event with respect to a more localized mental space set up with respect
to the main subject alone (to express her wishes, desires, or reactions for example). Since
there is no putative location in reality for a subjunctive complement, the mood does not
contain tense predications (Sandoval 1984). In terms of its semantic function, a subjunctive
clause represents an arbitrary instance of a process type, specifically conjured up to occupy
the mental space of the subject’s wishes or desires (Achard 1993). This use is similar to the
non-specific meaning of the indefinite article in such examples as “Harry wants to marry a
dancer”, where the dancer represents an arbitrary instance, specifically conjured up for the
description of Harry’s desires (Langacker 1991).
3. Givón’s solution in terms of the binding force of the main verb is certainly a step in the
right direction, but it does not explain the cases where two kinds of complements can be
found following the same main verb.
4. The possibility for a sentence to include a viewpoint different from the speaker’s is first and
foremost a semantic characteristic of the main verb, independently of the construction it is used
in. For example, a cognitive verb such as croire ‘believe’ in Jean croit le professeur ‘John believes
the teacher’ presents the viewpoint of the subject, even though the construction is monoclausal.
The occurrence of those verbs in finite clause constructions allows them to make use of that
property systematically. By comparison, other verbs (among them the causatives and the
modals) never present the point of view of their subject, in monoclausal and biclausal
constructions alike. Typically, in French, these verbs do not occur with a finite clause. I am
grateful to an anonymous reviewer for this volume for pointing these facts out to me.
5. In Achard (1996), these constructions are called “sentential complement constructions”.
I believe that the term was confusing, and that the new terminology better reflects the
specificity of the constructions.
6. In Benveniste’s model, modality represents one type of auxiliary relation, along with
temporal relation (the relation provided by the auxiliary in compound tenses), and the
relation de diathèse ‘diathesis relation’ (found with the passive auxiliary être).
7. The argument is that the imperative forms of the four verbs savoir, devoir, pouvoir, vouloir
are not borrowed from the indicative mood, as it normally is the case, but from the
subjunctive (which is representative of the true auxiliaries avoir ‘have’ and être ‘be’). The
borrowing of a subjunctive form to express the imperative is a characteristic of a “verbe

puissanciel” (Guillaume 1929).
8. This analysis cannot obviously account for the impersonal constructions. These will not
be considered here.
9. Sentence (13b) is perfectly felicitous in the epistemic sense of pouvoir. However, (13b)
presents the ability sense of pouvoir.
10. Similarly, in (13a), John is also the locus of potency of the modal force. The difference
between the modal sense of savoir and the ability sense of pouvoir essentially pertains to the
nature of the potency. It is investigated in detail in Achard (1998).
11. The importance of the relative role of the speaker and the main clause subject in the
construal of certain aspects of the complement scene is shown by Delbeque (this volume) to
account for the cliticization of the complement clause in Spanish.
12. If we take seriously the claim that the subjective construal of S2 explains the presence of
an infinitival complement, two points need to be made specific. The first concerns the
correlation between the subjective construal of an entity and explicit mention. Langacker
(1985) notes that the objective construal of an entity is often indicated by the explicit
mention of that entity, whereas the subjective construal of an entity is expressed by the lack
of explicit mention of that entity. For example, in John is across the table from me, the
reference point (from) me is put on stage. It is thus construed objectively and explicitly
mentioned. In John is sitting across the table, the reference point is not put on stage, but
construed subjectively and not explicitly mentioned. Note that the French data presented in
this paper are very similar. The objective construal of S2 in the case of a finite clause is
accompanied by its explicit mention, whereas the subjective construal of S2 in the case of an
infinitival complement is reflected by its lack of explicit mention. The second point concerns
the correlation between the subject’s lack of explicit mention and the presence of an
infinitival complement in the subordinate clause. Recall that the main difference between
infinitival and finite complements is that infinitival complements are not grounded, i.e.
include no person, tense, or mood predications (Langacker 1991). Note that different levels
of grounding can be observed. An indicative clause can be said to be fully grounded because
it includes both person and tense markers. A subjunctive clause is only partially grounded
because it does not include tense markers, but it still contains person markers which relate
the complement to the speech situation. It therefore seems that in French (and quite possibly
in several other languages), person markers constitute the minimum grounding, and that
other verbal grounding predications cannot occur unless person markers are also present. If
the subordinate subject is not explicitly mentioned, person marking on the verb is not
needed, and the complement is expressed as a process type, not considered with respect to
the speech situation. Note that the analysis presented here can easily account for the fact that
the subjective construal of the subordinate subject in CSC’s is only possible in French when
the main and subordinate subjects are coreferential. Recall that CSC’s describe the conceptu-
alization of C1. If C1 and S2 are different, the subordinate subject has no conceptualizing role.
It is only when C1 conceptualizes her own participation (as S2) in the complement process
that she can adopt S2’s vantage point to construe CL2.
13. For the sake of brevity, I am not considering the verbs of cognition which can be followed
by both kinds of complements. With those verbs, the complement choice is determined by
</TARGET "ach">
120 Michel Achard
pragmatic factors, ultimately decided by speaker choice. I refer the reader to Achard (1996)
for further detail.
References
Achard, Michel. 1993. Complementation in French: A cognitive perspective. Doctoral Disserta-
tion, UCSD.
Achard, Michel. 1996. “Perspective and syntactic realization: French sentential comple-
ments”. Linguistics 34: 1159–1198.
Achard, Michel. 1998. Representation of cognitive Structures. Syntax and Semantics of French
Sentential Complements. Berlin: Mouton de Gruyter.
Beck, David (this volume). “Nominalization as complementation in Bella Coola and
Lushootseed”.
Benveniste, Emile. 1974. Problèmes de Linguistique Générale. Paris: Gallimard.
Cervoni, Jean. 1987. L’énonciation. Presses Universitaires de France.
Delbeque, Nicole. (this volume). “Cognitive constraints on complement clause cliticization
in Spanish”.
Farkas, Donka. 1992. “On obviation”. In Ivan A. Sag, and Anna Szabolcsi (eds), Lexical Mat-
ters. Stanford: CSLI, 85–109.
Language 4: 333–377.
Guillaume, Gustave. 1929. Temps et Verbe. Paris: Champion.
Haiman, John. 1985. Natural Syntax. Cambridge: Cambridge University Press.
Horie, Kaoru. (this volume). “Complementation in Japanese and Korean: A contrastive and
Langacker, Ronald W. 1985. “Observations and speculations on subjectivity”. In John
Haiman (ed.), Iconicity in Syntax. Amsterdam: John Benjamins, 109–150.
Langacker, Ronald W. 1987. Foundations of Cognitive Grammar. Vol. 1: Theoretical Prereq-
uisites. Stanford: Stanford University Press.
Langacker, Ronald W. 1988. “An overview of cognitive grammar”. In Brygida Rudzka-Ostyn
(ed.), Topics in Cognitive Linguistics. Amsterdam and Philadelphia: John Benjamins, 3–48.
Langacker, Ronald W. 1990. “Subjectification”. Cognitive Linguistics 1: 5–38.
Langacker, Ronald W. 1991. Foundations of Cognitive Grammar. Vol. 2: Descriptive Appli-
cation. Stanford: Stanford University Press.
Polinsky, Maria. (this volume). “Variation in complementation construction: Agreement in
Tsez”.
Ruwet, Nicolas. 1984. “Je veux partir/*Je veux que je parte. A propos de la distribution des
complétives à temps fini et des compléments à l’infinitif en Français”. Cahiers de
Grammaire 7. Toulouse-le Mirail, 75–138.
Sandoval, Maria. 1984. “A propositional classification of Spanish sentences”. Coyote Papers
Vol. 1: 120–152.
Suzuki, Satoko. (this volume). “De dicto complementation in Japanese”.
Verspoor, Marjolijn. (this volume). “Iconicity in English complement constructions:
Conceptual distance and cognitive processing levels”.
<LINK "bec-n*">
<TARGET "bec" DOCINFO
AUTHOR "David Beck"
TITLE "Nominalization as complementation in Bella Coola and Lushootseed"
KEYWORDS ""
SIZE HEIGHT "220"
WIDTH "150"
VOFFSET "4">
Chapter 6
Nominalization as complementation
in Bella Coola and Lushootseed*
David Beck
University of Toronto
1. Introduction
The Salishan languages Bella Coola and Lushootseed,1 like all languages of the
family, have earned a great degree of notoriety in the literature for the tremen-
dous degree of flexibility that they show in allowing a wide range of elements,
including complement clauses, to fill virtually any syntactic role in a sentence.
This flexibility has led a number of researchers (e.g. Kuipers 1968; Kinkade
1983; Jelinek and Demers 1994) to argue that these languages lack an underlying
distinction between nouns and verbs. One of the strongest arguments against
this position is the existence of a pan-Salishan prefix, s-, commonly glossed as
a nominalizer (van Eijk and Hess 1986). The most transparent function of this
prefix is to change a word denoting a state or process such as Lushootseed ‘6S6d
‘eat’ to a word denoting a related object or thing — s‘6S6d ‘food’ — which has
a number of typically nominal morphosyntactic properties that the “s-less”
form does not have (e.g. the ability to take possessive prefixes — ds‘6S6d ‘my
food’ vs. *d‘6S6d). In the absence of a verb-noun distinction, the proper
characterization of this prefix’s meaning is unclear. In addition to its function
in forming lexical nominals, however, s- can be applied to more complex
expressions corresponding to entire English clauses. While these constructions
have often been treated as simple cases of morphological complementation
(particularly in Bella Coola — see, for example, Kroeber 1991), s-clauses show
a wide range of nominal properties. In the paper that follows, I will present
evidence that s-clauses in Bella Coola and Lushootseed are, in fact, instances of
nominalization and that, in spite of certain compositional or syntactic differ-
ences in the properties of s-clauses in the two languages, all instances of s- —
including formation of lexical nouns — make use of a single related set of
122 David Beck
meanings consisting of a prototype and two subschema. These meanings of s-

are applied in both languages to bare verb stems to form lexical nominals and,
when used in more complex expressions, also form clausal complements,
although the two languages differ in that Lushootseed applies s- at one level of
composition where it has scope over only the verb and its objects, whereas Bella
Coola applies it at the next level where it has scope over the verb, its objects, and
the syntactic subject. Taking these two factors — the schematically-related
meanings of s- and its compositional/syntactic properties — into account, it is
possible to offer a unified analysis of all uses of s- in two of the most distantly-
related Salishan languages. This analysis raises important questions about the
cross-linguistic relationship between complementation and nominalization
discussed in a number of the papers in this anthology (e.g. Horie, this volume;
Schüle, this volume), as well as the conceptual distinction between finite and
non-finite clauses, discussed in the final section of this paper.
2. Nominalizers and nominalizations in Cognitive Grammar
In both Bella Coola and Lushootseed, morphological nominalizations, includ-

ing lexical nominalizations, are realized by means of nominalizing prefixes —
s- and d6xw- in Lushootseed and s- and si- in Bella Coola. The focus of the
discussion here will be s-, both because it is the morpheme shared by the two
languages (and by the Salishan family as a whole) and because it is the one
whose use is most frequent and most resembles that of a syntactic complement-
izer. As a nominalizer, the primary function of s- is to shift the conceptualiza-
tion or profile of a verb away from that of a process or event (that is, a sequence
of component states over time) towards that of an noun, which prototypically
is an object occupying a specific, identifiable location in conceptual space. The
contrast is illustrated in Figure 1, which gives representations for the English
sentence “The plane approached the runway” (a) and the participial “The
plane’s approaching the runway” (b). In (a) the event is presented as a series of
component states (distances between the aircraft and the runway) which change
successively (diminish) over a specific period of time profiled by the expression
(as shown by the thickening of the arrow representing the event’s temporal
extension). Here, the plane is considered to be the trajector or “primary clausal
figure” (Langacker 1987) of the event, which is characterized in terms of the
change in distance between the trajector and a secondary object or landmark
(the runway).2 In (b), the same event is presented not as a series of states
Nominalization as complementation in Bella Coola and Lushootseed 123
(a) tr (b) tr
tr tr
tr tr
tr tr
lm lm
Figure 1. Verbal versus nominal profiling of events
“scanned” sequentially, but instead as a set of states whose temporal extension is

not at issue — instead, the profile of the expression is the event as a whole and the
region of conceptual space which it occupies (thickened oval, only latent in (a)).
The meanings of nominalizers, however, frequently include other semantic
parameters. In English, for example, the meanings of the nominalizing suffixes
-ee and -er include not only the reification of the process designated by the verb
to which they are applied, but also involve a shift of profile to one of the event-
participants: thus, interview >> interviewee shifts the profile from the process of
interviewing to the person being interviewed, while interview >> interviewer
shifts the profile to the person doing the interviewing (Langacker 1991). In
Lushootseed and Bella Coola, morphological nominalizations involve a variety
of meanings, three of which are expressed by the s-prefix. These are shown in
Figure 2. In its prototypical and most schematic use, s- is simply a nominalizer
which serves to reify an event, thereby suspending sequential scanning of the
process and delimiting a region of conceptual space — that is, it construes a
process atemporally as an object or a thing (cf. van Eijk and Hess 1986). Of the
two subschema, Subschema I shifts the profile of the nominalization away from
the reified event to a particular participant in that event — specifically, to a
participant which would be realized in a non-nominalized clause as an oblique
actant contained within a prepositional phrase (see the examples in (2) below).
Subschema II, which is peculiar to Bella Coola, shifts the profile, not to a
participant in the reified event, but to the time-span over which the event
occurs, as in (1):
(1) ‘as+‘uSqn+ak+s ‘uS ti+qla+tÑ ayx
use+pail+hand+3sg prep deic+water+deic
wa+s+‘as+qupÑ tn+ak+a+kw
deic+nom+use+drum+hand+3pl+quot
‘she took her pail to the water, the drums sounding.’
(Nater 1984: 103)
124 David Beck
Schematic Prototype
prototype reified event
subschema
Subschema I Subschema II
oblique actant time-span
Figure 2. The meanings of the s-nominalizer
Both of these subschemas will be discussed in more detail below.

In addition to such conceptual or semantic issues, a full discussion of the
nominalization patterns in a language must take into account certain compo-
sitional or syntactic factors which govern how the nominalizers are combined
with verbs and clauses to form larger constructions. The most important of
these factors is the scope of the nominalizing morpheme — that is, which
elements of the clause undergo conceptual reification as a result of the mor-
pheme’s application. This gives us three possible types of nominalization:
Lexical Nominalization applies to a verb with no accompanying actants (argu-
ments), creating a new lexical item which designates a generic type of action or
event;
Participial Nominalization applies to a verb and all of its actants except the
subject, creating an instance of an event not uniquely identified or located in
conceptual space;
Sentential Nominalization applies to a verb along with all of its actants including
the subject, creating an instance of an event distinguished from all others with
a unique and specifiable spatial and temporal location.
The distinction between the three depends on the level of composition — the
stepwise elaboration of the clause from its component elements — at which the
nominalizer is applied. In both Bella Coola and Lushootseed, the s-prefix can be
applied to bare verbs to form lexical nominalizations. Additionally, both
languages may apply s- to more complex expressions, including some or all of

the verb’s actants. Lushootseed makes use of the participial compositional
pattern, applying the nominalizer at the level of composition after the elabora-
tion of the objects and before the elaboration of the clausal trajector or subject;
Bella Coola, on the other hand, creates sentential nominalizations, applying s-
after the elaboration of both subject and objects.
This three-way distinction in nominalization patterns closely parallels a
proposal made by Langacker (1991), whereby three types of nominalization are
contrasted (action, factive, and nominalization of finite clauses) based on their
relative groundedness — in other words, on the specificity with which they are
identified with particular entities or instances of events that are known to, can
be seen by, or are presupposed to exist by the speaker (see also Achard, this
volume). However, while the nominalization types proposed by Langacker
parallel those proposed here in scope (bare verb stem, verb plus objects, verb
plus subject and objects), they differ in that in Cognitive Grammar (CG)
differences in scope correspond to differences in finiteness: action nominaliza-
tions create non-finite forms which designate types of events and factive
nominalizations create non-finite forms designating ungrounded instances. The
difficulty lies in that in CG a finite clause is considered to be fully grounded by
the specification of tense, whereas both Bella Coola and Lushootseed lack
inflectional marking for tense and so, in a certain sense, seem to lack what is
generally considered to be a finite clause, although of course they can not be
said to lack clauses which are fully grounded (i.e. which refer to specific events
located precisely in space and time). In the absence of tense, clauses in these
languages are grounded through full specification of their actants — in other
words, events are known to happen when and where their participants are
located. Given that Lushootseed participles — which realize their subjects as
possessors — also surface with all their actants, these do not appear to differ
conceptually or distributionally from nominalizations of “finite” clauses in
Bella Coola. Still, even though there does not seem to be a cognitive-functional
distinction, the scope distinction is preserved in the compositional/ syntactic
properties of the two types of nominalization. Some of the implications of this
will be discussed below, following a more detailed account of the
nominalization patterns in the two languages.
126 David Beck
2.1 Compositional pattern I: Lexical nominalization
Both Bella Coola and Lushootseed make extensive use of the s-prefix to create
lexical nominalizations — that is, to create word-level lexical items which
signify some type of reification of the meaning of the verb-stem to which the
prefix is attached. By far the most common schema of s- used for this purpose
is Subschema I, which both reifies the process signified by the stem and shifts
the profile to a syntactically oblique actant. This is illustrated in (2), where the
oblique object (italicised) in the first member of each pair of examples corre-
sponds to the profile of the lexical nominalization in the second
Lushootseed
(2) a. ‘u+‘6S6d tsi cÒ acÒ as ‘6 ti b6sqw
punc+eat deic.fem youngster prep deic crab
‘the girl ate crab.’
(Hess 1993: 39)
b. s+‘6S6d ‘food’
nom+eat
(Hess 1993: 202)
Bella Coola
c. ‘a†Ña+yuks+aw x+a+s+a†Ñ+aw+c
canoe-making+pl+3pl prep+deic+nom+canoe-making+3pl+deic
‘they were building their own canoes.’
(Davis and Saunders 1980: 183, line 91)
d. s+a†aÑ ‘canoe’
nom+canoe-making
(Nater 1984: 101)
In (2a), the single non-oblique (direct) actant of the intransitive root ‘6S6d ‘eat,
dine, feed on’ is the eater, while the single actant of ‘a†Ña ‘canoe-making’ (also
intransitive) in (c) is the canoe-builder.3 In both cases, it is the oblique actant
contained in the prepositional phrase which surfaces as the meaning of the
lexical s-nominalizations in (b) and (d). (3) contains additional examples of
lexical nominals formed with the s-prefix that shift the profile of the nominali-
zation away from the process and its trajector to a syntactically oblique object:
Lushootseed
(3) a. s+tÑ ilib ‘song’
nom+sing
(Hess 1993: 254)
b. s+‘uladxw ‘salmon’
nom+fish-for-salmon
(Hess 1993: 204)
Bella Coola
c. s+qwilac+S ‘juice, wine’
nom+crush-berries+perf
d. s+Sx̌wta ‘spittle’
nom+(to)spit
(Nater 1984: 101)
In each of these examples, Subschema I of s- is applied to a verb representing a

type, rather than an instance, of an event (i.e. ‘fishing’ in general as opposed to
a specific instance of ‘fishing’) and thus derives a new lexical item (‘fish’, not
‘what was fished [at a particular time]’). Once again, all of these verbs have only
a single direct actant, the syntactic subject, and require the use of a preposition
or some sort of transitivizing morphology to realize their semantic endpoints.
The fact that in simple, unaugmented verb stems these semantic endpoints
are only realized as oblique actants might lead us to expect that conceptually as
well as syntactically they are not direct objects in the usual sense of being
secondary, individuable event-participants. Rather than being participants, the
nominal semantic arguments inherent in meanings of words like ‘a†Ña ‘canoe-
making’ and ‘uladxw ‘fish for salmon’ seem to define the meaning of the verb
stem by delimiting a particular relation in which the trajector participates,
much as relative scales of size, colour, and so on characterize the trajector of
stative and atemporal predications (i.e. stative verbs and adjectives). Langacker
(1987) characterizes this type of entity as a relational landmark — that is, a
landmark which characterizes the trajector but which is not in itself an individ-
uable, conceptually autonomous entity. Thus, as relational landmarks, the
canoe in ‘canoe-making’ and the salmon in ‘fish-for-salmon’ serve to define a
particular type of ‘making’ or ‘fishing’ with respect to the typical object which
is made or fished but without evoking that object as a specific individual or
instance of a thing carved or fished. When the expression of such an entity as an
event-participant is required, it must be realized as a syntactically oblique object
contained within a prepositional phrase, the preposition serving as a semantic
link between the non-individuable landmark in the profile of the verb and the
fully individuable, grounded NP serving as the oblique object.4
Although by far the majority of productive uses of s- to form lexical
nominals make use of Subschema I, there are a number of other patterns found
128 David Beck
in the data and, as is typical of word-level processes in general, the meanings of

lexical s-nominalizations are often highly grammaticized. In a number of cases,
it is the verb’s trajector — or, more precisely, a specific trajector which repre-
sents an instance of the process designated by the root — rather than an oblique
actant which is profiled, as in:
Lushootseed
(4) a. s+qÑ axw ‘ice’
nom+frozen
(Bates et al. 1994: 181)
b. s+6li‘ ‘soul, life’
nom+alive
(Bates et al. 1994: 108)
Bella Coola
c. s+x̌um+aqÑ ws ‘tears’
nom+stream+eye
d. s+‘usqa+licÑ ‘smallpox’
nom+come-out+skin
(Nater 1984: 101)
The situation in Bella Coola is further complicated by the existence of some

inherently transitive roots, highly atypical of Salishan languages. These, when
nominalized, profile either an instrument or a direct object:
(5) a. s+qÑ Skw ‘tools’
nom+fix [something]
b. s+lqÑ ‘mind, brains’
nom+think [of something]
c. s+ulix+c ‘provisions’
nom+choose+food
d. s+kicÑ ‘laundry’
nom+wash
(Nater 1984: 102)
In the case of instrument-profiling (5a, b), this is clearly another application of

Subschema I (the instrument being oblique and realized as a PP); (5c, d), on the
other hand, profile a direct actant (direct object), and thus represent an
exceptional use of the s-prefix peculiar to Bella Coola with its “un-Salishan”
transitive roots. The examples given in (4) and (5) may indicate the need for
additional subschemas of s-, at least to the extent that these uses of the
nominalizing prefix can be shown to be productive, rather than conventional-

ized meanings associated with specific lexical items.
2.2 Compositional pattern II: Participial nominalizations
Although absent from Bella Coola, the second compositional pattern —

nominalization of a verb along with its objects but excluding its subject — is
one of the most distinctive features of Lushootseed grammar. Such construc-
tions are commonly referred to in Indo-European languages as “participles”, a
term I will extend here to cover the analogous structures in Lushootseed,
although it is not one that is commonly used in the Salishan literature.5
Lushootseed participles are illustrated in (6):
(6) x̌wulÑ pÑ a†Ña†Ñ ti‘iS s+‘abyid+s ti‘iS cÒ †Ña‘
only worthless deic nom+give+3poss deic stone
‘what he gives to Stone is only junk.’
(lit. ‘his given Stone [is] only junk.’)
(Hess 1993: 187, line 32)
This example shows a participle, s‘abyid ‘giving’, with an overt direct object,
ti‘iS cÒ†Ña‘ ‘Stone’ (an anthropomorphized boulder who is given worthless gifts by
Coyote), and a subject/ trajector (Coyote) realized by the possessive suffix -s
‘his/ hers/ its’.6 Being an example of the use of Subschema I from Figure 2
above, the participle in (6) has shifted its profile from the event of giving to the
object that is given — an oblique actant — which in turn serves as the syntactic
subject of the sentence as a whole. The formation of this type of complement
clause is the most frequent use of the s-prefix.7
In addition to acting as clausal complements, s-nominalizations in Lushoot-
seed, like participles in many languages, can also serve as attributive modifiers.8
(7) ti‘6‘ tu+s+6s+č6ba‘+s kwagwič6d
deic past+nom+stat+backpack+3poss elk
‘this elk he’d been backpacking.’
(Hess, 1998: 87, line 253)
In sentences like (7), Lushootseed participles take on the function of the English
oblique-centred relative clause, as they do in many languages around the world
(Comrie and Thompson 1985).
As well as allowing clauses to function as subjects, objects, and modifiers, s-
also creates more ordinary complement clauses through the application of the
schematic prototype, as in (8):
130 David Beck
(8) a. ‘6s+hay+dxw ti‘iS †Ñu+s+6+‘ib6š ‘6 ti‘iS dukwib6S

stat+know+lc deic hab+nom+stat+walk prep deic Changer
‘[he] knew that Changer would be travelling.’
(lit. ‘[he] knew [of] Changer’s travelling’)
(Hess 1993: 143)
b. ‘6s+x̌icil ti‘iS cÒ †Ña‘ ‘6 ti‘iS s+u+čala+d+s
stat+angry deic stone prep deic nom+punc+chase+caus+3poss
ti‘iS sbiaw
deic coyote
‘Stone [was] angry as he chased Coyote.’
(lit. ‘Stone [was] angry on/with his chasing of Coyote’)
(Hess 1993: 186, line 30)
Like the previous examples, the possessors of the nominalized clauses in (8)
represent subjects/ trajectors. In (a) the trajector of the nominalized clause
surfaces in a possessive “of”-phrase, ‘6 ti‘iS dukwib6S ‘of Changer’, while in (b)
it surfaces as the possessive suffix -s ‘his/her/its’.9
It is this realization of the trajector of the participle (the subject of the
corresponding finite clause) as a possessor that is the most obviously participle-
like feature of this type of nominalization. This feature is also illustrated in (9),
where the trajector of the nominalization is marked by the first-person posses-
sive d-:
(9) x̌wulÑ pÑ a†Ña†Ñ ti‘6‘ d+s+‘abyid
only worthless deic 1poss+nom+give
‘what I give [to him] is only junk.’
(lit. ‘my given [to him] [is] only worthless’)
(Hess 1993: 185, line 14)
In the next example in (10), the trajector of the participle s6sqÑ wu‘ ‘gathering’ is
an overt third-person NP, ti‘iS ‘iiš6ds ‘his relatives’:
(10) ‘u+šudxw ti‘iS s+6s+qÑ wu‘ ‘6 ti‘iS ‘iiš6d+s ‘al ti‘6‘
punc+see deic nom+stat+gather prep deic relatives+3poss prep deic
hikw cÒ †Ña‘
big stone
‘he saw that his relatives were gathered by the big stone.’
(lit. ‘[he] saw the gathering of his relatives by the big stone.’)
(Hess 1993: 185, line 3)
As in (8a), the possessor-status of the NP ti‘iS ‘iiš6ds ‘his relatives’ in (10) is

marked by the preposition ‘6 (cf. the corresponding finite clause ‘6sqÑ wu‘ ti‘iS
‘iiš6ds ‘his relatives are gathered’, in which the subject is not set off by a preposi-
tion). Consider also (11), which contains a participle, s‘6S6d ‘eating’ formed on
the intransitive verb ‘6S6d ‘eat [of]’.
(11) ¦Ju‘il+6xw ‘6 ti‘6‘ Su+s+‘6S6d ‘6 tsi s†Ñalq6b ‘6 ti‘6‘
enjoy+now prep deic irr+nom+eat prep deic.fem monster prep deic
qa wiwÑ su.
many children
‘he enjoyed [the thought] that the monstress would eat the many children.’
(lit. ‘[he] enjoyed [the thought of] the monstress’s eating the many
children.’)
(Hess 1976: 657)
Here the participle occurs with both an overt possessor — ‘6 tsi s†Ñalq6b ‘the
monstress’ — and an oblique object — ti‘6‘ qa wiwÑ su ‘these many children’ —
contained in prepositional phrases headed by ‘6. Note that both (10) and (11)
make use of the schematic prototype of s- to create a nominal actant which
would be complementized syntactically in a language like English.
The requirement that the trajector of the participle be realized in the role of
possessor bears close comparison to a similar feature of participles in a wide
range of languages (Comrie 1981; Comrie and Thompson 1985), as seen in this
example from Uzbek:
(12) a. men+i] yoz+gan xat+im
I+gen write+past-part letter+1sg
‘the letter that I wrote.’
(lit. ‘my having-written letter’)
(Comrie 1981: 82)
The same is true of English participles/gerunds, as in I gave Mary a ball >> my

giving Mary a ball.10 This use of the genitive/possessive to realize the trajector
of participles can be linked to Langacker’s (1991) reference-point analysis of the
possessive construction. Under this analysis, the possessive is seen as a relational
predication which profiles one object (the possessor) as a point of reference
with which another (the possessed) can be located in conceptual space and
identified as the particular object to which the speaker is referring, as shown in
Figure 3 (based on Langacker 1991: 171), which illustrates the English phrase
Bill’s father. Here, the speaker uses the identity of the possessor, Bill, to identify
132 David Beck
the particular father being referred to as that father which is located within the
possessor’s dominion — that is, as a member of the set of objects which the
possessor can be used to locate. Thus, the possessor becomes a reference point
and serves an indexical function (e.g. my country, my favourite food), identifying
the possessed item by dint of its unique association with its possessor (Bill’s
father being the only father whose identity can be uniquely established with
reference to Bill).
possessed
‘father’ ‘father’
‘father’
lm tr
‘Bill’
speaker possessor
dominion
Figure 3. The deictic function of possessor
Similarly, Taylor (1994) argues that the possessor of a deverbal noun in

English has the same function; for example, Harry in Harry’s love is used to
single out a particular instance of love for the hearer’s attention — that instance
of love of which Harry is the protagonist. Whether or not the possessor refers to
the underlying verb’s subject (as in Harry’s love) or object (as in Harry’s fear)
depends on their relative utility for correctly identifying the entity being singled
out by the speaker, in particular on their “topicality” and their “informativity”
(the precision with which they allow the hearer to pinpoint a specific event).
The details of Taylor’s argument are not directly relevant to Lushootseed
participles, in which the possessor always refers to the clausal subject. However,
these two criteria do offer an explanation of why it is that participles select for
possessor-subjects. In finite clauses, subjects are known to be highly topical
cross-linguistically (Keenan 1976; Li and Thompson 1979; Givón 1979) and, in
the case of Lushootseed and Bella Coola, subjects almost invariably correspond
to discourse topics (Beck 1996b). Because of this topicality, subjects often serve
the function of anchoring new information in discourse (see also Langacker
1991, where “topic” is defined in precisely these terms) and, thus, they serve as
the reference-point against which new information can be located in discourse
space. The same type of argument can be made for participles, which retain
more of their clausal properties than other deverbals and are thus, like finite
clauses, grounded by the location of their subject/ trajectors. Objects are not
available for this function, as they are included within the scope of the nominal-
ization (and so can not be used as a reference point for their own location).
Another common feature of participles typical of Lushootseed nominaliza-
tions is the retention of marking for voice and aspect. In Lushootseed there are
two classes of affix that express aspect or aspect-like concepts. The first —
which includes the stative (see (10) above), the punctual (8b), the continuative,
the progressive, and the stative-progressive — appears only on verbal roots and
derived stems. When verbs undergo nominalization, they retain this type of
aspectual marking, which appears “inside” (closer to the root than) the s-prefix:
(13) ti‘6‘ s+6s+qÑ w6l+b ‘6 ti lu†Ñ ‘6 ti‘6‘ s‘uladxw
deic nom+stat+roast+mid prep deic old prep deic salmon
‘an old man’s roasting salmon’
(Hess 1996: 47)
The second class of aspectual marking, which appears further from the root
than the nominalizing s-, covers a wider semantic range than aspect alone and
includes prefixes for the subjunctive, the habitual (see (8a)), the irrealis ((11)),
the past, and the additive. These prefixes can be applied to both verbs and
nouns (including participles), as in (14), and so are unaffected by participial
nominalization.11
(14) a. tu+qÑ iya†Ñ6d ti tu+scÒ istxw+s
past+slug deic past+husband+3poss
‘Slug had been her former (i.e. deceased) husband.’
(Hess 1993: 84)
b. hay lax̌+du+b+6xw ‘6 ti‘iS cÑ ix̌cÑ ix̌
then remember+lc+mid+now prep deic fish-hawk
ti‘iS tu+s+cut+t+6b+s ‘6 ti‘iS sč6txw6d
deic past+nom+speak-to+caus+mid+3poss prep deic bear
‘then fish hawk remembers what bear said to him.’
(lit. ‘then his was-spoken-by-bear [is] remembered by fish hawk’)
(Hess 1993: 194, line 46)
Note also that in (14b), the participle tuscut6bs ‘[past]+being spoken to’
contains the passivizing affix combination -t+6b, and the possessor of the
participle (shown by the third-person possessive suffix) is the object of the verb
cut ‘speak to’. Any grammatical voice, just as any type of aspect, found in
Lushootseed finite clauses can also be found in Lushootseed participles. This is
134 David Beck
particularly significant in that the hallmark of the finite clause in other languag-
es is the presence of markers for tense and mood, yet in Lushootseed — to the
extent that either of these are marked at all in the morphology — such markers
apply equally to verbs and nouns, and so may also be applied to participial
nominalizations. Thus, while the cross-linguistic distinction between participial
and sentential nominalization seems to be conserved in the syntax, the standard
CG distinction between the nominalization of finite and non-finite clauses does
not seem to hold, at least as long as the finiteness (or, more accurately, the
groundedness) of a clause depends on the inflectional marking of tense.
2.3 Compositional pattern III: Sentential nominalizations
The third compositional pattern, found in Bella Coola but not Lushootseed,
nominalizes a verb and all of its actants, creating a sentential nominalization. In
English and many other languages (e.g. Japanese and Korean — Horie, this
volume), sentential “nominals” are not realized morphologically, but are
created through the use of complementizers ([That he walks in his sleep] drives
her crazy; I see [that you have found your own way home]), which allow a finite
clause to serve a nominal role in a sentence. In Bella Coola, however, such
clauses are realized by the s-prefix, as in (15):
(15) a. pÑ wi ti+s+pu†Ñ+aylayx+aw
halibut deic+nom+(to)fish+lc+3pl
‘what they caught [is] a halibut.’
b. wic ‘ac wa+s+‘aSps+tu+m qwax̌w
idn deic deic+nom+eat+caus+3sg-pass raven
‘what Raven was fed [was] this’
(Nater 1984: 102)
c. ‘aSnap+is+kw+cÑ ta+qiiqtii+tx̌
know+3s-3s+quot+perf deic+baby+deic
wa+s+kwacta+tu+m x+ti+man+S
deic+nom+call+caus+3sg-pass prep+deic+father+1pl-poss
‘the baby knew what he had been named by our father.’
(Davis and Saunders 1980: 108, line 121)
In sentence (a), the nominalized clause is based on the intransitive verb
pu†Ñaylayx ‘to fish’, which takes the endpoint of the action (the fish) as an
oblique object marked by the prepositional clitic x- (cf. (15a) and pu†Ñaylayxaw
xpÑ wi ‘they caught a halibut’). Because the nominalization represents an oblique
actant of the non-nominalized clause, (a) represents an application of Sub-
schema I. Likewise in (b), an emphatic copular construction, the subject of the

sentence is an s-clause representing the food that was fed to Raven, the oblique
object of the passivized verb ‘aSpstum ‘s/he was fed’ (lit. ‘s/he was caused to eat
[of]’); in (c), the direct object of ‘aSnap ‘know’ is an s-nominal representing the
name the baby is called (‘baby’ being the subject of the passivized kwactatum
‘s/he was called’). Like lexical nouns, these sentential nominalizations are
marked with deictic circumclitics indicating number, gender, demonstrativity,
and relative spatial location (cf. taqiiqtiitx̌ ‘non-fem.sg.non-dem.dist baby’ in
(15c)) and serve as actants of the matrix verb.
Sentential nominalizations formed with the schematic protoype of s- often
serve as a more familiar type of clausal complement, as in (16):
(16) a. ‘anayk+c s+ka+†Ñap+c
want+1sg nom+irr+go+1sg
‘I want to go.’
(Nater 1984: 104)
b. ‘aSnap+it s+‘inus ti+ka+†Ñap
know+3sg-3pl nom+2sg-emph deic+irr+go
‘they know that it is you who will go.’
(lit. ‘they know that the going one [is] you’)
(Nater 1984: 103)
Bella Coola also uses the s-schematic prototype to form a wide variety of
adverbials such as purposives and circumstantials, as in:
(17) a. ‘ustxw+aw ‘ula+suS+aw s+kS+s ti+snx̌+tx
go-in+3pl prep+house+3poss nom+set+3sg deic+sun+deic
‘they go into their houses when the sun sets.’
(Nater 1984: 102)
Ñ
b. taws+ulmx+cÑ s+‘aSwlaa+SS+Ø
damp+ground+perf nom+rain+past+3sg
‘the ground is damp, it having rained.’
(Nater 1984: 105)
In other cases, the s-nominalizer is used simply to subordinate the nominalized
clause, as in wh-questions, clefts, and other predicate-complements:
(18) a. stam+naw+ks ‘ac s+smSk+aw
what+3pl+int these nom+fish+3pl
‘what kind of fish are these?.’
(lit. ‘what are these, they being fish?’)
(Nater 1984: 116)
136 David Beck
b. ‘ala‘aykÑ +Ø+kw s+qwlx̌uS+cut+aw

long-ago+3sg+quot nom+gather+refl+3pl
wa+xnas+uks+c
deic+woman+pl+deic
‘long ago the women would get together.’
(lit. ‘[it was] long ago that the women gathered’)
(adapted from Nater 1984: 103)
c. ya+Ø ta+‘imlk+tx̌ ti+s+‘aSukwala+Ø
good+3sg deic+man+deic deic+nom+Indian-doctor+3sg
‘the man was a good Indian doctor.’
(lit. ‘the man was good [at] being an Indian doctor.’)12
(Davis and Saunders 1984: 223)
Uses such as these represent particularly clear cases of the morphological process
of nominalization performing the equivalent of syntactic complementation.
As with Lushootseed participles, Bella Coola sentential nominalizations are
marked for the full range of aspectual and voice distinctions found in finite
clauses (cf. the passive in (15b, c), the irrealis in (16), and the past in (17b)).
However, unlike Lushootseed participles, sentential nominalizations do not
mark the trajector of the nominalized clause as a possessor, but instead realize
it as an ordinary subject NP, the predicate taking regular matrix-clause agree-
ment suffixes. Compare the matrix clause in (19a) with its nominalized counter-
part in (b) (a relative clause — cf. the Lushootseed example in (7c) above).13
(19) a. nap+is ti+†Ñmsta+tx ti+staltmx+tx
give+3sg-3sg deic+person+deic deic+chief+deic
x+ti+qÑ lsxw+tx
prep+deic+rope+deic
‘the person gives the chief the rope.’
b. ti+qÑ lsxw+tx ti+s+nap+is ti+†Ñmsta+tx
deic+rope+deic deic+nom+give+3sg-3sg deic+person+deic
ti+staltmx+tx
deic+chief+deic
‘the rope that the person gives the chief.’14
The verb in both of these sentences is marked with object-subject agreement

suffixes from the plain active transitive paradigm, the affixation of s- not
triggering any difference in the marking of syntactic subject in the matrix and
nominalized subordinate clauses. Because of this, sentential nominalizations in

Bella Coola are not obviously nominal, except in those cases like (15) where
they function as actants of a predicate and profile event-participants. Indeed, as
Kroeber (1991) points out, in Bella Coola “the term ‘nominalized’ loses much
of its meaning, since ‘nominalized’ clauses in fact display no distinctively
nominal properties: one might almost as well simply label s- a subordinative
prefix …” (p. 59). Nevertheless, it is striking (as Kroeber himself goes on to
point out) that it is the same morpheme (s-) which is used in lexical nominali-
zation, in oblique-centred relative clauses, and in those cases where a sentence
is required to fill a syntactic role usually filled by a noun or a lexical word; cross-
linguistically, s- serves a nominalizing function in all Salishan languages and, as
we have seen, in Lushootseed it is used to form participles. Where Bella Coola
differs most — and where Bella Coola sentential nominalizations seem most to
resemble the more familiar syntactic complementization found in languages
like English — is that the scope of the s-prefix in this language, when applied to
clauses, is sentential. In other words, s- includes the sentence predicate, its
object(s), and its subject within the scope of nominalization and so, like
that-clauses in English, sentential nominalizations bear no mark of subordina-
tion other than the presence of the nominalizer itself.
So far, of the three meanings of the s-nominalizer given in Figure 2 above,
we have seen only two — the s-prototype, which profiles an event as in (17),
and Subschema I, which profiles a oblique actant, as in (15). The third meaning
comes into play when s-nominalizations in Bella Coola are used as temporal
adverbials. This subschema profiles the temporal extension of the event rather
than the event itself, as in (20):
(20) a. ‘as+‘uSqn+ak+s ‘uS ti+qla+tÑ ayx
use+pail+hand+3sg prep deic+water+deic
wa+s+‘as+qupÑ tn+ak+a+kw
deic+nom+use+drum+hand+3pl+quot
‘she took her pail to the water, the drums sounding.’
(Nater 1984: 103)
b. †Ñiliwa+s s+‘mt+s
quick+3sg nom+get-up+3sg
‘he [was] quick getting up.’ (Nater 1984: 104)
In these examples, the relationship of the nominalized clause to the matrix

clause is that of an adverbial expression — specifically, an expression of
temporal extension: what is profiled by the s-nominalization here is the time
138 David Beck
over which the event in the embedded clause took place. This profiled slice of
time modifies the matrix clause, whose own temporal extension it serves to
define. Langacker (1991) offers a similar analysis of English expressions such as
(21) Working in the garden, I saw her go by.
Here the participial phrase serves to specify the time at which ‘I saw her go by’
by identifying this time as being the same as (or some part of) the time during
which ‘I’ was ‘working in the garden’. The major difference between the English
pattern and the Bella Coola adverbial s-clause is that English and a number of
other Indo-European languages seem to prefer (or, as in Russian, require) that
the trajector of the subordinate clause be coreferential with the trajector of the
main clause, whereas Bella Coola has no such restriction.
What is particularly interesting about Langacker’s analysis of the participial
phrase is the link it makes between the process of syntactic subordination and
the -ing morpheme which serves both to form attributive participles and more
overtly nominal gerunds. Subordination of one clause to another (i.e. comple-
mentation) means that, in effect, the profile of the subordinate clause is
subsumed in some way by that of the matrix clause. The profile of the sentence
in (21), for example, is that of the event of seeing, while the act of working,
realized in non-finite form as a participle by the application of -ing, is reduced
to additional information about the matrix clause. Clearly, the s-prefix in Bella
Coola and Lushootseed has precisely this function as well: by nominalizing
clauses and thereby transforming events into abstract nouns that can serve as
actants or modifiers, s- in effect allows for the nominalized events to be sub-
sumed into the profile of the matrix clause.
Thus, Kroeber’s observation cited earlier that Bella Coola s- often functions
as a complementizer can not only be extended to cover all cases of sentential
nominalization in Bella Coola, but can be applied to Lushootseed as well, since
— according to our analysis here — nominalization of a clause is, in fact, a sign
of subordination. Of course, whether all cases of s-complementation in Bella
Coola are nominalizations is another matter. It is certainly not out of line to
suppose, as suggested by Halliday (1994), that adverbial and other types of
adjuncts to a predicate might indeed be nominals — nouns can play this role in
many languages (e.g. He knocked three times on the door) and many Bella Coola
lexical adverbs are derived from nouns. Similarly, s-clauses that function as
adverbials in one sentence may well surface in a more noun-like role in another.
(22) a. ta+s+‘ali+nu
deic+nom+be-located+2sg
‘while you were here’
b. ta+s+ali+nu+tx̌
deic+nom+be-located+2sg+deic
‘the time you were here’15
On the other hand, it is not clear that Subschema II, which profiles a temporal
extension, meets the conceptual criteria for nounhood, although it certainly
could be argued that was‘asqupÑ tnakakw ‘as/while the drums sounded’ consti-
tutes a reification of a time in the same way that an English word like day (‘the
time over which the sun passes/is passing from horizon to horizon’) does. In
any case, it is also unclear to what extent this is a meaningful question — for the
purposes of our discussion: it is enough to note that the creation of adverbial
complement clauses is yet another function of the nominalizing prefix and that
these complements are in no way structurally distinct from more obviously
nominal constructions with s-.
3. Conclusion: Finiteness, non-finiteness, and complementation
The primary difference between the uses of s- in Bella Coola and s- in Lushoot-
seed has not so much to do with the meaning of the prefix itself as it has to do
with its compositional or syntactic properties. In both languages, s- serves as a
formative prefix creating new lexical items (lexical nominalizations) at the word
level; however, at the clausal level, s- is used in Lushootseed to form participial
nominalizations, whereas in Bella Coola it is used to form sententials. Although
Lushootseed shares two of the three meanings of s- with Bella Coola, it applies
this prefix at the “second” level of composition — that is, after the elaboration
of the direct object and before the elaboration of the subject — to form con-
structions which resemble analogous structures (participles and gerunds) in
other languages and which realize the subjects of the corresponding finite
clauses as possessors. Bella Coola, on the other hand, applies the prefix after the
elaboration of both object and subject, forming sentential nominalizations. The
historical source of this difference is likely twofold: on the one hand, there was
the development in Bella Coola of portmanteau subject/ object transitive
agreement affixes, while on the other there was the convergence of the intransitive
140 David Beck
subject agreement and possessive paradigms, eliminating any morphological

contrast between possessors of participles and subjects of intransitive clauses
(Davis and Saunders 1997). Synchronically, however, the distinction between
the two compositional patterns raises an important question from the point of
view of cognitive-functional approaches to language and linguistic structure.
Given the compositional/ syntactic differences between the two languages, is
there any reason to suppose that the difference in compositional patterns
corresponds to a difference in function or construal of s-nominalizations — in
other words, has grammatical divergence resulted in a divergence in the use or
conceptual structure of complement clauses?
In languages like English that make use of both participial and sentential
nominalization (John’s being drunk vs. That John is drunk), the conceptual
distinction between the two is attributed to the absence/presence of tense and
modality, which are said to be absent from participles but to be included in the
scope of a sentential nominalization (Langacker 1991). In tense-marking
language like English, then, finite clauses can be said to be grounded in space
(that is, identified as a unique event at a specified spatial location relative to the
speech act) by their subject/ trajectors whereas they are grounded in time
(identified as a unique event at a specific temporal location relative to the
speech act) by tense and mood. In Bella Coola and Lushootseed, however,
neither tense nor mood are expressed in the verbal morphology and, more
importantly, to the extent that they are expressed at all, they can be applied
equally to verbs and nouns (Nater 1984; Hess 1993). Thus, in Lushootseed,
participles bear as much marking for tense and mood as do “finite” clauses,
which in Salish seem not to be temporally grounded so much by tense/mood as
they are by the temporal extension of their participants, particularly their
subjects (Davis and Saunders 1975; Demirdache 1996). As a consequence,
sentential nominalizations in Bella Coola must be grounded in both space and
time by their subjects, which implies that Lushootseed participles are grounded
by their possessors, given that the possessors of participles — like the possessors
of ordinary nouns and of Taylor’s (1994) English deverbals — serve to locate
specific instances of events. When their trajectors are specified, then, Lushoot-
seed participles become fully grounded, making them the equivalent of Bella
Coola sentential nominalizations, with the exception that in Lushootseed the
grounding element (the subject) remains outside the scope of nominalization.
Thus, a Lushootseed expression like s6sqÑ wu‘ ‘6 tsi‘iS sSadSad6y‘ ‘the gathering of
the women’ is conceptually indistinguishable from its Bella Coola counterpart
sqwlx̌uScutaw waxnasuksc ‘that the women gather/gathered’ other than that, in
a particular context, the participle might be construed generically as an un-

grounded instance of “gathering” rather than a specific instance thereof.16
Just as many Lushootseed participial nominalizations seem best glossed by
English sentential nominalizations (cf. the glosses for (10) and (11)), Bella
Coola sentential nominalizations seem in some cases to serve as expressions of
ungrounded instances of events, as in some of the examples in (18) and (20)
above, which are most accurately glossed in English as participles. Thus, where
English employs two strategies of complementation (syntactic complementation
and participial nominalization) to distinguish two portions of a semantic field (the
reification of grounded, specific versus ungrounded, generic events), Bella
Coola and Lushootseed each employ different syntactic variants (sentential and
participial nominalization, respectively) of a single strategy (morphological
nominalization) to cover the same field as a whole.17 Given the absence of a
semantic contrast between participial and sentential nominalization in either
one of the two languages, it is hardly surprising that, in functional terms, their
uses of s-nominals are also highly parallel, the primary syntactic function of the
s-prefix in both being that of complementation — the subordination of the
profile of one clause to that of another. In both languages, s- allows complex
verbal expressions to serve subordinate roles as actants, modifiers, and predi-
cate-complements. A minor difference in the uses of s-nominalizations is that
Bella Coola forms temporal adverbials with s- while Lushootseed does not, but
given that Lushootseed forms other types of adverbial from s-clauses, this does
not seem to amount to a major functional distinction between the two languages.
Perhaps a more significant difference is the widespread use of s-nominals in
Bella Coola, illustrated in (18) above, to form wh-questions, clefts, pre- and
post-posed topics, and similar expressions. In these roles, as noted by Kroeber
(1991), s- most closely resembles the that-type of syntactic complementizer
used in English and similar languages, particularly given the fact that the
syntactic scope of the nominalizer is sentential (includes the entire clause and
all of its actants) and the meaning of s- used in these cases is invariably the
schematic prototype (pure nominalization). However, even in those cases
where the use of the s-prefix accomplishes an additional shift in profile from an
event to a nominal semantic argument (i.e. when Subschema I is applied), s-
continues to be reminiscent of syntactic complementizers such as what (cf. the
glosses of (9), ‘what I give [to him]’ and (15c), ‘what Raven was fed’), which
affect a similar shift in profile. The typological contrast that this offers between
languages that use a syntactic means versus languages that use a morphological
means of complementation is an interesting one, and it is of particular note that
142 David Beck
in Salishan languages the preferred strategy for complementizing clauses is

nominalization. What this seems to indicate is that in these languages the
notion of “noun” or “nominal” is in some way linked to the notion of “subordi-
nate” or “syntactic dependent” and that the reifying properties of the s-prefix,
by creating constructions from verbs and verbal expressions that are conceptu-
ally nouns, create “conceptually autonomous” (Langacker 1991) entities which
can then be incorporated into the profile of a larger expression by turning them
into (or making them more like) nouns, which prototypically have this func-
tion. To a certain extent this observation also applies to English, which uses a
morphological means to form nominal gerunds, but it remains to be seen to
what extent this relationship between nominalization and complementation
can be extended to other languages, particularly those that, like Bella Coola and
Lushootseed, make heavy use of morphological processes of complementation.
Abbreviations
1 first person mid middle

2 second person nom nominalizing prefix
3 third person non-dem non-demon-
acc accusative strative
add additive non-fem non-feminine
caus causative part participle
deic deictic pass passive
dist distal past past
emph emphatic perf perfective
fem feminine pl plural
gen genitive poss possessive
hab habitual prep preposition
idn identifier (copula) procl proclivity
int interrogative punc punctual
irr irrealis quot quotative
lc limited control refl reflexive
sg singular
stat stative
<DEST "bec-n*">
Notes
* The author wishes to acknowledge Thom Hess, Tom Hukari, and Leslie Saxon for their
advice and help getting this research off the ground. I also owe thanks to Kaoru Horie and an
anonymous reviewer for helpful comments, and to Gary Palmer for constructive criticism of
this paper leading to its current incarnation. This work has been supported in part by a
Doctoral Fellowship from the Social Science and Humanities Research Council of Canada.
1. The Salishan family encompasses some twenty-three languages spoken in an area
stretching from the Pacific Coast of British Columbia and Washington State eastward to the
Rocky Mountains and onto the plains of Montana. Languages of the family are notable for
their relatively rigid predicate-initial (roughly, VSO) word-order and share a number of
areal-typological features such as sentence-second clitics, elision of third-person participants,
lack of tense and copula, and frequent use of nominal and other non-verbal predicates.
Lushootseed belongs to the numerous Coast Salish group while Bella Coola, an isolate in a
branch of its own, likely represents the oldest offshoot of the family, whose time depth is
estimated at between three and six millenia (Kroeber 1999).
2. In the alternate case, where the runway is the trajector or primary figure, would require
the sentence to be realized in the passive, as in “The runway was approached by the plane”,
where the sentence takes the runway’s viewpoint on the event rather than the plane’s.
3. Verbs in Lushootseed and (for the most part) Bella Coola, as in Salish in general, are
almost invariably formed on such intransitive roots, transitive verbs being morphologically
derived (Beck 1995a, 1996a; Davis to appear).
4. The idea that oblique actants represent relational landmarks which may be profiled by
Subschema I also finds some support in the fact that in Bella Coola, s-nominalizations of
adjective-like words tend to profile their relational landmark, as in ‘n†Ñ ‘dark’ > s‘n†Ñ ‘night’;
note, however, the Lushootseed form Sax̌ ‘dark’ where the nominal is formed by combining
the bare radical with a deictic — ti Sax̌ ‘night’.
5. In the discussion of English grammar, a distinction is traditionally made between
participles and gerunds, the former filling an attributive role in a sentence and the latter
acting as a nominal (Trask 1993). Aside from the facts of their distribution, however, the two
categories seem to be identical and most likely reflect two uses of the same type of lexical
item. In Russian, the term “participle” is used to refer to the attributive usage of this class of
nominalization, coinciding with the English usage, whereas in some discussions of Altaic
languages such as Turkish (e.g. Comrie 1981) “participle” is used for both substantive and
attributive roles; the term “gerund” is more often used in Altaic (Comrie 1981; Poppe 1970),
Spanish (Solé and Solé 1977), and in traditional Russian grammars (e.g. Pulkina 1982) to
refer to what are more accurately described as “deverbal adverbs”; for this reason I have
chosen the term “participle” to use here.
6. In Lushootseed (and Bella Coola) recipients/beneficiaries are the direct objects of acts of
giving.
7. Non-oblique or “direct” actants — that is, subject and direct object — of a clause can be
profiled by a process of syntactic nominalization, wherein a clause without s- or an overt
144 David Beck
complementizer appears associated with a determiner-like deictic element, as in the

Lushootseed sentence
(i) put †Ñu+b6+‘itut ti‘iS ‘6s+dxw+pakw+ah+6b
only hab+add+sleep deic stat+procl+lie+ass+mid
‘this He-Lies-with-his-Ass-in-the-Air would only sleep’ (Hess 1993: 183, line 57)
These constructions, headless relative clauses, are discussed in more detail in Beck (1995b).
8. Ordinary nouns in Lushootseed are also used as attributively, as in
(i) ti‘6‘ kiyuuqws st6tudeq
deic seagull slave
‘these seagull slaves’ (Hess 1993: 117)
The plural meaning is contextual, plural marking of nouns being optional in Lushootseed
and most other Salishan languages.
9. Note the use of the preposition along with s- in (8b) to subordinate the clause ‘učalad ti‘iS
sbiaw ‘[he/she/ it] chased Coyote’, forming an adverbial phrase (cf. English ‘while chasing
Coyote’). This contrasts with Bella Coola’s use of plain s-clauses to form adverbials,
illustrated in (17) and (20) below. Plain s-clauses, or s-clauses introduced by deictics, do
appear in adverbial roles in Lushootseed texts, albeit infrequently. The conditions on their
use await future investigation.
10. English participles, however, depart from the Lushootseed pattern in that they are able
to take non-possessor subjects, which are (more or less) synonymous with possessor-
subjects, e.g. John/ John’s giving Mary the ball. Interestingly, Comrie (1981: 82) reports the
same pattern in Tatar:
(i) min kür+gän+ne bel+de
1sg+acc see+past-part+acc know+past
‘he found out that I had seen’
(lit. ‘he knew me seeing’)
(ii) min+em kür+gän+em+ne bel+de
1sg+gen see+past-part+1poss+acc know+past
‘he found out that I had seen’
(lit. ‘he knew my seeing’)
Here the trajector of a “verbal noun” is expressed either as an actant in the accusative case or
as a possessor in the genitive. In Lushootseed, however, as in many languages, the “subject”
of the participle is obligatorily a possessor.
11. See also example (7) above, which bears both types of aspectual marking.
12. Because nominal predicates with third-person pronominal subjects such as ‘aSukwalas ‘he
[is an] Indian doctor’ are, in effect, one-word sentences consisting of a noun plus (optional)
agreement features, this often leads to the rather bizarre situation where the nominalizing
prefix is applied to what looks like an ordinary noun, cf.
(i) paax̌+ic x+a+s+John+Ø
name+3sg-1sg prep+deic+nom+John+3sg
‘I will name him John’
(lit. ‘I will name him the one who is John’)
(Nater 1984: 101)
13. Just as in Lushootseed, Bella Coola can use simple nouns as modifiers in precisely the
same manner that the sentential nominalization in (19d) is used:
(i) a. kÑx+ic ti+staltmx ti+‘imlk+tx
see+3sg-1sg deic+chief deic+man+deic
‘I see the man [who is] chief ’
b. *kÑx+ic ti+‘imlk ti+staltmx+tx
see+3sg-1sg deic+chief deic+man+deic
Unlike other modifiers (which may either precede or follow their head), noun-modifier
constructions are obligatorily head-final, almost certainly as a means of disambiguation.
Davis and Saunders (1978) offer this as a diagnostic of a noun-verb distinction.
14. The gloss of this construction depends crucially on whether or not the gift, the rope, is
thematic or rhematic — if it is thematic, then the construction is interpreted as an oblique-
centred relative clause, as it is glossed here. On the other hand, if the rope is rhematic, then it
is interpreted as a predicate nominal, giving the construction a gloss of ‘what I gave to the chief
[was] a rope’. Such constructions — and the grammatical constraint that rhematic elements
in a sentence be syntactic predicates — are discussed in more detail in Beck (1997).
15. According to Davis and Saunders (1984), in (22a) “the absence of the deictic suffix Y
indicates a failure to identify the time boundaries, and the result is a span of time” (221).
16. See Demirdache (1996) and Matthewson (1996), who argue against the existence of
inherently generic expressions in St’at’imcets (Lillooet); many of their arguments could be
extended to Lushootseed.
17. See Achard (this volume) for a discussion of complementation-types in French from a
slightly different perspective, and Horie (this volume) for a comparison of the rich English
arsenal of these to two languages — Japanese and Korean — which, like Bella Coola and
Lushootseed, have a more restricted set of options.
References
Achard, Michel. this volume. “Construal and complementation in French: The perspective
dimension”.
Bates, Dawn. Hess, Thomas M. and Hilbert, Vi. 1994. Lushootseed Dictionary. Seattle:
Washington University Press.
Beck, David. 1995a. “A conceptual approach to lexical categories in Bella Coola and
Lushootseed”. In Papers for the 30th International Conference on Salish and Neighbouring
Languages. Victoria: University of Victoria, 1–31.
Beck, David. 1995b. A comparative conceptual grammar of Bella Coola and Lushootseed. MA
thesis, University of Victoria.
Beck, David. 1996a. “Transitivity and causation in Lushootseed morphology”. Canadian
Journal of Linguistics 41(2): 109–140.
Beck, David. 1996b. “Subjecthood, agency, and topicality in Lushootseed”. Toronto Working
Papers in Linguistics 15(1): 1–29.
146 David Beck
Beck, David. 1997. “Theme, rheme, and communicative structure in Lushootseed and Bella
Coola”. In Leo Wanner (ed.), Recent Trends in Meaning-Text Theory. Studies in Lan-
guage Companion Series 39. Amsterdam: Benjamins, 93–135.
Comrie, Bernard. 1981. The Languages of the Soviet Union. Cambridge: Cambridge Uni-
versity Press.
Comrie, Bernard and Thompson, Sandra A. 1985. “Lexical nominalization”. In Timothy
Shopen (ed.), Language Typology and Syntactic Description, Vol. III: Grammatical
Categories and the Lexicon. Cambridge: Cambridge University Press, 349–98.
Davis, Henry. to appear. “Deep unaccusativity and zero syntax in St’at’imcets”. In M. Uribe-
Etxebarria and A. Mendikoetxea (eds), Supplements to the Basque Journal of Linguistics.
Davis, Philip W. and Saunders, Ross. (1975). “Bella Coola nominal deixis”. Language 51(4):
845–58.
Davis, Philip W. and Saunders, Ross. 1978. “Bella Coola syntax”. In Eung-Do Cook and
Johnathan Kaye (eds), Linguistic Studies of Native Canada. Vancouver: University of
British Columbia Press, 37–65.
Davis, Philip W. and Saunders, Ross. 1980. Bella Coola texts. Victoria: British Columbia
Provincial Museum.
Davis, Philip W. and Saunders, Ross. 1984. “Propositional organization: The ‘s-’ and ‘si-’
prefixes in Bella Coola”. International Journal of American Linguistics 50(2): 208–31.
Davis, Philip W. and Saunders, Ross. 1989. “Language and intelligence: The Semantic unity
of -m- in Bella Coola”. Lingua 78: 113–58.
Davis, Philip W. and Saunders, Ross. 1997. A Grammar of Bella Coola. Missoula: University
of Montana Occasional Papers in Linguistics.
Demirdache, Hamida. 1996. “‘The chief of the United States’ sentences in St’at’imcets
(Lillooet Salish): A cross-linguistic asymmetry in the temporal interpretation of noun
phrases and its interpretation”. In Papers for the 31st International Conference on Salish
and Neighbouring Languages. Vancouver: University of British Columbia, 101–118.
Givón, Talmy. 1979. On Understanding Grammar. New York: Academic Press.
Halliday, Michael A. K. 1994. Introduction to Functional Grammar. London: Arnold.
Hess, Thomas M. 1976. Dictionary of Puget Salish. Seattle: University of Washington Press.
Hess, Thomas M. 1993. Lushootseed Reader, Volume I — Four Stories from Edward Sam.
Victoria: Tulalip.
Hess, Thomas M. 1996. “Lushootseed grammar in an annotated text”. In Crisca Bierwert
(ed.), Lushootseed Texts: An Introduction to Puget Salish Narrative Aesthetics. Lincoln:
University of Nebraska Press, 44–54.
Hess, Thomas M. 1998. Lushootseed Reader, Volume II. Four stories from Martha Lamont.
Missoula: University of Montana Occasional Papers in Linguistics.
Horie, Kaoru. this volume. “Complementation in Japanese and Korean: A contrastive and
Jelinek, Eloise. and Demers, Richard. 1994. “Predicates and pronominal arguments in Straits
Salish”. Language 70(4): 697–736.
Keenan, Edward L. 1976. “Towards a universal definition of ‘subject’”. In Charles N. Li (ed.),
Subject and topic. New York: Academic Press, 303–333.
Kinkade, M. Dale. 1983. “Salishan evidence against the universality of ‘noun’ and ‘verb’”.
Lingua 60: 25–40.
</TARGET "bec">
Kroeber, Paul. 1991. Comparative syntax of subordination in Salish. Ph.D. dissertation,

University of Chicago.
Kroeber, Paul. 1999. The Salish Language Family: Reconstructing Syntax. Lincoln:
University of Nebraska Press.
Kuipers, Aert. 1968. “The categories of verb-noun and transitive-intransitive in English and
Squamish”. Lingua 21, 610–26.
Langacker, Ronald W. 1987. Foundations of Cognitive Grammar I: Theoretical Prerequisites.
Stanford: Stanford University Press.
Langacker, Ronald W. 1991. Foundations of Cognitive Grammar II: Descriptive Application.
Stanford: Stanford University Press.
Li, Charles N. and Thompson, Sandra A. 1979. “Third-person pronouns and zero-anaphora
in Chinese discourse”. In Talmy Givón (ed.), Syntax and Semantics, Volume 12:
Discourse and Syntax. New York: Academic Press, 311–35.
Matthewson, Lisa. 1996. Determiner systems and quantificational strategies: Evidence from
Salish. Ph.D. dissertation, University of British Columbia.
Nater, Hank F. 1984. The Bella Coola language. Ottawa: National Museum of Man.
Poppe, Nicholas. 1970. Mongolian Language Handbook. Washington: Center for Applied
Linguistics.
Pulkina, I. M. 1982. Breve prontuario de la gramática rusa. Moscow: Russkij Jazyk.
Solé, Yolanda R. and Solé, Carlos A. 1977. Modern Spanish Syntax: A Study in Contrast.
Lexington: D. C. Heath.
Taylor, John. 1994. “‘Subjective’ and ‘objective’ readings of possessor nominals”. Cognitive
Linguistics 5(3): 201–242.
Trask, R. L. 1993. A Dictionary of Grammatical Terms in Linguistics. London: Routledge.
van Eijk, Jan P. and Hess, Thomas M. 1986. “Noun and verb in Salishan”. Lingua 69:
319–331.
<LINK "del-n*">
<TARGET "del" DOCINFO
AUTHOR "Nicole Delbecque"
TITLE "Cognitive constraints on complement clause cliticization in Spanish"
KEYWORDS ""
SIZE HEIGHT "220"
WIDTH "150"
VOFFSET "4">
Chapter 7
Cognitive constraints on
complement clause cliticization in Spanish*
Nicole Delbecque
University of Louvain (K.U. Leuven)
1. Introduction
This paper is concerned with Spanish verbs of saying, perception, and proposi-
tional attitude that may take a subordinate clause introduced by the conjunc-
tion que ‘that’ as complement. The purpose is to analyze the relationship
between such a direct object clause and the clitic object pronoun with which it
may alternate. To refer to this phenomenon, I will use the term complement
clause cliticization; however, no generative perspective will be implied.
I will attempt to account for the following facts. As illustrated in (1), the
neuter clitic lo ‘it’ may alternate with the complement clause if it is topicalized.1
(1) a. Nadie {cree /comprende} que haya hecho esto
nobody {believes/understands} that have-3sg.pres.subj done that
‘Nobody {believes/understands} that he has done that.’
b. Nadie lo {cree /comprende}
nobody it-neut.cl {believes/understands}
‘Nobody {believes/understands} it’
c. Que haya hecho eso, nadie lo
that have-3sg.subj done that, nobody it-neut.cl
{cree /comprende}
{believes/understands}
‘That he has done that, nobody {believes/understands}it.’
As (2) and (3) illustrate, these verbs may have a noun phrase as their object,
which may also alternate with a clitic construction, provided the clitic desig-
nates a known or predictable argument.
150 Nicole Delbecque
(2) a. Nadie {cree /entiende} esta-fem historia

nobody {believes/understands} this-fem history
‘Nobody {believes/understands} this history.’
b. Nadie la {cree /entiende}
nobody her-fem.cl {believes/understands}
‘Nobody {believes/understands} it’
(3) a. Juan {rechazó /descartó} la oferta
Juan {rejected/dismissed} the-fem offer
‘Juan {rejected/dismissed} the offer.’
b. Juan la {rechazó /descartó}
Juan her-fem.cl {rejected/dismissed}
‘Juan {rejected/dismissed} it.’
However, as (4) illustrates, not all clausal complement constructions may

alternate with clitic constructions.2
(4) a. Juan {rechazó /descartó} que la aceptara su mujer
Juan {rejected/dismissed} that her-fem.cl would-accept his wife
‘Juan {rejected/dismissed} that his wife would accept it.’
b. *Juan lo {rechazó /descartó} (as counterpart of (4a))
Juan it-neut.cl {rejected/dismissed}
‘Juan {rejected/dismissed} it.’
The question is what distinguishes constructions exemplified in (1a) from those

of (4a). What makes (1a) compatible with the clitic construction and (4a), not?
I will argue that this different behavior, which to my knowledge has hitherto
passed unnoticed, can be explained in cognitive terms, more precisely by taking
‘construal’ into consideration.
Not only verbs with a que ‘that’-clause may take the clitic construction, but
also verbs that may take clauses with another subordinate indirect discourse
structure, namely those introduced by for example si or por qué. As (5a)
illustrates, the verb rechazar, which cannot take a si or por qué complement
clause, is not compatible with a clitic construction either. Other verbs that
behave in this respect like rechazar ‘reject’ (cf. (3)–(4)), include admitir in the
sense of ‘allow’, aprobar ‘approve’, conceder ‘concede’, condenar ‘condemn’,
(de)negar ‘deny’, descartar in the sense of ‘reject’, desmentir ‘deny/ refute’,
perdonar ‘forgive’, ratificar ‘ratify’, reconocer ‘recognize’, reprobar ‘disapprove’.3
Conversely, as (5b) illustrates, when the predicate is compatible with such
an indirect discourse structure, nothing seems to hamper the reduced expres-
sion of the content of the subordinate clause by means of the neuter clitic.
Cognitive constraints on complement clause cliticization in Spanish 151
(5) a. *Juan rechazó {si /por qué} ha aceptado su mujer

Juan rejected {whether/why} has accepted his wife
‘Juan rejected {whether/why} his wife has accepted.’
b. No sé {si /por qué} han dicho eso. Realmente,
not know-1sg {whether/why} have-3pl said that really
no lo sé
no it-cl know-1sg
‘I don’t know {whether/why} they have said that. I really don’t know
it.’
I will try to analyze what this correlation tells us about the function of the
neuter clitic as grounding device.4
A last observation is in order here to show what is at stake. Many predicates
display a hybrid behavior when it comes to licensing the clitic construction. For
example, compare (6a) and (6b).
(6) a. Le dijeron que tuve un accidente. Se lo
to-him / her said-3pl that had-1sg an accident to-him / her it
dijeron rápidamente
said-3pl quickly
‘They told him/her that I had an accident. They told it to him/her
quickly.’
b. Y luego dicen que el pescado es caro. *Lo dicen
and then say-3pl that the fish is expensive. *It say-3pl
rápidamente
quickly
‘And then they say that the fish is expensive. *They say it quickly.’
Although, at first sight, (6a) and (6b) both present canonical expressions of
propositional arguments with the verb decir ‘say’, only (6a) allows the clitic
construction. In my view, this has to do with a difference in construal. In (6a),
the propositional argument (tuve un accidente ‘I had an accident’) is directly
grounded in the speaker’s actual mental space, whereas the proposition
embedded in (6b) (el pescado es caro ‘the fish is expensive’) corresponds to a
saying that the speaker is unable or unwilling to assert. Whereas in (6a) the
speaker “steps inside the subject entity’s mind”, so to say, and thus endorses the
role of conceptualizer, he refrains from doing so in (6b). This di erence is reflected
in the di ering acceptability of the clitic lo ‘it’ in (6a) and (6b) respectively.
On the basis of the facts exposed thus far, the clitic lo ‘it’ appears to func-
tion as quite a specific grounding device in relation to the que ‘that’-clause.
Therefore, it seems to me that whether a predicate that takes a complement

clause can also take a clitic fulfilling the same role is a question that has to be
dealt with in terms of mental space conflation. My hypothesis is that, given the
deictic nature of the clitic,5 the constraints on cliticization crucially hinge on the
role assumed by the speaker, in combination with that of the subject entity of the
main clause. Basically, I hold that cliticization reveals the relation between these
two conceptualizers, not about the scene evoked but rather how people “see” it.
Predicates of saying, for instance (6), introduce a speech event, or ‘put on
stage’ a grounded predication. Yet, when its content is integrated into a
complex sentence structure, this description begs the question how much of the
speech event is left, and how much of it is omitted. In principle, the syntactic
object function warrants the assignment of a nominal status to the complement
clause, and makes it behave as a clausal participant, susceptible of becoming a
candidate for topicalization and, hence, for cliticization. However, examples
such as (6b) suggest that this is only part of the story and that cliticization also
implies a conceptual shift that incorporates the representation of the proposi-
tional content into the ground of the speaker, i.e., his actual Mental Space.
On the other hand, examples such as (4b) suggest that to adequately analyze
this conceptual shift, it is not sufficient to take into account only the type of
relationship holding between the main clause and the subordinate clause, but
also the range of possible object types. Insofar as the extension of the object
paradigm gives a measure of the flexibility of the predicate and of the diversity
of choices available to the speaker, it may have a positive impact on the accept-
ability of object cliticization (cf. (5)). This idea can be put to test not only by
comparing different object types and their behavior with respect to cliticization
(Sections 2 and 3), but also by distinguishing various semantic classes of
predicates to see whether there is a relation between the semantic structure of
the verb and the relative acceptability of complement clause cliticization
(Sections 4, 5 and 6).
Last but not least, the internal complexity of the subordinate clause also has
a role to play. If the hypothesis is correct, i.e. if cliticization indeed is the overt
manifestation of mental space conflation in the sense indicated above, the
presence in the subordinate clause of elements that manifestly belong to a
mental space alien to that of the speaker will preclude cliticization, since they
will form an obstacle for the integration of the propositional content into the
speaker’s actual mental space (e.g. ay Dios mío ‘oh my God’ in (7)).
(7) a. Mi mamá decía que ay Dios mío su hija perdió la cabeza

my mother said that oh God mine her daughter lost the head
‘My mother said that oh my God her daughter lost her head.’
b. *Lo decía siempre que me ocurría algo raro
it said-3sg always that-conj to-me occurred something strange
‘She said it every time that something strange occurred to me.’
(starred as counterpart of (7a))
Summarizing, the analysis has to simultaneously examine the relative import of

the process type instantiated by the matrix predicate and the internal structure
of the complement clause in order to relate it to the position taken on the scene
by the speaker and by the overt conceptualizer of it, viz. the main verb’s subject
(cf. Achard, this volume).
The paper is organized as follows. The next two sections contain what I
consider to be necessary background information on the paradigmatic classes
of the neuter clitic lo ‘it’ and the conjunction que ‘that’. The three subsequent
sections contain the verification of the working hypothesis presented above.
In Section 2, I summarize what should be borne in mind about the behavior
of the nominal objects with respect to cliticization when tackling the phenome-
non of complement clause cliticization. Section 3 situates the conjunction que
‘that’ vis-à-vis other subordinating operators that also introduce an object
argument but, unlike que ‘that’, stand in a one-to-one relationship with
cliticization. Once the clitic construction and the complement clause construc-
tion are delineated according to this paradigmatic axis, full attention can be
paid to the function of the semantic class of the matrix predicate and of the
relative homogeneity or heterogeneity of the viewing arrangements within the
subordinate structure. In Section 4, the behavior of causative matrix predicates,
which score high on the binding scale, is analyzed in comparison with other
predicates, such as that of verbs of volition and emotive reaction. Section 5 is
devoted to the much larger class of viewing processes and mental attitude
predicates in general. And Section 6 further explores the impact of converging
and diverging perspectives that characterize declarative processes, where the
focus can bear alternately on the transfer or on the expressive dimension.
2. Background information on clitics
The clitic is an unstressed pronoun that occurs only in bound construction with
the verb and is marked for gender, number and case of the complement it
instantiates. The masculine and feminine clitics represent the reduced verb-
adjacent expression of an unfocussed, referentially specific, predictable and
familiar direct or indirect object nominal. Cliticization is thus a form of pro-
nominalization whereby the corresponding nominal loses its semantic specifici-
ty, while general discourse-grammatical information such as case, gender,
number and deixis are retained and incorporated in the verb, next to inflection
marking. This phenomenon is widespread among the Romance languages.
When the clitic stands in complementary distribution with an overt
nonclitic argument, it is used anaphorically and marks person/number agree-
ment with objects. The Accusative clitic (lo ‘him/it’/la ‘her’) is obligatory (8b),
that of the Dative clitic is also automatic with monotransitive verbs (9), but
variable with ditransitive ones (8c–d).
(8) a. He dado el libro a mi hermano
have-1sg.pres.ind given the book to my brother
‘I have given the book to my brother.’
b. {Lo/*Ø} he dado a mi hermano
{it /*Ø} have-1sg.pres.ind given to my brother
‘I have given it to my brother.’
c. {Se /Ø} lo he dado
{him-dat/Ø} it have-1sg.pres.ind given
‘I have given it to him.’
d. {Le /Ø} he dado el libro
{him-dat/Ø} have-1sg.pres.ind given the book
‘I have given him the book.’
(9) Eva es azafata. {Le /*Ø} gusta viajar
Eva is air hostess {she-dat/*Ø} please travel
‘Eva is air hostess. She likes traveling.’
The clitic also may co-occur with a non-clitic. This phenomenon, known as
“clitic doubling”, is, informationally speaking, pleonastic, and obeys specific co-
occurrence conditions. It normally echoes the definite nominal when the latter
precedes the verb (10a–b), but it can also occur with a postverbal indirect
complement (10c). Verbs that take only a dative require the presence of the
duplicating clitic (11a), unless the subject precedes the verb (11b).
(10) a. La culpa {la /*Ø} tiene el diablo
the fault {her-acc/*Ø} has the devil
‘It is the fault of the devil.’
b. A la novia {le /*Ø} hacen los regalos

to the bride {her-dat/*Ø} make-3pl.pres.ind the gifts
‘For the bride are the gifts.’
c. Los invitados {se /Ø } los hacen a la novia
the guests {her-dat/Ø} them make-3pl. pres.ind to the bride
‘The guests offer them to the bride.’
(11) a. A mis colegas no {les /*Ø} conviene {esta solución/Ø}
to my colleagues not {them-dat/*Ø} suits {this solution /Ø}
‘To my colleagues {this solution/it} is inconvenient’
b. Esta solución no {les /Ø} conviene a mis colegas
this solution not {them-dat/Ø} suits to my colleagues
‘This solution does not suit my colleagues.’
Among the factors that favor its variable use (10c, 11b), figure the semantic
category of the referent, viz. [+human] and [+definite] and the linear position
of the argument, viz. [preverbal]. Notice that clitic doubling is obligatory with
pronominal objects (12a), but remains optional with postverbal nominal ones
(12b). Given the grounding function of the personal pronoun (ella ‘she’ in
(12a)), this can be seen as evidence for the fact that the clitic not only enhances
the participant status but also the grounded character of the argument.
(12) a. Los vecinos {la /*Ø} llamaron a ella
the neighbors {her/*Ø} called to she
‘The neighbors called her.’
b. {La /Ø} escucharon a la mujer
{her/Ø} listened-3pl to the woman
‘They listened to the woman.’
As illustrated in (1b–c), the use of the neuter clitic — which is formally syn-
cretic with the masculine singular accusative — seems to fulfill the same
anaphoric and duplicating functions when the direct object is clausal instead of
nominal. By virtue of the iconicity principle (Haiman 1985; Givón 1985), the
fact that the complement clause is mapped on to the same clitic paradigm as
nominal objects is a signal that its conceptual domain is being thought of in the
same way. The structures (1b) and (1c) resemble those of (8) and (10) in that
they also reflect the mutual relevance and dependency of object and verb. The
conversion from the clausal to the clitic form implies the loss of essentially
verbal categories to the advantage of discourse-grammatical information, which
is cast in nominal categories (case, gender, number, deixis).
However, the relation between the clausal paradigm and the pronominal
one is not so straightforward as the one holding between the nominal paradigm
and the pronominal one. The pronominal paradigm is, indeed, directly propor-
tional to the nominal one, since it presents the same categorial distinctions, viz.
gender, number and case. A greater conceptual distance separates the clausal
complement from the clitic counterpart, since substituting the latter for the
former implies converting a prototypically time-unstable relational conceptual-
ization into a noun-like, i.e. a prototypically time-stable non-relational entity
(Givón 1979: 320ff.).6
The ability to undergo such a categorial shift crucially hinges upon the
presence of a subordinating conjunction that endows the clause with the ability
to take on nominal functions (cf. Section 3): the kind of “nominalization”
involved does not imply a recategorization or “transcategorization” but, rather,
a “decategorization” (Moignet 1974: 190ff.), since the modal and aspectual
dimensions of the clause remain intact.7 Yet, for cliticization to apply, recate-
gorization must come in, since the pronoun necessarily yields a global, unitary
view of the complement. Therefore, the relation of proportionality between the
content of the complement clause and the clitic is an inferential one, which
requires specific motivation.8
Moreover, of the different factors that favor clitic expression for nominals,
viz. [+human], [+definite], [preverbal], only linear ordering is definitely
involved as a possible motivating factor for complement clause cliticization.
The question is whether the notion of definiteness also applies and, if so, how
it should be stated. Assuming clitizication is the outcome of a metaphorical
process, the question is, how powerful must the metaphor be in order for
cliticization to be possible?
The Cognitive Grammar framework avoids the pitfalls of ontologization by
accounting for the assignment of a nominal profile to the complement clause in
the following way. The subordinate finite clause profiles a grounded instance of
a process type. Yet, conceptual reification applies to it in its entirety, so that the
result is a complex noun that can stand alone as a nominal and function as the
object of another clause. Within the superordinate relationship expressed by the
matrix predicate, the subordinate process is thus construed holistically and
manipulated as if it were a unitary, monolithic entity, conceptually subordinat-
ed and viewed at some distance (rather than in close-up) (cf. Langacker
1991: 148–149 and 446–450, and Boone 1994 and 1998). This favors its reifica-
tion as an abstract region, or, at least, the summary scanning of the component
states. In the words of Langacker (1991: 35):
the effect of reification (marked by that) is to ‘step back’ from the situation —
including both the event and its relation to the ground — and construe it as an
abstract object or proposition capable of being manipulated, evaluated and
commented on. Instead of being asserted, this proposition is taken as one
participant in a higher-order relationship (e.g. a relationship of belief, denial,
evaluation, etc.), whence its role as a clausal subject or object.
3. Background information on the subordinating conjunctions
The function of the subordinating conjunction que ‘that’, henceforth called

“subordinator”, can be better understood by comparing it with the other
subordinating operators, viz. the wh-forms, on the one hand, and the typically
Spanish el (hecho de) que ‘the (fact of) that’, on the other.
As we have just seen, the content of the complement clause can be under-
stood in terms of an abstract object and treated as a true participant in the event
structure of the main clause. As such, it becomes cliticizable, under conditions
similar to those observed for nominal participants.
This prediction also holds true for the complement clause introduced by a
wh-operator: both the clitic pronoun (13b) and the demonstrative pronoun
(13c) can substitute for any of the indirect discourse structures (13a). This
seems to support the idea that if a proposition is not asserted, it may be con-
strued as a participant in a higher-order relationship.
(13) a. Juan {no/ø} sabe {si /por qué/cuándo} vienen
Juan {not} knows {whether/why /when} come-3pl
‘Juan {doesn’t know/knows} {whether/why/when } they come.’
b. {No lo/Lo} sabe
{not it /it} knows
‘He {doesn’t know/knows} it.’
c. {No/ø} sabe eso
{not} knows that
‘He {doesn’t know/knows} that.’
However, the conditions of “definiteness”, “familiarity”, “specificity”, are not met
here. On the contrary, the wh-operator marks an information gap, signalling
that (part of) the proper ground of the process type instantiated by the comple-
ment clause may not be recoverable at all. Therefore, the definiteness condition
must operate at the level of the speaker’s personal deixis, i.e., beyond the
supposed (in)ability of the matrix predicate’s subject referent to fill the infor-
mation gap.
The indirect discourse structure manifests the speaker’s ability to relate the
different meaning components of the subordinate propositional content to his
personal spatiotemporal setting: whether (momentarily) suspending the global
(si ‘whether’) or partial (por qué ‘why’, cuándo ‘when’) projection of that
content, the wh-operator signals the availability of a corresponding slot at the
moment of speech. The speaker has a specific category in mind and presupposes
an instantiation for it. Whether he is able to identify the instance is irrelevant.
Therefore, cliticization is always possible.
Unlike the wh-operators, the que ‘that’-subordinator does not presuppose
a question but an assertion.9 Since the event structure is presented as identifi-
able by the speaker, this construction brings us closer to what habitually counts
as “definite”. Yet, que ‘that’ seems to occupy an intermediary position between
the wh-operators and el (hecho de) que ‘the (fact of) that’ used as subordinating
operator.
In the latter, the presence of the definite article el ‘the-masc’ makes the
complement clause isomorphic to the definite nominal argument, and the
addition of the interpretive nominal hecho ‘fact’ warrants the inference that the
intended reading is an “evidential” or “factual” one, i.e. not simply the repre-
sentation of the statement of an event or of a propositional content (14).
Epistemically, the added value of “credibility” makes this content unquestion-
able; in other words, communicatively “definite”, “familiar” and “specific” are
conceived of as taking precedence over the actual speech event. This is, without
exception, a sufficient motivation for cliticization (14b).
(14) a. Lamenta {que /el (hecho de) que} hayan tardado
regret-3sg {that/the (fact of) that} have-3pl.pres.subj delayed
en avisarles
in notify-them
‘He regrets that they delayed notifying him/her.’
b. Lo lamenta
it regrets-3sg
‘He regrets it.’
Since the el (hecho de) que ‘the (fact of) that’-operator yields a presuppositional
reading, if not a perfective one, it is incompatible with predicates of causation
(15a), volition (15b), perception (15c), propositional attitude (15d), and verbal
expression (15e). Yet, except in the case of causatives (15a) (cf. Section 4), cliticiza-
tion is available to echo an immediately preceding clausal complement (15b–e).
(15) a. Esto hizo {que /*el que} perdieran tanto tiempo.

this made {that/*the that} lost-3pl so-much time
*Lo hizo.
*It made-3sg
‘This caused them to lose so much time. *It made it.’
b. Queremos {que /*el que} todo sea claro. Lo
want-1pl {that/*the that} everything be-3sg.pres.subj clear it
queremos.
want-1pl
‘We want everything to be clear. We want it.’
c. Vi {que /*el que} había un coche parado en la acera.
saw-1sg {that/*the that} there-was a car parked on the pathway
Lo vi.
it saw-1sg
‘I saw that there was a car parked on the pathway. I saw it.’
d. Duda {que /*el que} le hayan dicho la verdad.
doubts-3sg {that/*the that} to-him / her have said the truth.
Lo duda.
it doubts-3sg
‘He doubts that they told him/her the truth. He doubts it.’
e. Dice {que /*el que} todos están conformes. Lo dice.
says-3sg {that/*the that} all are-3pl conform it says-3sg
‘He says that they all agree. He says it.’
Although analogy can be invoked with the regular cliticization of preverbal

nominals ((10a–b), (11a)), in the present case cliticization turns out to be rather
the exception than the rule. Moreover, less common, less central or less
representative predicates appear to be incompatible [*] or hardly compatible [?]
with the clitic: compare procurar ‘intend’ (16b), vislumbrar ‘perceive’ (16c),
considerar ‘consider’ (16d), exclamar ‘exclaim’ (16e), with the far more proto-
typical querer ‘want’ (15b), ver ‘see’ (15b), dudar ‘doubt’ (15c), decir ‘say’ (15d).
Conversely, using an implicative predicate rather than a strictly causative one
renders cliticization acceptable, e.g. (16a) vs. (15a).
(16) a. Esto implica {que /*el que} perderán mucho tiempo.
This implies {that/*the that} lose-3pl-fut much time.
Lo implica.
it implies-3sg.
‘This implies that they will lose much time. It implies it.’
b. Procuramos {que /*el que} todo sea claro.

intend-1pl {that/*the that} everything be-3sg.pres.subj clear
?Lo procuramos.
?*it intend-1pl
‘We intend everything to be clear. We intend it.’

c. Vislumbré {que /*el que} había un coche parado en la
perceived-1sg {that/*the that} there-was a car parked on the
acera. ?Lo vislumbré.
pathway ?it perceived-1sg
‘I perceived that there was a car parked on the pathway. I perceived it.’
d. Considera {que /*el que} le han dicho la verdad.
considers-3sg {that/*the that} to-him / her have said the truth
*Lo considera.
*it considers-3sg
‘He considers that they told him/her the truth. He considers it.’
e. Exclama {que /*el que} todos están conformes.
exclaims-3sg {that/*the that} all are-3pl conform
*Lo exclama.
*it exclaims-3sg
‘He exclaims that they all agree. He exclaims it.’
On the one hand, the que ‘that’ subordinator combines with the widest range of
predicates; on the other hand, it is the only subordinating operator for which
the relationship with cliticization turns out to be problematic.
Unmarked with respect to the use of the other subordinators, it is doubtful
whether que ‘that’ by itself imposes a propositional reading for the complement
clause. If the following complement clause were always conceptualized in an
atemporal, and even in a nominal way, it should always be possible to mark it
as highly specific or “definite” and, hence, to signal its topicalization by using
the clitic. In other words, cliticization should, in principle, be available at least
as a discourse device. Yet, there are quite a few cases in which this correspon-
dence does not hold, e.g. (17), and (4) above. Although, at first sight, the
percept of enjuiciar ‘judge’ seems to be equivalent to that of creer ‘believe’, cf.
(1), it resists being cliticized, even in the most favorable pragmatic conditions.
(17) *Que haya hecho esto, no lo enjuicio
that have-3sg.pres.subj done that, not it-cl judge-1sg.pres.ind
‘That he has done that, I don’t judge it.’
This shows that the subordinator does not always take the clause all the way
along the path leading from a processual to a nominal profile. In other words,
the shift from a relational to a non-relational profile is sometimes hampered.
This observation supports the idea, also put forward in Langacker (1991:
Chapter 10.2), that it is unclear whether the subordinator is an actual nominal-
izer or merely renders the predication atemporal. In this respect, the que-clause
may, indeed, well be indeterminate.
Should we interpret the incompleteness of the analogy with the (definite)
nominal argument in terms of restrictions on the application of the metaphori-
cal process that leads to the conceptualization of events, actions, activities and
states as objects (cf. Lakoff and Johnson 1980: 30ff.)? Or is it more accurate to
assume that this metaphorical mapping can vary in strength, and that cliticiza-
tion reflects its final outcome?
In order to answer these questions, it is worth examining the parameters
along which the content of the que-clause becomes less accessible from the
actual ground and resists losing its processual character, maintaining its own
viewing arrangement, independently from that of the main clause, or — at least
— keeping its grounded predication parallel to that of the main clause.
If it is right to hypothesize that the clitic testifies to the propositional status
of the subordinate clause, then the impossibility of using it becomes a criterion
of the greater conceptual independence of the content of the subordinate
clause. Whether this implies that this content is non-propositional, keeping a
proper spatio-temporal setting, remains to be seen. In the next sections, I will
successively analyze cliticization in relation to volitional and emotive reaction
predicates (Section 4), perception and propositional attitude predicates
(Section 5) and declaratives (Section 6).
4. Cliticization in relation to volition and emotive reaction predicates
This issue can be tackled from different angles. First, attention can be paid to
the semantic structure of the matrix predicate. This approach has led Givón
(1980) to establish a “binding scale” in complementation: the syntax of the
complement codes the degree of semantic integration of two single events into
a single complex event structure. So, the stronger the influence exerted by the
matrix predicate’s agent over that of the complement clause, the more the latter
is structurally integrated into the main clause and will fail to express indepen-
dent tense-aspect-modality.10
Since the clitic represents the ultimate reduction, testifying to the comple-
ment’s incorporation in the matrix predicate, it could be expected to freely
occur with verbs that are high on the binding scale. Yet, we will see that this
prediction does not hold, and that the burden of proof for licensing the
monoclausal next to the biclausal structure, cannot be put exclusively on the
matrix predicate.
The predicate classes situated on the highest and second-highest level of the
binding scale are those of causation, volition and emotive reaction (Givón
1980). The relationship between both clauses is most clearly one of dependency
when the state of affairs or the event described in the subordinate clause cannot
be thought of as holding simultaneously with the action expressed by the matrix
predicate, but as derived from it and, hence, following it. This is best exempli-
fied by means of causatives. Yet, as will be seen below, the causative predicates
turn out to be incompatible with complement clause cliticization.11
An explanation might be sought in the temporal relationship between the
main clause and the subordinate clause. The que ‘that’-clause depending on
causation and volition predicates is necessarily situated in the future perspective
of the matrix predicate (18). This is not necessarily the case with emotive
reaction predicates (19).
(18) a. volition predicates: ansiar ‘to desire’, desear ‘to desire’, querer ‘to want’,
etc.
b. que se {vaya /*haya ido}, todo el mundo lo
that refl {go-subj.3sg/*has-subj gone}, everybody it-cl
desea ardientemente
desires strongly
‘That he {leaves/*has left}, everybody wants it strongly.’
(19) a. emotive reaction predicates: aprehender ‘to apprehend’, deplorar ‘to
deplore’, lamentar ‘to regret’, temer ‘to fear’, etc.
b. que se {haya ido /vaya}, todo el mundo lo
that refl {has-subj gone/go-subj.3sg}, everybody it-cl
lamenta profundamente
regrets profoundly
‘That he {has left/leaves}, everybody regrets it profoundly.’
The latter predicate class also differs from the former two in that it is compati-
ble with the “factive” subordinator (Section 3, (14) vs. (15a–b)). Both the
retrospective view on the content of the subordinate clause and the possibility
of having an overtly “factive” reading imposed on it are reminiscent of the
“familiarity” and “definiteness” conditions encountered with nominal cliticiza-

tion (cf. Section 2) and seem to be fully congruent with the conceptualization
that licenses complement clause cliticization. However, if these were to be seen
as the explanatory factors, complement clause cliticization should be blocked
whenever they do not apply.
This prediction holds for causation predicates, but not for volition predi-
cates, as shown by the cliticization in (18). In other words, neither prospectivity
nor the impossibility of using the factive operator is by itself an obstacle for
cliticization. We, therefore, need to analyze more closely what is at stake with
causation predicates.
Construed with a que-clause, the verbs mentioned under (20) are interpret-
ed as causative verbs, generating a process or event. By default, the instigation
of an event does not coincide in time with the event’s occurrence, but is
anterior to it. This seems to imply that the temporal dimension of the subordi-
nate clause cannot be abstracted away from.
(20) causar ‘to cause’, dejar ‘to let’, determinar ‘to determine’, disponer to
dispose’, eliminar ‘to eliminate’, eludir ‘to elude’, fijar ‘to fix’, hacer ‘to
make’, imposibilitar ‘to make impossible’, intentar ‘to attempt’, ocasionar
‘to actuate’, procurar ‘to attempt’, producir ‘to produce’, resolver ‘to
decide’, vetar ‘to veto’, votar ‘to vote’
Rather than being interpreted as the counterpart of the que ‘that’-clause, e.g.
(21), the clitic lo accompanying the causative predicate, e.g. (22), is conceived
of, instead, as substituting for a masculine nominal (cf. Section 2). The latter can
represent either a first order entity (person, thing) (23a), or a third order entity
(abstract concept, proposition) (23b).
(21) Intenta dormir, le dije, y deja que yo decida.
try-imp sleep-inf, to-her said-1sg, and let-imp that I decide
‘Try to sleep, I said to her, and let me decide.’ [Marsé 1970: 88]
(22) *Déjalo (as counterpart of (21))
let-imp it-cl.neut
‘Let it.’
(23) a. Si este pastel no te gusta, déjalo
if this-masc cake not to-you pleases, let-imp him-cl.masc
‘If you don’t like this cake, don’t touch it.’
b. Es un asunto demasiado complicado. Déjalo
is a-masc matter too complicated. Let-imp him-cl.masc
‘This matter is too complicated. Drop it.’
To situate these alternative conceptualizations of the object argument instan-

tiated by the clitic lo in (22) and (23), with respect to that instantiated by the
que ‘that’-clause in (21), we may turn to the clausal complementation typology
proposed by Hengeveld (1989) and Dik and Hengeveld (1991), summarized in
Table 1. For the time being, we are concerned only with the spatio-temporal
deixis that corresponds to the first three layers of this four-fold model, currently
used in Functional Grammar. We will come back to the fourth one when
discussing the behavior of saying predicates (Section 6.2).
Table 1. Typology of entities in terms of spatio-temporal deixis (based on Dik and

Hengeveld 1991)
Entity types Spatio-temporal deixis
(x) : First order entity (person, thing) “existence”, situated primarily in space
(e) : Second order entity (state of affairs) “reality”, situated primarily in time
(X) : Third order entity (proposition) “truth”, potential facts, not situated
(E) : Fourth order entity (speech act) “felicity”, situated in time/space
What matters here is that, unlike emotive reaction and volition predicates, the
causative ones cannot take as second argument a higher order entity that
conveys a propositional content (third order entity in Table 1), compare e.g.
(24) and (25). This makes us conclude that the que ‘that’-clause introduced by
a causative predicate necessarily has a processual reading, i.e. corresponds to a
second order entity.12
(24) a. los estudiantes {hacen/temen/desean} que las universidades lleguen
a un acuerdo
‘The students {cause/fear/wish} that the universities come to an
agreement.’
b. los estudiantes {#hacen13 /temen/desean} un acuerdo
‘The students {#make/fear/wish} an agreement.’
(25) a. {hizo /aprehendió /quiso} que le dijeran
{made-3sg/apprehended-3sg/wanted-3sg} that to-him said-3pl
la verdad
the truth
‘He {caused/apprehended/wanted} them to tell him the truth.’
b. {*hizo /aprehendió /quiso} la verdad
{*made-3sg/apprehended-3sg/wanted-3sg} the truth
‘He {*made/apprehended/wanted} the truth.’
This may explain why, with causative predicates, the passage from the biclausal
to the monoclausal structure yields a non-trivial alteration of the meaning
structure of the predicate: with que ‘that’-clause, they depict “staged” causation,
prospectively oriented towards an aspectualized (not necessarily perfective)
resulting event, i.e. a second order entity. Construed with clitic object, however,
they depict another kind of event: the global, monolithic, “unstaged” interpre-
tation required by the use of the clitic, can be brought about either by a first
order entity (23a) or a third order entity (23b), but not by a second order entity
(22). With dejar ‘let’, the second most frequent causative after hacer ‘make’, the
instruction triggered by the clitic lo runs as follows: “look for an entity accessi-
ble from within the ground of the speaker”. Therefore, dejar ‘let’ comes to mean
‘leave untouched’ (e.g. food (23a)), ‘leave untreated’ (e.g. a problem (23b)), or
‘leave alone’ (e.g. a person).
With hacer ‘make’, the switch to immediate causation (viz. production,
manufacturing, performing) is even more striking. But an interesting ambiguity
shows up, due not only to the masculine-neuter syncretism of the clitic lo, but
also to the ability of hacer ‘make’ to anaphorically refer to an action verb, very
much like the English ‘do’ (26a–b). Even when hacer ‘make’ takes a que ‘that’-
clause in the preceding context (27a), the second occurrence of hacer, with
incorporation of the neuter clitic, is inapt at echoing the causative matrix
predicate (27b), but invariably forms a “pro-clause”, thus reflecting the embed-
ded predicate and its argument (retirar el ejército ‘retire the army’ in (27c)).
(26) a. Está acabando su libro. Lo hace al
is-3sg finishing his book-masc. It-cl.{masc/neut} makes to-the
estilo del primero
style of-the first
‘He is finishing his book. He {makes it (viz. the book)/does it (viz.
finishing the book)} in the same style as the first one.’
b. Le aconsejaron que acabara su libro, pero no
to-him advised-3pl that finish-subj.imperf.3sg his book, but not
lo hizo
it-cl made
‘They advised him to finish his book, but he didn’t do it.’
(27) a. La reacción generali hizo que retiraranj
the reaction general made that retired-subj.imperf.3pl
suj ejército,
their army,
‘The general reaction made them retire their army,…’
b. *pero no lo hizoi enseguida

but not it-cl.neut did-3sg immediately
(following (27a)) (with la reacción general as subject)
‘but it did not do it immediately.’
c. pero no lo hicieronj enseguida
but not it-cl.neut did-3pl immediately
‘but they did not do it immediately.’ (following (27a))
Although in (27c) the object argument instantiated by the neuter lo ‘it’ refers
back to a “reality”, situated in time (second order entity), it is now lifted to the
more abstract propositional level, viz. that of a non-situated “truth” (third
order entity), and the perspective on the scene crucially differs from the
“binding” involved in the biclausal causative structure of (27a): whereas in the
latter the scene envisaged is fully dependent on the matrix predicate’s subject, in
the former the speaker grounds its conceptualization in his proper mental space.
To further substantiate this analysis, let us take up again examples (16a) and
(16b), presented in Section 3 as having a “less prototypical” profile than the
causative and the volitive predicates, respectively.
If it is correct to state that cliticization is blocked with “staged” causation,
as defined above because this causative relationship imposes a single perspective
on the scene, viz. that of the Causer, embodied in the matrix predicate’s subject,
then it follows that clititization should become possible as soon as we move to
a different kind of causation, e.g., the one depicting an implication relationship
(16a). Construed with que ‘that’-clause, implicative predicates (28a) typically
have an inanimate subject (28b).
(28) a. acarrear ‘yield’, entrañar ‘entail/yield’, implicar ‘imply’
b. {Esta cláusula/Esta firma /*Ana} implica que compra la casa
{this clause /this signature/*Anna} implies that buys-3sg the house
‘{This clause/This signature/*Anna} implies that {he/ she} buys the
house.’
Yet, unlike the causation relationship stricto sensu, such a logical operation relies
on the intervention of the speaker as conceptualizer. This analysis is sustained
by two other facts, independent of complement clause cliticization. On the one
hand, the speaker can indicate on whose account the expressed view goes (29a),
eventually marking his own (dis)agreement (29b). On the other hand, the
implicative relationship also differs from the “staged” causation in that it allows
for the reversal of the causality chain, as illustrated in (30).
(29) a. Según Ana, esto implica que compra la casa

According to Anna, this implies that buys-3sg the house
‘According to Anna, this implies that {he/she} buys the house.’
b. Esto lo implica en opinión de Ana, pero no
this it-cl implies in opinion-fem of Anna, but not
en la mía
in the-fem mine-fem
‘This implies it in the opinion of Anna, but not in mine.’
(30) Esto {supone/indica/significa} que compra la casa
‘This {presupposes/indicates/signifies} that he/she buys the house.’
Furthermore, some predicates oscillate between a causative and a volitive

reading (31), i.e., they alternately occupy the highest and second-highest
position on the binding scale. In the former case, they preclude complement
clause cliticization (32); in the latter, however, they license it ((16b) and 33).
Cliticization can thus become a clue for the kind of interpretation involved.14
(31) volitive reading: intentar/procurar in the sense of ‘intend’ (instead of
‘attempt’), eludir/vedar in the sense of ‘avoid’ (instead of ‘prevent’)
(32) a. Procuramos que nos den otra oportunidad,
attempt-1pl that to-us give-subj.3pl another opportunity,
*lo procuramos
it attempt-1pl
‘We attempted to get another chance, this we attempted.’
b. El Consejo {vedó /vetó} que subiera el impuesto;
the Council {empeached/vetoed} that would raise the budget,
#lo {vedó /vetó}15
it-cl {empeached/vetoed}-3sg
‘The Council prevented the budget from being raised, this it did.’
(33) a. Procuramos que todos estén contentos,
intend-1pl that all are-subj.3pl satisfied,
lo procuramos
it-cl.neut intend-1pl
‘We intend for everybody to be happy, this we intend.’
b. Juan {vedó / eludió} que le implicaran en el
Juan avoided that him implied-subj.imperf.3pl in the
asunto; #lo {vedó /eludió}
matter; it-cl.neut {avoided/eluded}-3sg
‘Juan avoided being implied in the matter, this he avoided.’
The above observations show that there are strong limits on the parallelism
between nominal cliticization (cf. Section 2) and complement clause cliticiza-
tion. Moreover, the latter appears to be incompatible with the predicate class
that scores highest on the binding scale, viz. the causatives. What matters is not
the type of matrix predicate as such, but the interpretation of the relationship
between the main clause and the embedded clause. For cliticization to be possible,
the data suggest that the speaker has to ground the propositional content of the
embedded clause in his actual mental space. This turns out to be possible with
interpretations that range over implicative, volitive and emotive reaction.
5. Cliticization in relation to perception and propositional attitude

predicates
Let us now have a look at cliticization with perception and propositional

attitude predicates. Construed with clausal complement, verbs of viewing and
sensing do not quite refer to sensory perception, but to a wide range of cog-
nizing processes. This renders them particularly flexible (cf. Horie 1991, this
volume, Verspoor, this volume). As analyzed in detail for Spanish by Rodríguez
Espiñeira (1996) for the prototypical verb ver ‘see’, they can depict primary or
secondary perception of a state of affairs: e.g. (34a) illustrates immediate
perception, whereas (34b) illustrates mediated, or inferential, perception. The
perception verbs can also express mental perception of propositional content
(34c) and reception of the propositional content of a speech act (34d). Howev-
er, these differences do not affect the possibility of cliticizing the respective
complement clauses (35).
(34) a. ¿Has visto que ahí montan una tienda?
¿have-2sg.pres.ind seen that there install-3pl.pres.ind a shop
‘Did you see that they installed a shop there?’
b. ¿Has visto que Ana se ha cortado el pelo?
¿have-2sg.pres.ind seen that Ana to-her-refl has cut the hair
‘Did you see that Ana had her hair cut?’
c. ¿Has visto que Juan consiguió reparar
¿have-2sg.pres.ind seen that Juan managed repair-inf
la máquina?
the machine
‘Did you see that Juan managed to repair the machine?’
d. ¿Has visto en el periódico que habrá elecciones?

¿have-2sg.pres seen in the newspaper that be-3sg.fut elections
‘Did you see in the newspaper that there will be elections?’
(35) Claro que lo he visto
clear that it-cl have-1sg.pres.ind seen
‘Of course I have seen it.’
The shortcut answer lo he visto ‘I have seen it’ (35) accords with any of the
statements given in (34).16 Rather than encompassing any of the semantic layers
instantiated by clausal complementation as illustrated in (34a) to (34d), the
clitic seems to compress, transcend and schematize the complete deictic
organization of the complement clause, i.e. not only its spatiotemporal, but also
its personal and discourse deixis. In reducing the semantic import of the object,
the grammatical unification is not apt at retrieving anything but the most
abstract relational representation, devoid of any deictic anchoring distinct from
that of the speaker’s here-and-now. In Cognitive Grammar terms, cliticization
thus signals that the grounding of the perceived scene or image is overriden by
that of the actual speech event, i.e. the viewing frame ascribed to the matrix
predicate’s subject entity is entirely taken up in the scope of the speaker, who
subsumes it, as it were, into his actual mental space (cf. Fauconnier 1984).
Notice that cliticization suppresses the ambiguity inherent to (34d):
whereas (34d) does not draw a boundary between reception of a speech act, on
the one hand, and reception of the propositional content of a speech act, on the
other, cliticization imposes the latter reading. This matter will be taken up again
in Section 6, devoted to declarative predicates.
To further substantiate this analysis, I will now first try to define more
precisely the clitic’s deictic function in relation to the speaker’s vantage point,
drawing on the parallel with nominal cliticization (5.1). To explain the fact that
some propositional attitude predicates preclude clausal complement cliticiza-
tion, we have to examine more closely how exactly the ground of the comple-
ment clause relates to that of the speaker. The mental space approach offers an
interesting way to go about this kind of relationship (5.2).
5.1 Cliticization: uncovering the speaker as conceptualizer
As pointed out by Lyons (1977: 445), it is the third layer of the schema, given in
Table 1 (Section 4), that corresponds to “entities of the kind that may function
as the objects of such so-called propositional attitudes as belief, expectation and
judgement”. Yet, this does not prevent the possibilities illustrated in (34a) to
(34c) for ver ‘see’ from applying also to creer ‘believe’, esperar ‘expect’, or pensar
‘think’, e.g. (36a)–(36c). The corresponding cliticization applies, no matter
whether the content of the complement clause is conceived of as a second
(36a–b) or as a third order entity (36c).
(36) a. ¿Piensas que ahí están montando
¿think-2sg.pres.ind that there are-3pl.pres.ind installing
una tienda?
a shop
‘Do you think that they are installing a shop there?’
b. ¿Piensas que Ana se ha cortado el pelo?
¿think-2sg.pres.ind that Ana to-her-refl has cut the hair
‘Do you think that Ana had her hair cut?’
c. ¿Piensas que Juan consiguió reparar la máquina?
¿think-2sg.pres.ind that Juan managed repair-inf the machine?
‘Do you think that Juan managed to repair the machine?’
(37) Sí, lo pienso
yes, it-cl think-1sg.pres.ind
‘Yes, I think so.’
So far, I have focused on cliticization being a syntactic correlate of the upgrad-
ing of the propositional nature of the clausal complement and of the downgrad-
ing the other semantic dimensions it may convey, e.g. the progressive aspect in
(36a), the perfective in (36b), the terminative in (36c), the indefiniteness (36a),
or the definiteness of the agent (36b–c). All that makes up the internal structure
of the subordinate event (tense, aspect, mood, participant roles and relation-
ships) disappears behind one single deictic form, bound to the matrix predicate.
If only the conversion of a (possibly) non-propositional content into a
propositional one were at stake, the neuter demonstrative es(t)o ‘this’ and the
manner adverbial así ‘so’ would also do the job, as in (38). The difference seems
to be situated at the presuppositional level: by uttering (37), the speaker
indicates that he instantiates a propositional content as part of his own actual
mental space, and thus readily identifiable to the hearer. When (38) is used, the
propositional content is also presented as identifiable to the hearer, but not
necessarily from within the subject entity’s personal deictic setting at the
moment of speech. The clitic thus seems to be a more specific grounding device
than the demonstrative. It could be said to yield a higher degree of “definite-
ness”, were it not that the notion of “definiteness” is not habitually used on
such an abstract level. Rather, it usually relates to the referential level, which is
involved when the complement is a first order entity (cf. Table 1 in Section 4,
and also Section 2).
(38) Sí, pienso {esto/así}
yes, think-1sg.pres.ind {this /so}
‘Yes, I think so.’
The first person subject of (37) and (38), by default, designates the speaker.
Therefore, these examples obscure the fact that the ground element that is being
taken as point of reference for the clitic is the speaker, and not the “I” concep-
tualizer. In order to see whether it is the speaker (and not the conceptualizer as
expressed by the subject) who presents the object as identifiable to the hearer by
using the clitic, let us therefore take third-person subjects and complement
clauses that uncontroversially convey a propositional content.
If the propositional interpretation were a sufficient condition for cliticiza-
tion, then it should apply equally well to verbs such as pensar ‘think’ and opinar
‘think’ in the sense of ‘holding a view’.17 However, while pensar allows for
cliticization, opinar does not, e.g. (39) vs. (40).
(39) a. Montse pensaba que era muy triste vivir así
Montse thought that was very sad live-inf this way
‘Montse thought that it was very sad to live this way’
b. Montse lo pensaba
Montse it-cl thought
‘Montse thought so.’
(40) a. Montse opinaba que era muy triste vivir así
Montse held-the-view that was very sad live-inf this way
‘Montse thought that it was very sad to live this way’
b. *Montse lo opinaba
Montse it-cl thought
This means that the object entity is not a simple discourse deictic marker tied to
the subject entity. Moreover, it seems to be tied to the speaker’s vantage point
in a non-trivial way, i.e. not just by the fact that it is ultimately the speaker who
invests the subject entity with the perceptual abilities required to instantiate the
percept expressed by the complement. If the clitic form only served to tie the
object entity to the subject entity as instantiator of the propositional attitude
expressed by the matrix predicate, then it would not be expected to cooccur
with other deictics. The only reason why this cooccurrence is not perceived as
redundant is that the speaker can simultaneously integrate an argument in his

own setting by means of lo ‘it’, while relating it also to the specific on-stage
conceptualizer (así ~ Montse in (39c)) or to an off-stage collective one, beyond
the latter (esto ~ Montse + others in (39d)). Notice that opinar ‘have the opi-
nion’ is incompatible with this generic interpretation (40d), and although it can
combine with así ‘so’, clitcization remains impossible (40c).
(39) c. Montse lo pensaba así
Montse it-cl thought so
d. Esto Montse lo pensaba
This Montse it-cl thought
‘This Montse thought.’
(40) c. Montse {*lo / Ø} opinaba así
Montse {*it-cl thought so
d. *Montse opinaba esto
From this asymmetrical patterning it can be inferred that, although both pensar
and opinar introduce a propositional content, only pensar opens it up in such
a way that it becomes accessible to the speaker, and can be assimilated to his
here-and-now. With opinar ‘hold a view’, the content of the complement clause
is, of course, also tied to the subject entity as instantiator of the main process
type, but it does not fall within the scope of the speaker the way definite
nominal objects typically do. Situated outside the realm of the speaker, it yields
a mental representation that cannot quite be equated with that of a standard
argument. Another way of putting this is to say that the argument does not
seem to acquire a fully nominal profile. With pensar, on the contrary, it is the
speaker who functions as reference point: he invests his own vantage point into
the relationship with the object entity, next to the vantage point of the subject.
To sum up, (39a) and (40a) have in common that the subject entity
endorses the way the propositional content of the complement clause is
accessed. They differ, however, as to whether it is also accessible to the speaker,
or not. This difference is reflected by the acceptability vs. unacceptability of
cliticization (39b vs. 40b). In terms of argument structure, this difference testifies
to the fact that the argument of pensar can acquire full nominal status, whereas
that of opinar cannot. The verbs mentioned under (41) behave like opinar.
(41) calcular ‘to calculate’, considerar ‘to consider’, contemplar ‘to contem-
plate’, encontrar ‘find’, enjuiciar ‘judge’, especular ‘speculate’, evaluar ‘to
evaluate’, meditar ‘to meditate’, opinar ‘hold a view’, ponderar ‘ponder’,
razonar ‘to reason’, reflexionar ‘to think’
(42) a. Reflexionó, sin embargo, que, a pesar de su vejez y
thought-3sg, nevertheless, that, notwithstanding of his old-age and
abandono, aquel salón trascendía a grandeza grave y
abandonment, dem room trascended to greatness solemn and
a rancio abolengo
to old descent
‘He thought, nevertheless, that, notwithstanding its old age and
abandonment, that room was reminiscent of solemn greatness and
old-fashioned descent.’
[Larreta, Gloria de D. Ramiro, 2, p. 189, quoted in Cuervo (1994)]
b. *lo reflexionó
it-cl thought-3sg
‘He thought it.’
As illustrated by the context of (42), these verbs yield a subjective modality: they
express a specific way of looking at things, so that the vantage point is fully
determined by the subject entity. As a consequence, it is impossible to ground
the content of the complement clause otherwise than vis-à-vis the conceptual-
ization assumed by the subject entity, i.e. the on-stage conceptualizer, distinct
from the speaker.18
To summarize, (i) the content of the complement clause need not be
propositional to be cliticizable (cf. (36a–b)); (ii) it can be propositional without
being cliticizable (cf. (40a–b)); (iii) in the latter case only the manner deictic así
‘so’ can take it up, tying it to the matrix predicate’s subject’s vantage point (cf.
(40c–d)); (iv) cooccurrence of the clitic with así ‘so’ testifies to the coexistence
of the speaker’s vantage point (cf. (39c)); (v) the latter can even manifest itself
together with the representation of a collective, generic vantage point (cf.
(39d)). These observations confirm that it is indeed relevant to distinguish
between the mental space of the speaker and the mental space of the concep-
tualizer, as advocated by Fauconnier (1984: 20).
5.2 Cliticization and mental space conflation
Construed with complement clause, the verbs of propositional attitude intro-

duce a space that is dependent upon another space but leave undecided whether
the mental space of the speaker (MSS) coincides with that of the conceptualizer
(MSC), or whether it is separated from it, partially or entirely. The clitic
pronoun lo ‘it’, on the contrary, performs a reduction of space and signals that
the image belonging to the reality space of the speaker S and that of the space
emanating from the conceptualizing subject C are joined. This analysis is
represented schematically in Table 2.
Table 2. The mental space representation corresponding to the different complex

clause event structures and its effect on cliticization
Complex clause Mental space representation Cliticization
(i) VP+{si/qué/etc.} [MSC Õ MSS] lo VP
(‘wh’-cl.)
(ii-a) [MSC Õ MSS] lo VP
(ii-b) VP+que (‘that’ clause) Partially [MSC Õ MSS], partially lo VP
[MSC π MSS]
(ii-c) [MSC π MSS] *lo VP
(iii) VP + “direct discourse” [MSC π MSS] *lo VP
On the basis of the form of the complement clause, we distinguish three

types of complex clause event structures: (i) with wh-clause, (ii) with que ‘that’-
clause, and (iii) with direct discourse. What interests us here is their conceptual
organization as defined by the inclusion/ exclusion relationship between the
mental space of the conceptualizer (MSC) and that of the speaker (MSS). The
inclusion relationship licenses cliticization, the exclusion relationship precludes it.
With direct discourse (iii), as will be illustrated below, the MSC and the
MSS remain separate, since the conceptualizer C, symbolized by the subject
instance that does the thinking, sensing or saying, is by definition different from
the speaker S. Hence, cliticization does not apply at all. On the contrary, as seen
in Section 3, it systematically applies with wh-complementation (i): here, the
speaker has access to the mental space of the conceptualizer (MSC) and
integrates it into his own mental space (MSS).
With que ‘that’-clause, however, both representations of mental space are
possible. As shown in (ii), the MSC can be included in the MSS or coincide with
it (MSC Õ MSS) (ii-a), or, else, it can be different from it (MSC π MSS) (ii-c).
The intermediate situation is also possible, viz., part of the grounding of the
clausal complement can be integrated, whereas another part can remain outside
the MSS, as indicated in (ii-b). As a matter of fact, such a mixed grounding is
more common than might be thought. Example (43), inspired by Fauconnier

(1984), illustrates this point.
(43) a. Juan cree que la hija de Ana es más inteligente de lo que
Juan believes that the daughter of Ana is more intelligent of what
es en realidad
is in reality
‘Juan thinks that Ana’s daughter is more intelligent than she in fact
is.’
b. Juan lo cree
Juan it-cl believes
‘Juan thinks it.’
(44) a. Juan cree: “la hija de Ana es inteligente”
Juan believes“the daughter of Ana is intelligent
‘Juan thinks: “Ana’s daughter is intelligent”.’
b. *Juan cree: “la hija de Ana es más inteligente de lo que
Juan believes“the daughter of Ana is more intelligent of what
es en realidad”
is-3sg in reality
‘Juan thinks: “Ana’s daughter is more intelligent than in fact she is”.’
(45) a. la hija de Ana es inteligente (MSC)
the daughter of Ana is intelligent
‘Ana’s daughter is intelligent’
b. la hija de Ana no es tan inteligente (MSS)
the daughter of Ana not is that intelligent
‘Ana’s daughter is not that intelligent.’
The comparison contained in the complement clause cannot originate in the

MSC: the corresponding proposition can only be (44a), since (44b) would be
inconsistent. The complement clause thus contains a comparison between two
conflicting mental spaces, one ascribed to Juan (MSC), viz. the one containing
the propositional content expressed in (45a), and a second one that contains the
opposite propositional content, viz. the one expressed in (45b). It is clear that
the speaker can only identify (MSS) with the latter.
At the inferential level, the construction with complement clause (43a)
occupies the intermediate position between the independent statement of (44a),
on the one hand, and the cliticized formula, given in (43b), on the other hand:
the latter levels out the boundary between the competing mental spaces to the
advantage of that of the speaker (MSS). By taking in a complex statement, the
speaker does more than simply report on a propositional content: he integrates

into his proper mental space (MSS) the conflictive proposition that belongs to
the mental space of the conceptualizer (MSC), while overriding it at the same
time, viz. (45a, b).
In view of the foregoing, the condition on complement clause cliticization can
thus be restated as follows: to be cliticizable, the content of the que ‘that’-clause
must be(come) propositional and entirely retrievable from within the mental
space of the speaker (MSS). This is automatically the case when the comple-
ment clause already contains an element that is grounded in the actual mental
space of the speaker, e.g. the comparison más… de lo que ‘more… then’ in (43).
According to this analysis, verbs like those mentioned under (41) preclude
cliticization because they are lexically marked for instantiating a view that cannot
be incorporated within the scope of the actual mental space of the speaker.
It follows that cliticization can be used as a test to verify whether a verb
predicates a propositional attitude from the vantage point of the speaker, or
whether it introduces a judgement or a reaction confined to the vantage point
of the conceptualizer. Compare (46) and (47): only in (46) can the complement
clause be made to function independently as an assertion on the part of the speaker.
(46) a. Todos piensan que este mundo va mal
all think-3pl that this world goes bad
‘They all think that this world is in bad shape.’
b. Que este mundo va mal, todos lo piensan
that this world goes bad, all it-cl think-3pl
‘That this world is in bad shape, they all think so.’
(47) a. Todos piensan que va a llover
all think-3pl that goes to rain
‘They all think that it will rain.’
b. *Que va a llover, todos lo piensan.
that goes to rain, all it-cl think-3pl
‘That it will rain, they all think so.’
For the same reason, cliticization is uncommon with verbs that are used to
express the subjective reaction towards an event or process, e.g. (48).
(48) aborrecer ‘to abhor’, adorar ‘to adore’, detestar ‘to detest’, odiar ‘to hate’
Here, the fact that the mental space of the conceptualizer is cut off from the
speaker’s actual mental space can be brought to the fore by switching from the
que ‘that’-subordinator to the cuando ‘when’-subordinator, e.g. (49).
(49) a. Ana adora {que le traigan /cuando le

Ana adores {that to-her bring-subj.3pl/when to-her
traen} flores
bring-ind.3pl} flowers
‘Ana adores {that/when} they bring flowers.’
b. *Que le traigan flores, Ana lo adora
that to-her bring-3pl flowers, Ana it-cl adores
‘That they bring her flowers, Ana adores it.’
The temporal clause corroborates the experiential character of the scene and
indicates that the grounding of the complement clause cannot be superseded by
that of the main clause.
At the other end of the scale of the propositional attitude predicates, we find
verbs that express a minimally cognizing event, one that could perhaps be
labeled “factual reception”, viz. (50).19
(50) aceptar ‘to accept’, aprender ‘to learn’, aprobar ‘to approve’, con-
statar ‘to ascertain’, reconocer ‘to recognize’
They depict the way the conceptualizer, embodied in the matrix predicate’s
subject, receives the representation of a state of affairs that is already included
in the mental space of the speaker (MSS). When the reception turns out to be
more emotive than cerebral, the complement clause comes in the subjunctive
and the conceptualizer’s attitude can surface in manner adverbials (51a).
Whether the image of the situation found is truly propositional in nature is not
quite clear, since it sometimes appears to be temporally situated. Still, cliticiza-
tion naturally follows from its being grounded in the MSS (51b).
(51) a. Acepta de mala gana que le cambien el
accepts-3sg of bad mood that to-him change-subj.3pl the
horario a partir del segundo semestre
schedule starting from the second semester
‘He reluctantly accepts that they change his schedule from the second
semester on.’
b. Lo acepta de mala gana
It-cl accepts-3:sg of bad mood
‘He accepts it reluctantly.’
Since what matters is the way the speaker looks upon the content of the
complement clause, the idea that the latter has to be propositional does not
seem totally accurate. It may be more adequate to consider as a necessary
condition that the speaker conceives of it as a “fact” at a more pragmatic level.20
It remains to be seen whether a similar analysis can be applied when the
conceptualizer is maximally agentive, i.e. in declarative processes. This is the
object of Section 6.
6. Cliticization with declaratives
With declarative predicates, the action can be viewed basically in two ways: as
speech production, or as the transfer of some content. Although the latter seems
to be a necessary condition for complement clause cliticization, it is not a
sufficient one. Only when the subordinate predicational relationship is con-
strued from within the mental space of the speaker, is cliticization possible;
otherwise, it is blocked (6.1). The existence of a cognitively motivated autono-
my/dependency scale in terms of “separate” vs. “mixed” grounding of the main
and the subordinate clausal event structure, as presented in Section 5.2, is
further corroborated by two complementary phenomena.
On the one hand, illocutionary markers in the subordinate clause enhance
the predominance of the on-stage conceptualizer over the speaker, thus yielding
a separate grounding for the subordinate event structure, which makes it
incompatible with cliticization. On the other hand, modality markers adjoined
to the matrix predicate that relate to the speaker’s viewpoint, corroborate the
mixed grounding of the subordinate event structure, which makes them
compatible with cliticization (6.2).
Since these two types of markers do not cooccur in the complement clause,
and cliticization only patterns with the latter, this correlation can count as
evidence that the behavior of the declaratives regarding complement clause
cliticization is not haphazard, but obeys the same rationale as the one observed
with propositional attitude predicates (cf. Section 5).
6.1 Verbal expression with the focus on transmission
In verbal action, there is both an expressive and a communicative dimension.

The declarative predicates do not all convey these two dimensions in equal
proportion; compare, e.g. the English say, speak, talk, tell. Whether they focus
more on the utterer, i.e. on the productive side, or draw attention to the
message conveyed and to its transfer, thus privileging the notion of transmission,
also depends on the context and, in the first place, on the construction used.
Pursuing the line of argument developed in Sections 4 and 5, the hypothesis
is that the more the declarative is meant to express self-disclosure on the part of
the utterer, the less the complement clause will be susceptible of cliticizing,
because the mental space of the conceptualizer (the utterer in this case) will be
conceived of as standing apart, separate from the mental space of the speaker.
As already indicated in Section 5.2 (Table 2), the extreme case is that of the
direct discourse structure: unlike the complement clause introduced by the
subordinator que ‘that’, the direct discourse structure is incompatible with
cliticization. The reason is that the direct discourse constitutes the nucleus
bearing the central focus of the complex event structure, and not the matrix
predicate. The direct instantiation of the MSC preserves the illocutionary,
evaluative or affective strength of the source utterance.
Therefore, as indicated by the downmost position of the direct discourse
structure in Table 2, the very status of source-utterance makes the mental space
of the conceptualizer (MSC) alien to that of the speaker (MSS). The verbatim
reproduction is iconically separated from the MSS: graphically, by the use of
quotation marks or italics, and also syntactically. The deictic markers are
anchored in the original discursive space and combine with a non-restricted
paratactic syntax, as in example (52).
(52) “Eso no es nuevo”, protestó Pablo
“this not is new protested Pablo
‘“This is not new”, Pablo protested.’
The autonomy of the utterer’s voice corresponds to the dialogical function of

the direct discourse, which makes it impossible to attribute the interpretation
of the cognitive activity to the speaker (S), next to the utterer, who is the
original conceptualizer (C). The latter is on-stage, and the declarative verb only
serves to qualify the assertion of the conceptualizer.
Notice that the conceptualizer’s role as instantiator of the declarative event
can be endorsed by the speaker himself, as shown in (53).
(53) a. Protesté que me casaban por fuerza
protested-1sg that me married-3pl by force
‘I protested that they married me by force.’
[Rojas Zorrilla, El Caín de Cataluña, quoted in Cuervo (1994)]
b. *lo protesté
it-cl protested-1sg
‘I protested it.’
Putting himself on stage, the speaker combines two perspectives, viz. that of
himself as conceptualizer (first person subject) and that of his actual deictic
grounding at the moment of speech. A kind of internal splitting up is in order
to combine the two instantiations. However, when a verb like protestar ‘protest’
is used, the mental space of the conceptualizer (MSC) is presented as separate
(if not different) from that of the speaker (MSS).21
As reflected in Table 2, direct discourse remains outside the realm of
cliticization. However, it is not incompatible with the deictic adverbial así ‘so’
next to the matrix predicate, e.g. (54). This type of deictic marker situates the
declarative content within the mental space of the conceptualizer. Although it
more often characterizes intransitive predicates than transitive ones (Delbecque
1994), it sometimes co-occurs with the clitic lo ‘it’, e.g. (55). This formulation
allows the speaker to signal that although the content originates in the mental
space of the conceptualizer (cf. así ‘so’), the speaker is not prevented from
having access to it from within his own mental space (cf. lo ‘it’).
(54) El jefe {habló /dijo} así: “[…]”
the boss {spoke/said} so: “[…]”
‘The boss {spoke/said} so: “[…]”’
(55) El jefe lo {*habló/dijo} así
the boss it {spoke /said} so
‘The boss said it this way.’
However, when the focus of attention falls on the communication of the message,
its successful transmission precisely implies that the content cannot be confined to
the mental space of the conceptualizer, but must become accessible from within
the mental space of the speaker. Hence, the complement clause is cliticizable.
The que ‘that’-clause (53a) links the externally grounded point of view
(MSC) to the internally grounded point of view of the speaker (MSS).22 The
adjunction of a seemingly direct complement clause dedramatizes the formula-
tion and renders it monological, as if it were a regular indirect discourse
structure.23 Yet, it does not by itself guarantee the projection of the content of
the mental space of the conceptualizer (MSC) onto that of the speaker (MSS).
As indicated in Table 2 (ii), the construction is opaque as to the degree of
integration of the MSC into the MSS.
In other words, it is not clear up to which point the speaker renounces

assertion when echoing somebody else’s wording. Yet, the performative
character of the clausal complement prevents it from being cliticized, e.g. (53b).
Here are some examples of verbs that tend to be used only as performatives (56).
(56) apostar ‘to bet’, argüir ‘to argue’, argumentar ‘to argue’, aventurar ‘to
venture’, clamar ‘to shout’, fallar ‘to judge’, negociar ‘to negotiate’, pactar
‘to negotiate’, protestar ‘to protest’, sentenciar ‘to judge’, suplicar ‘to im-
plore’
(57) anunciar ‘to announce’, contestar ‘to answer’, decir ‘to say’, declarar ‘to
declare’, pedir ‘to ask’, replicar ‘to reply, reponer ‘to reply’
Again, it has to be stressed that in most cases it cannot be maintained that the
matrix predicate determines the semantic layer at which the content of the
complement clause is conceptualized. With many declarative verbs, the content
of the complement clause can be processed both ways, viz. as speaker-depen-
dent or as speaker-independent, i.e. either from within the speaker’s actual
mental space (MSS), or not. This is clearly the case for the most frequent
declaratives, e.g. those mentioned in (57).
Table 3. Complement clause cliticization as a reflection of speaker dependent process-

ing, as defined by the inclusion relationship between the mental space of the
conceptualizer (MSC) and that of the speaker (MSS)
Biclausal construal From binary to unitary Cliticization
Mental space approach Autonomy / dependency scale Prototypical cognition and
utterance predicates
[MSC Õ MSS] speaker dependent processing lo acepta/lo proclama
mixed grounding ‘he accepts/exclaims it’
[MSC {Õ/π} MSS flexible lo piensa/lo dice (mixed)
mixed grounding or ‘he thinks/says it’
separate grounding *lo piensa/*lo dice (separate)
[MSC π MSS] speaker independent processing *lo razona/*lo clama
separate grounding ‘he reasons/shouts it’
As shown in Table 3, this flexibility can be rendered schematically in terms

of a conceptual autonomy/dependency scale defined by the inclusion/exclusion
relation holding between the mental space of the conceptualizer (MSC) and
that of the speaker (MSS).
Table 3 summarizes the possibilities encountered so far for the semantic

class of cognition and utterance predicates: some prototypically convey the
image of a maximally speaker-dependent information processing, while others
depict speaker-independent behavior on the part of the conceptualizer. Clit-
icization is blocked only when the latter applies. Yet, the most common verbs
of cognition and verbal expression can go either way. With declaratives, the
presence of some specific markers is revealing for the choice made by the
speaker between a separate or a mixed grounding of the subordinate event
structure. Again, only the latter licenses cliticization, as will be shown in
Section 6.2.
6.2 Marking the choice of a separate vs. mixed grounding
This last subsection is devoted to a brief survey of the different modifiers that
are indicative of the type of processing involved in the MS grounding of the
subordinate clause. Modality markers that point towards the speaker’s interven-
tion are compatible with cliticization. However, the reverse is true of markers
that symbolize the speaker’s withdrawal in favor of the utterer as grounding
instance; it is sufficient that there be one single illocutionary marker in the
subordinate clause for cliticization to be blocked.
Modality markers are easily found in the que ‘that’-clause depending on
declarative predicates such as the ones mentioned in (58).
(58) aducir ‘to adduce’, alegar ‘to allege’, aplaudir ‘to applaud’, atestiguar ‘to
attest’, conmemorar ‘to conmemorate’, destacar ‘to highlight’, difundir ‘to
diffuse’, divulgar ‘to divulgate’, enfatizar ‘to emphasize’, evocar ‘to evoke’,
insinuar ‘to insinuate’, mantener ‘to maintain’, pretextar ‘to pretende’,
proclamar ‘to exclaim’, realzar ‘to emphasize’, subrayar ‘to underline’,
testificar ‘to testify’
Since modalization adds a nuance as to the choice of the matrix predicate, it has
to rely on a viewpoint external to that of the matrix predicate’s subject (cf.
Achard, this volume). Therefore, these predicates can be said to cluster at the
dependency end of the autonomy-dependency scale depicted in Table 3. The
speaker’s vantage point can be brought to the fore by adjoining an evaluative
modifier to the subject (59) or to the object (60).
(59) que nadie lo había avisado, lo pudo pretextar {en su

that nobody him-cl had informed, it could-3sg pretend {in his
calidad de invitado}
quality of visitor}
‘That nobody had informed him, this he could pretend {being a visitor}.’
(60) que tiene demasiado trabajo, lo aduce como {excusa/pretexto}
that has too much work, it-cl adduces-3:sg as {excuse/pretext}
‘That he has too much work, he adduces it as an {excuse/pretext}.’
Adding a modal auxiliary to the matrix predicate also highly favors the interpre-
tation of the verbal action as communicative behavior and, hence, as dependent
upon the speaker, e.g. (59). Similarly, adding the recipient together with the
specifically involved body part, as in (61), induces a clearly telic reading of the
matrix predicate, thus focusing also on the notion of successful transfer: from
the presence of al oído ‘to the ear’ in (61), one infers that transmission comes to
completion. Simplifying what is going on, we might say that if a third party is
susceptible of receiving the message, this a fortiori holds for the very speaker,
since the latter can play the role of omniscient storyteller without having
witnessed the scene as overhearer.
Such a precision makes both the biclausal construction and cliticization
more acceptable with verbs that are not usually construed with a dative but are
used to express verbal behavior without implying the transmission of a specific
content, e.g. those mentioned under (62).24
(61) finalmente se atrevió a murmurárselo {al oído},
finally refl dared-3sg to murmur-to-him.cl it-cl {at the ear}
que tenía miedo de la oscuridad
that had-3sg fear of the obscurity
‘He finally dared murmur it into his ear, that he was afraid of the dark.’
(62) balbucear ‘to stammer’, balbucir ‘to stammer’, farfullar ‘to stammer’,
gemir ‘to wail’, gimotear ‘to wail’, gritar ‘to shout’, gruñir ‘to grumble’,
murmurar ‘to murmur’, susurrar ‘to murmur’
From examples (59)–(61) it appears that indirect discourse with declaratives

can be thought of as an instantiation of the MSC that is mediated by the speaker,
i.e., the speaker proceeds to a reinterpretation of the MSC. In paraphrasing, he
compresses and schematizes whatever mental representation that originates in
the MSC. In terms of the layered approach (cf. Table 1, Section 4), this means
that the speaker abstracts away from the concrete wording of the underlying
speech act, viz. the fourth order entity grounded in the MSC, and brings its
content back to the level of a third order entity.
Globally speaking, expressions that concord with the limits of the MSS, that
are not overtly alien to it nor marked for being held at some distance (quotation
marks, italics, and the like), tend to be integrated into it. This inferential
tendency corresponds to the typical function of indirect discourse, viz., to
translate an utterance that originates in another mental space, so as to make it
accessible to the actual mental space of the speaker.
In prototypical indirect discourse, the MSS incorporates the MSC, i.e. the
speaker is in full control, which makes the complement clause cliticizable. The
more the matrix predicate is modalized and aspectualized, e.g. by the presence
of an auxiliary or an adverbial modifier, the more it seems apt at integrating the
MSC into the MSS, since its content is then removed from the original dia-
logical setting and, therefore, somehow “derealized”, in the sense of being
detached from its origin. But it is far from being a matter of all or nothing, and
many attested biclausal constructions with que ‘that’-clause amalgamate direct
and indirect discourse.
As expected, the presence of an element that marks the illocutionary
character of the subordinate clause blocks cliticization. Adverbial markers such
as (des)afortunadamente ‘(un)fortunately’, francamente ‘frankly’, ojalá ‘God give
that’, por supuesto ‘of course’, etc., convey a de dicto reading that is predicational
instead of propositional (cf. Suzuki, this volume). They are overtly mimetic,
since they (pretend to) reproduce literally part of the original speech event (the
way direct discourse does). They thus align the content of the complement
clause on a discursive space that remains inaccessible to the speaker. Such
attitudinal, argumentative and interpersonal markers are grounded in an
interlocutional setting that remains off limits with respect to the actual mental
space of the speaker.
Although these markers do not actually instantiate the original interlocu-
tion, they are oriented towards the original setting in a way that makes the
original speech event escape the actual control of the speaker. Therefore, they
are incompatible with cliticization, e.g. (63) vs. (64). In (64), the presence of
francamente ‘frankly’ blocks cliticization.
(63) a. Eva pretextó que nunca había visto nada parecido
Eva pretended that never had-3sg seen nothing similar
‘Eva gave as pretext that she had never seen anything like that.’
b. Eva lo pretextó
Eva it pretended
‘Eva gave it as pretext.’
(64) a. Eva pretextó que francamente nunca había visto
Eva pretended that frankly never had-3sg seen
nada parecido
nothing similar
‘Eva gave as pretext that, frankly, she had never seen anything like that.’
b. *Eva lo pretextó
Eva it pretended
‘Eva gave it as pretext.’
In written language, quotation marks indicate that the formulation goes back to
direct discourse. Whether it does so entirely (65) or only partially (66), it is
incompatible with clausal reduction.
(65) a. Aznar subrayó que ‘el mejor blindaje contra el terrorismo es la su-
perioridad moral del Estado de Derecho’. (El País 23.4.95/16)
‘Aznar underlined that “the best protection against terrorism is the
moral superiority of the Right State”.’
b. *Lo subrayó
It underlined-3sg
‘He underlined it.’
(66) a. El líder del PP pidió a los candidatos de su partido que asuman el
firme compromiso de “no aumentar los impuestos en ningún nivel
de la administración” (El País 23.4.95/16)
‘The PP leader asked to the candidates of his party to assume the
firm engagement of “not to raise the taxes at any level of the administra-
tion”.’
b. *Lo pidió a los candidatos de su partido
It asked-3sg. to the candidates of his party
‘He asked it of the candidates of his party.’
With the usual communication verbs, i.e. those that imply a hearer, even when
the recipient of the transmitted content is not overtly expressed, e.g. those
mentioned in (67), the reading of the complement clause is propositional by
default, precisely because the focus is not on the performativity but on the
transmission. So, if a message supposedly gets across to other persons, it must
a fortiori be accessible to the speaker.
Therefore, even multiple embedding does not necessarily preclude
cliticization, since the clitic appears to operate as a reduction and schematiza-

tion device that testifies to the full integration of the complement into the
valency of the matrix verb, assigning it a bare propositional interpretation in the
first place, e.g. (68). The clitic lo ‘it’ signals retrievability of this representation,
while neutralizing the internal structure of the embedded clause. If it were
merely a matter of topicalization, cliticization would not be possible.25
(67) advertir ‘to inform’, avisar ‘to inform’, comunicar ‘to communicate’,
informar ‘to inform’
(68) a. Juan anunció que su madre no llega antes del domingo porque el
sábado todavía trabaja
‘Juan announced that his mother not arrives before Sunday because
the Saturday still works.’
b. Juan lo anunció
Juan it announced
‘Juan announced it.’
The analogy with the predicates of propositional attitude, treated in Section 5,

can be rounded off by mentioning that the complement clause can, again,
contain elements that are already grounded in the domain of the speech event
expressed by the matrix predicate, e.g. (69). The clitic then symbolizes the
incorporation of a complex propositional content.
(69) a. Ana afirma que Paco llamará otra vez a las nueve
Ana affirms that Paco call-fut another time at the nine
‘Ana affirms that Paco will call again at nine o’clock.’
(cf. Fauconnier 1984: 135ff.)
b. Ana lo afirma
Ana it-cl affirms
‘Ana affirms it.’
In (69a), afirmar ‘affirm’ does not simply report on a speech event, transforming
it into a propositional content. Here, the matrix predicate introduces a state of
affairs (viz. ‘Paco’s calling at nine’), to which a predication modifier is added (viz.
‘again’) that conveys a presupposition. The state of affairs referred to is given in
(70a), and the presupposition in (70b). Whether this presupposition is shared
by the speaker and the conceptualizer, viz. Ana, cannot be inferred from (69a).
(70) a. Paco llamará a las nueve
Paco call-fut at the nine
‘Paco will call at nine o’clock.’
b. Paco ya llamó antes

Paco yet called before
‘Paco already called before.’
In other words, more than one mental space can be actualized in the comple-
ment clause and, crucially, these spaces are not necessarily shared by speaker
and conceptualizer. This can be seen by turning to the corresponding direct
discourse formula: it can be both (71a) and (71b), since Ana did not necessarily
know that Paco had already called, when she affirmed that he would call at nine.
This means that the mental space of the conceptualizer (MSC) can either
coincide partially or entirely with the mental space of the speaker (MSS),
reflected in the indirect discourse of (69a). Notice that the cliticized formula
(69b) subsumes the whole content of the complement clause, hence eclipsing
the potential contribution of the MSC. As soon as part of the complement clause
can be identified as stemming directly from the MSS, cliticization is possible.26
(71) a. Ana afirma: “Paco llamará a las nueve”
‘Ana affirms: “Paco will call at nine o’clock”.’
b. Ana afirma: “Paco llamará otra vez a las nueve”
‘Ana affirms: “Paco will call again at nine o’clock”.’
Finally, the complex interplay and superposition of alternating viewpoints
allows for both ambiguity and vagueness as to whether the speaker simply
echoes the content of the MSC, or whether he combines elements of his own
mental space (MSS) with elements of the MSC. There are, for instance, evalu-
ative adnominal qualifiers, e.g. el bobo/idiota/estúpido de N ‘the idiot of N’ that
can stem either from the speaker or from the conceptualizer. The complement
clause of (72) contains two propositions, viz. (73a) and (73b). On the one hand,
formulation (72) leaves undecided whether Eva literally said (74a) or (74b), i.e.
(72) can be a paraphrase on the part of the speaker.27
(72) Eva ha dicho que el bobo de su marido se jubila este año
Eva has said that the stupid of her husband refl retires this year
‘Eva has said that her stupid husband retires this year.’
(73) a. El marido de Eva es bobo
the husband of Eva is stupid
‘Eva’s husband is stupid.’
b. El marido de Eva se jubila este año
‘Eva’s husband refl retires this year.’
(74) a. Mi marido se jubila este año

‘My husband refl retires this year.’
b. Mi bobo de marido se jubila este año
my stupid of husband refl retires this year
‘My stupid husband retires this year.’
(75) Eva lo ha dicho
Eva it-cl has said
‘Eva has said it.’
On the other hand, it is not clear, either, whether the attribution reformulated
in (73a) goes on the account of Eva (MSC) or emanates from the speaker
(MSS). From the cliticized counterpart of (72), viz. (75), we still cannot be sure
whether Eva said anything like (73a) or (73b). However, one kind of indetermi-
nacy disappears. Lo ‘it’ signals that whatever comes from the MSC (Eva), it is
taken into the MSS, i.e. the two propositions, (73a) and (73b), are now ground-
ed in the speaker’s mental space, even if they did not arise in it to start with.
With examples (69) and (72) in mind, we are now able to refine the
conditions on clausal complement cliticization as follows: the presence in the
complement clause of a speaker-oriented presuppositional predicate modifier
(69), or that of an evaluative nominal qualifier (72), do not at all hamper
cliticization. These modifiers, which manifest the relevance of speaker deixis for
the interpretation of the complex event structure, even appear to facilitate
cliticization. They thus warrant the claim that the clitic symbolizes full integra-
tion into the mental space of the speaker (MSS) of a possibly complex proposi-
tional content that originates, at least partially, in a mental space distinct from
the MSS, viz. the MSC (the conceptualizer being Ana in (69), Eva in (72)). This
also corroborates the idea that cliticization erases the boundaries between the
different mental spaces.
7. Conclusion
There is a structural analogy between complement clause cliticization and

nominal cliticization, since the neuter clitic used for the former belongs to the
same paradigmatic set of object pronouns that is used for nominal arguments,
and its discourse function is also an anaphoric and topicalizing one. The
analogy is partial, however for the following reasons.
Unlike nominal cliticization, complement clause cliticization is never
obligatory. It is always possible with the typically interrogative set of wh-opera-

tors and with the compound one el (hecho de) que ‘the (fact of) that’, which
triggers a factual reading of the complement clause. However, with the most
commonly used subordinator, the unmarked que ‘that’, it often appears to be
problematic. Some propositional attitude predicates even appear to be incom-
patible with it.
Syntactically, substituting a clitic pronoun for a complement clause boils
down to modelling a biclausal sentence structure after the monoclausal one.
The mapping of a clausal event structure on to the pronominal paradigm can be
iconically interpreted as a signal whose conceptual domain is being thought of
in the same way as that of a nominal argument.
However, converting a time-unstable relational entity into a time-stable
non-relational one requires a strongly inferential metaphorical process. If it
were only a matter of abstraction, it would be sufficient to have a propositional
reading for the complement clause in order for cliticization to be unproblem-
atic and to add some “specificity”, “definiteness” or “familiarity” conditions,
analogous to those that operate as motivating factors in the cases of variable
nominal complement cliticization.
To unfold the cognitive mechanisms underlying this phenomenon of
variable clause reduction, we have drawn attention to (i) the role of the matrix
predicate, (ii) its relationship with the complement, (iii) the internal structure
of the complement clause.
The factors that present a negative correlation with cliticization can be
summarized as follows: (i) the more the cognitive processing expressed by the
matrix predicate is conceptually saturated, the more it refers to a self-contained
performative action, the more problematic complement clause cliticization will be;
(ii) the less transitive the subject-object relationship is, the less receptive the role of
the subject, the less likely the clausal complement is likely to surface as a clitic;
(iii) finally, the more independent and the more elaborated the deictic organi-
zation of the complement clause, the less it is susceptible of being cliticized.
These conditions cannot be defined solely in lexical and referential terms,
since they operate with regard to a unitary, presuppositional diagnostic repre-
sentation of an entire event structure. Since both the indirect discourse struc-
tures introduced by a wh-operator and the factual ones, marked by el que ‘the
that’, yield such a reading, they are easily cliticizable.
With que ‘that’-subordinator, however, the matter appears to be epistemic-
ally quite more complex. On the one hand, it combines with a far broader array
of process types and with a much larger number of matrix predicates than the
other subordinating operators. On the other hand, whether the complement

clause is cliticizable or not, is only partially determined by the type of matrix
predicate. Indeed, predicates can belong to the same semantic class, instantiate
a similar type of process and, yet, react differently with respect to cliticization.
The predicates that are incompatible with complement clause cliticization
typically situate the subject-object relationship outside the scope of the speaker: the
grounding predication of the complement clause is represented as inaccessible
from the actual ground. This fourth parameter, viz. (iv) the position of the
speaker, next to that of the conceptualizer, with respect to the content represented
by the complement clause (event or proposition), allows us to offer a principled
and unified account of the variable cliticizability.
This is not to say that the relevant factors of the other three parameters,
(i)–(iii) above, would not have symptomatic value. But they fail to uncover the
driving force behind cliticization.
Indeed, the flexibility of the semantic structure of many verbal predicates
allows them to alternatingly express a view from the outside or one from the
inside of the mental space of the speaker. The difference in conceptualization
that matters for cliticization cannot be satisfactorily captured in terms of
semantic dependency and integration of one event structure in another.
Variable cliticizability is not only a function of the semantic binding potential
of the matrix predicate, but crucially involves person, speaker and discourse
deixis, next to the spatiotemporal deictic organization of main and subordinate
clausal event, since the clitic’s deictic function primarily hinges on the speaker’s
vantage point.
In contrast, the demonstrative eso ‘that’ and the manner adverbial así ‘so’
do not necessarily relate an instance to the speaker’s personal viewing arrange-
ment, i.e. his own deictic setting at the moment of speech. The clausal reduction
operated by means of the clitic lo ‘it’, however, implies a coincidence, at least
partially between the speaker’s ground and that of the conceptualizer.
With respect to the underdetermined profile of the que ‘that’-subordinator
regarding the relation of the grounding of the subordinate clause with that of
the speaker, the disambiguating effect of using the clitic can be properly defined
in terms of mental space: what is at stake is the speaker’s ability and willingness
to integrate in his actual mental space (MSS) a propositional content that
originates in another mental space (MSC).
The data show that complement clause cliticization cannot be treated
straightforwardly in mere syntactic and discursive terms, essentially because the
verbs of propositional attitude and communication are susceptible of introducing
spaces that do not entirely depend on the MSS. The unstability of the cliticization
of the direct complement clause is due to the fact that the complement clause is
underdetermined regarding the construction and representation of mental
spaces. Hence, the necessity to analyze the relation between the matrix predicate
and the complement clause in cognitive terms that take into account the mental
space dimension.
The acceptability of the cliticization does not depend exclusively, nor even
primarily, upon the matrix verb, but it is essentially a function of mixed
grounding of the content of the complement clause. Instead of being confined
to the mental space of the overt conceptualizer, viz. the subject of the matrix
predicate, it also directly relates, at least partially, to the speaker’s deictic
organization.
This superposition of viewing arrangements is further manifested by the
possibility of adjoining speaker-oriented presuppositional and evaluative
modifiers to the complement clause, as well as by the possibility of further
elaborating upon the aspect and modality of the matrix predicate. Since these
elements pattern with an inferential reading of the complement clause, they
favor its cliticization.
The reverse holds for illocution or “direct speech” markers: by reflecting
part of the original wording, these markers attribute the clausal content
exclusively to the subject-conceptualizer, thus reducing the position of the
speaker to that of an external witness, a simple reporter or overhearer.
Within the limits of the present contribution, the difference between
cliticization and other forms of clausal reduction has only be alluded to in
passing.28 Yet, the basic findings set out in this paper suggest a track to follow in
further study of the coding of complex event structures. The mental space
approach has proved its utility for uncovering the interaction between speaker
deixis and conceptualizer deixis as far as clausal complement cliticization is
concerned. There is reason to believe that it will also turn out to be a powerful
analytic tool in the study of other variable phenomena observed at the level of
complex sentence structure and of larger text structures.
<DEST "del-n*">
Abbreviations
acc accusative neut neuter

cl clitic pl plural
cond conditional pres present
conj conjuction sg singularrefl reflexive
dat dative subj subjunctive
dem demonstrative Ø zero
fem feminine 1 first person
fut future 2 second person
imp imperative 3 third person
imperf imperfective * ungrammatical
ind indicative # inappropiate
inf infinitive ? of doubtful acceptability
masc masculine
Notes
* I owe a special word of thanks to Kaoru Horie, to Marjolijn Verspoor and to two anon-
ymous referees for their very helpful comments on an earlier version. Moreover, I especially
thank Marjolijn Verspoor for copy editing the text. For a discussion of French data, see
Delbecque 1998.
1. The examples encountered in written texts include the following:
(i) “Ya comprenderás que yo no voy a presentarme en Palacio, ni ganas. Téllez tam-
bién lo comprende.” [Marías 1996: 140–141]
‘You will readily understand that I will not present myself in Palace, no way.
Téllez also understands it.’ [it = that I will not present myself in Palace]
2. Outside the context of (4a), (4b) is, of course, not formally ungrammatical since the form
lo, which also represents the masculine clitic, can correspond to a masculine noun.
3. When these verbs take on other senses, they behave differently, e.g. admitir and reconocer,
which can both also mean ‘admit’ or ‘accept’ respectively.
4. “Ground” and “grounding” are used here in the Langackerian sense of “speech event, its
participants, and its setting” and “establishing the location vis-à-vis the ground of the thing
or process serving as the nominal or clausal profile” (cf. Langacker 1991).
5. Like the personal pronouns, the clitic form stems from a Latin demonstrative, and thus
shows a similar evolution from space to discourse deixis.
6. For further references on the idea that formal distance corresponds to conceptual
distance, see Haiman (1985: 106). The same notion is at the heart of various contributions to
the present volume, particularly those of Achard and Verspoor.
7. As seen above, clitic marking is not entirely predictable (cf. (8c–d), (10c), (11b), (12a)).
Moreover, it alternately occurs right before the finite verb or right after the non-finite verb
(i). Therefore, it is only partially comparable to inflectional marking (e.g. tense and mood).
(i) {Lo pienso hacer /Pienso hacerlo}
{it think-1sg.pres.ind do.inf/think-1sg.pres.ind do.inf-it
‘I consider doing it’
8. If the assimilation to a nominal were analogous, instead of inferential, a verb like pensar
‘think’, which does not take any nominal as direct object, would not allow cliticization either:
(i) Pienso {que tienes razón /*una razón}
think-1sg.pres.ind {that have-2sg.pres.ind reason/*a reason}
‘I think that you are right’
(ii) Lo pienso
it think-1sg.pres.ind
9. Unlike in English, the conjunction que ‘that’ of the Romance languages does not go back
to a former demonstrative, but stems from the Latin relative pronoun.
10. This is shown by predicate raising with implicative/successful manipulation verbs, e.g. hacer
‘make’ (Givón 1980:349) and subjunctive mood with verbs of emotional involvement, e.g.
esperar ‘hope’ (Givón 1980:366). A more general account of the isomorphism between the
semantic and the syntactic dimension of complementation is presented in Givón (1990: Ch. 13).
11. Needless to say, we are only concerned here with those causatives that take a finite
complement clause; verbs that only take an infinitive, e.g. tratar de ‘try’, are left out of
consideration.
12. The complement clause is the canonical manifestation of a propositional argument, but
not all complement clauses represent a propositional argument, nor are all propositional
arguments embodied in a complement clause. Next to deverbal and abstract nouns, even
some concrete nouns can occasionally stand for a propositional argument, e.g. (i) can mean
something like (ii) or (iii):
(i) They saw the stop sign
(ii) They saw they had to stop
(iii) They saw there was an instruction to stop
13. Hacer un acuerdo ‘make an agreement’ is marked as unfelicitous (#), rather than as
ungrammatical (*), since acuerdo ‘agreement’ can also refer to a document, in which case
hacer can mean ‘write up’, ‘produce’.
14. Some verbs, e.g. ocasionar ‘bring about’, impedir ‘prevent’, (im)posibilitar ‘make (im)pos-
sible’ even seem to be susceptible of extending their senses in both directions. Interpreted as
implicative or as volitive, they will allow cliticization, but not when they are conceived of as
strictly causative.
15. In Lo {vedó/vetó} (it {empeached/vetoed} it) the clitic lo is taken to be masculine and to
stand for el presupuesto (the budget), cf. examples (22) and (26a).
16. A demonstrative pronoun, e.g. esto ‘this’ or eso ‘that’, could be used instead of the clitic
(i). Unlike the clitic, the demonstrative does not necessarily occur in bound construction
with the verb, and need not have participant status (ii). This constructional difference has
far-reaching consequences as far as deixis and perspective are concerned. But this matter falls
outside the scope of the present contribution.
(i) Claro que he visto eso.
clear that have-1sg seen that
‘No need to say that I have seen that.’
(ii) Por eso, no he contestado.
for that nog have-1sg answered
‘Therefore, I did not answer.’
17. Notice that as nominal argument both pensar ‘to think’ and opinar ‘to hold a view’ only
take an indefinite (pro)noun (algo ‘something’, nada ‘nothing’, una cosa ‘a thing’).
18. Even when the subject is first person, this conceptualizer cannot be assimilated to the
speaker, since the latter anchors the utterance into his actual speech reality, independently
from the possibly different spatial and temporal setting in which the ‘I’-subject is situated.
19. Although here the “factive” operator el (hecho de) que ‘the fact (of) that’ cannot be used,
the conceptualization comes pretty close to the one yielded by emotive reaction predicates
(cf. Section 4, (19)).
20. Whether or not this implies the conversion of the representation of an event structure
into that of a proposition, is not clear to me. On this issue, see also Horie (1991).
21. This explains why the complement clause construction is restricted to a non-present
tense when this kind of predicate has a first person subject:
(i) Protesto {*que /: } me casan por fuerza
protest-1sg {*that/: } me marry-3pl by force
‘I protest: they marry me by force.’
22. When the conjunction is left out, i.e. when there is no nominalizing marker, the
complement clause is not functionally subordinated to the matrix verb and, hence, brings in
a space that falls outside its scope. When que ‘that’ is omitted, the introductory verb does not
really govern the clause (i). Due to its paratactic format, this event structure precludes
cliticization (ii).
(i) Una señora esperaba se le concediera pensión por ser
A lady hoped refl to-her concede-3sg.cond pension for be-inf
nieta legítima de J. M. Díaz
grandchild legitimate of J. M. D.
‘A lady hoped a pension would be conceded to her for being a legitimate grand-
child of J. M. D.’
[Marroquín, Blas Gil, 17 (p. 177), quoted in Cuervo 1994 under the entry of the
verb esperar ‘hope’]
(ii) *Lo esperaba
It-cl hoped-3sg
‘She hoped for it.’
23. Unlike direct discourse, indirect discourse is characterized by its monological function:
the paraphrastic reproduction only admits modes of expression found in hypotactic syntax.
Thus are excluded all forms that are typical of the dialogue, e.g. interrogatives, vocatives,
exclamatives, incomplete clauses. Also excluded are operations that are typical of the
independent clause, e.g. cleft, pseudo-cleft, topicalization and dislocation, or speech markers
such as reduplications, hesitations, discursive connectives and adverbial modifiers oriented
towards the hearer. Moreover, deictic markers, including the tenses, the pronouns and the
spatiotemporal adverbs, adjust themselves to the MSS, since the speaker-reporter mediates
the instantiation attributed to the conceptualizer C.
24. By extension it is possible to adjoin a complement clause even to verbs that qualify a
direct discourse or, else, are used intransitively, i.e. without further elaboration, e.g. (i). In
contrast to the verbs mentioned under (56), (57) and (58), these verbs are conceptually
“saturated”: the behavior they express is so specific that no further precisions are needed.
(i) chismear ‘to gossip’, chismorrear ‘to gossip’, cotillear ‘to gossip’, desembuchar
‘to relieve one’s feelings’, divagar ‘to digress’, epilogar ‘to summarize’, exagerar
‘to exaggerate’, farolear ‘to parade’, ironizar ‘to ironize’, monologar ‘to hold a
monologue’, moralizar ‘to moralize’, perorar ‘to hold a peroration’, polemizar
‘to polemize’, rajar ‘to talk rubbish’, rezongar ‘to grouse’, rugir ‘to grumble’,
tronar ‘to thunder’, vociferar ‘to roar’
Since they refer to verbal behavior as self-contained activity, and only take a clausal complement
in quite particular contexts (ii), they preclude its cliticization, as shown in (iii). The verbalization
that is being produced cannot be interpreted as a proposition, since there already is a
propositional dimension present in the verbal meaning structures, viz. the notions of gossip,
digression, summary, exaggeration, etc. are part of the corresponding predicates.
(ii) Gabriel Cañellas, presidente de Baleares, tronó que “la Unión Europea mantiene
a España de rodillas, castigada, siendo el hazmerreir del mundo.”
‘G.C., president of the Baleares, thundered that “the European Union keeps
Spain on its knees, castigated, as the butt of the world”.’ [El País 23.4.95/16]
(iii) *lo tronó
it-cl thundered-3sg
‘He thundered it.’
25. Needless to say, in (68a), both the references to Saturday (in the complement clause) and
to Sunday (in the embedded causal clause) are not computed from the vantage point of the
conceptualizer (Juan) but from that of the speaker who instantiates the main process type
(viz. the declaring (anunció)).
26. Hence, in Table 2, the cases of “mixed grounding” (partially [MSC Õ MSS ], partially
[MSC π MSS]) occupy position (ii.b) above the borderline that separates the cliticizable from
the uncliticizable complement clauses.
27. Adding an evaluative modifier to the subject, e.g. (i), favors the inference that the
proposition given in (73a) is also grounded in the mental space of the speaker (cf.59):
(i) Eva lo dijo en tono desaprobador
‘Eva it said in a disapproving tone.’
28. The picture has to be further completed by comparing cliticization with alternative types
of clause reductions, i.e. not only eso ‘that’ (i), así ‘so’ (ii), and the combination lo… así ‘it…
so’ (iii), but also que sí/que no ‘that yes/that no’ (iv):
(i) Juan cree eso (cf. (38))

‘Juan believes that.’
(ii) Así cree Juan (cf. (38))
‘So believes Juan.’
(iii) Así lo cree Juan (cf. (39c))
So it believes Juan
‘That is what Juan thinks.’
(iv) Pero yo creo que no
But I believe that-comp not
‘But I don’t think so.’
References
Achard, Michel. This volume. “Construal and complementation in French. The Perspective
dimensions”.
Boone, Annie. 1994. “La complétive: Un cas de nominalisation externe?” Travaux de
Linguistique 27: 29–42.
Boone, Annie. 1998. “La pronominalisation des complétives objet direct”. In Bilger, Mireille,
van den Eynde, Karel and Gadet, Françoise (eds), Analyse Linguistique et Approches de
L’oral. Leuven: Peeters, 103–114.
Cuervo, Rufino José. 1994. Diccionario de Construcción y Régimen de la Lengua Castellana,
continuado y editado por el Instituto Caro y Cuervo. Santafé de Bogotá.
Delbecque, Nicole. 1994. “Las funciones de así, bien y mal”. Revista de Lingüística Española
24 (2): 435–466.
Delbecque, Nicole. 1998. “Les limites de la pronominalisation: les complétives directes”. In
Ruffino, Giovanni (ed.), Atti del XXI Congresso Internazionale di Linguistica e Filologia
Romanza, Volume II. Tübingen: Niemeyer, 205–215.
Dik, Simon C. and Hengeveld, Kees. 1991. “The hierarchical structure of the clause and the
typology of perception-verb complements”. Linguistics 29: 231–259.
El País. Madrid, Spain.
Fauconnier, Gilles. 1984. Espaces Mentaux. Aspects de la Construction du Sens dans les Langues
Naturelles. Paris: Les Editions de Minuit.
Givón, Talmy. 1979. On Understanding Grammar. New York: Academic Press.
Language 4 (3): 333-377.
Givón, Talmy. 1985. “Iconicity, isomorphism and non-arbitrary coding in syntax”. In Haiman,
John (ed.), Iconicity in Syntax. Amsterdam/Philadelphia: John Benjamins, 187–219.
Givón, Talmy. 1990. Syntax. A Functional Typological Introduction. Volume II. Amsterdam/
Philadelphia: John Benjamins.
Haiman, John. 1985. Natural Syntax. Iconicity and Erosion. Cambridge: Cambridge University
Press.
Hengeveld, Kees. 1989. “Layers and operators”. Journal of Linguistics 25: 127–157.
Horie, Kaoru. 1991. “Cognitive motivations for event nominalizations”. CLS 27 (1): 233–245.
</TARGET "del">
Lakoff, George and Johnson, Mark. 1980. Metaphors We Live By. Chicago: Chicago Uni-
versity Press.
Langacker, Ronald W. 1991. Foundations of Cognitive Grammar. Volume II. Descriptive
Application. Stanford, CA: Stanford University Press.
Lyons, John. 1977. Semantics, Vol. 2. Cambridge: Cambridge University Press.
Marías, Javier. 1996. Mañana en la Batalla Piensa en Mí. Madrid: Alfaguara.
Marsé, Juan. 1970. La Oscura Historia de la Prima Montse. Barcelona: RBA Editores.
Moignet, Gérard. 1974. Etudes de Psycho-systématique Française. Paris: Klincksieck.
Rodríguez Espiñeira, Mª José. 1996. “Esquemas semánticos y sintácticos de los verbos de
percepción”. Paper read at the XI Congreso Internacional de la Asociación de Lingüística
y Filología de la América Latina, Las Palmas de Gran Canaria, 22–27 July 1996.
Suzuki, Satoko. This volume. “De dicto complementation in Japanese”.
Verspoor, Marjolijn. This volume. “Iconicity in English complement constructions:
Conceptual distance and cognitive processing levels”.
<LINK "ver-n*">
<TARGET "ver" DOCINFO
AUTHOR "Marjolijn Verspoor"
TITLE "Iconicity in English complement constructions"
KEYWORDS ""
SIZE HEIGHT "220"
WIDTH "150"
VOFFSET "4">
Chapter 8
Iconicity in English
complement constructions
Conceptual distance and cognitive
processing levels*
Marjolijn Verspoor
University of Groningen
1. Introduction
As the articles in this book have already shown, many terms are used to describe
the conceptual import or value of complementizers and complement construc-
tions. Horie distinguishes between event and proposition, based on Vendler’s
distinctions between object, event, and proposition, and among events those that
are directly and indirectly perceived; he also mentions distinctions between realis
versus irrealis, factive versus non-factive predicates, concrete object versus abstract
matter, and realis/ actuality versus irrealis/ potentiality. Beck distinguishes
complement clauses with the terms sentential, which refers to linguistic form,
and factive, which refers to conceptual import. Suzuki distinguishes between de
re (the domain of reality) and de dicto (the domain of speech).
Some of the problems that emerge in trying to relate these terms to each
other is that they are not mutually exclusive, relate to different ontological
levels, and may involve human interaction to different degrees. Some of these
terms denote the ontological status. Entities at their most ‘objective’ level are
entities in the world such as concrete objects, dynamic events or static states of
affairs.1 Entities may also be abstract matter and occur in a mental world. It may
even be argued that spoken words, that which is said in our communicative
world, the symbolic representation of an entity by means of verbal expression,
are ‘objects’ in a metaphoric sense and therefore also belong to this objective
level. Other terms refer to a more subjective level: terms like directly perceived
200 Marjolijn Verspoor
and indirectly perceived (=conceived) denote the kind of interaction between a

person and an entity. At this level, we may also include cognition, which, like
perception, refers to the type of evidence for what is known. At yet another level,
terms like realis versus irrealis and factive versus non-factive imply not only that
there may be some kind of direct or indirect interaction between a person and an
entity, but also that the person evaluates the conception of this entity.
What these terms do show and what the findings of the papers have in
common is that languages often code the type of entity (concrete versus abstract
matter, event versus state of affairs, or proposition)2 and the type of interaction
(direct or indirect) with such an entity. Also, languages may code how the entity
is evaluated (realis versus irrealis or factive versus non-factive). And at a more
pragmatic level, languages may code whether an entity is construed as a topic.
We may also conclude that linguistic forms may code conceptualizations
from different points of view. Achard has shown that the notion of perspective
is important as it distinguishes between the different conceptualizers that may
interact with an event or state of affairs. The conceptualizer is either the speaker
(the one who utters the whole sentence), who imposes his/her construal on the
whole situation, or the subject of the sentence, who may impose his/ her
construal on the event or state of affairs denoted by the complement clause. In
a similar vein, Delbecque has shown that integration of different mental spaces,
the speaker’s mental space and the subject’s mental space, may account for
complement clause clitization. Table 1 gives a summary of what I assume to be
the pertinent variables discussed so far.
Table 1. Summary of variables at a conceptual level discussed so far
linguistic realizations
subject main verb complement

construction
referring to persons, things, and relations in the world:
construal
relation construal relation
direct perception
indirect perception a person or thing
speaker/ (concrete/abstract)
conceptualizer a conceptualizer knowledge
assessment (true?) an event or state
mode of speech a proposition
Iconicity in English complement constructions 201
Other, partially overlapping, dimensions that need to be taken into account are
those imposed by the role of context. Suzuki argued that the degree of speaker’s
conviction regarding the reality of the information expressed in the comple-
ment, the speaker’s distance from the information, and the rapidity of the
speaker’s processing of the information are involved in the selection of a
complement structure, and Polinsky argues that information structure can
account for the phenomenon of variable agreement in Tsez.
In different types of languages, distinctions along these various dimensions
may be coded by means of a nominalization process and by finite versus non-
finite complement constructions, clitic climbing or agreement climbing. Other
languages code the difference by means of different complementizers, which
may get extended uses (as to in Japanese).
For languages like English, with the finite/ non-finite distinction, several
papers have supported Givón’s claim that iconic principles related to conceptu-
al distance are involved in the use of complement constructions:
The more the two events coded in the main and complement clauses share
their referents, the more likely they are to be semantically integrated as a single
event; and the less likely is the complement clause to be coded as an indepen-
dent finite clause (Givón 1990: 527).
And Horie concludes that “[t]he linguistic distance between expressions
corresponds to the conceptual distance between them” (1993: 216).
In this paper, I would like to go one step further. According to Langacker
“conceptualization resides in cognitive processing” and therefore, “our ultimate
objective must be to characterize the types of cognitive events whose occurrence
constitutes a given mental experience” (1991b: 2). If we accept that the main
verb’s meaning plays an important role in complement use, it is useful to relate
these meanings to actual mental states and processes. To do so, I will make use
of findings from neuro-biology and present a cognitive processing model with
variables that overlap to a great extent with the ones presented in Table 1. The
main differences are that in the cognitive modes expressed by the main verb, we
must also include ‘categorization’ and ‘feeling’. Moreover, we must make the
distinction between ‘indirect perception’, which may be considered ‘inference’,
and ‘knowledge’, which is based on rational thought, more clear. I hope to show
that if we include these extra cognitive modes and distinctions, a greater
number of English complement constructions can be accounted for.
I will first briefly review well-established findings concerning English
complement constructions. Then, after presenting an expanded list of variables
based on a neuro-biological cognitive processing model, I will relate these

conceptual constructs to the analysis of English linguistic complement con-
structions. More particularly, I will try to show that linguistic distance can
indeed be related to conceptual distance.
2. English complement constructions: An overview
Complementation has been a prolific area of research for many years and the
interface between the syntax and semantics of English complement clauses in
particular has been explored by a great many linguists from different schools
(cf. Akmajian 1977; Bolinger 1974; Borkin 1973 and 1984; Bresnan 1979; Croft
1991; Dik and Hengeveld 1991; Dirven 1989; Dixon 1984, 1995; Duffley 1992;
Givón 1980; Horie 1993; Karttunen 1971; Kirsner and Thompson 1976;
Langacker 1991a and b, Newman 1981; Riddle 1975; Quirk, Greenbaum and
Leech 1985; Wierzbicka 1988 to name a few).
Especially within the functionalist/cognitive paradigm it is generally
accepted that if one verb can be followed by more than one type of complement
clause, there must be some meaning difference between the sentences, usually
with a different pragmatic effect. Also several linguists (e.g. Bolinger, Dixon,
Langacker and Wierzbicka) have attempted to give general semantic character-
izations for each type of complement — that, -ing, to, zero and predicate
adjunct — even when they occur after different types of main verbs. Implicit in
many of the insightful observations is that there are iconic principles involved
in the use of the different complement constructions. A few studies that have
dealt explicitly with iconic principles in complement constructions are briefly
presented below.
Givón’s overall prediction is that the stronger the semantic bond between
the events described by the main and subordinate clause is, the more extensive
the syntactic integration will be of the two propositions into a single clause
(Givón 1993: 2). The grammatical coding devices signaling the degree of event
integration include (a) co-lexicalization, (b) case-marking, (c) verb morpholo-
gy, and (d) physical separation of the complement clause from the main clause.
To conclude he states: “The higher a main verb is on the scale of event integra-
tion, the more nominal-or less finite-will its complement verb be” (Givón
1993: 27). The scale of finiteness of verb-forms in English is shown in Table 2.
He claims that implicativity, co-temporality, and control for manipulative
(deontic) verbs determine the semantic bond. For example, in the following
Table 2. Finiteness scale (Givón 1993: 27)

least finite
lexical-nominal (removal)
bare-stem (remove)
-ing -infinitive (remov-ing)
to infinitive (to remove)
modal (may remove)
aspectual (having remov-ed)
most finite
examples (1)a–c the complement clause becomes more finite and less integrated
as the two events are further removed in time and space and consequently in
implicativity and control.
(1) a. He made her leave. two co-temporal events; strong
control and implicativity
b. He wanted her to leave. events are not co-temporal,
weaker control and not implicative
c. He told her that she should leave events are not co-temporal, weak
control, no direct contact
necessary, not implicative
For epistemic verbs he argues that similar principles play a role. When one sees
an event, the perception is co-temporal with the perceived event and therefore
by nature implicative and/ or factive. On the other hand, when one reflects
upon an event, all manner of temporal gapping may exist between the reflection
and the reflected-upon event (Givón 1993: 14).
(2) a. He saw her leave/leaving.
b. He believed her to have left.
c. He knew that she had left.
To conclude, Givón (1993: 23–27) argues that the various semantic properties
such as (a) successful causation, (b) intended manipulation, (c) preference/
aversion, (d) epistemic anxiety, (e) epistemic certainty/uncertainty, (f) quota-
tion under speaker’s perspective and (e) quotation detached from speaker’s
perspective interact with syntactic properties.
I agree with Horie (1993: 39) that Givón’s binding scale seems to work well
for the ‘manipulative verbs’ where the main clause subject exerts influence over
the complement clause subject, but a brief look at complement clauses after
‘perception’ and ‘cognition’ verbs indicates that, at least for English, the scales
are parallel rather than placed along one continuum because similar non-finite
forms of the verb and finite clause occur after both types of verbs. Moreover, in
a perception or cognition event, the main clause subject does not exert any
control over the complement clause subject.
(3) a. I made/saw him leave.
b. I persuaded him to leave/I believed him to be honest.
c. I told him that he should leave/I believed that he had left.
What seems to be at issue with perception and cognition verbs is whether an event
was directly perceived or not. Langacker’s (1991:450) discussion of such finite
versus non-finite complement clauses — that clauses, to infinitive structures,
and zero complements — also implies that such iconic principles are involved:
(4) a. Susan found that the chair was uncomfortable.
b. Susan found the chair to be uncomfortable.
c. Susan found the chair uncomfortable.
He agrees with Borkin (1973) that the different complement constructions

reflect semantic contrasts: (a) might be used if Susan had searched through her
files to learn the results of consumer tests, (b) would be more appropriate if
Susan had conducted such tests herself, and (c) implies that Susan had tried the
chair herself and had directly experienced the discomfort.
He argues that the type and number of grammatical elements reflect
semantic contrast between indirect and direct experience. The more grammati-
cal elements the complement has, the more abstract and propositional the
conceptual import of the complement is. The most indirect relationship
between Susan and the experience of discomfort is conveyed by the comple-
mentizer that and the fact that the clause is functionally elaborated with a finite
verb, viz. a grounding predication, which locates the event in time relative to
the current speech situation. Therefore, the conception imputed to Susan is
abstract and propositional.
A simpler and less abstract conceptualization is conveyed by the other two
constructions: (b) suggests an overall judgment that goes beyond immediate
experience as the verb be implies that she conceives of the chair’s uncomfort-
ableness as extending through some span of time. However, to conveys that
there is no specification of temporal coincidence: (c) is quite naturally inter-
preted as being induced by direct perceptual experience (cf. Newman 1981) as
the absence of be conveys an assessment of uncomfortableness tout court as
opposed to uncomfortableness extending through time.

Horie, who investigated the coding of directly and indirectly perceived
events cross-linguistically in perception and cognition verb complements,
concludes that as in English, in other selected European languages and in non-
European languages such as Modern Japanese, Korean, and Classical Japanese,
iconic principles tend to play a role: “Especially remarkable has been the cross-
linguistic tendency to use (non)finiteness and/or different complementizers to
encode the ontological distinction between Directly Perceived Events and
Indirectly Perceived Events” (1993: 225).
It is a well-established fact that many verbs taking finite/non-finite comple-
ments in English are verbs that denote an act involving an intentional state or
a human mental process: traditionally seen as the persuade versus believe class,
or the ‘deontic’ versus ‘epistemic’ class. ‘Deontic’ verbs are verbs of doing,
thinking, or planning. ‘Epistemic’ verbs are verbs of thinking, believing, or
knowing. Other complement-taking verbs are verbs of declaring, evaluating, or
feeling. I will also show that some verbs denote an act resulting in or establish-
ing a category. My own analyses (Verspoor 1990) have resulted in the following
general classes as illustrated in Table 3 below.
Table 3. Semantic classification of English complement-taking verbs

Type Definition Subtypes
Deontic Verbs expressing a ‘want’ or ‘wil- Verbs denoting the subject’s intentional
lingness’ towards a potential event state and simultaneous act resulting in
or state of affairs; also called ma- an event (make, crown) or verbs denot-
nipulative or causative verbs. ing a subject’s prior intention and a
possible future event involving the
subject or object (persuade, encourage,
plan).
Epistemic Verbs expressing a ‘belief ’ or Verbs denoting the mental state of
‘opinion’ towards a state of affairs knowing or thinking (know, think,
or a mode of perception. The believe) or verbs denoting a mode of
latter are also called evidential perception involving perceived or con-
verbs. ceived images (see, imagine, remember).
Other Verbs expressing a mode of Verbs denoting a mode of speaking
speech, a feeling, or assessment, (say, whisper), verbs denoting a feeling
also called de dicto, evaluative, towards an event (regret, enjoy, love,
and declarative verbs. abhor, endure), or verbs involving an
assessment or (verbal) judgment act
(consider, declare, prove).
It is also a well-established fact that English has the complement types shown in
Table 4 below.
Table 4. Syntactic forms of English complement constructions3

1. that clauses with indicative, subjunctive or modal verbs
2. to infinitives in various forms and with different syntactic properties
3. plain infinitives in various forms and with different syntactic properties
4. -ing forms with different syntactic properties
5. predicate adjunct constructions
Without reviewing detailed syntactic arguments here, I will follow several

linguists (cf. Givón 1980, Horie 1993) in claiming that (a) to (e) represents a
‘finiteness scale’ that strongly interacts with a ‘binding’ hierarchy of event
integration. I will assume that iconic principles are involved not only for
manipulative and perception and cognition verbs, but also for verbs of assess-
ment, feeling and evaluation (cf. Table 2). As mentioned earlier, Givón argues
that the scalar dimensions are along one single continuum. However, I do not
agree. In particular, Givón’s single linear scale fails to explicate the striking
parallel complement structures after the four different types of verbs: those
containing an element of will, those expressing the type of evidence for what
one says, those expressing an assessment or judgment, and those expressing a
feeling or attitude as exemplified in (5)–(8).
(5) Manipulative verbs (containing an element of will)
a. He told her that she should leave. (that clause)
b. He told her to leave. (to infinitive)
c. He made her leave. (zero)
d. He made her upset. (predicate adjunct)
(6) Verbs of perception or cognition
a. He knew/saw that she had left. (that clause)
b. He believed her to have left. (to infinitive)
c. He saw her leave/leaving. (zero)
d. He saw her leaving (-ing)
e. He saw her upset. (predicate adjunct)
(7) Verbs of assessment or judgment
a. He found that the chair was comfortable. (that clause)
b. He found the chair to be comfortable. (to infinitive)
c. He found the chair comfortable. (predicate adjunct)
(8) Verbs expressing a feeling

a. He regrets that she has to leave. (that clause)
b. He loves to swim. (to infinitive)
c. He loves swimming. (-ing)
d. He prefers his coffee black. (predicate adjunct)
For different types of verb, separate explanations have been given for these
iconic principles. For example, in the case of manipulative verbs (5), implicativ-
ity and co-temporality seem to play a role (Givón 1980). In the case of percep-
tion or cognition verbs (6), direct or indirect perception has been argued to play
a role (Akmajian 1977). For assessment and judgment complements (7) concep-
tual distance seems to play a role (Borkin 1973), and for verbs expressing a
feeling (8) factivity has been mentioned (Karttunen 1971).
However, even though all these types of verbs express cognitive modes, it is
not clear what these cognitive modes may have in common to account for the
four rather similar hierarchies in complement structures. Would it not be
possible that more general conceptual links may account for these similar
hierarchies?
What I will argue is that there are more general conceptual links that
underly the linguistic links and that these general conceptual links are experi-
entially motivated. Our cognitive processing of things, events and situations
may involve different levels of consciousness and interaction. When we
perceive, move, feel, or categorize, we interact directly with an ‘object’. When
we conceive of an ‘object’ that is not in the here and now, we interact indirectly
with it. And finally, when we conceive of an ‘object’ in terms of its symbolic
representation, i.e. in terms of a linguistic expression, we do not interact with the
‘object’ directly or indirectly, but with the symbolic representation of that object.
Of course, our ‘objective’ experience does not translate directly into how we
linguistically represent this objective experience. This is due to the fact that
humans have the “capacity to structure or construe the content of a domain in
alternate ways” (Langacker 1991b: 5) and in the case of finite and non-finite
complement clauses, I will assume that “[w]hen we use a particular construc-
tion or grammatical morpheme, we thereby select a particular image to struc-
ture the conceived situation for communicative purposes (Langacker 1991b: 12).
According to Langacker (1991b, Chapter 3), when an ‘object’ is construed
as a concrete or abstract thing, it is linguistically realized as a noun and when an
‘object’ is construed as an event or a state of affairs, it is linguistically realized as
a clause containing a verbal. An event or state of affairs that is grounded is
linguistically realized as a finite clause. An event or state of affairs that is not

grounded is linguistically realized as a non-finite clause. My claim is that the
English language, in the case of non-grounded events or states of affairs, codes
even more subtle differences in construal. When an event or state of affairs is
accessed and construed as experienced directly (as denoted by the main verb),
it is linguistically realized as a predicate adjunct, plain infinitive, or -ing
complement clause. When the event or state of affairs is accessed or construed
as experienced indirectly, it is linguistically realized as a to infinitive comple-
ment clause. And finally, when the event or state of affairs is accessed or
construed in terms of its symbolic representation, it is realized as a that comple-
ment clause.
To make clear the subtle differences between directly and indirectly
experienced in terms of symbolic representation, I will briefly discuss Edelman’s
biological theory of consciousness (Edelman 1989) as it pertains to making the
neuro-physical relation between the various mental states and processes
towards events and situations more explicit.
3. A cognitive processing model
Edelman’ theory of consciousness (1989) may give us some facts and hypothe-
ses about the neurological and physical processes that take place in our brains.
Edelman’s theory takes (1) intentional motor movement, (2) perception in the
widest sense of the word, (3) attributing value, and (4) categorization as the
basis upon which higher order categorization, conceptualization, and finally
symbolic presentation (language) is built. Basically, there is a hierarchy of levels
of cognitive processing, related to the reception of signals, each dependent upon
and interacting with the other, but also each one further removed from the here
and now and therefore more abstract.
Simply put, the ‘levels of consciousness’ are the following. They are cogni-
tive processes interacting with an ‘object’. The ‘object’ of these cognitive
processes can be things or persons, events, states of affairs, or propositions
(defined as linguistic realizations denoting an event or state of affairs) and can
be in the outside world or within the mind. The type of interaction with an
‘object’ can be at different levels of directness:
Level 1, which Edelman calls ‘the level of perceptual categorization,’ includes
perception, the need for motor movement, categorization and attributing value.
At this level the interaction between a human and an ‘object’ is direct. In other
words, a human can directly see, feel, do (= physically interact with), or assess
an ‘object.’
Level 2, which Edelman calls ‘primary consciousness’ involves relating present
‘input’ to events/states of affairs not in the here or now. The type of interaction
is indirect. For example, in the case of planning or believing, a human can
interact with something directly and relate it mentally to another event or state
of affairs not in the here or now.
Level 3, which Edelman calls ‘higher order consciousness’ involves symbolic
representation, propositional thought and verbalization. There is no cognitive
interaction between subject and object. For example, in the case of saying or
knowing something, a human interacts with the words and propositions that denote
events or states of a airs, but not with the events or states of a airs themselves.
Actually, Edelman’s model has some correspondences with the one concerning
the relationship of different entities to Time, Space and the World by Vendler
(1967), the one concerning intentionality developed by Searle (1983)4 and the
folk model of the mind developed by D’Andrade (1987). However, one main
difference with these models — one that should not surprise anyone conversant
with cognitive linguistic theory — is that it makes explicit that ‘attributing value’
and ‘categorization’ are as basic as ‘perception’ and ‘intentional movement.’
I will now describe the neurological theory in more technical detail. As
mentioned above, primary consciousness involves four interdependent process-
es that interact with each other almost simultaneously. The four processes are
listed below together with the English verbs illustrating each process:
1. Intentional motor movement [do],
2. Perception [see],
3. Attributing value [feel],
4. Categorizing [assess].
The cognitive processes of perceiving objects, attributing value to them,
categorizing them, and intentional states related to smooth motor movement
involve a direct, and therefore contemporaneous, interaction between subject
and ‘object’. It must be kept in mind that these cognitive processes are intrically
related through reentrant mapping and occur almost simultaneously; therefore,
they are not clearly delineated. Simply put, seeing something involves at least
some intentional motor movement (directing my eyes towards an object),
attributing value and categorizing it. Intentional smooth motor movement, in
turn, would also necessarily involve seeing where one is going, which in turn
would involve attributing value and categorizing.
Value refers to the most basic sensations of good/bad. It is an evolutionarily

determined assignment of adaptive or homeostatic function and it can be
altered by synaptic changes resulting in preference of one state over another.
For example, at birth an infant will have ‘value’ concerning body warmth, food,
and touch. The infant is able to perceive (see, hear, feel) and to categorize his
perceptions (good, bad) (far, near). The major premise upon which this
development is built is that of reentrant mapping. That which is perceived and
categorized is matched with ‘value.’ For example at the most basic level, the
infant may learn that food is ‘good.’ However, according to the theory of neural
group selection (TNGS), categorization of object and events is relative, depend-
ing on cues, context, and salience. Therefore, categories are not fixed but can be
changed by the current state of an animal. For example, we all know that food
is not always found ‘good’ by an infant, who may reject it and go on crying.
Primary consciousness emerges from the interaction of the value-category
memory and ongoing (current or present) perceptual categorization. The
categorization of smooth movements and the development of short-term
memory are the bases of ordering in space and time. For primary consciousness
to appear, two sets of structures are required: a special memory system concep-
tually relating past categorizations to value and reentrant connections relating
the conceptual system to the current immediately categorized signals. In other
words, memory is recategorization. Possessed with primary consciousness
alone, animals are bound to the small time intervals mediated by short term
memory; they have no concept of the past.
Before proceeding to level 2, we must realize that one of the most extraordi-
nary properties of animals that are capable of perceptual categorization is that
they can generalize. After an encounter with a few instances of a particular
category under learning conditions, they can recognize a great number of
related but novel instances. For example, an infant may learn soon that the
sound of nearing footsteps means someone is on the way to take care of her. She
will learn to recognize her caretaker’s voice and face. Basically, recognition is
recategorization: to match new instances with past instances. Neurobiologically
what happens is that the selection of neuronal groups in one map can be
coupled to the selection of neuronal groups in another, yielding a classification
couple (cf. Edelman 1989: 48).
Much of matching new instances with past instances is conceptual, not
linguistic. According to Edelman, concepts are not necessarily properties of
language, nor should they be regarded as merely mental images or even worse
‘language of thought’. He clarifies the notion of concepts as follows:
An animal capable of concepts is able to identify a particular thing or action

and control its future behavior on the basis of that identification in a more or
less general way. It must act as if it could make judgments based on recognition
of category membership or integrate ‘particulars’ into ‘universals.’ This
recognition rests not just on perceptual categorization (although a concept
may have a highly sensory content) but, to some degree, must also be relational.
It can connect one perceptual categorization to another event in the absence of
the stimuli that triggered these categorizations (1989: 141).
Level 3 involves language. With the evolution in hominids of the vocal tracts, of
the supralaryngeal space, and of Broca’s and Wernicke’s areas a means for
symbolic categorization emerged. Also one of the chief concepts to emerge
along with language is that of the ‘self ’ not merely as a biological individual but
as a subjective identity demarcated from conspecifics. Edelman suggests that
when sufficient phonology emerges, words and sentences become symbols for
concepts. Symbolic categorization and speech permitted concepts of past and
future to be developed along with time-independent models of ‘self ’ and
‘world.’ With such models, the more developed forms of higher order con-
sciousness became possible: The entire speech system evolved by way of
affective and observational learning in a social scene. The actual signs and
symbols of language are arbitrary, but they serve to link conceptual categoriza-
tion to a specialized system of phonological recognition and memory.
According to this theory, there is a semantic bootstrapping via linkage of
phonological and lexical systems to already existent conceptual systems.
Finally, when a lexicon is sufficiently developed, the conceptual apparatus
may recursively treat and classify the various productions of language them-
selves — morphemes, words, and sentences — as entities to be categorized and
recombined without any necessary further reference to their initial origin or to
their bases in perception, learning and social transmission. (This is related to
what is known in cognitive linguistics as metaphorical and metonymical uses of
linguistic expressions.)
The theory holds that both during evolution and in the individual, concepts
precede language and meaning. They are driven by the perceptual apparatus
and are constructed by the brain as it models its own classes of activity. Con-
cepts are nonetheless coherent and can correspond to things and motions. In
their most elaborate form, they may serve as the bases for image schemata
(object, motion, barrier, container) summarizing a variety of general physical
situations. Once embellished and fortified by the enhanced powers of a lexicon,
they retain these metaphorical properties but can also recombine and modify
each other in an explosive fashion.

To conclude this brief, simplified overview of Edelman’s theory, we can
distinguish among different types of ‘objects’: concrete and abstract things/
persons, events, states of affairs, and propositions. These ‘objects’ are interacted
with by a combination of interacting cognitive processes: perception, the need
for motor movement, evaluation and categorization. These processes may occur
at different levels of interaction between conceptualizer and ‘object’: directly (in
the here and now), relational (not in the here and now), and symbolically (not
directly but through symbolic representation of the ‘object’).
Table 5 gives an overview of these cognitive modes and their respective
objects, with a solid-lined arrow heuristically standing for direct physical
interaction with an ‘object’, a dotted-lined arrow standing for mental interac-
tion with an image of an object, and a lined-arrow standing for a rational
interaction with the symbolic representation of an ‘object’.
Table 5. A cognitive processing model based on Edelman’s findings

human cognitive modes kind of interaction objects
perception
need for motor movement a person, thing, event, or
evaluation (good/bad) state in the here and now
categorization
perception/belief (inference)
need for motor movement (planned) mental image of a person
evaluation (good/bad) (remote) or thing, an event or state
categorization (tentative) not in the here or now
symbols (words) representing

rational thought persons, things, event and
speech states that may or may not
be in the here or now
In the next section, I will show how these conceptual distinctions may relate to
the linguistic analysis of English complement constructions.
4. From cognitive processes and ‘objects’ to verbs

and complement constructions
A typical English sentence with a complement construction can be characterized

as follows: Within a certain pragmatic context and with a certain communicative
intention, a speaker/conceptualizer utters a sentence containing a
main subject + main verb + complement construction
and where the complement construction, in turn, consist of the following ele-
ments:
(complementizer) + (subordinate subject) + subordinate verb + (X) + (Y))
By using such a grammatical construction at the sentence level, the speaker

selects a particular image to structure the conceived situation (Langacker 1991b,
12), but as Achard (this volume) has pointed out, this construal relation may
allow the main subject, too, to be construed as the conceptualizer of the event
or state of affairs denoted by the complement construction.
The main verb typically denotes a cognitive process (want, see, think, believe,
feel, consider, and so on) or a mode of speaking (claim, say, whisper, and so on).
Often, verbs denoting cognitive processes profile one particular cognitive aspect
more than another, e.g. see profiles the act of perception and enjoy profiles the
mental state of attributing value; however, many of these verbs denote mental
events or states involving a combination of processes, but they may be used
metonymically in the sense that they profile one aspect more than another.
For example, avoid would involve an intentional state towards a smooth
motor movement (not explicitly named in the complement construction),
which often involves perception, and naturally involves attributing value and
categorization. Especially, ‘value’ is profiled somewhat as the meaning of avoid
clearly implies a ‘do not want’ attitude to the event named by the complement.
Also find in a structure like I find him nice could involve elements of all four:
some intentional action, perception, categorization and assessment. Actually, it
is not surprising that a lexical verb like find in particular has developed an array
of metonymically related senses, which in turn may be compatible with
different types of complement constructions, as with each different type of
complement, a different sense may be profiled.
The meaning of the main verb also relates to the kind of interaction (direct,
indirect/relational, symbolic) between main subject and event or state of affairs
denoted by the complement clause. For example, the central meanings of verbs
such as see, feel, make, and find imply that a ‘subject’ directly interacts with an
‘object.’ The central meanings of verbs like want and believe imply that the
‘object’ is not in the here and now of the subject. The central meaning of say
implies that words are uttered, but even though the ‘subject’ directly interacts
with these words, these words may describe an event or situation that is not in
the here and now of the subject.
The complement construction denotes the type of ‘object’ with which the
subject interacts. In the case of complement constructions, the ‘objects’ are not
single entities such as concrete things or persons or entities construed as things,
as these would be denoted in English by nouns or noun-like constructions, but
they are typically events or states of affairs, which can be construed along at
least two dimensions. First of all, the event/ state of affairs itself can be con-
strued along a continuum from dynamic to static. Secondly, they can be
construed along a continuum of temporality. They may occur at the exact same
moment that the main event occurs (co-temporality) or at a moment before or
after the main event (non-co-temporality). Temporality, in turn, is related to
the type of interaction between ‘subject’ and ‘object’: direct, indirect, or
symbolic. In the next sub-sections, these different dimensions as they pertain to
English complement constructions will be treated separately.
4.1 Verbs denoting ‘direct interaction’ and -ing, plain infinitive

and predicate adjunct complements
As Langacker (1991a: 442–449) has argued, the different types of complement

constructions can be regarded as conventionalized symbolic units, each
imposing a different construal on the event or state of affairs and therefore each
more or less compatible with main verbs.
For example, just as in a finite construction with an -ing verb like ‘John is
building a canoe’, a non-finite -ing construction imposes a construal of an
imperfectivized bounded event, as the beginning and end of the event are not
within the scope of the conceptualizer’s viewing frame. Since the -ing symboliz-
es an event construed as ongoing, it is natural that it be compatible with main
verbs that express co-temporality of main event and subordinate event.
The difference between an -ing and plain infinitive construction is that the
latter imposes a construal of a perfected bounded event, where the beginning
and end of the event coincide with the scope of the viewing frame; after
perception verbs, the plain infinitive also expresses a co-temporal event. The
difference between -ing and plain infinitive in sentences such as I saw the ship
sinking or I saw the ship sink is that the -ing profiles a part of the event and the
plain infinitive the event as a whole. After causative verbs such as make, the
event expressed in the plain infinitive construction is not co-temporal, but one
that immediately (without time interval) follows the causative event (cf.
Verspoor 1997a).
Along the continuum of dynamic to static, the predicate adjunct construc-
tion is the most static. It denotes a static state of affairs rather than an event and
expresses the state or category in which a ‘thing’ is. As mentioned above, a
sentence such as Susan found the chair uncomfortable means that Susan makes
a judgment, which applies to the moment at which the judgment is made,
which is then naturally co-temporal and induced by direct experience (cf.
Verspoor 1997b).
What -ing, plain infinitive and predicate adjunct constructions have in
common is that they denote co-temporal events and states of affairs; therefore,
it is not surprising that they are compatible with main verbs that denote
cognitive processes which entail that a subject directly interacts with an ‘object’:
those denoting direct perception of world or mental objects, intentional motor
movement involving an ‘object’, attributing value to an ‘object’, and categoriza-
tion of an ‘object’.
(9) I saw him walking Subject perceives event directly.
I saw him walk. Subject perceives event directly.
I made him walk. Subject causes event directly.
I avoided walking. Subject attributes negative value
to (any instantiation of) event.
I enjoy walking Subject attributes positive value
to (any instantiation of) event.
I found the chair comfortable Subject attributes categorical
status to thing.
4.2 Verbs denoting ‘indirect interaction’ and to infinitive complements
Whereas the -ing, plain infinitive and predicate adjunct constructions prototyp-
ically denote co-temporal events and states of affairs, the to infinitive may have
a future sense, especially after manipulative verbs as in He wanted to go.
However, after an epistemic main verb, the potential event or state of affairs
denoted by the to-infinitive may very well be co-temporal with the main verb or
even past in combination with have, as in I believe him to be/to have been in his
room. I have argued elsewhere (Verspoor 1998) that to with its original direc-
tional sense (source, path, goal schema) has lent itself well for not only the
‘futurity’ sense but also for another type of relational sense, a direction in which
thoughts go, namely towards a location in a mental space. This sense would be
very similar to the one of a modal auxiliary like may, which according to
Langacker (1991a: 277–278) may have a ‘present’ but ungrounded sense in
potential reality.
I argue that it is this relational sense of to that makes it especially compati-
ble with verbs denoting ‘relational thinking,’ which is not necessarily based on
language but on concepts, relating an event/ state in the here and now to
another event/state in the not here and now. On the one hand, a to infinitive
can express an event involving ‘controlled future behavior’ as in I wanted/
promised to go and on the other hand it can express a state of affairs or event
that is not directly perceived, but ‘connected’ to some perceived stimuli as in I
believe him to be in the house.
Verbs like want and promise express an intentional state, usually one that
includes a ‘want’ or ‘prior intention’, that is directed towards a future event.
The want occurs at the moment the main verb refers to, but the event or state
of affairs expressed by the to infinitive occurs at a moment future to that.
Normally speaking, the event or state of affairs is a direct result of the want or
prior intention. For example, (10b) doesn’t make much sense because the
semantics of force does not include a want on the part of the agent of the action
expressed by the to infinitive. In other words, the to infinitive may express
‘controlled future behavior.’
(10) a. I wanted to go; therefore, I packed my bags, booked my ticket and
left. (= I went because I wanted to)
b. ??I wanted to go, and I went because I was forced to.
Verbs like believe also express that some relational thought is involved. An
example of such relational thought would be seeing particular sandwich
ingredients on the counter and relating them to someone fixing a sandwich.
Even if we have not directly seen the person eating a sandwich, we may relate
the directly perceived evidence (sandwich ingredients on the counter) to an
event of a person fixing and eating a sandwich. What happens in Edelman’s
terms is that the directly perceived event is matched with other such events in
the past and generalized. From previous experience, the subject knows what
sandwich ingredients on the counter means; someone has fixed/is fixing/will
fix a sandwich. This event is not directly perceived or experienced, but inferred
and therefore it is not directly related to the here or now. It can occur now, but
not here, or it can occur not now and not here. Since there is always a degree of
speculation involved, it is also possible that the link made to what cannot be
seen is incorrect. In any case though, it is possible to conceive of such a possible
or hypothetical event or state.
A less prototypical use of a to infinitive construction is found after some
verbs of speaking such as claim and profess as in He claims to be an artist. I
believe that what is at stake here is that the to infinitive profiles the subjective
construal from the subject’s perspective. The subject wants others to believe
that he is an artist; with such a sentence the speaker, by attributing the subjec-
tive construal to the subject, indicates that he may not share this belief.
4.3 Verbs denoting ‘symbolic representation’ and that clauses
Because of the fact that it is not grounded, a to infinitive construction imposes

a more subjective construal on the contents of a clause than a that complement
construction would. A that clause, which per definition is finite, serves to
objectify the conception of the proposition expressed (Langacker 1991a: 447)
and therefore is conceptually the most distant of the complement constructions.
I argue that a that clause represents propositional thought or speech, which may
occur at the highest level of consciousness, and which is a ‘symbolic’ representa-
tion of an event or state of affairs. Therefore, when using a sentence with a that
clause, a speaker construes the main event as an event in which the subject does
not directly interact with the event or state of affairs itself, but with the symbolic
representation thereof.
Therefore, verbs of speaking and verbs of thinking in propositions are
compatible with that clauses. Also, some verbs of perception, which may be
used in the sense of cognition, are sometimes followed by that clauses. It is also
not surprising that ‘pure’ speech verbs like whisper and shout occur only
exclusively with that clause complements. Only those verbs of speaking that
may have a ‘directive’ sense or those that contain an element of speaker’s will to
direct a hearer to a certain belief are compatible with to infinitive constructions.
(11) a. He whispered that he was going to leave.
b. *He whispered to leave.
c. He ordered that I should leave.
d. He ordered me to leave.
4.4 Summary: verbs denoting cognitive processes

and their complements
Admittedly, I have only shown a limited semantic analysis of the different

complement constructions, and to be convincing, each type will have to be
analyzed in much more detail, showing how peripheral members of a construc-
tion are related to the more central ones and under what pragmatic conditions
and with what subtle meanings they are used. However, I do think I have shown
that there are four general classes of verbs denoting cognitive processes that are
compatible with complement constructions: manipulation, perception /
cognition, assessment/ judgment and feeling. I do not believe it is accidental
that these coincide quite neatly with basic cognitive processes (intentional
action, perception, attributing value and categorization) recognized by Edelman
in his theory of consciousness.5
We might conclude that, besides being quite consistent with findings in
cognitive linguistic research, Edelman’s theory of consciousness clearly shows
that some mental states and processes are more basic than others, one involving
signals in the here and now, the other matching present input with projected or
potential reality not in the here and now. Finally, only at the highest levels, in
rational thought and speech, language is involved. The verbs denoting these
cognitive processes are followed by different types of complement construc-
tions, each of which imposes a different kind of construal on the event or state
of affairs. (See Table 6.)
What is interesting is that verbs from almost every semantic class of verb —
sometimes only one or two verbs from that class — may be used with almost
every complement type (ignoring different syntactic properties).
The following examples (repeated from (5)–(8)) give an overview, and with
each type of verb, similar iconic principles seem to be involved. As the discus-
sion shows, the observations of which are very much in accord with
Wierzbicka’s observations (Wierzbicka 1988), the greatest conceptual distance
occurs with that clauses and the least with predicate adjuncts, plain infinitives,
and -ing forms. At the intermediate level, where events and states of affairs are
not interacted with directly, the to infinitive occurs.
(12) Manipulative verbs (containing an element of will)
a. He told her that she should leave. (that clause) level 3
b. He told her to leave. (to infinitive) level 2
c. He made her leave. (zero) level 1
d. He made her upset. (predicate adjunct) level 1
Table 6. Verbs denoting level and mode of cognitive processes

Levels of processing Time / space of ‘object’ Verbs denoting such levels
1. perceptual categorization here/now
perception see
intentional state towards action want
(good/bad) value feel
categorization assess
recategorization remember
hallucination imagine/dream
2. primary consciousness not here/not now
conceptualization plan
generalization believe
3. higher order consciousness not here/not now
symbolization say
(propositions) think
know
According to several linguists (cf. Wierzbicka), a that clause (12a) may occur
even when the indirect object and subject of the complement clause are not co-
referential. The to infinitive does imply that there is some direct interaction
between the subjects of the main clause and complement clause, but the event
‘to leave’ is not directly caused by the telling and is somewhat future to the
moment of ‘telling.’ In other words, the leaving event is not in the here and now
of the subject of the sentence. This construction may also occur with verbs that
have no speech involved as in He persuaded her to go or even He nudged her to
go. Such verbs may not occur with a that clause. The plain infinitive (12c) and
predicate adjunct (12d) constructions denote events and states of affairs that are
in the here and now of the subject of the sentence. There is definitely some
direct interaction between the subjects of the sentence and the complement
clause. As the causation is direct and immediate, there is also some direct
interaction between the subject of the sentence and the event itself (her leave) or
state of affairs (her upset). The subject of the sentence directly perceives this
event or state of affairs. To conclude, the conceptual distance between main
clause subject and event or state of affairs in the complement clause is greatest
with the that clause and shortest with the plain infinitive or predicate adjunct
constructions. Also the experiential distance between the subject of the main
clause and the subject of the complement clause is greatest with the that clause.
(13) Verbs of perception or cognition

a. He knew/saw that she had left, (that clause) level 3
b. He believed her to have left. (to infinitive) level 2
c. He saw her leave/leaving. (zero) level 1
d. He saw her upset. (predicate adjunct) level 1
A that clause (13a) may occur even when the subject has not directly perceived
or experienced the event expressed in the complement clause. The speaker
construes the situation as follows: the subject knows about this event through
a process involving rational thought or hearsay. The to infinitive implies that
there is some directly experienced evidence upon which the claim for the truth
of the event or state of affairs expressed in the complement is based. The plain
infinitive, -ing and predicate adjunct construction imply that the subject has
perceived the event or state of affairs expressed in the complement directly.
These complement constructions occur exclusively with verbs of perception like
watch or look at that naturally imply direct perception and do not have a
metaphorical sense of ‘understanding’. To conclude, the conceptual and
experiential distance between main clause subject and event or state of affairs in
the complement clause is greatest with the that clause and shortest with the
plain infinitive or predicate adjunct constructions.
(14) Verbs of assessment or judgment
a. He found that the chair was comfortable. (that clause) level 3
b. He found the chair to be comfortable. (to infinitive) level 2
c. He found the chair comfortable. (predicate adjunct) level 1
With verbs of assessment or judgment, the that clause (14a) may occur when
the subject has not directly perceived or experienced the state of affairs ex-
pressed in the complement clause. The speaker construes the situation as
follows: the subject knows about this state of affairs through a reasoning process
involving secondary evidence. Rather ‘objective’ reasoning verbs like calculate,
demonstrate, and reason occur with that clauses exclusively. Like with percep-
tion verbs, the to infinitive may imply that there is some directly experienced
evidence upon which the claim for the truth of the event or state of affairs
expressed in the complement is based, but it may also imply that the subject
tentatively places an item in a category. The predicate adjunct construction
implies that the subject has had some direct interaction with the subject of the
complement and mentally put the item into a category. To conclude, the
conceptual and experiential distance between main clause subject and event or
state of affairs in the complement clause is greatest with the that clause and
shortest with the predicate adjunct constructions.
(15) Verbs expressing a feeling
a. He regrets that she has to leave. (that clause) level 3
b. He loves to swim. (to infinitive) level 2
c. He loves swimming. (-ing) level 1
d. He prefers his coffee black. (predicate adjunct) level 1
Verbs expressing a feeling or attitude towards an event or state of affairs seldom

occur with that clauses. Regret is an exception, but it has often been noted that
a structure like to say is then left implicit or understood and the emotion is
construed as not directly or strongly affecting the subject of the main clause.
The to infinitive occurs after verbs that have in their meaning next to ‘feeling’
an aspect of a ‘want’ towards a potential event. The -ing construction implies
that the subject has directly experienced such an event and that the subject
attributed (and still attributes) to the event a particular value. The same goes for
the predicate adjunct construction, which however expresses a categorial state
of a thing. Again, the that clause expresses the greatest conceptual and experien-
tial distance between subject and event or state of affairs expressed in the
complement.
If this provisional analysis were to prove adequate, it means that iconic
principles indeed play a role in the use of English complement constructions.
However, I have shown that the scalar dimensions are not along one single
continuum, as argued by Givón, but along four parallel continuums after four
different types of verbs: those expressing a willingness, a type of evidence,
assessment and feeling.
5. Conclusion
In this article I have argued that conceptual distance as realized linguistically in

English by the use of different complement constructions may very well have an
experiential basis. Those verbs that denote cognitive processes at level 1,
involving direct interaction between ‘subject’ and ‘object’ are compatible with
the ‘least finite’ complement constructions. Those verbs that denote cognitive
processes involving relational thinking and therefore indirect interaction
between ‘subject’ and ‘object’ are compatible with complement constructions
at a slightly higher level on the finiteness scale. Those verbs that denote cogni-
tive processes involving propositional thought and no interaction between
‘subject’ and ‘object’, are compatible with complement constructions highest on
the finiteness scale.
Whether every language iconically codes conceptual distance remains of
course to be seen, but what the papers in this book seem to indicate is that if
conceptual distinctions are coded, the distinctions made will be made along one
or more of the following dimensions:
1. The ontological status of the ‘object’. As several models (e.g. Vendler and
Frazyngier and the one presented here) have shown the ‘object’ may be con-
strued as a concrete or abstract thing or person, as an event or state of affairs,
which may be construed along a continuum of dynamic or static, or as a
proposition, defined here as the ‘symbolic’ (linguistic) representation of an
event or state of affairs. In English, this object may be realized as a nominal or
clause. Commonly used notions such as ‘event’ and ‘proposition’ are directly
related to this dimension.
2. The mode of cognitive interaction with the ‘object’. The modes of interaction
that are quite commonly mentioned are ‘direct and indirect causation’ (both
related to intentional action) on the one hand and ‘perception’ and ‘cognition’
on the other. Here I have argued that ‘attributing value’ and ‘categorization’ are
also modes of cognitive processing that play a role in complement selection. It
may be argued that the distinction between ‘factive’ and ‘non-factive’ comple-
ments is related to this dimension as specifying the mode of interaction may
specify the type of evidence for what is stated in the complement construction,
making the event or state of affairs ‘factive’ or ‘non-factive’. In English, the
mode of cognitive interaction may be realized as a lexical verb.
3. The construal relation. When a speaker within a given context with a given
communicative intention reports on an event or state of affairs, involving a
‘subject’ and an ‘object,’ which in turn is an event or state of affairs, the speaker
may impose a more or less ‘subjective’ image on the whole utterance from his/
her own perspective or from the subject’s perspective. The notions of perspec-
tive, viewing arrangement, subjectivity, and objectivity are related to this
dimension. Part of this ‘subjective’ construal dimension, however, is the
perceived physical and/or conceptual distance between ‘subject’ and ‘object’.
4. The level of cognitive interaction between ‘subject’ and ‘object’. The physical
and/or conceptual distance between ‘subject’ and ‘object’ may be construed as
involving direct, indirect/ relational interaction, or independent (no direct
<DEST "ver-n*">
interaction with event or sate of affairs). Commonly used notions such as direct
perception, indirect perception, and cognition are directly related to this
dimension. Notions such as co-temporality, implicativity are natural symptoms
of direct interaction. Notions such as futurity and the realis/irrealis/potentiality
distinctions are different manifestations of indirect/relational cognitive interac-
tion. Also the distinction between de re and de dicto is related to the distinction
between various degrees of interaction versus no interaction.
Notes
* I would like to thank Kaoru Horie, Bernard Comrie, and Michel Achard especially for their
insightful comments on earlier versions of this paper. Of course, all shortcomings remain my
own responsibility.
1. I will use the term ‘entity’ in Langacker’s sense: “A maximally general term: anything one
might conceive of or refer to for analytical purposes. An entity need not be discrete,
separately recognized, or cognitively salient” (1991a: 548)
2. I will define the term ‘proposition’ as a verbalized thought or utterance, that which is said
or thought in terms of symbolic representations.
3. The term ‘construction’ is here used in Langacker’s sense: “a symbolic unit that is both
complex and schematic” (1991b: 17)
4. Searle’s model served as a basis for my analyses of complement taking verbs in my
dissertation (Verspoor 1990). I now prefer Edelman’s model because it accommodates
categorization acts and is more explicit about the role of attributing value.
5. The editor of this volume even suggested that some Whorfian effects may have played a
role as Edelman is a native speaker of English!
References
Akmajian, A. 1977. “The complement structure of perception verbs in an autonomous

syntax framework”. In P. W. Culicover, T. Wasow and A. Akmaijan (eds), Formal
Syntax. New York: Academic Press, 427–60.
Bolinger Dwight L. 1974. “Concept and percept: Two infinitive constructions and their
vicissitudes”. World Papers in Phonetics: Festschrift for Dr. Onishi’s Kizyu. Tokyo:
Phonetic Society of Japan, 65–91.
Borkin, Ann. 1973. “To be and not to be”. Papers from the 9th Regional Meeting of the Chicago
Linguistic Society, 44–56.
Borkin, Ann. 1984. Problems in Forms and Function. Norwood, NJ: Ablex.
Bresnan, Joan W. 1979. Theory of Complementation in English Syntax. New York: Garland.
Croft, William. 1991. Syntactic Categories and Grammatical Relations. Chicago: University of
Chicago Press
D’Andrade, Roy. 1987. “A Folk model of the mind”. In Cultural Models in Language and
Thought (eds) Dorothy Holland and Naomi Quinn. Cambridge: Cambridge University
Press, 112–148.
Dik, Simon C. and Hengeveld, Kees. 1991. “The hierarchical structure of the clause and the
typology of perception-verb complements”. Linguistics 29: 231–59.
Dirven, René. 1989. “A cognitive perspective on complementation”. In Jaspers, D. et al.
(eds), Sentential Complementation and the Lexicon: Studies in Honor of Wim de Geest.
Dordrecht: Foris, 113–39.
Dixon, Robert M. W. 1995. A New Approach to English Grammar, on Semantic principles
Oxford: Clarendon Press.
Dixon, Robert M. W. 1984. “The semantic basis of syntactic orientations”. Proceedings of the
Annual Meeting of the Berkeley Linguistics Society10. (583–595)
Duffley, Patrick. 1992. The English Infinitive. London: Longman.
Edelman Gerald M. 1989. The Remembered Present: A Biological Theory of Consciousness.
New York: Basic Books. Inc.
Language 4.3. 333–77
Givón, Talmy. 1990. Syntax II. Amsterdam: John Benjamins.
Givón, Talmy. 1993. English Grammar II. Amsterdam: John Benjamins
cognitive perspective. Unpublished dissertation. University of Southern California
Karttunen, Lauri. 1971. “Implicative verbs”. Language 47: 340–58
Kirsner, Robert S. and Sandra A. Thompson. 1976. “The role of pragmatic inference in
semantics: A study of sensory verb complements in English”. Glossa 10.2: 200–40
Langacker, Ronald W. 1991a. Foundations of Cognitive Grammar II. Stanford: Stanford
University Press
Langacker, Ronald W. 1991b. Concept, Image, and Symbol. Berlin: Mouton de Gruyter
Newman, John. 1981. The semantics of raising constructions. Unpublished dissertation,
University of California, San Diego
Quirk, Randolph, Sydney Greenbaum, Geoffrey Leech and Jan Svartvik. 1985. A Compre-
hensive Grammar of the English Language. London: Longman
Riddle, Elizabeth. 1975. “Some pragmatic conditions on complementizer choice”. In Gross-
man, R. E. et al. (eds.), Papers from the 11th Regional Meeting of the Chicago Linguistic
Society, 467–74
Searle, John R. 1983. Intentionality. Cambridge: Cambridge University press
Vendler, Zeno. 1967. Linguistics in Philosophy. Ithaca, N. Y.: Cornell University Press.
Verspoor, Marjolijn. 1990. Semantic criteria in complement selection. PhD dissertation.
University of Leiden, The Netherlands
Verspoor, Marjolijn. 1997a. “The story of -ing: A subjective perspective”. In Pütz, M. and R.
Dirven. (eds.), The Construal of Space in Language and Thought. Berlin: Mouton de
Gruyter, 417–454.
</TARGET "ver">
Verspoor, Marjolijn. 1997b. “Predicate adjuncts and subjectification”. In Verspoor. M. et al.

(eds.), Lexical and Syntactical Constructions and the Construction of Meaning. Amster-
dam: John Benjamins Publishing Company, 433–449.
Verspoor, Marjolijn. 1998. “To infinitives”. In L. de Stadler (ed.), Issues in Cognitive Linguis-
tics. Berlin: Mouton de Gruyter, 505–526.
Wierzbicka, Anna 1988. The Semantics of Grammar. Amsterdam. Benjamins
<TARGET "gloss" DOCINFO
AUTHOR ""
TITLE "Glossary"
KEYWORDS ""
SIZE HEIGHT "220"
WIDTH "150"
VOFFSET "4">
Glossary
absolutive: in an ergative language, the required for the well-formedness of

case form marking the subject of an in- structures containing this verb.
transitive verb and the direct object of a
transitive verb. autonomy/dependency: an expression
that does not presuppose another for its
actant: a syntactic dependent of a predicate full manifestation is autonomous; the
which is subcategorized for by that predi- more structures are needed for its pro-
cate; roughly, the argument of a verb. cessing, the more dependent it will be, e.g.
a clitic is more dependent than a nominal
adjunct: a syntactically and/or semanti- or clausal complement; most independent
cally optional category which in principle are the subject and the main clause.
could be removed without affecting well-
formedness. case-marking: one of the coding devices
in the grammar of complementation. For
agreement: the grammatical phenomenon example in a finite clause, the subject of
by which the appearance of one item in a the complement verb will have nomina-
sentence in a particular form requires a tive case, but in a non-finite clause often
second item which is grammatically linked accusative case.
with it to appear in a particular form.
clitic(ization): a clitic is an unstressed
agreement climbing: the grammatical pronoun that occurs only in bound con-
phenomenon by which the main predi- struction with the verb and is marked for
cate in a finite clause agrees with a con- gender, number and case; cliticization
stituent within the clause embedded in represents the reduced verb-adjacent ex-
the matrix. pression of an unfocussed, predictable
and familiar argument.
alienation: discord between the speaker’s
intended message and the ostensible co-lexicalization: one of the coding de-
meaning of words s/he uses. vices in the grammar of complementa-
tion (cf. Givón 1993: 23). This term refers
argument: an NP bearing a specific syn- to the use of the complement verb right
tactic and/or semantic relation to a verb; after the main verb as in let go which
the overt or covert presence of this NP is expresses strong subject control and little
228 Glossary
object agentivity. For example in She let as the subject may know about this event
go of him, the subject’s control is much through a third party.
stronger than in She let him go. In She let
him go, let go is considered a co-lexical- conceptualizer (role): the central partici-
ized verb. pant role with propositional attitude
predicates (e.g. think, believe, hope), con-
comparative semantic typology: a theo- strued as grammatical subject.
retical framework proposed by Hawkins
(1986) which captures cross-linguistic conceptualizing subject construction:
variability in form-meaning mapping. grammatical construction in which the
main clause subject acts as a conceptualizer
complement construction or clause: a with respect to the complement clause.
finite or non-finite clause functioning as
direct object of verbs like make, see, be- deixis: deixis is the generic term for the
lieve, encourage, say, and so on. In cogni- experiential structure relating to the
tive theory, the complement construction grounding instances; those categories that
is assumed to make semantic contribu- are considered ‘deictic’ situate the scenes
tion to the sentence. and their components in terms of prox-
imity with respect to the grounding in-
complementizers: grammatical elements stances, and always with respect to the
that link complement clauses to main speaker, who constitutes the ultimate
clauses such as subordinate conjunctions point of reference of the speech event.
(like that), verb particles (like to), prepo- Deictic anchoring operates at different
sitions (like for) or verb endings (like levels of coding: space, time and dis-
-ing) that express some logical or seman- course, referring both to the coded and to
tic connection between a dependent and the coder as reference point for locating
independent clause. In cognitive theory entities and scenes.
they are assumed to contribute meaning
to a complement construction. deontic: traditionally, deontic modality is
concerned with the necessity or possibili-
composition: the process by which lexical ty of acts performed by morally responsi-
items are combined to form complex ex- ble agents. (Lyons 1977: 823). In
pressions. Verspoor’s paper, the term “deontic
verbs” refer to verbs that express actions
conceptual distance: in complementation or mental states causing actions or states
research, conceptual distance refers to the in others. For example, in I made him
degree of interaction between the subject leave, the verb make is a deontic verb.
of the sentence and the situation or event
expressed by the complement clause. For distance: refers to viewing distance or to
example in I saw the ship sinking, the conceptual distance. Viewing distance
subject I directly interacts (by means of means the distance between the perceiver
perception) with the event denoted in the and the percept; it is the smallest when
complement clause. In I know that the the percept is taken in “close-up”. At a
ship sank, no direct interaction between more abstract level, the notion of distance
the subject and the event itself is implied is used metaphorically along the following
Glossary 229
lines: the more categorial distinctions are event: a dynamic state-of-affairs which
shared by (members of) two paradigms, happens in real spatio-temporal dimen-
the smaller the conceptual distance is be- sion and is perceptually observable
tween them, e.g. the distance between the through sensory organs.
pronominal and the clausal paradigm is
greater than that between the nominal evidentiality:
and the pronominal paradigm. a. coding of the source of information
(in the narrow sense).
domain de dicto: domain of speech. b. coding of the speaker’s attitude toward
information (in the broad sense).
domain de re: domain of reality.
exceptional case marking (ECM): a par-
egocentric viewing arrangement: an ar- ticular phenomenon in case assignment,
rangement in which the objective scene is under which a clausemate verb fails to
expanded beyond the region of perceptu- govern its argument and a matrix verb
al optimality to include the observer (or governs it instead.
analogously the conceptualizer) and his
immediate surroundings (Langacker experiential basis for language: this is the
1991: 543–566). cognitive linguistic notion (cf. Lakoff
1987) that the senses of linguistic symbols
embedding: a structure in which one can in many cases be directly related to
constituent is contained inside another human experience. For example, it is fact
constituent. that in many languages a word like up
usually has a positive sense and down a
epistemic: traditionally, epistemic modal- negative sense as in He has moved up in
ity is concerned with matters of knowl- the world versus He feels down today. The
edge, belief or opinion rather than fact idea is that these senses are natural in that
(Lyons 1977: 793 and 681–82). In Ver- they relate to human experience. Some-
spoor’s paper, the term “epistemic verbs” one who stands straight up has more self-
refer to verbs that express mental actions confidence, power, and positive feelings
or states to denote the type of evidence than someone who is in a “down” posi-
the subject has for what is expressed in tion (e.g. head down or lying position),
the complement clause. For example, in which is associated with subjection or
the sentences I saw him leave, I believe depression. In Verspoor’s paper, the term
him to have left or I know that he has left experiential basis refers to the notion that
the verbs see, believe, and know are episte- the degree of cognitive processing (direct
mic verbs. perception and inference versus a linguis-
tically realized proposition) is related to
ergative: in an ergative language, the case the complement constructions used.
form marking the subject of a transitive
verb (referred to as ergative) is different factive/non-factive: a semantic distinction
from the case form marking the subject of between two types of complement-taking
an intransitive verb and the direct object of predicates proposed by Kiparsky and
a transitive verb (referred to as absolutive). Kiparsky (1970); a factive predicate (e.g.
regret) includes speaker’s presupposition
230 Glossary
that the proposition expressed in the The locates an object in space, and is
complement is true, whereas a non-fac- (present tense) locates the event in time
tive predicate (e.g. think) does not in- (here the moment of speaking).
clude such a proposition.
iconic principles: in linguistics this refers
finite clause: an expression that desig- to the fact that a “sign” in some way may
nates an epistemically-grounded process resemble “conceptual import”. For exam-
that is further construed as being unitary ple, the order in which clauses occur
(i.e. a single event or a single situation) often reflects the order in which events
(Langacker 1991: 543–566). occur (e.g. He ran across the street and got
hit by a car) or the number of words may
finite verb: a verb that shows agreement reflect the degree of “distance” (e.g. very
and tense. For example, in “John is walk- wordy sentences are usually more polite
ing”, the verb “is”, which is marked for than very short sentences). In com-
present tense and third person singular is plementation research, iconic principles
a finite verb. refer to the notion that the less integrated
two clauses are, the more morphemes
focus: the semantic component of a and words are used and the more concep-
pragmatically structured proposition tual distance there is between two clauses.
whereby the assertion differs from the
presupposition. implicativity: in the grammar of comple-
mentation this traditional term refers to
grounding: a semantic function that con- the notion of successful manipulation: If
stitutes the final step in the formation of a that which is said in the main clause
nominal or a finite clause. With respect refers to an event that took place, then
to fundamental “epistemic” notions (e.g. that which is expressed in the comple-
definiteness for nominals, tense/modality ment clause must have taken place. (cf.
for clauses), it establishes the location vis- Givón 1993, Vol 2, 30) For example, in
à-vis the ground of the thing or process She made him leave we will assume that if
serving as the nominal or clausal profile she made him leave he actually left, but in
(Langacker 1991: 543–566). She asked him to leave he may or may not
have left. Make is therefore considered an
grounding predication: a term used by implicative verb. Implicativity has often
Langacker (1987). The term “predica- been argued to play a role in complement
tion” refers to the semantic pole of a selection. Verspoor argues that
linguistic expression, which could be a “implicativity” is in itself not a “selection
morpheme, word, phrase or more com- feature” but should be seen as a correlate
plex expression. The term “ground” of direct perception or causation.
refers to the speech event, its participant,
and its setting. The term “grounding (in)direct discourse: when a cognized
predication” refers to a morpheme or event structure comes in the form of a
word that helps locate a situation or event syntactically independent clause, this
in time and place. For example, in the paratactic representation is called “direct
sentence “The boy is writing” there are discourse”; this cover term subsumes
two grounding predications: the and is. sayings (“quoted speech”), thoughts and
Glossary 231
other sorts of mental processes attributed marked for tense and person. Non-finite
to a conceptualizer. A cognized event verb forms in English are present partici-
structure can also have the form of a ple (writing), past participle (written),
syntactically dependent clause: this “indi- plain infinitive (write) and to infinitive
rect discourse” is introduced by a subor- (to write).
dinating conjunction (that-complement-
izer or wh-complementizer). noun class: same as gender, a grammati-
cal category by which nouns are divided
isomorphism: concept in which identity into mutually exclusive groups requiring
of form is regarded to reflect similarity in different agreement forms on determin-
function. ers, modifiers, verbs.
landmark: a secondary figure in a predi- objective (construal): an entity is con-

cation whose relation to the primary strued objectively to the extent that it
figure or trajector is the meaning of the functions asymetrically as the object (as
expression. opposed to the subject) of conception in
an optimal viewing arrangement. (Lang-
matrix agreement: the grammatical phe- acker 1991: 543–566)
nomenon by which the main predicate in
a finite clause agrees with an embedded oblique actant: an actant of a verb which
sentential constituent. is not subject or direct object.
medial verb form: a non-finite form of optimal viewing arrangement: a situation

the verb which is used to indicate that the in which the roles of the observer and the
predication formed by it is subordinate to observed are fully distinct, the latter being
a higher predication. sharply differentiated from its surround-
ings and situated in a region of maximal
monoclausal/biclausal: a complex event perceptual acuity. (Langacker 1991: 543–
structure can be construed alternatingly 566)
with one or with two verbal predicates.
The latter structure, which is the biclausal participanthood: unlike entities that pro-
one, presents a higher degree of concep- vide the setting of an event structure,
tual elaboration than the monoclausal those that have participant status occupy
structure. positions of the setting and interact with
each other. Prototypical participants are
multiple topics: the occurrence of several small and mobile. For example, the
topics within one sentence. participanthood of the complement
clause is untypical.
nominalizer: a nominal word or suffix
which serves to turn a clause into a nomi- perspective: an aspect of construal that
nal argument of a sentence. subsumes such factors as vantage point
orientation, and subjectivity/objectivity.
non-finite verb: a verb that does not (Langacker 1991: 543–566)
show agreement and tense. For example
in John is writing, the verb writing, is not
232 Glossary
profile: the entity that an expression speaker control: with respect to a com-
designates; in a clause, the profile is that plement construction, indicates the
set of participants and relations that are speaker’s direct conceptual relation with
included by the speaker in the utterance the complement process.
and which the expression is intended to
designate. subject: the conceptualizer of the propo-
sitional attitude predicate is construed as
proposition: an epistemically-grounded its grammatical subject.
process as conceived of beyond any spa-
tiotemporal setting, i.e., as a statement or subjectification: a semantic extension in
assertion consisting of a predicate and a which an entity originally construed objec-
set of arguments. tively comes to receive a more subjective
construal. (Langacker 1991: 543–566)
prospective/retrospective: the viewing
direction on the temporal axis as defined subjective (construal): an entity is con-
in function of the chosen reference point strued subjectively to the extent that it
(the moment of speech or that of the ma- functions asymetrically as the subject (as
trix predicate in a complex event struc- opposed to the object) of conception.
ture): forward projection into futurity is (Langacker 1991: 543–566)
prospective, backwards projection into
anteriority is retrospective. subschema: a particular, more specific,
instance of an abstract schematic mor-
psychological distance: distance that the pheme; for instance, baseball, basketball,
speaker feels towards information. and football are all subschemas of the
more abstract category “ball”.
raising: the appearance of a linguistic ele-
ment in a higher clause than is seman- theory of territory of information: theory
tically appropriate. developed by Kamio (1997) (Information
falling into the speaker’s territory of
realis/irrealis: the distinction between what information is characterized as informa-
actually happens in the real world (typically tion that the speaker considers close to
past event) and what remains a possibility him/herself)
in the real world (hypothetical event).
topic: the semantic component of a prag-
relational landmark: an entity in a predica- matically structured proposition which
tion whose relationship to the primary the proposition is about. If a proposition
clausal figure or trajector constitutes the has more than one topic, they will subdi-
meaning of that expression but which is vide into main and non-primary.
not itself an individuable entity or event-
participant. topicalization: the grammatical phenom-
enon which has as its output a structure
sentence-focus: a structure in which no in which a certain element is marked as
pragmatic presupposition is formally topic, either by fronting or by morpho-
evoked, and the entire proposition corre- syntactic means.
sponds to an assertion.
</TARGET "gloss">
Glossary 233
trajector: the primary figure in a predica- verb morphology: one of the coding
tion, the most salient entity in a clause. devices in the grammar of complemen-
The subject. tation. The complement verb may be
finite or non-finite (e.g. participle or
upgrading/downgrading: according to
infinitive form).
the construal chosen, focal prominence is
differently assigned to the different com-
viewing arrangement: the full circum-
ponents of an event structure, including
stances of a speech event together with
its viewing arrangement; e.g. using a
the relationship between the ground and
clause, a nominal or a clitic, respectively,
the situation under description, including
yields the upgrading, neutralization or
both content (e.g. whether some facet of
downgrading of the time deixis.
the ground is itself being described) and
vantage point: the position from which a construal (in particular vantage point and
scene is viewed, including the orientation objectivity/subjectivity).
of the viewer-conceptualizer. (Langacker 1991: 543–566)
<TARGET "NI" DOCINFO
AUTHOR ""
TITLE "Name Index"
KEYWORDS ""
SIZE HEIGHT "220"
WIDTH "150"
VOFFSET "4">
Name Index
(Excluding self-citation)
A F
Achard, M. 6, 66, 88, 125, 146, 153, Fauconnier, G. 7, 169, 173, 175,
182, 193, 197, 200, 213, 223 186, 197
Authier, J. M. 2, 8, 66, 88 Frajzyngier, Z. 2, 6, 9, 28, 29, 33,
35, 36, 43, 51, 52, 54
B
Beck, D. 1, 7, 91, 199 G
Bolinger, D. 1, 8, 12, 13, 28, 202, Givón, T. 2–4, 9, 13, 27, 29, 71, 88,
223 95–97, 115, 116, 118, 120, 132,
Borkin, A. 1, 8, 202, 204, 207, 223 147, 155, 156, 161, 162, 194,
Bresnan, J. 1, 8, 202, 223 197, 201–203, 206, 207, 221,
Bybee, J. 3, 5, 8, 9, 54 224, 227, 230
C H
Comrie, B. 12, 27, 28, 60, 61, 81, Haiman, J. 10, 38, 44, 51, 54, 95,
86–88, 129, 131, 144–146, 223 120, 155, 193, 197
Coulmas, F. 34, 52, 54, 55 Hawkins, J. 16, 29, 228
Croft, W. 3–5, 9, 202, 224 Hengeveld, K. 164, 197, 202, 224
Hess, T. 121, 123, 126, 127,
D 129–131, 133, 140, 143, 144,
D’Andrade, R. 209, 224 146–148
Davis, P. 126, 134, 136, 139, 140, Hooper, J. 1, 9, 13, 29
144–147 Horie, K. 34, 52, 54, 80, 86, 88, 91,
Delbecque, N. 1, 7, 149, 180, 193, 120, 122, 134, 143, 146, 147,
197, 200 168, 193, 195, 197, 199,
Dik, S. 164, 197, 202, 224 201–203, 205, 206, 223, 224
Dirven, R. 3, 9, 202, 224
Dixon, R. M. W. 2, 9, 16, 26, 28, I
29, 202, 224 Inoue, K. 30, 37, 54
Duffley, P. 2, 9, 202, 224
J
E Josephs, L. 11, 17, 30, 34, 54
Edelman, G. 208–212, 216, 218, 223,
224
</TARGET "NI">
236 Name index
K P
Kamada, O. 52, 54 Palmer, G. 3, 9, 10, 143
Kamio, A. 3, 9, 37, 38, 54, 55, 232 Polinsky, M. 6, 8, 91, 117, 120, 201
Karttunen, L. 1, 9, 13, 30, 202, 207,
Q
224
Quinn, C. 30, 45, 54, 55, 224
Keenan, E. 75, 88, 132, 147
Kemmer, S. 5, 9 R
Kim, N. K. 11, 19, 20, 30, 54 Ransom, E. 2, 10
Kiparsky, P., and C. Kiparsky 1, 10, Riddle, E. 1, 10, 202, 224
13, 19, 30, 55, 229 Rizzi, L. 2, 10
Kirsner, R. 1, 10, 36, 55, 202, 224 Rosenbaum, P. 1, 10, 11, 30
Kudoo, M. 43, 55 Ross, J. R. 1, 10, 14, 30, 146, 147
Kuno, S. 11, 17, 29, 30, 34, 54, 55 Ruwet, N. 94, 120
S
L
Saunders, R. 126, 134, 136, 139,
Lakoff, G. 2, 3, 10, 161, 197, 229
140, 145–147
Lambrecht, K. 3, 10, 60, 71–73, 75,
Searle, J. 209, 223, 224
77–81, 85–88
Shinzato, R. 17, 30
Langacker, R. 2, 3, 10, 66, 88, 91,
Slobin, D. 37, 50, 56
92, 95–99, 106, 108, 109, 115,
Suzuki , S. 6, 27, 28, 31, 91, 120,
118–120, 122, 123, 125, 127,
184, 198, 199, 201
131, 132, 138, 140, 142, 147,
156, 161, 193, 197, 201, 202, T
204, 207, 213, 214, 216, 217, Taylor, J. 3, 10, 132, 140, 148
223, 224, 229–233 Thompson, S. 1, 3, 10, 36, 55, 89,
Lee, M. S. 11, 23, 30, 54 129, 131, 132, 146, 147, 202,
Lehmann, C. 2, 10 224
Li, C. 128, 132, 147 Tomasello, M. 2, 10
U
M
Uehara, S. 5, 10
Makino, S. 17, 24, 30, 53, 55
Martin, S. 9, 11, 21, 23, 30, 49, 50, V
53, 55 Van Valin, R. 3, 10
Maynard, S. 38, 39, 52, 55 Vendler 13, 31, 33, 56, 199, 209,
McCawley, N. A. 11, 17, 30 222, 224
McGloin, N. 22, 30 Verspoor, M. 3, 7–9, 27, 28, 31, 91,
Munro, P. 44, 55 117, 120, 168, 193, 198, 199,
205, 215, 223, 224, 225,
N 228–230
Nakau, M. 11, 30 Vološinov, V. N. 38, 56
Noonan, M. 2, 10
W
Wierzbicka, A. 2, 10, 202, 218, 219,
O
225
Ohori, T. 5, 10, 22, 29, 30
Okamoto, S. 53, 55
<TARGET "SI" DOCINFO
AUTHOR ""
TITLE "Subject index"
KEYWORDS ""
SIZE HEIGHT "220"
WIDTH "150"
VOFFSET "4">
Subject index
A
action 30, 124, 125, 134, 162, 165, components 60, 86, 100, 158, 228,
178, 183, 189, 211, 213, 216, 233
218, 219, 222 composition 122, 124, 125, 139, 228
activated 60, 92 concept 1, 8, 13, 28, 91, 163, 210,
agreement 3, 6, 17, 59–80, 82–89, 211, 223, 224, 231 (See also
120, 136, 139, 140, 145, 154, percept)
164, 166, 194, 201, 227, 230, conceptual distance 7, 31, 113, 114,
231 117, 120, 156, 193, 198, 199,
alienation 38, 51, 54, 227 201, 202, 207, 218, 219, 221,
aspectual categories 84 222, 228–230 (See also
assertive 9, 13, 29 (See also factive, psychological distance)
non-assertive) conceptualizer 6, 92, 95, 97–101,
autonomy 9, 178, 179, 181, 182, 227 103, 104, 112–114, 118, 151,
153, 166, 169, 171, 172–174,
B 176–182, 186–188, 190, 191,
biclausal 118, 162, 165, 166, 181, 194–196, 200, 212–214,
183, 184, 189, 231 228–230, 232, 233
binding 2, 9, 27, 29, 89, 116, 118, conflation 152, 173
120, 153, 161, 162, 166–168, construal 6, 88, 91, 92, 96–99, 101,
190, 197, 203, 206, 224 103, 106–110, 112, 115–119,
semantic binding 2, 190 140, 146, 150, 151, 181, 197,
200, 208, 213, 214, 217, 218,
C 222, 224, 231–233
categorization 4, 7, 201, 208–213, control 109, 110, 143, 184, 202–204,
215, 218, 219, 222, 223 211, 227, 228, 232
clitic 87, 117, 134, 149–157, copula 25–27, 65, 143
159–165, 169–171, 173, 174,
180, 186, 188–191, 193, 194, D
201, 227, 233 de dicto 6, 9, 28, 31, 33–37, 39, 43,
cliticization 7, 117, 119, 120, 149, 45, 46, 48–52, 54, 120, 184, 198,
152–154, 156, 158–163, 199, 205, 223, 229 (See also de
166–174, 176, 177, 178–191, re)
193–196, 227 de re 28, 33, 34, 44, 52, 199, 223,
comparative semantic typology 16, 229 (See also de dicto)
228
238 Subject index
deictic 33, 100, 132, 135, 143, 144, finite clauses 67, 112, 122, 125,
146, 152, 169–171, 173, 179, 132–134, 136, 139, 140
180, 189, 190, 191, 195, 228 finiteness 27, 125, 134, 139, 202,
deixis 146, 154, 155, 157, 164, 205, 206, 222 (See also non-
169, 188, 190, 191, 193, 194, finite)
228, 233 focus 10, 33, 34, 60, 72–74, 78–80,
deontic verbs 228 (See also 84, 86–89, 91, 122, 153,
epistemic verbs) 178–180, 185, 230, 232
dependency 155, 162, 178, 181, 182, focus-marking 72
190, 227 fronting 69, 72, 81, 83, 85, 89, 232
direct discourse 174, 179, 180, 184,
185, 187, 195, 230 (See also G
indirect discourse) grammaticalization 5, 9, 10, 29, 30,
directly perceived 15, 28, 199, 204, 54, 56
205, 216, 220 (See also ground 97, 100, 106, 109, 135, 143,
indirectly perceived) 152, 157, 161, 165, 168, 169,
171, 173, 190, 193, 230, 233
E grounded 96, 100, 118, 119, 125,
Egocentric Viewing Arrangement 127, 133, 140, 141, 151, 152,
(EVO) 98, 99, 108, 113, 114, 155, 156, 161, 176, 177, 180,
116, 154, 184, 185, 187, 188, 184, 186, 188, 196, 207, 208,
196, 229 217, 230, 232
(See also Optimal Viewing groundedness 125, 134
Arrangement (OVA)) grounding 96, 97, 116, 118, 119,
epistemic verbs 203, 229 (See also 140, 151, 155, 169, 170, 174,
deontic verbs) 177, 178, 180, 181, 182, 190,
ergative 62, 69, 86, 87, 227, 229 191, 193, 196, 204, 228, 230
event 2, 7, 13, 15–24, 26, 28, 29, 34,
37, 47, 54, 95–97, 114, 118, H
122–124, 127, 129, 132, 137, hierarchy 2, 5, 9, 27, 29, 54, 120,
138, 140, 141, 143, 152, 157, 197, 206, 208, 224
158, 161, 162, 163, 165, 169,
170, 174, 176–179, 182, 184, I
186, 188–191, 193, 195, 197, iconic principles 201, 202, 204–207,
199–209, 211, 213–223, 228–233 218, 221, 230
(See also proposition) iconicity 4, 7, 31, 120, 155,
evidentiality 38, 54, 56, 229 197–199 (See also
Exceptional Case Marking 82, 229 isomporphism)
identifiable 60, 122, 158, 170, 171
F implicative 9, 13, 30, 159, 166, 168,
factive 13, 19, 20, 125, 162, 163, 194, 203, 224, 230 (See also
195, 199, 200, 203, 222, 229, non-implicative)
230 (see also assertive, non- independence 2, 44, 95, 96, 116,
factive) 118, 161
feeling 39, 201, 205–207, 218, 221
Subject index 239
independent 36, 44, 65, 75, 80, 82, N

95, 96, 109, 110, 113, 114, 161, nominalization 7, 14, 17, 22, 27, 29,
166, 175, 181, 182, 189, 195, 201, 54, 91, 120–126, 129, 130, 133,
211, 222, 227, 228, 230 134, 136, 137, 138, 140–142,
indirect discourse 150, 157, 158, 144–146, 156, 201
180, 183, 184, 187, 189, 195, of finite clauses 125
231 (See also direct discourse) nominalizer 7, 15, 18–20, 22,
indirectly perceived 15, 28, 199, 200, 27–29, 37, 52, 54, 86, 121–125,
205 (See also directly perceived) 135, 137, 141, 161, 231
information structure 3, 6, 10, 59, non-assertive 13 (See also assertive)
60, 65, 71, 79, 80, 83–88, 201 non-factive 13, 199, 200, 222, 229,
intransitive 62, 63, 73, 77, 126, 131, 230 (See also factive)
134, 140, 144, 147, 180, 227, non-finite 12, 13, 28, 67, 96, 109,
229 122, 125, 134, 138, 193, 201,
irrealis 13, 16, 20–22, 24–26, 28, 204, 205, 207, 208, 214, 227,
133, 136, 143, 199, 200, 223, 228, 231, 233 (See also finite)
232 (See also realis) non-implicative 13 (See also
isomorphism 6, 35, 51, 194, 197, implicative)
231 (See also iconic principles, non-referring 75, 76
iconicity) noun class 6, 59, 61, 231
L O
landmark 122, 127, 144, 231, 232 objective 99–101, 104, 106–112,
(See also trajector) 115–117, 119, 148, 199, 201,
relational landmark 127, 144, 232 207, 220, 229, 231 (See also
levels of consciousness 207, 208 subjective)
lexical 7, 18, 19, 21, 23, 24, 47, 73, ontological levels 199
120–122, 124, 126–129, 135, Optimal Viewing Arrangement (OVA)
137–139, 144, 146, 189, 211, 98, 99, 107, 108, 112–114, 116,
213, 222, 225, 228 231 (See also Egocentric
local 65–67, 74, 76–79, 82, 84, 85, Viewing Arrangement (EVO))
87
Logical Form 84 P
Long-Distance 59, 64–80, 82–85, 87 participant 98, 99, 106, 109,
112–114, 123, 127, 152, 155,
M 157, 170, 194, 228, 230–232
mental space 7, 118, 151, 152, 166, participial 62, 102, 122, 124, 125,
168–170, 173, 174, 176–181, 129, 133, 134, 138–141
184, 187, 188, 190, 191, 196, participial phrase 138
200, 216 participle 12, 86, 129–131, 133,
monoclausal 118, 162, 165, 189, 231 138, 141, 143–145, 231, 233
morphological 21, 81, 87, 89, percept 1, 8, 13, 28, 160, 171, 223,
121–123, 136, 140–142 228 (See also concept)
morphological complementation 121 perspective 6, 9–11, 16, 17, 26, 29,
morphological process of 136 54, 88, 91, 92, 96–98, 100, 102,
240 Subject index
104, 111, 114–117, 120, 146, 151–153, 157, 158, 165, 166,
149, 162, 166, 194, 197, 200, 168–188, 190, 191, 194–196,
203, 217, 222, 224, 231 200, 201, 203, 213, 217, 220,
potentiality 1, 13, 20, 199, 223 (See 222, 223, 227–229, 231, 232
also reification) subject 1, 2, 7, 11, 16, 23, 33, 40,
presupposition 60, 186, 229, 230, 41, 52, 53, 59, 60, 71, 73, 77,
232 80, 81, 83, 85, 88, 91, 95, 96,
Principles and Parameters Theory 84 98–104, 106–108, 110–119, 121,
pro-drop 77 122, 124, 125, 127, 129–133,
profile 96, 97, 122, 123, 126–129, 135–137, 139, 140, 144, 145,
137, 138, 141, 142, 144, 156, 147, 149, 151–154, 157, 166,
161, 166, 172, 190, 193, 213, 169–174, 177, 180, 182,
230, 231 189–191, 194–196, 199, 200,
proposition 13, 15–24, 26, 28, 34, 203–205, 209, 213–217,
39, 40, 59, 60, 64, 65, 71, 73, 219–222, 227–229, 231, 232, 235
78–80, 118, 151, 157, 163, 164, subject-conceptualizer 191
175, 176, 190, 195, 196, 199, subjectification 106, 109, 120, 225,
200, 217, 222, 223, 229, 230, 232
232 (See also event) subjective 99–101, 104, 106,
psychological distance 6, 34, 35, 108–112, 115–117, 119, 148,
37–40, 43–45, 48–52, 56, 232 173, 176, 199, 211, 217, 222,
(See also conceptual distance) 224, 232 (See also objective)
subordinate 94–96, 98, 109,
R 111–116, 119, 135, 137, 138,
raising 66, 82, 88, 89, 95, 194, 224, 141, 142, 144, 149, 150, 152,
232 153, 156, 158, 161–163, 170,
178, 182, 184, 190, 202, 213,
S 214, 228, 231
scale 26, 28, 44, 100, 153, 161, 162, subordination 95, 137, 138, 141,
167, 168, 177, 178, 181, 182, 147
202, 203, 206, 222 syntactic complementation 136, 141
scene-setting 71, 80 syntactic integration 202
scrambling 66, 83
realis 13, 16, 20–22, 24–26, 28, 199, T
200, 223, 232 (See also irrealis) territory of information 38, 54, 55,
reification 1, 13, 123, 124, 126, 139, 232
141, 156, 157 (See also topic 10, 24, 27, 35, 44, 48, 49, 52,
potentiality) 60, 71–75, 77–81, 83–86, 88, 95,
sentence-focus construction 73, 74 132, 147, 200, 232
sentential negation 72 topic uniqueness 71
speaker 5–7, 24, 25, 28, 34, 35, topic-first principle 80
37–45, 48–54, 92, 95, 97, 98, topical 73, 83, 132
100–104, 106, 107, 108–111, topicalization 8, 71, 72, 81–83,
114, 117–120, 125, 131, 132, 85, 152, 160, 186, 195, 232
</TARGET "SI">
Subject index 241
trajector 95, 109, 112, 122, 125–131, viewing arrangement 6, 92, 97–101,
136, 138, 143, 145, 231, 232 104, 106, 107, 110–117, 161,
(See also landmark) 190, 222, 229, 231, 233
viewpoint 5, 7, 21, 37, 118, 143,
U 178, 182
ungrounded 125, 141, 216 (See also
grounded) W
wh-question 69, 70, 79
V wh-word 69–71, 73, 79
vantage point 6, 92, 97, 98, 100,
103, 104, 108, 112–117, 119,
169, 171–173, 176, 182, 190,
196, 231, 233

Complementation

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

Complementation

Uploaded by

Copyright:

Available Formats

<DOCINFO

TITLE "Complementation: Cognitive and functional perspectives"

SUBJECT "CELCR, Volume 34"

SIZE HEIGHT "220"

John Benjamins Publishing Company

of American National Standard for Information Sciences – Permanence

Library of Congress Cataloging-in-Publication Data